Because the price of electronics is so low today four- core Intel Atom processor cost 5 dollars and Allwinner A33 four- core cost 4 dollars and microcontrollers even most sophisticated ones cost less than 0,4 dollars a piece how these cheap prices could be used to manufacturing cheap devices for third world countries? It is not the cost of phone or tablet PC but the cost of a internet connection that prevents using internet in the third world. Cheap chinese phones at the price of 5 - 6 dollar factory sell out price can be used as cheap PDA (Personal Digital Assistant). These phones has graphics/ video processor so video games and movies can be watched at SD memory card, analogue radio is a standard feature and some have analogue TV (sold as “mobile phone with TV function”). So 7-8 dollar bazaar price phone can act as a game console a TV, a radio, a music MP3 player, a video player, a “computer” with educational etc. software etc. even without internet or telephone connection. Printed electronics is coming to real mass production. Roll printing makes possible almost no cost at unit- production and manufacturing scale of billions of units of some product. In India and China there are already large investments in roll printed electronics, so manufacturing cheap simple electronics there or if the tech is similar like printing books, country with book manufacturing can have own printed electronic plant. Price per unit manufactured would be very low so commercially sponsored products with no price to the end user are possible, and printed electronics products can be given away as free. Manufacturing displays, batteries, even a sort of magnetic memories and other electronic components is possible using roll printed electronics so ultra low cost television / computer monitor, radio- etc. equipment for third world is possible. Devices like hearing aid products are cheap bulk products already in China and elsewhere, but expensive ones like Bone Anchored Hearing Aid cost from 3000 euros to 4000 dollars. However with the cheap price of modern electronics BAHA device that cost only few dozen dollars is possible to build, using standard silicon chip processors that are cheap nowadays. And altough internet is not economically possible to majority of the population in the third world because that cost money, making free internet connection to end user with commercial (advertisement) paid, with limited access (data rate restrictions weekly / monthly scale) to internet and only for public domain / freeware material is possible. Perhaps video material is only for paying customers so text- based browsers like WAP are used, audio and text is left for cost free version, or even audio is forbidden in cost free net so only text- based information is in the free net for the third world. However, how much restricted the free internet connection would be, at least some kind of payment- free system in the third world for internet use must have. Because 4G networks are being made even in the poorest of countries, and 4G in there means few dozen megabits/ sec rate, that is however enough that several / several thousands of people share same “phone cell channel”. Every message received will be received by everyone in the same “radio cell” so messages must have some header code that only phone with proper header opens the message altough it is received with everybody in the radio cell. Now one telephone connection can be shared by thousands of people, which makes individual data rates very low but maximum data compression solves that problem. Using direct connection without base station like Motorola / Nextel iDEN / WIDEN or Sprint push- to- talk, or european OMA PoC (for that no phone manufacturer has ever made products of, unlike iDEN network that is in use in some South American countries, making possible telephones to communicate directly like radio telephones without base stations) and Gotenna etc. solutions for direct communication. Wifi ranges has many similar solutions but maximum range is 40 metres at wifi frequencies for phone to phone communication . Longer range is possible using athmospheric propagation. That is like Andrew Lippmann s viral telephone network or Ted Nelson s Xanadu. But if normal telephone and internet mobile network is using commercial sponsors and advertisement paid, like Jana / mCent app is doing, typical normal telephone / mobile internet network can offer free to end user internet connection in the third world. That is one suggestion for cost free internet connection for the third world. There are three businesses in the world that are mainly state owned and governmental control: oil business, weapons trade, and telebusiness. Oil is traditionally offered below budget price in those countries that have state- owned oil business. What if instead those billions used in cheap fuel that money is used to free internet for their citizens? State owned telebusiness is in around the world, in the third world some chinese state-owned company, or Orange (earlier France Telecom), or some investment company from oil rich middle east country (state owned) or Russian Vimpelcom (owned partially by Telenor Norway, which also owns Grameenphone), even TeliaSonera has telebusiness in Afghanistan etc. So third world telephone connections are under parlamental control at least partially, and for non-mobile internet telephone wire networks are mainly state-owned still in third world countries. So commercially paid by advertisements free internet for end user is possible to arrange using parlamental decision. What is possible to build with cheap electronics for third world market? Such as cheap musical instruments, or “educational tablet PCs” like Prasad in India or Cheertone in China, aimed for children, Personal Digital Assistants with educational programs but with lower prices than Cheertone 20 dollars and Prasad 40 dollar price today. Those have Android OS but no internet connection. And if the aim is to make cheap internet device / TV display / game console/ computer the best option is perhaps not to use graphical display at all because they are expensive, at least in tablet PC sized devices. Instead all graphical information goes to very cheap video glasses, and the tablet PC or phone has only simple touchplate in the place of touchscreen. Cursor is seen at the video glasses and same touch control that is in touchscreen is now seen as cursor in video glasses display. Using double click or double tapping the touchplate when cursor is in the right place in the video glass display is substitute for touchscreen display. Cheapest chinese video glasses were Elekworld / IVS VG260 model that was for sale at 2,5 dollar price at some chinese netshop year ago or so. So cheapest quality video glasses can be build with very cheap prices and with printed electronics even cheaper. The device itself is like ordinary phone or tablet PC but without touchscreen, instead it has simple touchplate from corner to corner. Some simple functions can perhaps have indicator lights, LED lights etc. or simplest LCD screen, and perhaps push buttons for primary functions. All other information goes to the cheap video glasses that are integral part of the device. Smaller and simpler would be “Android TV stick” -style device with few push buttons for primary functions and all other information is handled at the video glasses, navigation push buttons do the cursor navigation in the video glasses like navigation buttons in the ordinary phones without touchscreen. That would be even cheaper. If the phone is ordinary type with touchscreen display, making Google cardboard VR glasses type simply plastic support for phone that can be now used as virtual glasses is best solution for ordinary phones. Altough printed electronics is not suitable for large scale integrated circuits simple circuits like 4000 series CMOS or XR2206 perhaps can be built with it. Music synths using circuits like “Lunetta” type is suitable for third world market, made using printed electronics, almost no production cost at all. Altough if the synth is modular changing connection wires to some simple push-button type is best because otherwise connection wires cost more than synth itself. Microcontroller cheap modular synths like Midibox, OWL Modular, Lush projects One, Axoloti, Mako DSP, FreeDSP, Aquila DSP, Xibo RaspberryPi, Horus DSP and other. For devices there are Motus movement synth, Mogees app etc. Microcontroller synths with minimal prices (1-2 dollars) is possible to build for third world market with production runs of millions of units. Other Arduino, RaspberryPi, Onion Omega2 -style and other (Pine64, C.H.I.P) are also suitable for third world if their prices are small enough and they can be manufactured in large scale mass production. What other devices with ultra- low price are possible for third world? How to use modern cheap prices for electronics for the benefit of third world? What kind of devices should be manufactured? Using microcontrollers, both cheapest and more expensive ARM and Intel processors architechtures, DSPs, DSCs (Digital Signal Controllers), even old CMOS series 4000 and other chips with minimal price, what can be built? Even cheap microcontroller have a capacity of ordinary PC of 1980s, and their architechture (microcontroller) is based on some cases Apple PC, IBM PC or Amiga / Commodore and Z80 processors, so 0,4 dollar or less costing PC processor running old 1980s or 1990s programs is possible. Perhaps modern microcontreller with x86 architechture can run old first generation Pentium PC computer programs from early 1990s. Not only modern AMD Gizmo or Quark processor but some very cheap old 5X86 or 6X86 architechture microcontroller can perhaps run old Pentium computer programs from early 1990s, if this old style microcontroller is manufactured at modern narrow nanometer technique. And for 4 dollar price is available modern quad-core tablet PC processor. Perhaps for magnetic memory if printed electronics is used, manufacturing Richard M. Lienau s SHRAM (Sheet RAM) memory is usable. Manufacturing of devices should be in China, India, Indonesia etc. where cost is minimal because target price of product itself is minimal, perhaps less than one dollar street price if printed electronics is used, few dollars for microcontroller based products, and 5 - 10 / 20 dollars for phones and tablet PCs and other products for example. If TV sets and other displays can be built cheaply using printed electronics I don t know. And for other products: cheap “discman” CD / DVD players at china factory price is 5-6 dollars about, the player plays CD and DVD discs. In supermarkets are being sold DVD films at cardboard sleeve with price of 1,95 euros, but discount price of 1 euros. These DVD films have first been licensed to some local firm, that gives them to supermarket, and DVD factory takes its own profit for manufacturing DVD discs. So very little is being left for DVD copyright owner (film firm). There is also manufacturing technique Ecodisc, it makes single layer DVD (and CDs) s very cheaply. Because film firm receives very small amount of money from 1 euro DVD at supermarket sale in western country, the same firm can offer those DVDs for sale at third world even cheaper, if Ecodisc manufacturing, local DVD factories, and local distrubutors are used. And not only DVD films. Also music is cheap nowadays because of Spotify and other streaming services that offer music with minimal copyright payments to copyright owners, the amount of money generated is absurdly low for streaming an CD album from the internet using Spotify or other streaming services. So making the same amount of money selling Ecodisc CD in third world with cardboard sleeve and cheap manufacturing would bring same amount of money to music owner than that situation that someone in the third world listens the same album in the music streaming service. So ultra cheap music CDs and DVDs with films can be sold in third world, and that would help bring down piratism in DVD and CD trade there, because those DVDs and CDs are now sold in similar prices as pirated CDs and DVDs but now copyright owner gets royalty payments and not all money go pirated products (CDs and DVDs) like is nowadays case in third world. Much of nowadays music is self-owned, so copyright owner can himself order the amount of royalty payments per CD in the third world, and even large record companies would get the same amount of money that music streaming from the internet brings to them, in direct CD sale for customers in the third world, nowadays CD market in the third world is full of pirated products that rightful owners won t get any money. The cheap chinese “discman” players have either DVD or CD machinery, not both. Choosing only DVD machinery and then releasing music simply as DVD discs (12 or 8 cm) without a video is best solution, 8cm DVD with 4cm plastic ring to make 12cm disc even cheaper, DVD player can be used as watching films and for listening music. There are myriad audio and video standards, cheapest and no licence cost and simplest (for the cost of electronics) should be for the devices. As today s DVD players and portable music players are packed with different codecs (MP3, AAC, Microsoft) something like chinese DRA or ogg Opus for sound and chinese AVS for video (these cheap devices are propably made in china anyway) or libde265 or other effective like Google s video codecs but license free, so that single layer DVD can be packed with long video and audio material, if Ecodisc manufacturing is used. And using only one audio and one video codec for simplicity of electronics. One audio codec can handle both telephone voice and music content like ogg Opus, and the music “MP3 player” section can also be used as wavetable music synthesizer (Antti Huovilainen), and if several music codecs are used MP3 and AAC codecs can have shared internal circuity partally etc. to save costs. And using such as AAC filter banks or DAC:s delta-sigma modulator with ring oscillator/ modulator as musical instrument also. And for musical instruments: if DSPs are used then old (20 years old) Motorola 56000 or 68000 series or Texas Instruments TMS 320 or earliest SHARCs etc that are manufactered now in very cheap price, are an option, altough modern TMS 430 is cheap also. First “real” FPGA Xilinx Virtex 1 is from november 1997 about so it is only 19 years old. There are dozens of FPGA synths, from the netpage fpgasynth.beepworld.de/history or 96khz.org (Jurgen Schumacher) netpage. There are dozens of others too, almost all academic research projects. Available number of FPGA synths must be around 100 or more by now, but not commercial products, cheap ones, has never appeared based on these myriad studies. One FPGA can include dozens of different synths in one chip, so in this way FPGA is a cheap platform. The microcontroller- based devices itself would be small size like dspGplug or Mixtela synths within USB or MIDI cable, like Midi Vampire or Roman Sowa Midimplant, or such hand held products like Wrist Piano, swiss made Watch Out mini (Felix Bänteli) synth, and “Boowis-Kleinster synthesizer”, KeKePad, “Second Skin Synth”, Cornell University Aura gloves, and mi.mu gloves.1980s were “Coca Cola Coke music system” keyboard watch, and “Seiko Frequency” wrist watch drum machine. Nowadays “Click Watches” manufactures “Dip Switch” and “Turn Switch” watches with printed electronic boards. From these small and cheap wristwatch- like devices price range and device size can go up to 100 dollar sophisticated synth workstation with keyboard and many features etc. For video gaming 8 bit game consoles are already manufactured with only few dollar factory price in china. Also music generating instrument like Rhea Jeong designed Samsung “Colorsonic synthetic music device” or Giorgio Sancristoforo s “Tableau generative” that itself composes, using music generative algorithms simply with pushing a button “play”, with microcontroller design is an option. And how to use printed electronics roll printing to minimize cost? So what kind of different devices can be build using cheap electronics? For third world. And for more sophisticated products like phones speech synthesizer is a must because large amount people cannot read or write. Yamaha eVocaloid and old Micronas MAS 3515G had integrated speech synth, and so were discontinued Cyberworkshop Taiwan / Hong Kong sound chips. Also “Hidden Markov model speech synthesis” and Google s “WaveNet”. The user interface must be simple also with simple icons like “Food” , “Work” , “Clothing” so pressing that icon gives information on these subjects, and perhaps “info pack” with basic information on different subjects is in every phone or tablet PC. The price must be very low, 5- 10 dollars for phones and 10 - 20 dollars for tablet PCs, Perhaps phones can be made such a low price that they are commercially sponsored and paid like free net connection and given away by free. For other products perhaps less than one dollar dollar street price if printed electronics is used exclusively, and from few dollar - 5 / 10 dollar microcontroller products to 5 - 10 to 20 dollar “sophisticated” products, perhaps even 50 dollar or max 100 dollar “high end” products like sophisticated synth music workstation. Examples of microcontroller products are Piranha “soft synth” for microcontroller, and Gameduino and Arduino OctoSynth, and Goom and Groovesizer synth, a sort of microcontroller based music workstation with cheap components. There are myriad microcontroller projects but their developers are not interested in large- scale mass production. Selling license production rights to China or Indonesia or India to some electronic factory or simply itself transfer production to some cheap factory there and start mass production, instead of handcrafting one piece by piece microcontreller products and selling them at prices of over hundread dollars. Now third world people will get to use those products also and at price that is suitable for them. Chinese toy electric pianos like Peng Zhan / Peng Jia 168B cost 1,65 dollars and that price includes shipping to overseas. What if now replacing toy piano sound circuit with microcontroller and creating a “professional” music synth or keyboard controller from cheap toy piano? And manufacturing that product a million pieces like like toy piano. Manufacturing niche products in the western country is not so profitable as manufacturing large scale mass produced low price products for third world market. And for use of internet: Free net or similar should be as possible “semantic web” with “semantic computing” and “semantic desktop” etc, like MongoDB, OntoLinux etc. are using today, with semantic, ontological etc. databases and content. But if it becomes expensive solution etc. then perhaps not, but any solution for free internet is suitable if it is just simply cheap. And perhaps using roll printed electronic components such as old first generation sound chips like TI SN76489N, AY-3-8910 and different later derivates like Yamaha YM3439, YMZ284, YM2203 or YM2612 and Philips SAA1099, Atari and Gameboy sound chips. Also using roll printing some of the earliest speech synth chips from late 1970s/ early 1980s could be manufactured. Also earliest computer CPU processor from the seventies / early eighties with about 10 000 - 40 000 transistors or ARM 0 that has only 10 000 transistors can be build using printed electronics, cheap “computers” that have the capability of early Apple 2 or first IBM PC but cost is almost nothing and can be given away free as advertisement paid commercially sposnsored (the device itself is the “advertisement”) in development countries, production costs are 0,01 dollars per unit or something like that if production runs are billions of units, printed electronic display (black/white or other low quality), capacitave keyboard and perhaps solar panel and no batteries at all, all made using roll printing. There even are “soft synths” like Piranha and “Apple2 digital music synthesizer” that use these limited capabilites CPUs. And “The flexible sound synthesis using FPGA”, “VITA synth : a modular platform based on FPGA”. There were synth sound chips for phones manufactured by NEC, Oki, Rohm, Micronas, Macronix, Vimicro, Atmel SAM / Dream, Yamaha, and CS9236. These were mainly simple wavetable synths, late NEC and all Yamaha chips had FM modulation and some also have PWM. Only Yamaha and Dream remain in production. Even simply MP3 player circuit can be turned into wavetable synth as Antti Huovilainen has demonstrated. Even cheaper than these perhaps are old PC sound card synth chips from late 1990s, made by NEC, Ensoniq, EMU, Roland, Yamaha and many other manufacturers. Almost all old PC sound card models and types (from about 20 years ago) are still in production in China and other places, with their synth chips, and these old type PC sound cards are at very low price nowadays, factory prices are only few dollars or even less for large order of PC sound card of old design. These synth chips have PWM, FM and other features but require extra DAC chip etc. because they are about 20 years old designs. Sound cards are manufactured still today but synth chips from PC sound cards have completely disappeared, so old models are only available for OEM with synth chips. What if using cheap toy piano and putting old PC sound card or synth chip inside this toy instrument and create real professional keyboard that cost only few dollars and could be mass produced millions of units? There are also “real” synth chips for OEM production like EMU 8030. Some Yamaha chips have all-in-one solution where everything possible is integrated into single chip: synth, effects, DAC, headphone amp and loudspeaker amp of several dozen watts power in single chip. Some Yamaha chips have 0,9 Watt “loudspeaker”, is this loudspeaker amp or is really small loudspeaker with 0,9W power integrated inside chip? If so, not even loudspaker is needed if everything is in the chip, perhaps “horn” type structure that amplifies tiny 0,9 Watt loudspeaker voice is needed in the device. Printed electronics use roll printing and it is cheap but transistors are a problem, no efficient roll printed transistors are manufactured. Danish DTU university demonstrated roll printed solar panel and logic circuits in the same roll printed thin film, and logic circuits of 8 bits with thousands of transistors is manufactured and also DAC that works only in 4 bits is being made, enough to make clones of old 4000 series integrated circuits and other old semiconductor logic stuff. Something that reminds first computer CPUs from the seventies with their few thousand transistor count can perhaps be built, or ARM 0 core with printed electronics, but far slower than silicon chip ARM 0 core. Also altough full music synthesizer is perhaps too much for today s printed electronics, perhaps things like Theremin, Ondes Martinot, Trautonium, “Coupleaux-Givelet synthesizer” (from 1929) etc. can be manufactured using cheap printed electronics, simple electronic circuits, like BareConductive Touch Board. Already 20 years ago someone called John Rigg made multi-voice (12 voice) Theremin and Moog make 5- voice version and Midi controllers Air Piano (airpiano.de), “M!ltone”, and “MultiMultiTouchTouch” by Tim Thompson use Theremin principles. Catarina Mota Piano Box, Ototo by Dentaku or Digitalartsonline and Novalia are printed electronics projects. Cheapest phones or tablet PCs don t have motion sensors or even camera, outside device that has motion sensors and camera for musicmaking and other that connects to cheap phone, so apps like Beatfonic or LeapMotion or other 3D sensor and camera apps can be used, like Augumenta MARISIL or Innovega contact lenses video projection etc. And “Playing with constrants stylistic variation with a simple electronic instrument”. Midi controllers like Udar, Kyub, GEST, Hothand USB Midi, OWOW, Drawdio, Popboard Arduino foot controller, DEMO Midi control jacket, Piolin Raspberry Pi, Lumiphone, Takara Tomy (Air Guitar & Ningen Gakki), LightHarp (Garry Greenwood), Beatjazz Hands (Onyx Ashanti), FM3 Buddha Machine, Tapdrum (Urip Wisnuardi Indonesia),Karlax or Sylphyo, Tribal Tools Kadabra, SD Midi controller, Midify interface (division-6.com), JamBass Midi Controller that are unconventional and not simple keyboard controllers etc. Older generation controllers were Buchla Magic Flute / Thunder / Lightning / Wind, EMS Soundbeam, Max Matthews Radio Baton etc. There is also modular miniature synths like EmSynth Miniature Synth and Bastl Instruments Trinity and Patchblocks that are based on 4000- series simply circuits and microcontrollers. Also VST hosts like SM Pro Audio V-Machine, or Plugiator, could be build. And microtonal keyboard controllers for non-western musical scales, or H-Pi Tuning Box scale transformer. Also using printed electronics oscillators or waveform generators that are like old Exar 2206 or Intersil ICL8038 can be made, Maxim MAX038 has 855 transistors function generators are simpler than CPUs, have several megaherz range and can be used as oscillator for analogue sound. Multiple voices from single function generator are possible simply dividing 2 Mhz range to 100 x 20 kiloherz channels and using simply frequency splitter in 20 kiloherz intervals. Perhaps full polyphony is not possible but paraphony at least with 100 voices at 2 mhz analog or 4 megaherz digital if sampling rate is 4 Mhz. “XR-2206 function generator” is in vwlowen.co.uk/arduino netpage. Old AD9835 function generator is being sold less than one dollar price, full PCB board with AD9850 and associated circuits was for sale at 4,82 british pounds price at Hong Kong etc. So prices of single chip function generators (like AD9837, AD9838, AD9833) are cheap and these can be used as digital oscillators for cheap synths (analog ones) if not using dedicated digital oscillators or microcontrollers. Altough dedicated (old) synth chips or computer CPUs have very low prices also (like “real” digital synth chips Toshiba TC203C760HF-002, T6TJ3XBG-001, HG73C205AFD, and Roland R02782778, Toshiba chips were used 20 years ago in Tyros music workstation etc.) using ordinary silicon chip technology. Firms like CSEM and SEFAR are making electronic filters using printed electronics. Printed electronics oscillators are being also made. EDP Wasp / Gnat, SWTPC Psych Tone 1971 and Triadex Muse were early digital synths. There is also Hornet, analog/digital EDP Wasp hybrid synth. Even ARP 2600 and Arp Pro Soloist and other seventies synths were using integrated circuits extensively so cheap clones of them with printed electronics is possible with digital oscillators. Building EMS Synthi 100 clone that is super using Hornet components (printed electronics plus discrete components version) cheap replacement for super expensive synth. Nowadays synths that use old 4000- series and other simple circuits are Lunetta synths, CMOS synths, Thomas Henry circuits, www5b.biglobe.ne.jp/houshu, etc. Now there are manufacturers like Castle Rocktronics, Folktek, Ciat-Lonbarde, Gieskes and Error Instruments netherlands, Casper Electronics (Blue Box), Sismo synth brazil, sonodrome.co.uk, and circuit bending community is huge around the world. So products for printed electronics almost no cost versions of these devices are possible, because circuits are simple. Examples of products are The Mot-Box (Kinetik Laboratories), Critter and Guitari, The Nebulophone, Pimoroni Piano HAT, Hyve Touch, Stylophone, “Mini-modules: Minimoog DIY” (Julien Delgoulet) The Counterpointer (Luisa Pereira), Unearthed Circuits Drone Machine, Therapsid, Theresyn, GinSing Babblebot IC, Olegtron, The Cracklebox, noystoise.com, 4Klang soft synth, Nanoloop, Plogue Chipspeech CMU. Digital hardware can use 1 bit ADC like monobit, pipeline, flash, time-interleaved, tracking, and dual-slope, and modulation like A/DPCM, ODelta compression (patent by Gurulogic OY), Differential & Augmented Vector Quantization or Negative Beta Encoder. ADC can be used “reversed” as PWM synth sound source. With dithered signal 8 - 10 bits (60 dB) dynamic range improvement is achieved, dither noise is audible, but 60dB is enough for sound chips and even cheap and simple low bitrate audio codec for 1 bit logic: “Maximum likelihood estimation from quantized data” (Gustafsson, Karlsson). Pink noise/flat dither is “musical” so using it with music synth as part of synth sound is not annoying. Perhaps audio & video codecs with 1 bit logic and dither & logarithmic scale are possible also. Also printed electronics CPUs using 1 bit logic, dithered signal can have 15 - 16 bits precision improvement using noise shaping, and if dither signal becomes audible in signal chain 20, 22 or 24 bit dynamic range is achivied. So 1 bit logic gates that use dithered signal can have max 24 bit precision. Texts: “A single chip synthesizer generating a digital signal of tunable frequency with nearly no spurs”, “Turn DDS chip into low-frequency arbitrary signal generator” Munoz 2011, “Flash- ROM- based multichannel arbitrary- waveform generator” Woodward 1999, “Simple quadrature VCO covers two frequency decades with ultra low distortion” Gedde 2011, “Arbitrary waveform generator for 20$ -Instructables” for simple circuits for printed electronics. Old CEM and SSM digitally controlled analogue circuits for synths are also canditates for printed electronics roll printing. For ordinary silicon chip technology Anadigm FPAAs were used in cancelled Clavia Nord Modular “G3” about 2009, that used 4 Xilinx FPGAs and (4?) Anadigm FPAAs as filters, 128 voices polyphony that used (32 voices per one Anadigm chip?) five different filter types: Moog, Oberheim, Yamaha CS80 and MS20, and SSM filters. All five in 32-voice polyphony in one Anadigm FPAA filter chip that has price of about 6 dollars. Because Moog-style filters were used and are used today in many synths, Anadigm FPAA can replace all those filters. For printed electronics there are already factories in China and even India (Keentronics) that use roll printed electronics, so ordering mass produced cheap printed products from them would put cost to absolute minimum. For CPUs (Strong)ARM6- 9 or AMULET has 35- 111 000 transistors and ARM0+ about 50 000 minimum, or Hitachi SuperH now called J core and open source, could be roll printed. Malleable Technologies MECA Accelerator DSP, and configurable microcontroller ASIC made using FPGA block hardwired (economical). Roll printed processor has several square metres die area (manufactured using micrometer- class process) and is then folded like hankerchief to make it smaller, but optical (2 micrometer plastic optical fibre), neuromorphic, FPGA or FPAA & digital hybrid components, and approximate (Rice university) and reversible computing leads to small power consumption. ECL logic is fast and also low power versions exists, texts “Circuit idea / revealing the truth about ECL circuits”, “Software based Finite State Machine (FSM) with general purpose processor”. For silicon chip electronics there are projects like Novena platform, Nervana Systems, Numenta etc.If internet connection would be cost free at least in poorest countries of the world, people could participate in Loomio and other netpages, Loomio has Diaspora integrated in it, a social network that is distributed version of Facebook, and through internet people could get themselves “cognitive social capital” according to “Actor-Network Theory” (ANT) or Danilo Zolo s “teledemocracy”. And could do microwork or “advertisement watching work” and earn money, and watch openculture.com and MUBI.com free feature films. Felix Faire Contact is 3D enviroment for artistic expression, how that suits for development countries internet?
The sound chips mentioned on above text (“real” synth chips Toshiba TC203C760HF-002, T6TJ3XBG-001, HG73C205AFD, Roland R02782778) were for sale in china in alibababa.com and other places a year ago and at low price, only few dollars, except Roland chip that was 35 $. Now that is 40$ but another supplier sells it at 1,8 dollars (Shenzhen Winsome Electronics). I hope that is Roland sound generator chip, not Roland printer IC chip, it is R-zero (R02782778) not RO designation. Three Toshiba chips seems to have disappeared from production, (but perhaps will resume if someone just will start to order them, from china, indonesia etc). “XS725A00” which is TC203C760HF-002 in just different name is in Shenzhen Julixin electronics but shows “item not found” sign, so not in production (but if order will came perhaps production will resume). These sophisticated synth chips from about 20 year ago, used in Tyros music workstation etc,. now cheap price surplus chips at few dollar price, would offer sophisticated synth for just few dollars. It is possible to build 1980s / 1990s “synth collection” in one PCB board with very low price because those old chips that were used not only musical keyboards but in computer PCL cards etc. are still in production in form or another in low price. So compiling old Yamaha (inc. digital waveguide chips), EMU, Roland “linear arithmetic”, NEC (Casio) chips etc. together in one device is cheap. Even old Casio phase distortion synhesis (NEC D933 etc. series) chip was for sale in china year ago with keyboard scanning function. Now it seems to disappeared from production. So “classic” synth sounds from 1980s or 1990s are available at low price and in hardware using original chips, not software emulation. Also DSP chips are cheap nowadays, TI TMS320 is 1,95 $ at cheapest version, and at 6,30 $ quite sophisticated TI DSP is available. Sometimes DSP is coupled with (ARM) CPU in some processors, so perhaps Xenomai OS or similar that has Linux for CPU and RTOS for DSP is suitable for DSP sound synths and other applications, two operating systems coupled that have RTOS and Linux together (used in “music computer”, cheap tabletop mini-PC aimed for third world markets, now it would have soft synth platform ARM or x86 CPU and hardware synth DSP processsor in single chip). Earlier in the 1990s TI TMS320 DSPs were used in Clavia Micro Modular etc. synths, now similar will cost only few dollars. Also audio amplifiers that are integrated with single chip are cheap now. If keyboard or other instruments are manufactured to third world markets, including integrated amplifier in cheap integrated chip solution makes possible that no outside amplifier is needed. Texas Instruments (just one of integrated audio amplifier manufacturers) makes up to 600W audio amplifiers in single chip, both analog and digital integrated circuits. If every keyboard, or even guitar etc. instrument had its own 600 Watt or less internal inbuild IC chip amplifier no extra amplifier is needed in the band and instruments could be plugged to cheap loudspeakers directly. Yamaha makes synth sound chips with integrated amplifiers (aimed for consumer electronics like karaoke machines or TVs, but using them as keyboard musical instruments is possible also).
Not mentined are medical and agricultural possibilities of cheap electronics, printed electronics etc. Cheap medical tech in the form of printed electronics or microcontrollers or cheap processors etc. For example electronic thermometer manufactured with roll printed electronics and almost without cost per unit, that can be given away for free in third world and mass produced millions (or billions) of units. That is one example of health care cheap tech. Applications of agriculture also can benefit with cheap roll printed electronics and other cheap microprocessor tech. Third world countries need this kind of cheap (almost no cost per unit) tech that can be useful, health care / medical tech and agriculture are perhaps the most important sections of cheap electronics products and techiques but left unnoticed by me in earlier posts. PragmatIC is firm that has manufactured ARM cortex M0+ processor using roll printed electronics, at 1 micron process (transistors are 1 micron / 1000 nanometers wide). If microprocessors are possible to manufacture at 1 micron process using roll printing, that opens totally new possibilities of cheap microprocessors manufactured almost at no cost. Old x86 (Pentium) style processors from 1990s (Pentium Pro, Pentium 2 or 3), or old ARM designs from 1997 - 1999, old DSP designs, old FPGAs (or “coarse grained reconfigurable arrays”), microcontrollers, GPUs, GPGPUs etc. can be made effectively and almost no price. Only drawback is that because those old designs were made using 500nm, 350nm, or 250nm process, the new roll printed versions are 4 - 16 times larger and slower than their silicon chip counterparts. They are also aimed to be used devices that are charged with solar panel etc. and used in enviroment where electricity is rare, so mobile (low power) versions of these processors are needed which slowes down them even more. If solar panel (roll printed cheaply also) is used with these processors, perhaps small roll printed lamp (OLED or other cheaply manufactured lamp) is also given so small light can be used at nights in places where normally is not electricity. Other electric charging methods instead of solar panel like heat can also be used if they are more effective. Because these roll printed processors are large and slow efficiency improvements, like unum floating point units and other can be used. Unum is IEEE standard (unofficial) “extension”. And because unum offers accuracy improvement over standard floating point system using hardware and not complicated software accuracy improvement, that accuracy can be used to several FP numbers “inside each other” representation and increase information density even more, so that one (unum) FP number can have dozens other FP numbers inside itself (inside extented mantissa accuracy), these other FP numbers also other FP numbers inside itself, until accuracy potential is used, now “compression ratio” is hundreads or thousands FP numbers inside just one. In text “Hybrid hardware/software floating-point implementations for optimized area and throughput tradeoffs” 2017 is software/ hardware FP approach that can be perhaps used with unum etc. computing. Other ways to improve slow 1000 nanometer processor is to use manycore designs like Kalray, Adapteva Epiphany etc, chinese Loongson / Godson etc. that can use (almost) regular Linux or Windows. There are also designs that require special operating system and are not standard compatible, like “processor-in-memory” designs, Green Arrays processor, systolic arry processor, chinese UPU Harmony CPU/GPU hybrid etc. If they are more effective than standard x86 or ARM they can be used, perhaps in some special application. The Mill processor is Linux / Windows compatible and offer good effectiveness but is not released yet. If processor has big transistor count and it is manufactured at 1000nm roll printing, it would be very large. But roll printed big area but thin processors can be folded like hankercief so overall diameter is not much bigger than conventional processors. Processors must be designed or their design slightly altered so that folding is performed smoothly. One processor type other than ARM or x86 is Cell processor (Playstation 3). It was manufactured at 90nm process and has 240mm2 area so 1000nm version will be 120 times larger and 120 times slower 12 X 15 cm ( 4,7 X 6 inch) “chip” with only about 2 GFLOPS speed versus 200 GFLOPS of “real” Cell processor, and still consuming about 65 - 100 W power. But manufactured almost at no cost. Intel teraflops chip has 100 million transistors and 274mm2 area at 65nm process, so 1000nm version will 225 times slower and larger 25 X 25 cm (10 X 10 inch) and only 8 GFLOPS speed. Intel sigle chip cloud computer has teraflops speed, made at 45nm, 567mm2 area and 1,3 billion transistors. 1000nm will be almost 500 times larger, and slower, 53 X 52 cm about 2,5 GFLOPS. Folded to smaller size 4 X 5 foldings this “chip” has 13 X 10 cm diameter (4 X 5 inch). If Intel teraflops chip is manufactured with same area efficiency as Single Chip Cloud Computer perhaps 480 core version of it is possible instead of 80 cores, 600 million transistors at about 600mm2, same size and foldings like SCCC so it is about 50 X 50 or 60 X 60cm when roll printed at 1000nm process, speed about 50 GFLOPS, teraflops chip can be used in servers only because teraflops chip uses non- standard structure that has no operating system compability. Or using instead SCCC with Xeon Phi structure with 48 cores. That is like “poor man s Xeon Phi”, made at 1000nm process. Smaller roll printed ARM and x86 (mobile) chips that use less power can be used in hand held Personal Digital Assistant type device. Processor, display, and battery is all manufactured using roll printing, and solar panel also that brings electric power. If cost is 0,01 dollar per item that is 0,04 dollars together. Memory is also needed, but roll printing is not perhaps used because NAND flash RAM price is 4,8 dollars for 128 gigabytes according to dramexchange.com, or 0,037 dollars per gigabyte, so 300 megabytes is about 0,01 dollars. This traditional silicon memory can be used in hybrid conjuction with roll printed processor circuit as cache memory etc., slow memory speed is not a problem because processor is slow also. And use mask ROM or flash RAM cheap memory as main memory if it is cheap enough. Roll printing must use hybrid printing (ordinary silicon memory circuit printed to plastic surface together with roll printed circuits). If roll printed memory is better (cheaper) than traditional silicon memory then roll printing must be used also for memory. Mask ROM is cheaper than RAM, so using it, also in memory “card” or whatever outside memory this device uses. Using cheap mask ROM like old “ROM pack” of 1980s computers make cheap outside memory possible (together with RAM memory, if outside RAM memory is needed). Total cost (processor, display, perhaps only black and white display, keyboard simply roll printed also, no touchscreen perhaps, battery, solar panel, all roll printed, plus memory) is about 0,05 dollars, or even cheaper if manufactured billions of units. Audio is through headphones, headphones are made using roll printing also, acoustic elements roll printed, super cheap. Any electric contact surfaces are not with regular jacks but thin roll printed flat surface contact jacks, super cheaply made. 5 million dollars is enough for 100 million devices. These can be distributed in third world countries. These have no internet connection, no radio frequency or internet bus contacts, they are just personal digital assistant, game console, video / MP3 player type devices. Only way to bring information to them is cheap (perhaps roll printed) memory “card”, either RAM or mask ROM or both. These devices have large mask ROM memory and small RAM memory inside. ROM includes large “info pack” of healthcare, education, housing, clothing, governmental public information etc. Also educational section of device can be very large, like educational tablet PCs for children that are being manufactured in India and China. Better versions of these devices have internet (radio frequency) connection, but they are more expensive (“real” phones and tablet PCs, with “real” silicon chip CPUs and SoCs made with silicon wafers, but with roll printed displays, batteries etc.), and are connected to zeronet (cost free internet for development countries). Manufacturing cost of internet devices is about 5 - 10 dollars, but they also are commercially sponsored and no cost to customer, because commercially sponsored (zeronet) internet line costs more than these devices per year. Those cheaper PDA style devices can however be distributed to areas that have no internet connection, and they are much cheaper. Old abondonware (games etc.) can be used with them, like old VIC 20, Atari 2600 or Sinclair Spectrum games that have only few kilobyte or few dozen kilobyte memory requirement. (Software) emulator is needed so that those programs can be used. Using these almost no cost devices for entertainment and education would be great thing at third world countries. 1 micron manufacturing process makes possible very sophisticated processors to be made with almost no cost, only thing to worry is power consumption. Also optical components to make those 1000nm processor faster can be used, optical fibre core is 9 micron smallest standard, but 5 and 3 micron is also mentioned and firm Nufern makes 1,8 micron (1800nm) optical fibre mass production, and 1,3 and 1,55 micron is mentioned in “Tapered silicon optical fibres” and “Extruded single-mode index core fiber” 2005, so mass produced (plastic) optical fiber can be used as cheap optical connection together with roll printing to improve otherwise slow 1000nm processor performance. However license costs are much higher than manufacturing costs, so license that is granted free for these roll printed processors or then minimal price (0,01 dollars per chip), or without license at all (public domain or shareware and freeware processor licences) is needed. These roll printed processors will be mostly 20 year old designs anyway, and outdated because modern CPUs are made using 16nm process, 3500 times more area efficient. But roll printing can be made cheaply. GPUs, DSPs microcontrollers etc can be made cheaply. If efficiency must be improved, transputer (XMOS makes transputerlike microcontrollers), “systolic array” processors, “processor in memory” processors, VLIW-, CPU/GPU hybrid, old Coldfire processor/DSP (not just ARM or x86 processors), VIA is manufacturing x86 processor with low power and improved efficiency, roll printing it would perhaps make suitable 1000nm processor, and some exotic processor models can be used (if is any sense to use them, manycore/multicore and other exotic processors that small niche manufacturers manufacture use non-standard operating systems mostly, some manycore processor manufacturers, 64 -100 cores or more, use ARM cores, other use their own RISC designs that are non-standard). But efficiency is not perhaps the main point but cost, how cheap (or expensive) these roll printed electronics are. Cheap hand held PDA must be super cheap. These electronic products are perhaps for sale (or distributed free of cost by aid organisations etc.) only in development countries, not western countries at all, and specially aimed at third world market, roll printed electronics makes cheap products possible. Hitachi Super H or J Core is public domain core that uses ARM microcontroller code, other free cores are avialable (but is there programs and other software for them?). Enough programs, games etc. is needed so if old ARM programs are needeed only Symbian has those, x86 processor can use old Linux programs and abondonware from 1990s etc. Programs need to be freeware, public domain, shareware or abondonware also, or granted license that is free of cost for these almost no cost PDA devices. Memristors are coming (2018) to computers, so perhaps memristors can be roll printed also and increase efficiency of 1000nm slow processors. 8 bit game consoles are already made in only few dollar price at third world market in china, roll printed versions of them would be almost no price at all. Also utility electronics like health care and agriculture / food production can benefit from roll printed electronics. Power consumption of 1000nm electronics is thing to worry, but because those are slow electronics power consumption can be handled. Actually power consumptieon restricts powerful roll printed chips to servers or perhaps “home servers” can be used in private homes with electricity, acceleretor cards for cheap (printed electronics) PCs (these “accelerator cards” would be very slow comnpared to silicon chip, but they are used in printed electronics PCs that are slow also, if some other than ARM or x86 architechture makes possible fast operation, it is faster than ARM or x86 architechture processor, but can not be programmed like ARM or x86 processor, so it is limited to graphics accelerator card etc. duties) and cheap (printed electronics processor) tabletop PCs themselves. But electric power consumption makes futile the idea of cheap PC computer chip, altough processor is almost free of cost power consumption that user must pay makes the whole idea of almost free PC nonfunctional (or solar panels etc. must be large so that electric power requirement is filled). So perhaps roll printed CPU and GPU processors, DSPs, microcontrollers etc. are only used in mobile solar panel charged or otherwise charged devices where user can get electric power without paying for it, or otherwise small amount of electricity consuming devices. These devices can be given away for free, they are so cheap, paid by commercial sponsors or aid organisations. In india and china already exists PDA type children tablet PCs, no internet connection in them only internal programs and memory card slots. This educational / entertainment hand held “phone” (it is actually PDA without internet / radio frequency connection or any other connection than cheapest possible “memory card”) can be made almost no cost (if about 0,05 dollar is manufacturing price), but it will be useful for people who have no money to pay, and in areas where internet connection is not possible. In every village (or in every town if not in every village) can be “information loading station” where news etc. can be loaded by plugging memory card slot directly to another device (physical contact between devices is only way to share and load information, or use memory cards). PragmatIC is partially owned by ARM holdings, so cheap ARM licences should be possible. Even more cheaper would be 5 - 10 micron printed electronics, manufactured with inkjet printer with conductive inks or using almost standard book printing machinery. Because books are printed in third world countries also, same machinery that is used to print books there can be used to print electronics. 5-10 micron or more will make very slow processor. If 16 nanometer is current trend in silicon chips perhaps 16 micrometer is suitable for ultra-cheap printed electronics, manufactured at third world countries with book printing machinery or printed directly using inkjet printer. Perhaps ARM M0+ is suitable for large diameter process, it has less than 50 000 transistors. Other microcontroller - minimal transistor count but effective processors can be used, and microcontroller designs based on old 486 or 586 (Pentium) processors or 68000 (used in old Apple etc). Mass produced optical plastic fibre manufactured at 1,25 - 1,8 micron can be used to improve efficiency of long electric connections, if long connections are optical not electric. Displays (black and white perhaps, like electric ink), batteries, touch keyboards (not touchscreens) etc. can be manufactured with even primitive printed electronics. 1 micron (1000 nm) is needed to effective printed electronics processors that resemble late 1990s ARM or x86 PC processor models. Old processor had sometimes better than now standard x86 or ARM performance, like old Cray computer processors from 1980s/1990s that had floating point unit that was fast and using own Cray FP format but it was sometimes possible to get wrong results with FPU, or MIPS R18000 or R20000 processors that had different floating point unit also (but using standard IEEE), Apollo Computer PRISM was another old but effective processor. If manufacturing of them roll printed at 1 micron makes any sense, if they are more efficient than usual x86 or ARM processors, and if they have any usable software for general purpose PC. Or use modern design that is not x86 or ARM but more effective per transistor count, using it for special duties processor only perhaps if there is not enough software for it as regular PC. “Simplified floating point division and square root” Viitanen, “Multi-functioning floating-point MAF design with dot product support” Gök 2007, unum concept, “Hybrid hardware/software floating-point implementations for optimized area and throughput tradeoffs” 2017, and “Hardware-based floating-point design flow” 2011 (Parker) Altera FPGA are floating point efficiency improvements. In quinaplus.com netpage is “Qflop: an ARM cortex M0 floating -point library” that uses only 436 bytes so it can be inside cache memory of processor etc., and “Gal s accuracy tables methods revisited” 2004/5 is table lookup floating point method that is efective. Making optical computer using stacked mass produced plastic optical fibres at 1,25 - 1,8 micron will make very fast computer but very different of computers today, because no optical transistor or optical computer memory has been made. Roll printing plastic optical fiber to optical processor circuits is perhaps possible. TTA (transport triggered architecture), Dataflow processor, Synthesis kernel (Massalin) and asynchronous processors like ARM Amulet, Lisp machines from the 1980s (Texas Instruments), or lambda calculus direcly on the hardware like “Introducung the PilGRIM: a processor…”, and unikernel (Mirage), RAM machine, TRON / BTRON processor from 1990s, Unicore, or “NISD: a framework for automatic narrow instruction set” are other exotic methods. Not only building analog circuit logic signal processor or floating point unit or other ALU, but building whole analog computer using roll printing is perhaps possible. But there are no programs for them, so perhaps old Cray processor and Apollo PRISM based “microcontroller” designs can be made using roll printing, or use modern manycore microcontroller designs for roll printed products, like ePUMA, because only microcontrollers have such limited transistor count that roll printing is feasible. Modern manycore microcontroller design with most modern floating point ALU etc. can be made with roll printing. ARM and x86 designs however have programs that PC user could use, altough only earliest Symbian programs from early 2000s are available for slow ARM processors made with roll printing, 1000nm process makes such a slow processsor. How slow is modern VIA Nano processor at 1000nm roll printed, or latest ARM cortex A-55 processor if its transistor count is lowered removing some core memory transistors that processor can do without, and processor has couple million or copule dozen million transistors (10 million?). Because silicon memory is relatively cheap nowadays hybrid roll printing of ordinary silicon cache memory with roll printed other logic in processor is perhaps possible, making processor smaller and not much expensive than totally roll printed processor, silicon cache memory and roll printed parts are printed together to roll printed plastic surface. If mass produced optical fiber of 1,25 - 1,8 micron is possible to use, that can be used too, roll printed cheaply, even for making totally optical analogue or digital computer, but that is totally different from computers today. Also processor-in-memory designs like Venray Technologies and Micron Automata processor are possible to made cheaply, because the price of magnetic memory is low, 128gb “flash project price” is 4,8 dollars at dramexchange.com, so 1 gb costs 0,375 dollars. Because these PIM designs are using cheap magnetic memory as processor element instead of expensive CPU transistors, together with memory at least in some designs, using memory circuit cheap transistors (some PIM designs) as CPU processor that is 12 times faster than similar ordinary CPU in massively parallel tasks, cheap processor in memory computer made using 1 gb logic memory gates costs only 0,375 dollar using cheap and slow magnetic memory. Programming of these is different from PC s of today so software for them is needed, but cheap and reasonably fast PDA style device is possible to build at about the same price as roll printed CPU PDA, this another PDA uses silicon wafer made PIM chip as memory and processor and the rest is roll printed (display, battery, solar cell, touch keyboard, headphones). These two different (they require different software) PDA devices can also be used together so that they together are complete device so not displays etc. are needed to print twice for two devices if they have only one user. Also analog electric components such as analog signal processsors, GPUs, DSPs or floating point units can improve performance of otherwise slow 1000nm (or even 16 micron) electronic if roll printed instead of digital components. One way to improve slow processor is to use residue etc. number systems: “A residue number system reduced instruction set processor”, “A novel RNS-based SIMD RISC processor for digital signal processing”, “Direct residue-to-analog conversion scheme based on chinese remainder theorem” 2011, “Efficient converters for residue and quadratic residue number systems” 1993 Sturaitis. Inexact computing can also improve slow processor and decrease transistor count significantly, if processor is just designed so that program uses small snippets and when processing fails small snippet is rerun from memory (snippet is stored in memory until it is processed in CPU, when processing is complete another snippet takes is place in memory, if program snippet fails due to computing error it can be rerun again fast, not chrashing whole computer). Inkjet printers have 14400 dpi (dots per inch pixel) accuracy at least in Screen Jet 3100 and Epson Stylus, about 1,76 microns,and transitor can perhaps be 3,5 micron wide when inkjet printed in slightly modified inkjet printer. Book printing machinery can be 600 lpi (lines per inch) with 16 colour cells so accuracy is 9600 dpi, 2,65 microns, and some offset or gravure or flexigraph printing can reach 14 400 dpi. In third world countries electronics can be printed in place using inkjet printers and book printing machinery, price would be lower than using some western factory. But licensing cost of of almost no-cost chip is problem, chip should have very low license costor no license at all (Hitachi Super H series). Using asynchronous CPU like ARM AMULET would propably make chip faster, if not using 8-bit multicore or transputer etc. exotic solution. If 2540 dpi is 10 micron (10 X 10 micron dot) then 9600 dpi is about 1/4 or 5 X micron and 14400 dpi 1/6 or 4 X 4 micron dot. Also not only digital, but analogue electronics can be roll printed or inkjet printed, and they can be much larger, dozens or hundreads micron large process, and even square meter or more in size, like analog “configurable analog chip” FPAA (J. Hasler) that uses small amount of electricity and even square meter size chip is possible and can be folded to smaller size. High licence costs make cheap electronics for third world almost impossible, if manufacturing cost is below 0,01 dollar per chip but license much higher. Almost- no cost devices can be given away for free, paid by commercial sponsors and aid organisations in third world. Integrated chips does not need be licence free, but royalty payment free or almost payment free (0,01 dollar per chip payment?) so that they can be used in printed electronics in third world countries. Actually if manufacturing volumes are great licence of ordinary silicon chip about 1 dollars (Allwinner A33 SoC cost 4 dollars and Intel Atom 4-core CPU for tablet PC 5 dollars, couple a years ago, when they were new processors, so 1 dollar royalty payment for processor is realistic) if similar processor is manufactured using roll printing, it will be much slower and larger, and less useful, but manufactured at much larger quantities, so about 1 dollars for 10 000, 100 000 or million ordinary sislicon processors does bring same amount of royalty payments if using only 0,01 dollar payment per chip but manufacturing scale is million, 10 millions or 100 millions cheap but very slow roll printed same processors. So licence holder gets same amount of money as nowadays, altough processors are much slower and larger (hundreads or thousands of times perhaps if roll printed) and so less useful, but high volume manufacturing makes same amount of money as normal silicon chip licence/royalty payment. So 0,01 dollar royalty payment/ licence for roll printed electronics is realistic, concerning about thousand times or so less usefulness of roll printed chips. Profit margin/ royalty / licence payment can even be in par with chip size, hundread times larger roll printed chip than silicon version has only 1/100th license/profit price per chip, 1000 times larger 1/000th of ordinary silicon version etc. Large volume cheap roll printed production that is not possible with silicon wafer technology however brings about similar amount of money to license holder than typical silicon chip licence. However chips itself are very large and extremely slow compared to silicon chip version. Possible devices are personal digital assistant type, several different designs all about 0,05 dollar cost, third world market only, made 1: either western factory roll printed, (hybrid printing of silicon memory with other components and optical fibre components is possible, also analog roll printed electronic, memristors, even analog optical computer roll printed using mass produced optical fibre) or 2: inkjet printed or book printing machine manufactured in third world (in India, Brazil, Indonesia or Egypt and Nigeria perhaps are book printing machinery for fine grain gravure, offset or flexograph printing, turning these machines to electronic circuit production. Printed displays, even black and white, such as TV sets, or analog radios, batteries, touch keyboards but not touchscreens, can be made with much smaller accuracy and much coarse roll printing machinery than PC processors, memory can be roll printed or cheapest silicon version available, silicon memory needs hybrid printing tech with roll printed components, CPU needs roll printing with best accuracy). Western roll printing has 1 micron accuracy (PragmatIC), inkjet and book printing machine 4 microns, perhaps 5-7 micron transistor is posible. These devices have no internet connection, other than direct physical contact to another device, no loudspeaker, only headphones, no display, only Google cardboard virtual glasses style most primitive video glasses black&white roll printed, simple touch keybord, solar panel brings power, cheapest possible “memory card” can be used in some devices. All contact surfaces are roll printed flat connectors, not round connector jacks for memory card or headphone, video glasses etc. Different processors and operating systems 1: Linux and Windows (old, 20 years old?, Windows Mobile /CE? For free like Android), processor old Pentium / new Atom or Quark, or other x86 but not Intel, or other Windows compatible (Transmeta Efficeon, Elbrus, Kalray, Loongson, Mill processor). 2: ARM (old or new design), only old Symbian work in slow processor, Linux and Android perhaps but needs optimization for slow speed. 3: Processor-in-memory cheap silicon chip (Venray, Micron Automata), cheap and fast, perhaps made using cheap maskROM, but needs own operating system, same price, about 0,05 dollar per device, altough CPU and memory made using silicon wafer 4:Other (GreenArrays, Rex Computing REX 256, Knupath Lambda fabric, UPU Harmony CPU/GPU hybrid etc), needs own special OS. Speed needs optimizations so inexact computing (Rice university) should be used, 0,25% error rate in CPU and 8% error rate in GPU and signal procesing (computer graphics, audio, video, still images, 8% error propably makes bad audio quality but does not matter). Program must work altough small mistakes are inevitable (using error correction? non blocking data structures?). Asynchronous CPU or GPU is another possibility, but needs special OS so it is worser option than inexact computing, can be applied to Other than ARM or x86 or Windows compatible processors because they have different OS anyway, together with inexact computing. Device can have CPU + GPU combination where GPU is inexact and bigger than lesser inexact or non-inexact (standard) CPU, most transistors are in GPU not CPU, which is other way around than usual practice. Because several different 0,05 dollar or less costing devices can be made, video glasses and headphones are only visual / audio means, they can be changed from device to another so devices don t need loudspeaker or display. Some may lack battery, they must be connected to solar panel or another device s battery. Solar panel can also be one separate piece like headphones and video glasses, no solar panel is needed in every device. Second device gategory: internet receiver. These have analog and/or digital TV, analog radio (like chinese cellphones), radio frequency receiver for internet, but no transreceiver, so they cannot send information, only receive it from internet, like TV or radio. This is cheapest solution for no cost internet, they only can receive multicast internet information: internet radio and TV channels, and other internet multicast material. No subscription line payments or other line payments because they are only receiving devices. Near Video On Demand NVOD is old multicast technique, that can be used in internet also, for example Wikipedia has 5,5 million pages, if 0,01 second per page (no pictures, only compressed text) is used in multicast channel about 12 hours is time when all pages have been send. If user search Wikipedia interenet receiver has given information (page index) when page searched has its 0,01 second time slot in 12 hour time, device automatically downloads it when right page index mark has been received, in 12 hour time period (0-12 hours, avarage time is 6 hours to be waited). Several channels can be used in multicast transmission to speed up process, like NVOD. Multicast internet sends same content in same radio frequency to all internet users, all receive the same channel, but cannot respond. One frequency channel can send information to millions of devices. Several hundreads or thousands multicast radio, TV, and textual / graphical information channels can be on air at the same time. Data compression can be used. Internet receivers can be sold not given away, using cheap 90nm indian new silicon chip production plants for example, so not roll printed processors altough display, keyboard, battery, keyboard, solar panel can be roll printed. These have colour display, and internal loudspeaker and memory card slot for small, slow and cheap memory card. CPU perhaps 90nm manufactured in India? Or Chinese modern model.
Third device would be internet transreceiver, which is simply normal tablet PC or cell phone which comes with free internet (zeronet) connection. One person can have all these three kinds of devices and cheapest PDA style can devices can one person can have many, because they are almost no-cost made anyway. price is 0,03 -0,05 dollar for cheap PDA, 2- 4 dollar about for tablet PC or cell phone that only receives (“listens”) internet multicast transmissions, and 6-8 dollar for cheap zeronet tablet PC or cell phone. PDA style devices are sold i india and china, using version of Android OS, but these have no internet connection and are “educational tablet PCs for children” (Pradash in India and Cheertone in China). But using ultra cheap C-RAM / C-MROM (computational mask ROM) or printed electronics PDA for not only for children but for everybody as PC, game console, and video and audio player, but without internet connection. Internet receiver is like regular phone / tablet PC but no ability to send messages to internet, only listen. And no phone line connection of course. It can be made cheaper if analog TV and radio is not needed in device, if those analog transmission have transmissions in multicast internet also, so only multicast internet receiving is needed to see TV or audio channels (internet TV and radio and analog TV and radio as internet version). Because price of devices are low, 2- 8 dollars, simplifing assembly is possible if instead of several components (SoC, battery, touchscreen) are separately packed into PCB board and then this board is put inside phone and then put to plastic cover, all is wrapped up to one System-in-Package, Package-on-Package or Dual in line package. So the whole phone / tablet PC is one “microprocessor” that is covered in plastic package with touchscreen in one side and memory card slot and connection jacks points on the other side. Phone cannot be opened (because innards are covered in plastic, like PoP or SiP package) and PIN card put inside, so slot like memory card is needed for PIN. Phone cannot also be repaired if broken down, like microchip is not repaired when broken but changed to new one. Battery can be integral part of PoP or SiP package, just a hole in plastic casting, battery chemicals and parts poured in, hole closed with plastic, and now battery is just a part of phone s casted plastic cover. So phone or tablet PC is single “microchip” with touchscreen and memory card slot etc. connections. This can make cheap 2- 8 dollar device even cheaper when assembly is simplified. And cheap 0,03 dollar PDA can be made same way, but in this case device also have looks and the size of PC CPU microchip, but this microchip has rudimentary 6-9 key multitap or chorded keypad membrane keypad in one side of microchip and also connections to flat small connection jacks for video glasses and headphones, perhaps even small flat “memory card slot”. So this is in fact microchip with membrane keypad and connections, battery and memory is included in microchip package with CPU. This small couple square centimeter (2 - 4 cm X 2 - 4 cm?) microchip can be carried in pocket etc., and headphones, video glasses and memory card directly plugged to it. So additional package is not needed, the microchip itself is the device. If not new technology like memristor or 3D circuit design are used, improving efficiency by inexact-, asynchronous- or unusual computing (reversible-, chaotic- and stochastic computing) can improve (slow) performance, and also limiting cache memory to minimum puts transistor count down if chip is very large in size (printed electronics). Also using bipolar /ECL transitors or new transistors like NEC silicon germanium, IISc transistor, stacked GDF transistor, Intel 3D, MIT p-type, IBM double gate-, 4D transistor, subtreshold Schottky thin film-, nano particle Coulomb thin film-, Suvolta transistor, Avto quantum transistor, or using other semiconductors than silicon, or silicon mixed with them etc. New logic gate structures like “new 4-transistor XOR and XNOR”, “A new 6T full adder 2T XNOR”, “Transistor implementation reversible gate using novel 3…”,“Reducing the number of transistors in digital circuits using gate-level evolutional design”, " A new dynamic logic circuit design for an effective trade…", “A new state diferential design for for hybrid SET-CMOS logic gates”, “A new topology for low-voltage null convention”. In order to make cheap magnetic memory tape memory can be used, half of LTO standard tape 960 metres is 40 metres, if tape width is also halved from 12,7mm to 6,35mm data capacity is 1/4 or 1,5 terabytes not 6TB like LTO cassette. Double reel cassette size of 80-87mm X 107mm is large enough that LTO Accelis type tape 480 metres fits inside it. 8mm tape can be used if old 6,35mm tape width is difficult to find, Cassette thickness is only about 9mm using 6,35mm tape, and 6-6,4mm if using even narrower 3,81mm tape, same as C-cassette. Laptop PCs and even tablet PCs can use 6-9mm thin cassette, and even mobile phones. Altough electric motors etc. require power. So tablet PCs and laptobs would be like old cassette players and telephones like old Sony Walkmans. Mass produced magnetic tape, manufactured millions of units like old c-cassetes and videocassettes is cheap. But loading takes over 10 hours etc. tape has its drawbacks. But price of terabyte-class storage is super cheap compared to other memory types. Desktop PCs could use standard LTO tape with take up reel inside PC, but LTO cassette should be narrower like 16mm not 22mm like today to fit into home PCs. Even double reel cassette size 115 X 150mm can be used, same 960 metres tape than standard LTO. But for mobile devices even smaller “microcassette” that has LTO single reel structure (take up reel is inside mobile phone) and cassette size of only 76 X 75 mm is enough for 480 metres of tape. Tape width is 3,81 mm. Capacity is 15% of full LTO tape, or 900 gigabytes, and 23 terabytes if Fujifilm nanocube is used. Virtual reality needs lots of memory space, so cheap tape memory can be solution for that. Hybrid RAM / tape memory like hybrid hard disk storage with cheap slow RAM, 40 dollars is for one terabyte of cheapest and slowest RAM, combined with even slower tape memory can be used, faster RAM makes memory to function more smoothly with slow tape storage. Thinner version of standard one reel LTO cassette would fit inside tablet PC, 960 metres 3,81mm width tape and cassette is about 100 X 102mm wide and 6 -6,4 mm thick, but needs take up reel inside tablet PC. Also idea of analog memory to be used with analog electronics, electric or optical computer or analog sound processor. There are several analog magnetic storage technologies, like US pat 5504699A 1996 “Nonvolatile magnetic memory”, US 5339275A 1994 “Analog memory system” that is partly optical (?). If its optical storage it can be perhaps used like Blu-ray disc, but in analog form, “magnetic bubble memory” and “Improved magnetic information storage using return-point memory”. Optical analog computers can use optical analog memory. Computational RAM (C-RAM) is cheap way to make a CPU in its “DRAM-optimized” form. If 128 gigabytes RAM cost 4,8 dollars at cheapest (slow and large RAM), that means 0,0375 dollar for gigabyte and 0,01 dollar about for 300 megabytes. That is ultra cheap price per transistor, and if its possible to build CPU using DRAM-optimised method, price of that CPU is in the same class. Also mask ROM manufacturing technology can be used, making “computational mask ROM C-MROM”. C-MROM is not memory chip but CPU manufactured using cheap mask ROM manufacturing principles. It cannot compete with “real” CPU but at least it is super cheap, so even silicon chip technology can compete with printed electronics in manufacturing costs. If the price of cheapest RAM is 0,01 dollar for 300 megabytes, that is simeilar scale to printed electronics processor price if transitors are counted. Turning cheap RAM or mask ROM into CPU makes processor that has about same price as printed electronics but much smaller and faster circuits is possible. 40nm will be 625 times faster and smaller than 1000nm circuit, 65nm 240 times, 110nm 80 timesd and even 180nm 31-32 times faster and smaller, tremendous advantage when compared to printed electronics. Processors need new programs to work because DRAM-optimized C-RAM or computational mask ROM (it is not ROM memory but CPU that is made with simple transistors and layers of mask ROM factory production, much simpler than regular CPU but much cheaper) does not work like regular CPU, but advantage of great speed and small circuit space makes this worthwhile when price is about same level as printed electronics. Using inexact computing, asynchronous processor, unconventional computing, reconfigurable computing (HPRC, Xputer, iLAND project, PipeRench, CompactRio) etc. Also if silicon wafer size is 450mm or even 550mm instead of typical 350mm chips could be manufacured even cheaper, if 65nm, 90nm, 110nm or other old process is used. So cheap chips could be manufactured even cheaper, and large production runs (millions of units per order), even one billion per order. One way to improve slow processor is “abstract code compression” that means making program code data compressed some way, this is usually for embedded small processors but slow printed electronics processors can benefit from it also. Another way to make printed electronics better is lambda calculus in hardware, there are proposals like “Haskell to hardware” 2015 and “Professor Sherwood proposasl hardware based on lambda calculus” 2017 that when printed electronics is made using inkjet printing, instead of making program that work in some computer, making program that in itself is the computer is better. So different computer programs are lambda calculus computers that are printed using inkjet printing. SECD machine, CAM, FAM, ZAM, CLS, CCS, CEK, VEC2, even JVM (Java Virtual Machine), TyCO calculus, pi-calculus, Sequent calculus etc. are examples of working principles, “Fractional integro-differential calculus and its control-theoretic applications” 2013, and programming languages like OCAML, Hardcaml, Haskell, etc. make possible to that computer program is directly printed into transistors and integrated circuits, or then using hardware description language. Inkjet printers have 1,76 micron (14400 dpi) accuracy, so perhaps 3,5 micron transitors are possible to print, and laser printers 9600X9600 dpi (2,65 micron), so about 4 micron transistor is possible. One million transistor count like first Pentium processor, or 5-7,5 million like Pentium pro / Pentium 2 maximum, or 100 million transistors like Intel Quark with low power consumption, but processor is now square meter size. But size of processor is not an issue in printed electronics, but power consumption is. There are laboratory prototypes of inkjet printers of 100 - 65 nm accuracy specially made for printing electronics, but no commercial model is available. Narrow process inkjet & laser printing would make lanbda calculus hardware computers economical. So each computer program would be own individual hardware computer also. Also “Simultaneus sampling data acquisition architectures” 2015. “Introducing PilGRIM: a processor for lazy functional…”, “Pipelined graph reduction instruction machine”, “The Reduceron processor” (Naylor), “Programming a multicore architecture without coherency and atomic operations”, “Implementation and application of functional languages”. And one way to use electronics is heat pump system that takes heat from ground or air and for example heats water with that energy. Panasonic Aquarea is innovative heat pump system that takes heat from air and transfer that to water heating, and hot water can be used to warm houses in cold climate etc., and it works even in low temperatures. That kind of heat pump systems are suitable for third world countries, for household heating (if heat pump system uses air heating no water is needed in household heating and only air inside house gets warmer) providing hot water (in hot climate) or hot air inside house (suitable in cold climate) etc. “Energy harvesting cell phone”, and “passive Wifi” are systems that does not require battery or need additional energy for cell phone and other Wifi communication, altough backscattering cell phone works in analogue form, not digital. But passive Wifi is digital and can transfer 11 megabits / second. However distances are only 30 metres or so to Wifi link. But super cheap PDA style devices that normally does not have radiowave communication can benefit from energy harvesting (solar panel power, heat power such as body heat, or radiowave harvesting) to make possible radiowave communication without batteries or internal power without batteries, so super cheap PDA does not even need battery. But Wifi backscattering needs short distances so user must get near Wifi link. But super cheap devices that have radiowave communication for megabits per second and operate like Rfid tag are possible, and their price is about same as Rfid tag too (first class of of my classification of super cheap computer devices, 0,05 dollar about price range “chip computer”, CPU and other things in single small cheap package). If several of these super-cheap devices are chained to each other (by physical contact) only one of them needs to use Wifi backscattering, other devices can use that one together. “Alias-free short-time Fourier transform - a robust time frequency transform for audio processing” is efficient FFT form, but is aimed for audio only (Juha Vilkamo). Analog voltage controlled oscillators are used in some form of A/D conversion such as delta- sigma modulators coupled with VCO, or other VCO-ADC type solutions, PFM-ADC (frequency modulation), PWM-ADC, SAR-ADC with or without VCO etc. Perhaps analog VCO designs such as Harmony Systems inc. / Delora (delora.com) and Pigtronix VCOs that are exceptionally stable, or Malekko/Richter Oscillator 2, Zeroscillator (analog FM oscillator), several VCO designs by Rob Hordjik, or analogue additive VCO mentioned in electro-music.com netpage “Analogue gear news 9” 2006 (Bjorn J or BJ additive VCO that was in “synth-diy list” netpages) can make efficient A/D or D/A conversion circuit with VCO. Feedback amplitude modulation FBAM, and Vector Phaseshaping Synthesis can be used perhaps also with quadrature etc. DAC, delta- sigma or other. Some VCO-ADCs use “offset injection background calibration” technique, perhaps such can be used in analog music synths as DCO calibration. If internet connecting device only receives information and does not sent anything, so it is only “internet radio / TV receiver” MixRadio and Neverthink video (neverthink.tv) type solutions that are NVOD-style can be used, MixRadio used 300 channels and it was possible to skip track 6 times per hour. In NVOD style transmission 300 internet radio channels are possible, or perhaps 50 channels with ability to skip 6 tracks (6X50). Neverthink uses “curated video channels” or several channels with pre-set content so it is like NVOD (near video on demand). Cheap PC solution for development countries can be several cheap CPU chips overclocked and put to liquid cooler to about -50C temperature (even -90C is possible) using “refrigator” type cooling. For example 16 bargain priced 5 dollar Intel Atom chips stacked at 4X4 at PCB, three PCBs at top of each other, and fourth PCB with mid price GPU and cheap FPGA for computing acceleration (about 50 dollar FPGA), because shared FPGA - CPU computing can increase computing speed substantially in some cases. These four PCBs are inside cubic termos cooler (vacuum isolation from room temperature) and overclocked at -50C to -90C. This termos cooler can be assembled and plugged to normal PC board for example like GPU with air cooling fan. Using several cheap chips overclocked and put into cheap liquid cooler can be cheap but effective solution for efficient but cheap PC for development countries. Mid price GPU and cheap FPGA acclerator included. Price could be substantially lower than one expensive CPU and one expensive GPU.
Liquid coolants that are non corrosive freeze at -84 C, so -80 C is possible using “refrigator” cooling, and even -90 C or more without using liquid nitrogen etc, but coolants are corrosive. Cheap refrigator style cooler can be like GPU assembly card with air cooler and propeller in the PC, and about same size and assembly on the PC PCB like GPU with its air cooling unit for example, altough it is liquid coolant termos type unit. This cooler unit can be manufactured cheaply in some third world country, china etc., and inside it are cheap multiple CPU chips, cheap FPGA and cheap GPU overclocked. This is for third world market cheap PC. Concerns about internet security have arisen nowadays, already 1997 Tuomas Valtonen suggested different information network than internet which had “smarter technical core” and would be immune to such attacks that internet has suffered or at least cyber crimes would be much difficult to do. There is very little or no publications about this alternative to internet structure by him, altough “Security and privacy in a structured information network” 2004 Valtonen comes close (but it is about regular internet, not alternative to it). His internet replacement would be different to nowadays broadcast- based internet model, in his model information must be ordered and this someways makes cyber crimes almost impossible. Old TV or radio cannot be hacked but two-way communication makes it possible. In his model hacking is somehow avoided altough it is information network like internet and two way communication also, but not in the way internet nowadays uses two-way communication. However there is no precise information what this secure alternative to internet actually is by Valtonen, or I have not find it from his publications from 1997 onwards. However this secure and controlled information network would be ideal for low bitrate payment free information network in development countries (zeronet), because it is monitored and sensorized to prevent illegal material and also more secure than regular internet. Those who want anarchy and freedom of internet must pay for regular internet connection, but free to end user and safe against cybercrime network would be Valtonen s “smarter technical core” solution suitable for third world countries payment free “internet”. Also “Cell concentration in the autonomous error-tolerant cellular fabric” 2002 Valtonen is alternative to FPGA or similar structures. It can perhaps be applied to analog FPAAs too, (analogic blocks), not only digital FPGA-like structures. 3 D printing is promoted to developing countries, but is it more efficient than for example if some tool is needed, simply paying for skilled artisan to do the job (to make the tool) instead of printing it using 3 D printer is better way, and instead of building houses in third world using large 3 D printers simply paying for workforce to built that house would be better. Some applications like printing medical tools that are unavailable at third world countries etc. would be suitable for 3 D printing, but nowadys from china etc. is possible to order any kind of material and goods for low price so instead of making the same things with 3 D printer, is doubtful is it more economical. Making reprap machines using scrap metal parts from discarded copy machines etc., can make ecological and cheap 3 D printers however possible, together with ecological and cheap plastic filaments. About making cheap electronics: lowest FPGA prices are 3 dollars, 50 dollar is adequate FPGA about. So custom chips are possible. FPGA - based ASICs is possible to do and are sold in almost every FPGA manufacturer, base fabric is standard FPGA, upper metal layers are custom ASIC circuit. Price should be cheaper than “real” ASICs. Also no dscrete circuit analog electronics are needed no more in nowadays because analog “semi custom ASIC” manufacturers (there are dozens or hundreads of them) offer cheap FPAA, analog array, or semi custom chip (hardwired FPAA array) at cheap prices and complexity ranges from slightly over 20 transistors + opamps and resistors etc. to over million component FPAAs (Triadsemi VCA for example, from very simple to very complex is Triadsemi VCA range FPAAs). There is ALD EPAD (electrically programmed analog device), like EEPROM version of analog circuit, and it is easy to program and low price according to manufacturer. There are “mixed signal circuits” with both digital and analog parts from microcontrollers to large FPAAs, DSP (MSP 430 versions), Cypress pSOC, Maxim integrated PIXI, Microsemi Fusion etc. Manufacturing process of analog IC vary from below 100nm (as small as 16nm process is possible in analog circuits by MegaChips) to 24 micron by Universal Semiconductor semi-custom and custom ASICs. Prices are quite cheap and semi-custom or even custom analog ASICs are obtainable at small quantatities. So even eurorack analog module manufacturers could use analog FPAAs, small and cheap analog arrays , semi custom analog ASICs or even real application specific analog integrated circuit specifially designed for them. Analog music synth manufacturers (desktop models, models with keyboard etc.) use discrete components despite the fact that analog IC industry has hundreads of cheap analog array and ASIC products sold also at small quantities. Nowadays eurorack module manufacturers use digital processors which is contrary to nature of analog modular synths, but analog processing using cheap analog IC circuits is possible also, and cheap circuits does not mean Lunetta style 4000- series circuits but modern analog arrays and semi custom chips. There is even possible to build real analog ASIC cheaply, design tools are either free or cost about 1500 dollar cheapest for circuit design, and manufacturers like MOSIS, Europractice-IC or CMP offer real cheap price manufacturing in limited production runs. 350nm BICMOS analog ASIC IC at 1,25 mm2 area cost 1187 euros at CMP, that ASIC circuit can have less than million transistors, resistors, opamps etc. So it is possible to build 1000- voice analog synth with each voice having less than 1000 (several hundreads? 100-200?) transistors, resistors, opamps etc in VCO, VCF and VCA combined, in synth in chip solution. If production run is about 200 units price per chip is about 7 dollars. So 7 dollar analog chip that is specially designed to needs of analog synth manufacturer and contains 1000 analog synth voices and has only 200 unit production run or so is possible. Even eurorack limited production run manufacturers could use this ASIC design. So possibilites are pretty endless in analog ASIC market. Why analog synth manufacturers do not use this capacity, and instead keep building labour intensive analog synths with each transistor and resistor soldiered on PCB board individually etc. and analog synths cost 5000 dollars or more etc. With 1,25 square millimeter area chip is energy efficient too, perhaps 0,1 watts for synth voice, 100 watts in total. There are also mixed signal analog / digital arrays, (field programmable mixed array FPMA), “Field programmable array of analog and digital devices FPAAD” 2013, “Massively parallel inner-product array processor CIDDARM mixed signal parallel distributed array”. “DESA: A new hybrid global optimization method to analog integrated circuit sizing”, “Digitally-assisted analog and and analog-assisted digital integrated circuit”, “A powerful optimization tool for analog integrated circuits design”. Already 20 years ago or so Panasonic offered “Analog master” series of analog semi-custom ASICs. Production began 2016 of old analog synths chips (Curtis and SSM) using 8 micron process. These are mixed digital (control) and analog. Using digital circuits in 8 micron manufacturing is waste of circuit area, almost all digital functions could go to separate cheap microcontreller, that has D/A converter, only analog control information goes to synth chip (if that is possible and whenever it is possible). Only if circuit cannot be controlled other way than direct digital control digital logic can be in 8 micron chip. This extra digital microcontroller with DAC can be in separate chip or in same plastic package with analog chip. Or use real CPU to make digital control. Also ADSRs and LFOs can be digital and generated in microcontroller or CPU, only D/A converted information enters anolog chip (if possible). If needed LFO can be analog circuit modelled virtual, so nobody notices difference between real analog and digital signal. Also temperature compensation circuit is not needed in IC if chip is warmed up to about 60 C (VCOs), and heating can be done with separate heating element in PCB or inside plastic package of circuit itself. Also discrete music synth analog components do not need tempcos anymore if simple heating element is on the circuit board where transistors, resistors etc. are soldiered and that element heats components to 60 C about. Temperature sensors are manufactured by Moortec that makes analog ASICs too. More ways to improve is to use outside hearing range frequencies to send control information, if chip has 0-20 khz range 0-30 hz can be used to LFO (Alesis Andromeda uses this perhaps?) and 30-18 000hz to synth voice, and 18 000 - 20 000 to control information, analog control signals or analog telephone modem style digital signals, 2000 hz range that is filtered out of audio range otherwise. These signal ranges are separated in analog chip perhaps so only one 8 micron signal path is used instead of three for LFO, synth sound and control signal or signals. Several LFOs can be grouped also to frequencies over 18 000 hz and then separated using 30- 80 hz intervals etc, so 6 X LFOs use about 500hz from 18 000hz to 18500 hz, rest is for ADSR information , 1500hz available or cother control information, in single channel. Or use 2 -4 or 4-8 separate channels for LFOs and ADSRs but these are swtchable so they can both use those channels, if 1 channel is used to ADSR, LFO has 3 channels in 4 channel line, or if 1 LFO is used 3 channels is for ADSR. These 4 channels can go to VCO, VCF or VCA, they are routed according to switching information signal and when one channel uses ADSR in VCO “slot” LFO must use another channel to have VCO connection (out of 4 possible channels) and 4 channels are combined for ADSR and LFO for VCO, VCF and VCA, so 1 or maximum 2 connections is possible simultaneusly in them if all three are used to connect with LFO and ADSR (4 combined signal paths for ADSR and LFO together, and each one of them four can connect either VCO, VCF or VCA, so situation when VCO, VCF and VCA have each one connection of either LFO or ADSR, that leaves one extra either LFO or ADSR to connect with them). Also Electric Druid netpage has analog synth chips with filter connections. Those filter connections can be make internal inside IC circuit so no external discrete components are, to make complex filter connections straight inside IC chip, if the chip has synth in a chip complete solution or even without it. Also using reconfigurable IC connections like Serge modular uses using discrete components, where single circuit module can be used to 3 or even 4 different functions. This reconfigurable approach can be used to all analog or even digital circuit designs, however only in Serge modular I have noticed it. If it is possible to make analog circuit with 4 different functions if circuit is carefully designed to perform those 4 functions in single circuit, same can be used perhaps in digital circuits. Only reconfigurable computing uses two -way circuits that can be used in two differnet ways, but Serge modular uses about 3 or 4 functions in each (analog) module about. Making override in analog synth signal path so that digital CPU oscillators can be used instead of circuits own analog, or sending analog VCOs signal to CPUs digital filterbank instead of analog VCF is possible. VCOs of IC chip can be DCOs (digitally controlled) or autotuned analog, or “real” VCOs. Also new (digital) sound synthesis methods are: Trendline synthesis by Chris Pykett, Feature modulation synthesis (FMS) 2007, “Audio signal driven sound synthesis”, Waves organic resynthesis. Analog circuits can also be made using printed electronics (inkjet printing), 100 micron wide, and without poisoneus chemicals, but resulting transistors and resistors are several square centimeter wide, so they are larger, not smaller, than their discrete counterparts. However manufacturing is very cheap and can be done with modified inkjet printer. Resulting wide but thin circuit board can folded so that it fits inside eurorack module etc. So alternatives for making analog synths with discrete components are possible. There is 200 000 IC circuit and ASIC engineers in India, paying few thousand dollars for someone there to make analog ASIC music synth and then manufacturing it with cheap ASIC manufacturer with 1187 euro price production run like CMP is possible even smallest “boutique” eurorack or tabletop synth manufacturer. No need to spend rest of life soldiering iron anymore if cheap analog arrays, semi-custom ASICs or real custom ASICs are used, and even real custom ASIC is cheap. Anadigm FPAAs cost 6 dollar about a piece, and can be used as filters and VCAs, so only oscillators are needed to make cheap IC analog synth. If it has 1000 voices, for example 1024 voices in eurorack module with 16 outputs, 64 X 16 voices (64 in one output) and internal mixer so one output can have all 1024 voices if needed. This is just an example, even wildest analog synth circuits are possible using analog ASICs or arrays, and price is not high. If one synth voice is duplicated at 1024 times, the circuit is actually quite simple, and someone in India will propably do IC design in about 5000 dollars, 5000 dollars for ASIC production cost for production run of 1000 circuits, 10 000 dollars together, not bad for unique analog ASIC price. Amd up to 16 nanometers is manufacturing possible. 1000 ASICs instead of soldierin iron with discrete componete making hundreads of eurorack modules etc. is huge saving of worktime. Covering analog chip contact surfaces with gold or silver like analog plugs and wires are, would perhaps improve sound quality of IC chips. Even discrete component circuit boards if tin or copper wiring of them is covered with silver or gold layer would improve sound perhaps, like silver or gold is used to cover CD and DVD discs, gold or silver layering method for circuit boards can be similar to CD disc manufacturing, masket circuit boards to oven where small amount of gold or silver is put top of circuit board wiring surface, making “gold standard” boards. If analog circuit is made at 24 or 12 micron (like Universal Semiconductor does) or 8 micron process, that is very large, digital parts should be in minimum in circuit because any cheap microcontroller is better digital logic than 24 - 8 micron. 24-8 micron has cheap price and excellent analog quality. In India is soon analog semiconductor fab and another is proposed there, and in China there are more. Manufacturing coarse process 8 - 24 micron chips in large wafers , 450mm or even 550mm is very easy, even homemade VLSI circuits have 12,5 micron accuracy so 24 micron is cheap, if wafer is 450mm or even 550mm. Chips are large and slow, but if analog audio chip is the design goal, slowness does not matter. So analog ASIC, semi custom ASIC and other that is extremely cheap is possible to make using coarse 8 - 24 micron manufacturing and 450mm or 550mm large wafers and indian or chinese fab. That would be the first fab using 450mm or even 550mm wafers in the world. There are also very large wafers used in thin film transistor or solar cell manufacturing, so even larger than 550mm wafer is possible for coarse 24 micron process, but the larger the wafer is the heavier it gets and handling it and cooling it after it has been taken from the oven takes more time. It is possible to build even billion voice (1000 x 1000 x 1000) synth chip using very narrow tech (16 or 28 nm analog or digital), and designing it won t even cost much because one synth voice is just replicated billion times to wafer surface. Power consumption is several kilowatts if one voice consumes several millioneth-of watt power. So manufacturing must use power management IC technology so that current won t burn up the chip, nevermind how good it is cooled. Old 8 micron analog synth chips are cloned now, one way to improve them is taking most of digital logic away to outside microcontroller and replacing temperature compensation circuit to heating element inside package, and also making then multivoice chips, instead of 1-2 voice VCO or VCF or synth voice 4 or 8 voices in same chip is possible, multiple voices in place of digital and temperature compensation circuits in analog (mixed signal) chip. Making “Mighty Serge” synth in one IC chip and that chip has 9-16 Serge modules and those Serge modules are each capable of 2, 3 or 4 different functions would make compact but efficient analog synth IC chip, monophonic but when grouped to 4, 8 etc groups of multiple chips multi-voice Serge modular with IC chips is possible. Also sub-oscillators can be added to OSC analog chips, so outside discrete sub-OSC circuit is not needed, and filter designs like Chris Andrews “multiple identity filter” inside chip, not outside, is possible to build. Another improving would be cross modulating LFOs and ADSRs, digital LFO, two or three digital LFOs can crossmodulate to “3D” LFO signal and that is transferred to analogue form to analog chip, and two digital ADSRs can crossmodulate to one signal that goes to analog chip. Taking analog LFOs and ADSRs out of analog chip and making them digital with microcontroller or CPU generating them saves space again in analog chip, only D/A converted signal enters the analog IC, and now analog chip could have 4,6 or 8 voices in place of 1-2, altough it is based on old Curtis or SSM design. It would not be pin compatible to earlier 1980s chips but much modern version of them. Making analog chips using inkjet printing or laser printer is another option, but those circuits are still large in size and bit impractical to use in large quantities (transistor / resistor size is several square centimeters etc.). One source of cheap chips is old designs of more than 15 year old, that can be legally copied but if they have patent protected technology time is 20 years or so that they can be copied, or patented technology must not be used in younger copy. There are companies like IDMOS that make old circuits, altough almost all those old chips are manufactured in China. If analog synth has multiple voices like thousand or million, only way to manage them is virtual interface using video glasses. If modular synth concept is used multiple wires for multiple voices are difficult. Already in the 1970s were firms Polyfusion and Wavemakers specialising polyphonic modular synths, Wavemakers used semi-modular prepatched structure but used also wires that could be used to override prepatched connections. For ASIC manufacturing in India there are hundreads of firms, in netpage edaboard.com/showthread “ASIC companies in India” 2007 is list of them. Also: “Inexpensive analog isolation using digital isolation” 2012 at edn.com/design netpage and circuitstoday.com “Passive tone control circuit”, edn.com “The Wien bridge oscillator is reborn” 2009, and williamsonic.com “Developing a state variable oscillator for audio applications”. Sound synthesis methods like RayBlaster and CARL synth (Toggleaudio netpage) SPS “starfield parallax synthesizer” (parallax scrolling?). In beginning of text I proposed cheap PC with several cheap CPUs, a cheap GPU, cheap FPGA accelerator all together inside cooler. CPUs does not all need to be x86, x86 and ARM processors can both be together inside PC if they are cheap, now PC can use also Android like Remix OS etc. and not just Linux or Windows. also GPUs are included even cheapest CPU chips sometimes, these in-CPU GPUs can be used in mathematical duties, not GPU duties if one discrete GPU is included in PC package, so computing power is now greater. Analog gear can also be used to postprocessing digital sound, using analog circuits that are used in television sets for surround sound, dolby procesors, subharmonic analog processors (Peavey Kosmik Thing etc.), Three Dimensional Sound TDA2, Treia Sound processor, aural expander etc. analog gear. Even D/A converter like Wolson/Cypress or XMOS Audio Engine can be used as music synth, and Wolson A/D chip even as analog music synth because it is mixed circuit design. Old bucked brigade (BBD) designs can be used as delay / phaser / flanger in digital and analog synths and their price is cheap from China, one synth can have hundreads of BBD chips, one for each voice, and that would still not cost very much. Alesis Andromeda used analog chips that were much more sophisticated than old SSM or Curtis chips, if production of Alesis Andromeda analog chips would start again that would offer more possibilities than old 1980s designs. Curtis and SSM chips are back in production, why not Andromeda chips too, directly ordered from Alesis, or Alesis should make new products using them or put Andromeda back to production, in discount price. “LCD glass generation 10,5” uses glass (silicon) wafers of over 3 meter diameter, so semiconductor industry already uses oversize wafers, altough 3 meter wafer is impossible in photolitegraphy circuit like CPU and wafer would have enormous weight. But 450mm or 550mm wafer for very coarse 1 micron to 24 micron analog or mixed signal tech can be used to make very cheap analog and mixed signal ASICs. DEIAZ.com is another source of simple FPAAs from 20 transistors to about 600, and Flex Logix makes digital FPGA that can be bought as IP core that can be used as computing accelerator and it has won technology awards, like Achronix FPGAs. But instead of large wafer could be small wafer down to 1 inch (25mm) size. As modern 14nm tech is expensive and 10nm, 7nm and 5nm is coming, cheap manufacturing that does not need to be billion dollar factory is needed. 25mm wafer would be 12 times smaller than 300mm, and photolitegraphy machinery would be much smaller too and so cheaper, and factory would use workforce not automation (switching from 300mm wafers back to 200mm is partly because 200mm factories ndoes not need expensive automation but use cheap workforce from China and soon from India). 25mm wafer would fit about 10 billion transistors in 10nm process in same percentage of wafer area that now has 30 billion transistors (Cyclone FPGA chip, largest processor in use) in 32 nm process in 300mm wafer, but those 10 billion transistors will be 40 times faster than 32nm version. Manufacturing speed of factory using 25mm wafers could also be faster than using 300mm wafers if counted wafers per day production, altough overall chip volume would be smaller. As photolitography costs increase in smaller nodes and dominate cost of production of chips in smaller nodes, smaller photolitegraphy equipment could be much cheaper in 25mm than 300mm diameter. Also analog IC now is 16nm narrow, making analog IC at 10nm, 7nm or 5nm is perhaps possible if wafer size is 25mm or 50mm or more, the narrower the wafer gets more accurate photolitography is possible perhaps, and anolog chips could be made at same narrowness than digital 300mm wafer. Also wafer scale 24 micron analog chips are possible like Universal Semiconductor that uses 3 inch wafer and 24 micron manufacturing, altough 76mm scale chip is huge it is manufactured cheaply so propably low cost per wafer scale chip. However power consumption is perhaps kilowatt class. But no need to discard badly functioning chips in music synth chip production if one synth voice is duplicated thousands or million times to one IC chip, altough some of those synth voices won t work chip is still usable if majority of synth voices work, so those chips normally discarded in semiconductor chip production can be used, lowering price still. Same goes for digital and analog synth multivoice chips with huge polyphony. Altough target polyphony is not reached (from 1000 or million voices etc.) due to partly dysfunctioning circuit, chip is still usable. LCD glass generation 3 is 550 x 650mm, and even 3 meter is used in G 10,5 LCD. From 550mm to 3 meter scale IC wafer using folded optical path like diaprojector and optical aberration correction in litography can be used to manufacture 3 meter wafer photolitography in 24 micron process or smaller, manufacturing would be combination of normal IC chip litography and LCD manufacturing process in LCD display manufacturing factory. When IC chip is packed in chip scale package / wafer scale package, some of those packing methods offer “unlimited pin count” and if “chip” is wafer scale almost unlimited pin count is needed, also for million to billion voice synth circuits. Not one pin per voice but thousands (or millions?) of pins however is possible using chip scale package in IC chip. But printed electronics is coming and making mass produced electronics at same narrowness (micron scale) and much cheaper and enviromental friendly. Printed electronics can also be used to make cheap analog components, altough present silicon IC narrowness is not reached yet, making old transistor arrays, OTA arrays and similar simple old IC circuits at “new old production” but printed versions of old ICs with small price. Those old but discontinued OTA arrays etc. are in need to music synth building community etc. and can be used inside eurorack modules and other. Old 4000-series chips can also be done with analog printed electronics. Altough printed electronics made, these new old chips are perhaps packed in plastic like old IC circuits or at least are pin compatitible with them. “Mini-modules: Minimoog DIY” by Julien Delgoulet seem simple enough for printed electronics made. So Minimoog clone made at few printed electronics analog plastic sheets and price of few dollars is possible. Studio Electronics Boomstar has one standard VCO but different filters, similar system using Anadigm FPAA is cheaper than discrete filters and VCO even perhaps analog printed electronics made. “LVDT oscillator / demodulator design and build” 2014. Making music synths with “universal oscillator”, standard digital oscillator that is made using cheap x86 CPU, cheap ARM CPU (several ARM chips like those in Arduino, Synthduino, Raspberry PI, Zynthian, C.H.I.P, so at least Allwinner A33 and A13 is needed), FreeDSP chip, BBC Micro microcontrollers, and other open source circuits that can be used to music synthesis. FreeDSP chip is expensive (15 dollar if ADAU1452) so not perhaps needed. Microcontroller chips (Mako DSP, BBC Micro) can be made using printed electronics. x86 chip can be VIA Nano, Intel Atom or chinese new cheap x86 CPU. ARM chips could be more than just two (A13 and A33), A80 etc., and more than one x86 CPU for greater performance. MSP430 DSP is cheap and open source (Olimex) and partially analog in some versions. A/D conversion could be Wolson / Cypress DAC that is partially analog too. So even “analog synth engine” could be made using them. MSP430 is 0,54 dollar cheapest, so several of them (8-16) can be used or semi analog version of it at 5 dollar / piece. Even Coldfire or SHARC or Blackfinn processor can be added (Phoenix board, Use Audio Plugiator, Monome Aleph) in more expensive versions. This universal oscillator can be its own (multivoice) digital synth engine, or virtual analog / digital VCO / LFO / ADSR generator for analog synths, even in eurorack form. Open source software all through except x86 chip, but it uses Linux, and if user has Windows PC also Windows programs can be loaded to x86 processor. Using circuit boards from cheap used or broken iPads makes possible to use iOS programs. This universal synth engine that is open source can be in almost every synth as additional feature, if price is cheap. Universal filter could be Anadigm FPAA analog circuit. That FPAA can be added to any analog or digital synth as standard feature to complement its own analog or digital filter, even in eurorack modules. Anadigm FPAA can imitate multiple filters like Roland, CS80, MS20, Oberheim and Moog. Japanese synth manufacturers have sold professional audio keyboards as “home keyboards” or “toy instrument”. Using supermarkets and their toy departments instead of music shops as selling channel for very cheap 25 - 37 key small versions of analog and digital synths with cheap price can bring millions of units of sales with text “recommended for children of … age” altough they are pro instruments. Yamaha made DX100 that had small keys for children but was pro instrument. Critter and Guitari Organelle is example of simple ARM-processor Pure Data player. When old analog chips are rereleased cheap clones of old classic synths are possible, small keys for children and “toy instrument” supermarket sale would be success for them. Alesis Andromeda is build using few cheap IC chips, 25-key 16 voice small toy- Andromeda with cheap price would be success. Even Yamaha CS80 cloned with printed electronics voice cards or IC cloned, keyboard control modern microcontroller based, small size suitable for children. However it is pro instrument. Using roll printed tempcos outside chip but inside plastic package can be used instead of heating element in IC. In Nigeria TV sets have been made since 1960s and new electronic assembly plant opened. Cheap assembly of electronics is available in Nigeria, India, Malaysia, Mexico etc. Even eurorack module can be OEM made in India etc. using discrete components, sold at supermarket, if cheap enough. In Mexico Hard-Mod makes cheap discrete component cheap synths. If wafer is smaller, not larger than 300mm that could bring savings to IC production. Even 25mm wafer can be used, optical ultraviolet litography is very expensive at narrow nodes, 32nm - 5nm, but narrow node can pack large amount of transistors in small space and they are fast circuits. So if wafer is 25mm only, then 1/12th optical diameter is needed in optical machinery instead of 300mm, making cheaper optics and the narrower optic diameter goes more accurate it gets. So cheap 10nm, 7nm and 5nm circuits are possible, without billion dollar investments in machinery. Nanoimprint litography is cheap and 10nm narrowness is possible, and no large optical system is needed. Using LCD display factory and generation 10,5 glass size wafer in combination with nanoimprint system, ICs like CPUs could be made even from 3 meter wafers cheaply. Not only nano but microimprinting 1- 24 micron wide analog or mixed signal ICs using wafers of LCD glass generation scale up to several meter diameters. If 10nm is possible nanoimprintingin 300mm wafer, 110nm or narrower is possible using 3 meter wafer nanoimprint. Using 450 or 550mm wafers with nanoimprinting, or even regular optical litography, 90nm in 550mm wafer needs only about 1/4 optical accuracy of 14nm / 300mm size optical system and volume of production is 3,4 times greater than 90nm / 300mm. Using cheap mask ROM manufacturing techniques not to build memory circuits but simple and cheap CPUs brings silicon chip prices to roll printed electronics price range. Hierarchical modulation with Vector Phaseshaping Synthesis or Feedback Amplitude Modulation can be used with QAM or D/A conversion etc. Hitachi, TSMC & Mapper (for 450mm wafer), Multibeam, Mycronic and eASIC have developed cheap electron beam litography. Proton beam & direct laser writing can be used also. Fujitsu has FRAM LSI ASIC.
Altough already couple years ago roll printed electronic tech reached stage where simple devives can be made, including hybrid printing where cheap (small memory capacity and slow) ordinary VLSI memory components can be printed with roll printed logic corcuits, no products have surfaced. Roll printed electric thermometers, pocket calculators, Nintendo Game&watch style handheld game consoles, other old 8 bit game consoles from 1970s / 1980s, electronic logic for portable DVD / CD player or portable media player that uses memory card as storage. Those video codecs must be early 1990s MPEG format or other computationally simply video codec, and audio codec in portable media player simple also. Roll printing can make 10 micron or below manufacturing tech digital circuits, so they are slow, and MOS not CMOS etc. If display is printed also or it is electric ink or other cheap type, and battery printed also, device is really cheap. Roll printed thermometers, handheld game consoles, calculators and roll printed wristwatch logic are relatively simple to do. Portable media player (video/audio) more complicated. Those circuits can be grouped to telephone- lookalike PDA, simplest is thermometer, watch and calculator coupled with flat-file personal digital assistant memo, no operating system needed, only simple logic circuits. No video screen needed, simple LED-lookalike screen with number / alphabet display. Simple handheld game console can be integrated in, with Game&watch type LCD- lookalike display. More complicated version looks again like telephone, but has all those previous features plus media player (simple video/audio codec) and place for memory card and rudimentary video screen (black&white will do if colour screen is too expensive). Making telephone -like media player / PDA with roll printed parts makes manufacturing cost perhaps 0,05 - 0,1 dollar. So it is ideal for third world market. Earphones with roll printed parts can be included. And cheap video glasses with roll printed screen and logic. Also clones of old video game consoles can be made roll printed, their electronics is so old (1970s and early 1980s). If roll printed electronics need to be nearer to nowadys standards, large roll printed processors with square- meter size, and designed to run hot, high clock frequency, those macroprocessors cost almost nothing to make, they are designed to be folded like hankerchief so they fit in small place, and then they are put into liquid coolant, water or other coolant, liquid nitrogen or helium. They are designed to have very high clock frequency so they are use so much current that circuits almost melt, but they are inside very cold liquid coolant cooler. Altough those processors have micron (1000 nanometer) or more manufacturing MOS tech, they have such large area and such high designed current that they can be as fast as modern ordinary silicon processors, but consume much more electricity. They cost almost nothing, but have much higher electric consumption. Simplest digital integrated circuits are TTL, 7400- and 4000 series circuits from 1960s old MOS versions not modern CMOS. They were made using 20 - 10 micron tech, so are possible to make with roll printing today. With them can be build Lunetta- style music synths and other similar electronic projects. Large concentrations of those ICs can be in single roll printed sheet as Lunetta synth or other device, this roll printed sheet is large single-piece “IC”. Again suitable for third world market (minimal price). Another way to use roll printed electronics is dedicated music tech circuits. Not only early digital circuits from early / mid / late 1980s, but roll printed analog circuits also. Digital circuits can be used in “playable picture”, roll printed picture of Yamaha DX7 or old Casio, Korg or Roland 1980s early digital synth, that picture produces sound when picture of keys are pressed and changes modulation when picture of knob or slider is touched. Price is minimal, but features are same as those old 1980s synths. Again for third world market. Memory card slot for memory, or internal memory, and MIDI connection is needed in “playable picture”, or MIDI can be left without if price must be even cheaper. Also old computer sound card synth chips (Yamaha, Roland MT-32) can be made. Old arcade machines not just home game consoles can be cloned using roll printing. Old computer processors and old computers from late 1970s / early 1980s can be cloned in roll printed single sheet (whole computer circuitry is in single roll printed IC) like Apple 2, Vic20 or Commodore64, now also rare early PC computers can be cloned easily in single circuit. Roll printed analog circuits can also me made, they do not need to be integrated circuit size, they can be in same size like their discrete analog versions. So they are roll printed discrete analog components, but can be grouped in single sheet as single very large analog IC. Again cloning old analog music synths comes to mind. Moog Emerson modular costs 150 000 dollars, and it is monophonic. One dollar printed version of it as “playable picture”, it is full size picture of Emerson modular, inside this picture are analog components that produce sound in large picture-size IC sheet. Sound quality is not so important, analog roll printed sound can have noise and drifting out of tune, it can have bad sound compared to original, and voltage / current / ampers / watts may be different than original, main point is low price. Also digital sound chips when roll printed does not need to be same voltage / current / ampers / watts as originals, main point is that they produce sound. Cheap clones of analog synths, including large modular systems in single sheet IC, and smaller polysynths, can be made, and if sound is bad more modern IC analog circuits can be mixed with roll printed large one-piece IC with roll printed components, Anadigm FPAA, newly produced Curtis and SSM chips, and IC opamp and transistors arrays. “World s smallest computer” is IBM and then Michigan university made, and similar is Intel Quark. Those 0,1 dollar microcontrollers can be put inside those new production analog IC analog sound chips (Inside its plastic package), they have processing speed of old PC and low power consumption, now each analog sound chip can have extremely sophisticated digital control, analog oscillator chip can include softsynth digital extra oscillators and other control, analog filter chip its own digital softsynth oscillators inside chip etc. Because sound is analog if its signal path (digitally controlled) oscillators and filters are analog, but modulation can be digital and still it is called analog synth. Envelopes / ADSR can be digital, and LFO (if virtual analog LFO, sound can be made exatly same harmonics etc as analog), and white noise generator also, because nobody notices the difference. Modulation is usually ring modulator or amplitude modulation, and can be digital in analog synth if for example Moog synths are full analog altough they have digital modulators. But more digital modulation can be in analog synth, FM modulation, phase distortion modulation which leads to Vector Phaseshaping, and Feedback Amplitude Modulation, but that may require digitally controlled analog filter in “analog synth”. All those may be used digitally in analog synth, modulators are digital components. Roll printed analog components can be digitally controlled, DCO and filter also to prevent drifting. When analog sound IC silicon chips are now again available, coupling them for example 16 voices in single IC chip package, like processor cores are grouped or even several processors (Epyc) in single chip package. If one voice analog IC VCO or VCF costs 1,6 dollars, 16 of them in single package is not so costly, or if they are dual voice only 8 is needed in one IC package, additional circuits like “multiple identity filter” design (Elecric druid) in VCF etc can be included in IC package, and microcontroller or Intel Atom CPU for digital control of analog voices. All 16 voices, additional circuits and CPU inside same plastic chip package with large pin count. Circuits that work like Serge modular, where one circuit has many functions, can be made in IC or roll printed form. Modular synth that has pin, DIN, XLR, VHDCI or SCART type connector in module cables so one cable can hold 6-16 voices or more in modular synth not just one, or simple 4-voice cable that is simple to use can be used in modular polyphonic synths. EML Polybox or Ethenvar Patch Chord can be used to expand (poly-) paraphonic voices up to 25 X of polyphonic voices, up to 100 or 1000 or more voices using Patch Chord, integrated circuit is needed to make 1000 or more -voice Patch Chord. Alesis Airsynth and Roland D-beam used gesture control but suffered from too sudden changes in sound using gestures, russian ANS synth was uncontrollable due same reason. Erkki Kurenniemi in his harmonics theory solved this problem (Dimi H). Plastic foam controller, ball or brick -shaped that produces 3D shaped sound when it is squeezed, or hold in mouth like babys dummy so both hands are free in keyboard, or soft surface instead of foot pedals that shape sound when they are pressed with foot can be used. Aalto Flexible interactions and devices FLINT, Network Of Intelligent Sound Agents NOISA. Wablet scanned synthesis for music can be used. If digital roll printed electronics is used to make CPU, in industrial use or other use, new processor architectures, like manycore, Mill processor, XMOS transputer, Microsoft / Qualcomm EDGE E2 (improved RISC processor), Rex Computing, VISC processor, asynchronous- or one bit (serial) processor like 1970s PCs and 1980s supercomputer one bit processors etc., that increases efficiency and lowers electric consumption can be used, and operating systems like “the COSA operating system” or CoreOS. Processor can be streamlined using inexact computing (Rice university) and “bespoke processor”. Ferroelectric memory is made possible by german firm, but that is CMOS tech not perhaps suitable for printed electronics, altough it can be used as memory if hybrid printing is used. If new processor architectures cannot surpass already established CPUs, roll printed processors need every way possible to make them more efficient, they are so big and slow, so new processor architectures can be commercialised using printed electronics. Another candidate for roll printing is FPGA, FPGA prices are so high that FPGA or other reconfigurable logic with cheap price is needed. Roll printed FPGA would be extremely slow, but at least cheap. ASICs are other way to use roll printing. “Endless mini PC” is PC for third world market. Cheap double CPU (both cheap x86 and ARM for both Windows and Android mobile programs in PC without emulator) mini PC is needed in third world market, and cheap x86 processor. Hygon makes x86 but only server and only China market, Rockchip makes x86 telephone processor but cheap PC processor is needed in x86, perhaps VIA with its new chinese version is at last cheap x86 CPU that makes possible 20 dollar Windows mini PC like cheapest Android PCs, Pine64 and others, now cheapest Windows PC costs 79 dollars. Cheap x86 PC chip makes possible that Windows can compete with Android in cheap mini PCs, USB stick PCs, laptops and tablet PCs, for third world market. Cheap x86 PC chip is the one what is needed. Cheap PC for third world market can include roll printed old 1970s/1980s PC processors or complete PCs and game consoles, now old abondonware games and programs of several PCs and game consoles can be used in cheap PC without emulator. Those old PCs and game consoles used Motorola, MOS and Zilog chips so one old CPU can be used in several PCs or consoles if roll printed additional circuitry makes multi-system possible (several PCs and game consoles in single roll printed IC). Those three CPUs cover most old 1970s / 1980s PCs and consoles, so one multisystem roll printed IC with three CPUs is enough for almost all old programs and games. Modern Intel CPU can handle old MS DOS programs, and that is in main PC board of cheap computer for third world market. In printed electronics very large processor, perhaps square meter size, can be designed so that it can be folded so it fits in small space. Altough squaremeter size processor is slow and consumes lots of electricity, it can be designed to run high frequency because it is always cooled with liquid helium. In very cold temperatures speed of processor can be very high, so big and otherwise slow processor can be fast and in cold perhaps use less electricity than in room temperature. When this macroprocessor is in room temperature it burns out right away when power is turned on, because it is designed to low temperatures. It does not need to be special superconductive processor, but just cooled with helium. Processors can be designed to bepower effective, Intel Quark varies from 2,2 watts to 0,025, so another Quark version uses only 1/100th of electricity from other version. Intel Atom uses 0,65 watts at minimum version. Similar power saving can be made at squaremeter size roll printed processor, so when it is also designed to be helium cooled it does not use megawatts of power but only kilowatt or so or only hundreads of watts like ordinaray CPU. But it costs about couple of dollars to make unlike ordinary high tech processor or GPU. This roll printed processor can have billions or dozens billions of transistors, like most modern CPUs, GPUs and FPGAs. But it is roll printed made at more than micron size manufacturing tech, not 14 - 10 nm which is normal VLSI nowadays. Actually first Intel Pentium processsor from 1993 can be made using roll printing, altough it would be slower than 1993 Pentium and consume more power (MOS not CMOS made), in room tempereture. Simpler devices can be made using roll printing, electric thermometer (medical and air temperature versions, medical electric thermometer that cost almost nothing would have impact on health care in third world), clock (wristwatch and alarm clock versions, pocket calculator, handheld game console with changeable memory card. Memory cards need new new standard in roll printed electronics. Memory card can be slow and big mask ROM, 1980s were used “ROM packs”, similar principle but small memory card size ROM pack card with modern memory capacity. Also RAM memory card, but slow and cheap memory made at 500 -250nm tech or even micron size manufacturing (if it exists anymore). Those memory circuits are large but cheap, so larger size card than SD card size is needed. PDA which combines three of previous circuits without game, or game player can be included, and has no OS but simple flat-file for calendar and memos. Then audio player with computationally simple audio codec, then video player with audio player circuit, video also simple codec. ROM and RAM memory cards can be used in them. Finally portable media player with all previous features combined. Even more complex would be media player with analog audio or analog audio and TV integrated circuits (not roll printed but ordinary made, like memory is made with normal VLSI tech and then hybrid printed with roll printing, or memory is only in memory card and device is without it). Also internet receiver that uses IoT internet receiver but does not send anything, this IoT circuit is normal VLSI, device can now be cheap “internet radio” receiving internet TV and internet radio programs. This internet receiver includes roll printed media player with all previous features without analog radio and TV. Even the last one, most complex device can be manufactured at almost nonexistent cost if production runs is hundreads of millions or one billion units. If profit for manufacturer is only 0,001 dollar but one billion cheap products is sold, that is million dollar profit. When back catalague is thousands of products like memory card films, music and games, profit is billions of dollars altough product (media player and memory cards) cost almost nothing to buyer, it still makes billions of dollars to seller. Films offered are most likely old silent films that are in public domain or other public domain films. In supermarket I find Hollywood movies that are sold in cardboard DVD package at price of 0,95 euros. Those films were new, not old Hollywood productions, altough old animated fims and silent films were also included. Those cardboard DVDs are in limited sale in some supermarkets. If cheap copy protected memory ROM card is made billion units and sold in third world, it earns much more money to film industry than this 0,95 euro DVD, and memory ROM card films can be sold at very small price, including modern Hollywood films like those cardboard DVDs. Also several old public domain films and TV series can be grouped in one ROM card and sold like one night s TV programs, commercials included, and commercials in ROM cards lower price even more. Even those modern films sold in ROM card can have commercial breaks like TV, this “adware” brings cost down even more. For third world market. Spotify and other streaming services offer so small royalty payments that for people who have no internet connection copy protected RAM card with music album that after couple of listenings wipes out RAM memory, and artist is paid with Spotify prices, is solution. Those music albums in RAM card are super cheap. Artist and record company are paid music streaming service - class royalties per album. For third world market and in poorest countries only, not worldwide. Old computer or console games are mostly abondonware, but those games need CPU in media player to work. Old handheld Game&Watch games are simpler and need only cheap LCD- style display in media player. Audio / game media player can be roll printed made without sophisticated display, but if has video player option screen must have at least some pixel count, but colour is perhaps not needed, or colour displays in more “expensive” model (prices of these devices are less than one dollar about, between 0,1 -0,5 dollar media players made with extensive roll printing in components). Wristwatch & alarm clocks can be several different looking models altough internal circuit is same in them, simple plastic flexible one size closed wristband connected to roll printed watch for super cheap watches. Toy music instruments can also be made with roll printing, different looking models but with same internal circuit, ROM sample music content can vary between models. If back catalogue is hundreads of different roll printed electronic products, and every model is made 100 million to one billion units, and profit is 0,01 dollar for hardware manufacturer per unit, manufacturer have billions of dollars profit if several hundread or one hundred different roll printed products are on sale. Third world market products. Roll printed solar cells, batteries and lamps can be made. Even small amount of electric light from small roll printed lamp is suitable for areas where is no electricity, solar cell powered and simple printed battery. Simple PDA device looks like small phone or pocket calculator but is much thinner, membrane or other touch keyboard and no tochscreen even in more sophisticated models. PDA collects air temperature thermometer, clock and calculator with simple “LCD screen” game. Those four can be sold separately products at lower price. Electronic simple notebook included in PDA. More sophistacated model has also audio player, and then another model with audio / video player, video has low pixel count. All can be made roll printed (circuits, battery, display, keyboard, no CPU) so price is cheap, from 0,1 to below 0,5 dollars. Roll printed video glasses can be sold also, low pixel count and perhaps black&white. Then more expensive has either analog silicon LSI radio and TV receiver like some phone models, or digital internet IoT receiver (when device just “listens” internet radio and TV transmissions no line cost is neet to be paid), low cost IoT circuit. E-ink or other cheap TV displays can be made. Analog radio circuit in telephone is propably very cheap altough normal silicon LSI, so suitable for cheap media player also. If analog electronic roll printed circuits are done, they can be even larger than their discrete analog versions, and lower / different specs. Perhaps analog simple radio can be made using those circuits and sold. Roll printed analog circuits if used in analog music synth does not need to be sophisticated and not same voltage / ampere / ohm / watt class as normal discrete components, and not same specs, noisy circuits with 10 000 -12 000hz frequency range is enough, altough 15 000 - 18 000hz better, old ARP synths had only 12 000hz range and they have “classic” sound. High frequency reproduction is method where only about 7khz is needed for sound and higher frequencies are replicated using algorithms, usually two times base (7khz) frequency. Roll printed analog or digital music synth or other device needs only 7khz range and with ordinary digital silicon IC frequency replicator at mono or stereo output device can have 14khz range (or 6khz/12khz in ARP clone), needs DAC because of digital processing over 7khz output in analog device. Using ordinary silicon IC cheap BBD chip for phaser / flanger / chorus / delay / echo is possible, Panasonic clone or other, with roll printed additional circuits, in cheap music device. EXCTR aural exciter also. Using ordinary silicon IC even million voice analog or digital synth is possible, if 8 micron was analog and 4 micron digital synth tech in 1980s. 16nm digital chip with million voices has 2 X size of old 4 micron chip of 8 voices. 65nm analog chip with million voices has 32 X size of 2-voice synth chip of 8 micron and 45nm analog c.15 X size, but still acceptable price and power consumption, c.18khz audio range. 450 mm wafer size for analog circuits brings down costs, those ICs have larger manufacturing width than digital. John Vanderkooy: Simple square-law circuit gives 100 db dynamic range. Analog TV receiver can perhaps be build roll printed. Memory ROM and RAM cards for roll printed devices are also perhaps bigger than SD cards but much cheaper, using super cheap and very slow memory, perhaps even micron size manufacturing, normal silicon VLSI or roll printing if roll printed memory is possible and cheap. However slowest / cheapest RAM and mask ROM are almost as cheap as roll printed electronics. Normal microcontroller CPUs are cheap also so it can be inside most “expensive” 0,5 - 1 dollar price media player. TTL and ECL integrated digital circuits can be also done with roll printing, starting with their 1960s 10-20 micron manufacturing tech. Not perhaps same voltage / ohm / amperes and not same specs but same functionality in circuit, they perhaps cannot be used replacement parts of old 7400 / 4000 series etc chips but can be used in similar electronic projects like Lunetta synths. Analog roll printed discrete components or large groups of those in roll printed PCB does not need to be same specs as normal discrete analog componets, but they have same functionality. Those media players without internet connection or those who just “listen” internet are for people who have no money to pay for internet, When cost free zeronet internet comes to those countries normal cheap smartphone can handle same things better. But they can still be used for example in family only one has smartphone with net connection, others in family use media players. Music albums can also have commercial breaks before and after album like DVD films have commercials, and hit collection albums even commercials between tracks like streaming radio. So music ROM memory cards can be sold cheaper. Also TV digiboxes can be made with roll printed electronic components, 1990s MPEG TV or some other computationally simple TV codec. Other unconventional processor architectures are TTA transport triggered, dataflow, synthesis kernel (Massalin), unicore, unikernel (Mirage), RAM machine, TRON/BTRON, NISD: a framework for narrow instruction set, Knupath lambda fabric, GRVI Phalanx, systolic array, Lisp machine, SECD machine, Cell processor, Loongson, Kalray, GreenArrays, reconfigurable computing, Introducing the PilGRIM: a processor…, Pipelined graph reduction instruction machine, ePUMA, PULP, Parallel Random Access Machine PRAM (in TOTAL ECLIPSE version), Honeycomb: a application driven online adaptive processor, RICA reconfigurable instruction array, MATRISC, Spurbus symbolic processor, stack machine, ZPU, SuperH J-core, UPU, ARC 600, RePIC, COFFEE RISC, RISC-V, HP the Machine, WARP machine and Warp processors (two different things), 1TOPS/W software programmable media processor, p-VEX, Xputer, Elbrus, polymorphic processor the Molen, Janus - a gigaflops RISC + VLIW SoC tile, TSAR tera-scale processor. Prototyping and production is perhaps much cheaper using printed electronics, either roll printed or inkjet, so new processor types can be put in production cheaper than using silicon VLSI. And there are no CPU or DSP products using printed electronics before, so opportunities are great for manufacturers. Some COFFEE RISC related hardware runs graphic programs (old games possible?) without CPU (Markus Moisio), so CPU-less electronics for games is suitable for cheap printed elecronics? However SH-2 CPU is sold at 0,03 dollar factory price so those ultracheap CPUs are suitable for printed hybrid printing. Then is possible to use exotic number systems in computer arithmetic like unum / posit, quote notation (Hehner), different non-integer (logarithmic, golden ratio etc. like balanced ternary tau) bases, ternary (zero displacement ternary), quinary, pentanary base, signed integers, complex number systems (several different for computer arithmetic), U-value number system (Neuraloutlet wordpress netpage), Robert Munafo PT number system, multiple base composite integer (at MROB netpage), Paul Tarau s number systems, magical skew number system, mixed radix systems, index-calculus number system, fractional number system, folded number system etc. and all those are possible in computer arithmetic perhaps, even “infinity computer” that computes with (almost) infinite values and reaches almost infinite accuracy. Asymmetrical number system is used in data compression, and in the book “Dynamics of number systems” by Petr Kurka is almost all number systems and how to use them. Then is symbolic algebra, lambda calculus in hardware, lazy number systems, combinatorial logic / number systems, Fujitsu FPGA combinatorial problem solving system. “Promise library” is way to make processor more efficient. When CPUs are 0,03 dollar at cheapest, perhaps GPU can be at same price, modern GPUs have thousands of cores, 32 or 64 of them would be cheap, in GPGPU like CUDA configuration, so small microcontroller CPU can use GPU for also other computing than just graphics. Even one or or two CUDA or similar cores in ulltra cheap GPU version that costs 0,03 dollars. Those CPU and GPU can be hybrid printed with roll printed ultra cheap electronics. If roll printed slow processor can benefit from that that it uses some exotic number system in computer arithmetic. In third world are markets for billions of dollars worth for ultra cheap electronic products. However no firm in western world seems to be interested in manufacturing those things for third world. Why? Those roll printing firms could earn billions of dollars for making those ultra cheap products. Instead “printed intelligence” is used in wearables or “intellligent food packages” etc. in western countries. However billions of people in poor countries could use cheap roll printed electronics, in scale of billions or dozen or hundreads or thousands of billions of units of products, if all different products are counted, including ROM and RAM memory card content products, and all those billions of people that could use them, if those products just would exist. Why this market opportunity of billions of dollars profit does not interest western roll printed electronics makers? There are roll printed electronics firms in India and China also, but nothing so sophisticated like in west today. Is this another lost market opportunity that west loses to China just because nobody takes notice? Billions of dollars worth.
There is two-colour system principle that uses only two colour base instead of three. So only two colours is needed to represent all colours.Edwin Land invented 2-color system that used only two versions of yellow colour. So simple low pixel count roll printed display that has two almost similar yellow color layers can repsent colour picture. Colour quality may not be good but main point is that display is cheap and cheaply build. Practical Color Coordinate System is japanese colour system that uses only two values, not three like CIELAB etc. If PCCS is possible to put together with two -color system I don t know. Using only two colours as base makes displays cheaper, and display is most costly component of phone today. Also perhaps bitwidth can be saved 1/3 (or not). Old two color systems like Prizma color was used in 1920s when three color films were not made. US patent 6768815 “Color sensor” by Roger L. Woodall mentions “bichromatic theory” not trichromatic. “A VLSI neural network for color constancy”. “Improved Retinex image enchancement algorithm”, “Retinex image processing: improved fidelity to direct visual observation”. Perhaps bijective number system can be used with 2 - color system. Internet offers films for free, but billions of people are without internet connection so without those free films. They have no money to buy even TV receiver, analog or digital. So ultra-cheap media player is needed in third world market, with cheap display or video glasses. With perhaps analog TV and radio in in integrated circuit, silicon IC or roll printed, or digital TV if digital TV transmissions exists there, but minimum needed is internet receiver (not transeceiver like phone) that just “listens” internet TV- and internet radio programs. In Zero Dollar Movies is listed 15 000 free movies, in Tubi over 6000, Movietube, Yidio, Pluto TV, Snagfilms (10 000 films), Kanopy (30 000 films), Vudu (18 000 films + 5000 TV programs), Veoh, and Big Five Glories, Retrovision, Classic Cinema Online, Viewster, OpenCulture (1000 films altough not just film streaming site), Internet Archive (about 10 000 films) are other feature films free content sites. Pluto TV, Tubi and others are like ordinary commercial (advertisements) TV service, but it is in internet not TV channel. JustWatch, MoviesFoundOnline and Can I Stream It are services to search free film sites. Somehow all billions of people who have no internet or TV must get those free films. Ultracheap media player is solution. Free films (with commercial breaks in films) can be put to cheap copy controlled memory card and sold with memory card price, or in cheap DVD ecodisc discs with 5-10 hours or more per disc. Because people have no internet, instead of putting that free content to internet simply giving that content to third world state owned or state-wide TV channels so people can watch those films in TV, not in internet, with adverts included in TV transmission, is another solution, but it leaves those who have no TV outside. Mosfilm has 582 films in internet (with adverts), Sony Crackle has 150 films and 75 TV programs / series with rotation (new content comes and old goes), Paramount Vault is ad free with 150 old films, Troma and Maverick Entertainment has over 200 films both in internet for free. Then there is thousands of Bollywood films for free in internet. Then there is public domain old films. Because it is free, with or without ads, third world TV stations could show that content in their TV channels, with adverts when needed. Also cheap media players with memory cards can be used in third world, and internet TV / streaming. Maybe streaming that free content means that user of media player cannot download (store) that film, he/she can only watch it (stream) in intermet, if the film is not in public domain. Public domain laws vary from country to country. If in original country where content is first made and published and it then enters public domain later, but not worldwide, solution for problem is it in public domain or not is that copyright laws are in use worldwide, but price of that content which is in public domain of its original country, is zero. So that content can be “sold” with zero dollar price in certified server in internet or lawful physical distribution channel if memory card is sold, but price of content is 0 dollars, because it is in public domain in country where it originally come from. However copyright laws of that content in foreign land are in use, so it is not public domain, its price is just 0 dollars. United Kingdom, Japan and USA are those countries of different public domain laws than Berne convention / TRIPS, and India and China also, altough their laws are not so distinct from general death of author plus 70 years law which is in EU (but for films and sound recordings in India and China it is counted from publication year, not death of author). Instead of mess what can be published in what country that is the public domain worldwide system nowadays that gives only confusion and “grey area” where material that is not in public domain in some country but is in public domain in its original country, all that mess can be eliminated if it is in public domain in its original country it can be published without cost everywhere, but with all copyright laws included if it is not yet in international / that country where it is sold public domain, “sold” everywhere using certified lawful channels who “sell” that content with no cost and copyright attached. Now when they are in “grey area” they are illegally downloaded for free anyway, altough they are in public domain only in their originally country, for example if server is in country where content is in public domain. When they are downloaded legally they have copyright that prevents copying or reselling that content, and it can be counted and monitored for whom and how many units and what material is “sold” that way, in illegal downloads have no similar limitations, so situation is both clearer and better with legal downloads but no cost, if content is in public domain in original country. Because they download that material anyway, using illegal channels to get that “public domain” material. That material becomes public domain anyway in anywhere, but later than in its original country. For example in USA if copyright was not renewed that mean that owner of that material wanted that material to go to public domain. If now outside USA that material is still under copyright, it happens against the will of the principal owner of that material. That material can be under copyright outside USA, but no cost so it is free. When now it is used like public domain material outside USA anyway, which is violation of laws and in “grey area”, and becomes part of illegal distribution systems in internet. This can be avoided and at same time messy copyright / public domain system from country to another streamlined and simplified if material that is in public domain but outside that country still under copyright, stays copyrighted in foreign country but it can be “sold” like any product but without cost using legal channels, not illegal or “grey area” which is half legal / half illegal. In films that are in internet for free but are not in public domain, like indie films put in Youtube and other sites, and other legal films offered free, if they have adverts, and they are sold in cheap memory card or ecodisc DVD with minimal price, several films collected in one memory card or DVD perhaps, like “4 films in 2 DVDs” in one package DVD packages, but with simple cardboard DVD cover and one DVD only, they can be shown without commercials if similar amount of money that is earned from commercials per one internet streaming or download is put straight to memory card or ecodisc DVD price, because memory card and DVD without content costs also something. Mask ROM memory card made for example billion units and slow / large / ultracheap memory does not cost much, neither ecodisc DVD with billion unit scale. Actual costs of those (without cost of the content, the content in memory card or DVD costs also, if it is not in public domain or other cost free license) can be lower than 0,1 dollar, and nearer or below 0,01 dollar at billion scale manufacturing. Picture quality can be low pixel count low bitrate, so much material fits in one memory card or DVD. TidalTV has some films also, and TidalTV uses some geographically and user base based analyzing to maximize low costs, similar suits for third world internet streaming services. Simplest solution would be if third world TV stations would show those films that are now offered free in internet in their TV channels, with adverts in those films that have adverts. But many in those countries have no TV. But those who have, could see in TV what material in internet or in cheap media player memory cards or DVDs are to offer, if local TV shows them “highlights” of that material that now only internet offers for free, those films and TV programs. It is however questionable how for example afghan people react when they see for example Troma film in Afghanistan television first time. Films can have both dubbing and subtitles, and sound transmission in two channels, those who can read could read subtitles and turn dubbing off, those who don t read can listen dubbed film dialogue. There is also Tvchannelsfree and other internet services that offer just TV channels from around the world for free, aforementioned has about 4000 TV channels to watch without cost. Also free e-books can be put to memory cards or in internet, translated to local languages, public domain books or other written material offered for free or with very low license cost to author per unit sold (if author agrees that). Huge amount of written information in just plain text without fancy additions, just basic written text nothing more, fits in one memory card. If translations are paid, translator s wage can be added to memory card price. Huge amount of films are in internet for free today, but owners of that material can earn more money if that material is shown in third world TV stations with same adverts paid (or free) system like in internet today, and sold using memory cards of cheap media player or ecodisc DVD discs to people who have no internet connection (billions of people). They can watch films in media player s screen or using cheap video glasses. The amount of money generated with those movies that now are in internet only could be multipled. Media player must have self-sufficient electric power, using ultracheap roll printed solar cell or heat element / Peltier generator in device itself or separate, sold with device, so that no additional batteries are needed to be bought for media player or video glasses, and no electric network is needed. Electricity is stored in roll printed battery in device. Miro was media player, but it was for PC, not for simple devices. E-ducation and e-learning can be used with memory cards or DVD discs. DVD / DVD audio / CD player can also use solar cell or heat element charging. And change to different matter: previous post had million voice (expensive professional) music synth, if it has 32 times larger chip area than ordinary synth chip, because it has million voices that are individual, normal fault IC yield can be ignored and those voices that don t work inside chip don t make whole chip fault, some voices work and others don t but chip is sold anyway. In 1980/90s even less than 20% chips in wafer worked sometimes. But even 20% is enough voices in million voice synth that chip can be sold. Moreover old analog synth chips made at 8 micron had 5 inch wafer, and in that wafer less than 1000 chips perhaps. Instead of selling old synth chips with about 1,6 dollar price, whole wafer can be sold for about 500 -1000 voice synth, if it is synth- in - a chip solution. Fault circuits in wafer does not matter, it can be sold with fault circuits because it has hundreads of voices and not all can be broken. Even separate wafer size VCO connected to wafer size VCF etc solutions are possible, hundreads or thousands of voices. And still “cheap” price of hundreads of dollars (or thousands) per synth “chip” (wafer size IC). Using Surface Mount Device - size additional analog components in die size area for each voice for additional analog processing, placed in top of die and in gaps between dies in wafer is possible. SMDs are separate (discrete) analog and digital components like transistors and resistors but small size, and some are ICs. Trueno analog synth is USB- stick size and propably uses SMDs. Roll printed digital control logic or 0,03 dollar JH-2 microcontroller can also be used with each voice, or computer CPU in wafer package with analog synth wafer with its dies not separated from wafer. “Multiple identity filter” and Bastl BitRanger analog logic computer are examples of how to use aditional circuits with synth voices. Alesis Andromeda chips had 8 inch wafer size? 8 inch size analog synth chip is then possible if that wafer is used. And about 100 000 analog voices in one wafer-size chip? Or million? Old wafer masks can be used again and no new ones are needed. No need to separate dies from wafer and pack them individually, and no need to discard wafers that have too many fault dies anymore. Wafer is soldiered to PCB board, with channels for audio information and electric power in PCB board and / or in between gaps of dies. And SMD- size additional analog components if needed, and control logic digital microcontroller if CPU is not used in wafer package (CPU is inside same plastic package that also includes wafer size analog “chip”). PCB board connects different dies and audio channels and brings electric power to dies. SMD components are not individually packed to plastic packages, but all which are needed in wafer size analog music synth are put inside one large plastic package with wafer size analog synth “chip” and possible digital control logic. Analog SMDs can be discrete components but they just are smaller and cheaper than typical through-hole components and their size is between ICs and normal transistor or resistors etc if their size is without package about 100 micron (0,1 mm) or smaller. 100 000 voice wafer size synth needs over million SMDs for additional circuits, like ordinary analog synth needs additional circuits in addition to its analog siund IC chips. But because they are all put inside just one package price is perhaps acceptable. Or use then Anadigm FPAA or other FPAAchips for additional circuits and pack them inside wafer package. SMDs can be used to build analog synths of their own, like Trueno possible does, those synths would be cheap and small, if not newly re-released analog ICs are used. Anadign FPAA scan also be used as synths of their own, Anadigm FPAA filter chip has 128 filters, so multi-voice synth can be build cheaply using Anadign FPAAs only also, not old 1980s synth chips in wafer level scale. FPAAs or FPGAs can also be used wafer level scale, especially older designs that are cheap in wafer level price. Most of FPAA manufacturers went out of business but their old wafer masks are still somewhere. Some of dies in wafer do not work and some work only partially, but anything that works even partially can be used in wafer size processor (actually wafer size assembly of hundreads / thousands of processors connected to each other). SMDs can be used to make analog TV receiver also, like Android TV stick, but it would be small analog TV stick, or larger analog TV box like digibox but analog, connecting phones, media players and video glasses to analog TV transmissions. But is its price any cheaper than integrated circuit analog TV that cheap chinese phones use? If not, those analog TV ICs can be used in cheap media player, together with analog radio IC. SMDs can be used to make cheap radio also, but again not if IC version is cheaper. But still roll printed digital or analog muisc circuits / complete synths are cheaper. Noisy output can be cleared using Dynamic Noise Reduction / Limiter circuit. For expensive synths AC power conditioner, TIM / DIM distortion noise shaper, True Dimensional Audio TDS 2 and Neutral Audio Drei (not “Three dimensional audio TDA2” and “Treia sound processor”, writing mistakes in post no:5 ) can be used. Cheap musical devices: “Japanese Finger Piano” and Ototo by Dentaku are cheap devices that are suitable for third world market. Keyboard, even toy keyboard does not need keys when “piano gloves” -type solutions are used even in cheap toy musical instruments. Piano gloves are much cheaper than full key range keyboard musical instruments, there is no need for keyboard if instrument uses only piano gloves, and one piano glove can be used with all musical instruments, so it saves money that way. For third world market. Wafer size products can be for memory like RAM (DRAM, NAND etc) and ROM. Old manufacturing machinery that first costs billions but when used 4-5 years is changed to newer one and goes for “garage sale” thereafter, can be but to good use again, and old wafer masks recycled when they are used again, not making RAM or ROM chips but large wafer size memory stores. Old 350 nm, 250 nm and 180 nm manufacturing can be used. 250nm is 250 times larger than 16nm modern memory chip, so 250 such dies is needed in wafer to replace one 16 nm die. And they are slower than 16 nm die.But wafer size production of old tech can still be cheaper than modern memory chip when capacity of wafer size memory is compered to similar capacity modern chips (compared with price). This cheaper massive mass memory is suitable for third world cheap PCs and even laptop PCs, if memory designed for embedded or phone / tablet PC RAM and ROM are used, old designs from 10 - 20 years ago, using old wafer masks and machines. Most modern memory wafer costs 1600 dollars about (16 nm - 10 nm memory), so wafer sized old tech memory “chip” should be very cheap. Altough size of memory “chip” is 8 or 12 inches (20cm - 30cm) it still fits inside PC tower or inside laptop PC. Old analog synth chips mentioned earlier were made at 3-, 4-, 5- or 6 inch wafers (7,6 cm -15 cm). Even CPUs and GPUs can be made at wafer level package, not separating dies from wafer, but those are suitable for supercomputers only because wafer level processor compilation uses enormous amount of electric power, either CPU or GPU. Efficient liquid cooling must be used, but supercomputer is now very compact. But FPGAs, FPAAs, old ASICs and DSPs are suitable to be put production again, not separate chips but old wafer masks and production machinery used and wafer level package “chips”. So instead of selling separate chips from wafer complete wafer is sold in wafer level package or even larger size 20 inch (0,5m) package etc, all faulty dies or only partially working included in wafer. Only old 10 -20 year old machinery and wafer masks are used, so that price does not go too high. This wafer level “chip” business using old machinery, wafers and wafer masks is most suitable for memory, different RAM and ROM memory designs and versions, but old FPGAs and DSPs and ASICs can also be produced again, this time using whole wafers not chips that are sold. Modern supercomputer can use wafer level package also but In CPUs and GPUs. Those wafer level chips are not wafer level chips but groups of dies in wafer, because they are made using old wafer masks that were used to produce separate processor dies from one wafer, hundreads or thousands of them in one wafer. Recycling old wafer masks and using them for massive memory storage or other use like FPGA or DSP is perhaps more cost effective and cheaper than using modern tech. Suitable for third world cheap products perhaps. No need to separate faulty dies from the wafer or those which are only partially working, whole wafer is sold as one “chip”. But if memory is more expensive counted by dollar / gigabyte ratio using old chips, even in wafer size package, than using modern memory chips, then makes no sense to use old memory circuits again. But Programmable Logic Devices, different kind of, and ASICs and DSPs can beperhaps sold with old versions but in wafer size package with hundreads / thousands processors. In netpage factorandequilibrium “Wafer mounter equipment market worth $ 180 billion by” is graphic that shows that over 1/3 world s manufacturers use silicon wafers that are smaller than 300mm or 200 mm, about 1/6 is 150 mm, 1/10 100 mm and about 1/20 other size. Both TSMC and UMC still have 9% of their manufacturing capacity using 150 mm wafers. So wafer scale “chips” can be made and sold, recycling old wafer masks from 20 - 10 year ago. According to news semiconductor industry has overproduction capacity, factories are closed and from 300 mm back to 200 mm wafers is the trend. Wafer size “chip” can bring new production to semiconductor firms. Old analog synth chips were made perhaps at 3 inch / 76 mm wafer. Because there is plenty of 150 mm and 100 mm wafer capacity those sizes can be used to make wafer size chips. 200 mm wafer also when manufacturing process is old enough (180nm? 90nm? Or 65nm? Or 250nm?). Small wafers are also thinner, so they cool faster, and production can be faster that way. Modern wafers are thinner than old, so they cool faster, even 200 mm wafers? Production where one wafer means only one “chip” is then possible, and factory can always have maximum production speed. Cost cutting to maximum not only packaging whole wafer in plastic package so dies are not cut away and packed separately and fault dies discarded, but also if testing of circuits is not done, brings cost down even more. So manufacturer and buyer do not know how many dies are working and which work partially or not at all. When die separating process and separate packaging is not done and not even testing of them, wafer size chip can be sold cheaper. When wafer comes from oven it cools down and then it is packed directly to plastic package, no testing and no die separation. Altough wafer has hundreads of dies, manufacturing uses old machines and wafer masks, and production is simplified considerably, so price of wafer size chip is acceptable, still much higher than modern microprocessor, but wafer is large with hundreads of microprocessors so old tech can still be competetive with new tech. Also old wafer masks can be used to make multiprocessors where in one packakage can be 16, 32, 64 or 128 old model processor dies, some of dies are faulty some working only partially, but chip is sold anyway and bad dies not separated from working ones so dies are cut from wafer in 16 - 128 die blocks, not each one separately, or then separate dies, all cut from wafer are after testing put to one package (16 -128 dies), all dies working and bad dies discarded after testing. Those plastic packages would be larger than modern processor packages. Tier 3D or something like that was american firm that made 3D FPGA chip, but it was MPGA (mask programmable gate array)? If manufacturing scale is millions even MPGA can be used not FPGA if MPGA is cheaper. Old style FPGAs cost now only 3 dollars or so, and they can be used to imitate old 1980s wavetable and sampling based music synths, so few dollar priced copy of them is possible (Korg M1, Roland D50). But roll printed digital electronics is suitable for them also, price would be like 5 -10 dollars or cheaper for “playable picture”, plastic sheet roll printed picture of synth with keyboard etc that can be played with keyboard controller or touching picture of keyboard. Some additional cheap IC circuits can be put inside playable picture, altough music producing logic is propably roll printed. Surface Mount Device circuits that are analog discrete transitors or resistors etc can be grouped to one ASIC IC in one plastic package, and this ASIC with SMDs inside (SMDs do not need individual packages because they are all inside ASIC) can be used to make clones of early analog synths or modular synth modules in eurorack format or another, but because they are made using cheap single chip ASIC / SMDs inside (only this one single voice/multi voice ASIC is synth s only sound producing circuit) their price can be “toy synth” class and sold in supermarkets as toys etc. Rack mounted synth clones even do not need keyboard if hey have MIDI control. Anadign FPAAs are multi-synth voices themselves so they can be used to make cheap 128 -voice synths or even toy synths. New production Curtis and SSM chips are so cheap that they can be used to produce cheap clones of 1980s polysynths, but now thay are cheap price “toy synths”, with roll printed- or SMDs as additional circuits.
Problem with desktop PC is that it is big and not portable, and often expensive. However if using cheap x86 chip like Vortex86 or VIA or new chinese x86 CPUs build in licence, CPU price can be low. Windows is without price or low price possible to be in cheap device according to MIcrosoft. So real cheap PC would be PC that uses x86 programs in low power usage laptop PC x86 processor, but this PC has size of portable CD player, so this pocket - PC can use desktop PC programs but it fits in the pocket and is very cheap. It does not need display and keyboard. It uses cheap video glasses as display, so display size does not matter, so graphical programs intented for normal size PC display can be used. Video glasses can represented them so that graphics does not seem too small like watching normal PC prpgrams in tablet PC or phone screen. Instead of keyboard it uses cheap data gloves (very cheap data gloves, not expensive hundreads of dollars models) like cheap piano gloves for musical instruments.Because pocket-PC is small, and uses cheap video glasses and cheap piano gloves- price cheap data gloves this PC is very cheap and in the MIcrosoft class of cheap devices for which Windows is free. For third world market. With or without net connection. For people who have no net connection, cheap media player (MP4 player, portable media player) can have price of 2 - 5 dollars for “expensive” media pöayer, and it can have color screen, even touchscreen, and use old chinese processors like Rockchip RK28xx series, RK2906, Spreadtrum 6500, HiSilicon K3V2, Alwinner A10, F1C100, Upfront NS2316, Actions Semiconductor ASM 7021 or other outdated and cheap processor, without internet connection. If toucscreen is not used keyboard (most simple touch keyboard) can be backside of device so screen can be full size of frontside. Back-typing keyboards are Alphavi, T-Blade and Grippity and others. Really cheap media players can use cheap CPUs like SuperH2 / J2-core, or similar low price (0,03 dollar for microcontroller class J2) other cheap and outdated CPUs. If roll printing is used in CPU also, Cortex M35 or some other modern model can be used, because modern design are more efficient than old, and using “bespoke processor” tech CPU can be streamlined even more. RAM memory (DRAM, NAND) price is about 0,035 dollar / GB cheapest for gigabyte if 128 GB chip is used. If instead of 128 GB chip only 1 GB chip is used in device, manufactured using old tech so large die area, but price 0,03 dollar for GB, that can be used in super cheap media player. Additional FPGA circuits can be used with single kernel CPU like “Wave processors” to improve performance (over 100 X sometimes), and “Floating point adder design flow” (Michael Parker, Altera 2011). Even very small CPUs like microcontrollers could use those wave processor methods with some additional FPGA circuits, or Xputer, or Coarse-Grained Reconfigurable Array. Perhaps analog electronics can also use wave processors techique or Xputer in analog form, or Xampling. For making cheap musical instruments: digital roll printed products not necessirily need to be copies of 1980s digital synths, but new design, old phone sound chips used wavetables and samples with FM for simple sound circuits. Similar like phone sound chip is suitable for roll printing due to simplicity. Old 1980s / 90s tracker program type cheap and simple designs can be used (tracker was sampler with sequencer). Newer methods like digital waveguide (used in 1990s in sound cards), XOR synthesis, scanned synthesis, granular synthesis, “Computationally efficient music synthesis-methods”, vector phaseshaping, Feedback Amplitude Modulation, “sound synthesis using allpass filter chain” or other computationally simple method can be used in cheap roll printed music synth. Using Ogg Opus style “high requency hue (cue?) coding” or “stereo intensity coding in high frequency” makes possible that only 7khz sound is needed to make, it can be expanded to 14 khz using simple or complicated versions of high frequency reproduction (SBR), in separate silicon IC in roll printed device or roll printed high frequency replicator. Side information for high frequency can be stored in synth patches etc, so that replicator does not need to be working “blindly”. Controlling methods like “timeline-based modulation” (Progress Audio Kinisis) can be used. Silicon IC memory, gigabytes large, can be used with roll printed logic in still very cheap device. If normal silicon semiconductor must be done cheaply, methods like “homemade integrated circuits” can be used, but in industrial scale. Someone made integrated circuit in his garage 2018 using DIY methods in 50mm wafer. If in India or some other third world country garage- semiconductor firm can be started, manufacturing using simple methods made ASICs and other analog analog electronics, or digital, at low price because it uses simplified IC manufacturing and not expensive billion dollar machinery for simple ICs, and instead of billion dollar factory makes them in some garage etc. If garage IC manufacturing is possible, that 2018 IC was in 50mm wafer and smallest feature size was 2 micron about, so compatible to 8 micron newly produced sound chips. Garage IC production can give third world countries opportunities in IC business, no expensive machines and facilities, and low volume production run ASICs are possible. In previous text I wrote that after wafer is tken from oven it can be packed to plastic package, but etching and other process must be used to make circuits work, but still packing whole wafers without testing do dies work or not is cheap way to make larke capacity wafer size IC. It is called hybrid integrated circuit if normal discrete components are mixed with ICs and then whole thing is packed in epoxy or ceramic. Multi Chip Module is thing where several ICs are packed to thin film, thick film, ceramic base or to small miniature PCB board, and then whole thing packed in plastic or ceramic package. Those new produced sound chips or Anadign FPAAs can be packed with additional SMD circuits like transitors or resitors etc, or with analog gate arrays, in one package. So inside synth is only one voice circuit, multi voice. Old 8 micron dies can be cut from wafer in 4X4=16, 4X6=24, 6X8=48 or 8X8=64 blocks, resulting multi voices, packed with additional circuits, faulty dies not separated from working ones. Or use 3X3=9 for 8 voice, 5X5=25 for 24 voice, 6X6=36 for 32 voice, and 9X9=81 for 80 voice sound chips, extra dies are for faulty die replacement, because some dies not work anyway, so extra dies are used when some of the voices do not work. CEM 3310 sound chip has size of 1,8 X 1,8mm, and CEM 3340 and 3345 is 2 X 2,2mm. If they are made with 76mm wafer, about 1300 and 900 dies are in wafer. if yield is 67% about 850 and 600 dies are working. Price of those old sound chips is 1,6 dollar cheapest. But they are used only 8-16 in polysynth. If wafer size cheap chips are used maker could sell even about 1000 voice sound chip (but some of the voices are broke), so price is less than 1,6 dollar per voice when individual packaging per die is not used. Volume of production is much larger than using only 16 or less per musical instrument voices, so price per voice can be cheaper. Stacking multiple wafers on top of each other (VCO wafer to VCF wafer to VCA wafer etc) in package -on -package arrangement or multi chip module makes working synth with almost 1000 voices. Wiring die to IC package uses 20 -40 micron wide gold wires. Those wires and their assembly system can be used to make extra electronic analog circuits outside analog IC, if IC needs additional circuits, made of those wires. Those wire-circuits can occupy 100 micron or more wide areas between the dies in the wafer, if several dies are in same IC. Cutting 4 to 81 dies in blocks from wafer and making them one “chip” where dies are not separated from block, and block packed as complete chip (faulty dies included), and when block size is nearing wafer edge, dies from the edge are cut individually and packed like normal dies one at a time so wafer size is not wasted. Altough Anadign FPAAs have diffrent volts / watts than other musical instrument electronics, Anadigm FPAAs can be used with them anyway if they are within Anadigm FPAAs voltage / watt range, only concern is that using more watts / volts than FPAA destroys the FPAAs, so using other spec analog with FPAA must be careful or use sort of automatic level control that FPAA is not destroyed. But Anadigm FPAAs can be used with almost any analog musical instrument, toy instruments included. Trackers were 1980s / 90s software, but making hardware tracker using roll printed electronics and about 1 dollar price per item or less for third world market, with their early PC sound card chips also roll printed in device is possible, some kind of simple touch based user interface in device is needed. If CPU is needed in roll printed devices thay can be ESi-RISC, ZPU or even XMOS microcontroller or Xputer. Cheap musical instruments can be build using hybrid integrated circuits, like modular synth modules where inside eurorack module is only one hybrid IC as sound chip, this IC includes all components for sound producing circuits in that module, it has normal discrete components mixed with ICs. Also old monophonic synths can be cloned using only one hybrid IC. More complicated things can be build using Multi Chip Module with SMD and IC components, for example “Mighty Serge” 12 module synth in one MCM chip, build with SMD and IC components and all 12 modules packed in one MCM chip package, ceramic or plastic, so inside this 12 -module Serge synth is just one MCM chip. How good or bad sound those things have is irrevelant, main point is chap price. For digital control of analog sound J2 microcontroller can be included for example in every old CEM or SSM sound chip in MCM package, or outside MCM chip cheap Vortex86 or VIA x86 processor, which is used to make modulation, ADSRs and LFOs. Very old softsynth software like Armonyx PC synth (1985) that is fast and needs small processing requirement can be used with microcontroller CPU. Gigabytes of cheap memory can be available for (analog) synth patches. Instead / or together with x86 CPU also cheap open source ARM processor like Allwinner A33 (Pine64) for analog synth modulation is possible. Roll printed digital or analog sound circuits can use quadrature mirror filter bank in digital form to improve low frequency. If sound is produced using analog or digital circuit at 3,5 khz, QMF filter bank expands it in mono or stereo sound output stage up to 7 khz, that is upscaled to 14 khz using high frequency replication. So only 3,5khz is needed for 14 khz sound. Missing dynamic range of analog roll printed circuits can be upscaled using dynamic range compander as reverse to expander, “Simple square-law circuit gives 100db dynamic range” John Vanderkooy. Bland sound can be brightened using aural exciter, perhaps using silicon IC, Aphex aural exciter is in IC form, or Graig Anderton s cheap aural exciter circuit. In digital roll printed circuits using one bit serial logic, like 64 khz sampling rate that is 9 x oversampling 7,1 khz, 9 bit dynamic range, 3,5khz sound, after QMF bank and HFR it is 14 khz sound. Because it is sampling over 7khz it has sound information over 7khz, but deacrasing quality. That sound information can be used to reconstruct higher frequencies up to 14 khz, 9 bit dynamic range can be expanded like old Yamaha DX7 used from 12 bit IC chip to 15 bit output, it will be 9 bit to 12 bit. Old G.711 and G.722 and G.726 telephone standards used ADPCM (Wikipedia ADPCM page), and ADPCM is standard method also in telephone sound chips up this day, and in old synth sound chips. ADPCM samples where 4 bits almost equals 16 bit accuracy like DDPCM (Dynamic DPCM bistnbites netpage). So only 4 bit logic is needed not 16. Perhaps using sub band ADPCM like those G- series ADPCM with logarithmic companding (G.726), with 8 or 6 bit a-law or mu-law or other, and then that to 2 bit ADPCM is enough for synth voice, so only 2 bit roll printed logic is needed for 11 - 14 bit (depending on logarithmic system used) sound. System that expands 2 bit sound / wavetable samples to 11-14 bits can be ordinary silicon IC in final stage when sound comes out of roll printed device. G.726 standard used 8 bit to 3 bit to 2 bit ADPCM compression. Using recursively indexed ADPCM (RIQ-ADPCM and RIVQ-ADPCM) can be used. Asymmetrical numeral systems (FSE encoding) with differential representetation is possible? Like delta sigma with multi-bit asymmetrical numbers, or Takis Zourntos one bit, or ADPCM. 1 bit ADPCM is in opencores nepage, and “2 bit file format - John Comeau page”. Stackoverflow “8 bit sound samples to 16 bit” 2009 is many methods how to improve 8 bit sound to near 16 bit quality. So first 1 or 2 bit ADPCM or simlar is expanded to 8 bit integer, or to 4 or 6 bit logarithmic and that to 8 bit, and that to near 16 bit. Expanding from 1 or 2 bit upwards can be in silicon IC in roll printed circuit. If analog sound is expanded from 3,5 khz to 14 khz that makes it digital sound, because processing from 3,5khz upwards is digital. So roll printed analog synth has digital output. Roll printed logic can be 1 or 2 bit wide. “New aproach based on compressive sampling” 2014 Bonavolonta is 1 to 50 sampling ratio. Xampling is analogue sampling? Sparse fourier transform can be used. Because logarithmic system is version of floating point, perhaps unum, posit or valid can be used in place of logarithmic companding, or John G. Savard s EGU / HGU (Quasicrystals and data compression post). NICAM style ADPCM where instead of white noise dithering is used to improve accuracy. If hybrid integratd circuits are used to make eurorack modules, simplified systems like Frobesynth can be used to low cost. Dialog Audio 3244 software modulation etc can be used. Also Hammond B3 organ or Hammond Chord Organ can be cloned, using tube valves like original. Hammond sound in right pitch is Hammond trade secret, but can be emulated using digital conrol of tube valves. PARI.E made organ with plastic tonewheels, similar are suitable for cheap clones. Keyboard can be separated from organ, and actual organ is rackmounted, or keyboard is just looking like Hammond B3 console but it is empty inside and lightly build, sound machinery is in rackmount. Pocket size computer (PC) for third world market has that advantage that it can use desktop PC programs that require large screen, and desktop PCs are expensive. If instead low power requirement x86 processor with laptop PC battery are put to pocket size package like size of portable CD player, and used with data gloves and video glasses, this pocket size portable desktop PC can use desktop PC (Windows and Linux) programs. (Cheap) video glasses make possible to see desktop PC programs without large monitor screen, and cheap data gloves makes keyboard unnecessary. Price is both cheaper than desktop PC and now is possible to use desktop PC programs everywhere. Electric cable can be addition, in places where electric power is available. Real mini-PC with electric cable, pocket size desktop PC, can be also with data gloves and video glasses, it needs electric cable connection to work but can be carried in pocket etc. to one place to another. Cheapest mini PC would be “Android stick” size device that is size of USB stick and is plugged to phone or tablet PC, and it has x86 processor (or ARM), it has size of USB stick but it is working PC. Cheap chinese ARM processor or SoC can also be used in all of these cheap PCs, those SoCs have net connection but PC can be used without it, but desktop PC requires x86 processor and ARM can use PC desktop programs only with emulator. For third world market. More exotic processors like Loongson, Kalray in 16- core version, or Pezy in 16- core version, that are cheaper to make than full size Kalray or Pezy processors, can be used, if they are very computationally effective. Cooling method for computers is in US pat. 10020436 (2018), it is energy harvesting method for very low power requirement computers, it is power generator that uses heat of CPU for power, high watt amount processor can be used but because it harvests and recycles energy wattage used is actually low. Simplest use of media player can be cheap feature phone that chinese make at 5 dollar price cheapest, but without phone / internet connection, so user have no payment for phone line or internet. Instead cheap phone is used as media player, playing MP3 audio and videos, and software (freeware and other programs that are possible to use without payment) programs like games and other programs can also be used in phone (it has cheap ARM processor so Symbian is only solution? Or Android? Or Linux? Or old Windows Mobile?). Cheap chinese feature phone with no modification, ordered straight from factory line at millions of units, cheap feature phone, it is not connected to phone network or internet and uses memory cards as storage, that phone can be used as media player for people who have no net connection, they can still watch videos or listen music, play games, use e-learning and e-ducation programs, and use other software etc. Memory card is the way to load new content to this phone that is used as media player. Because billions of people are without net connection. If later free internet (zeronet) becomes reality they can simply switch net connection on. Or they can already now use that phone as device that only “listens” internet TV and internet radio stations, if those net TV and radio stations are free of cost, so it is like watching TV or listening to radio, but using phone instead of radio or TV. If roll printed cheap devices are used as media players, one bit serial logic can be used in massively parallel computing, or transputer style logic. Roll printed signal processor / GPU that uses one bit logic or transputer in parallel processing, if that is more easier to large and power hungry roll printed signal processors. One bit logic was used in supercomputers in 1980s, and ransputers also. Differential or logarithmic coding instead of linear makes bitwidth smaller. Inexact computing can be used in signal (video) processing also (Rice university). “Bespoke processor” is another way to make processor simpler. In XLNS research - overview netpage in articles section is many logarithmic number systems. Lambda calculus in hardware using roll printed 1 bit logic or transputer is perhaps possible, if it brings more efficiency to large, slow and power hungry roll printed circuits. Making cheap musical instruments that are “playable picture” type: also digital sound chips like those used in phones to early 2010s, now discontinued (Yamaha SMAF, NEC etc.) and Dream- and EMU 8000- series chips that were used in cheap keyboards, those cheap IC chips can be packed in plastic inside “playable picture” and create very cheap synth. Also analog filters ICs like used in Waldorf Wave, PPG Wave, Ensoniq or Fairlight mark 2 can be packed in playable picture, creating very cheap versions of those instruments. Oscillators are either wavetable or digital in those so they can be roll printed, or use silicon IC memory (wavetables), those are early / late 1980s designs. Several of those cheap phone sound chips / cheap keyboard chips can be packed in one keyboard, normal cheap music keyboard instrument like chinese cheap models, and altough it has many sound chips “collection” inside its price is cheap. Phone sound chips are not made anymore, but if they are ordered production propably starts again, or manufacturing is licensed to some other manufacturer. Old drum machines that had analog filters can also be made again with cheap price. Real cheap keyboards can be also be played with piano gloves without video glasses. Video glasses needs PC or phone, some kind of CPU with graphics to work. But just paper or plastic picture of synth keyboard will do, with large picture of knobs and modulation wheels, this is just picture without any electrinics inside picture. Piano gloves must have position sensing then like phones have, so moving hands aboard picture like playing keyboard is sensed in gloves as position movement. Or pictures sense only touch and transfer that information to actual synth, so those pictures are keyboard controllers that are just primitive touch sensing pictures of keyboards. Actual synth can be in small plastic box connected with wire to those playable pictures, using silicon IC or other inside. Or several dozens of those small plastic box synths can use one standardized keyboard controller model, that is normal keyboard controller with normal keys and knobs. But if plastic or paper sheet keyboards are used, some numerical simple display, for example simple LED number display or electric ink that shows numerical values of percentages when pictures of knobs or wheels are “turned” using piano gloves or hand if touch sensing picture is used, is needed perhaps. This cheap display can be one standard model shared between all cheap similar synths. Planar playing interface like Haken Continuum works with piano gloves with position sensing, or touch sensing picture that works like Continuum fingerboard. Roll printed synths can use cheap silicon CPU like J2 (cost 0,03 dollar) in hybrid printing. Guitar synth that can be connected to acoustic or electric guitar, and works like EML Polybox (similar acoustic guitar synths are already manufactured) can be made, polybox increases paraphonic voices to 28 or perhaps over 100 from one guitar note. For third world market. Because memory is cheap now, making roll printed synth with silicon memory that has for example 4096 presets ROM and 1024 user presets RAM can be made, and it costs almost nothing, for example DX7 roll printed copy. Playable picture can be simple picture of synth with only buttons of DX7 picture working, keys, knobs and modulation wheels of picture do not work. So playable picture needs additional keyboard controller to use other than button switches of picture. That would be simplest solution for playable picture. Electronics of synth can be inside picture, used with MIDI, or outside picture in small plastic box. Or picture is just plain picture of synth, and all functionality is in outside keyboard controller, synth electronics inside picture or in outside box, or inside keyboard controller which is then many synths in one itself, and picture only have synth patches and presets in silicon memory. Aerotaxy is new method for making cheap circuits: “New method of manufacturing smallest structures in electronics: discovery could revolutinioze semiconductors” 2012. Also garage manufacturing of ICs in some barn in africa or India, making small production run “boutique” components, analog or electronic, for example to high-end hifi systems, spare parts for discontinued old ICs that has market still altough they are not made anymore etc. 2018 first garage IC was build in america. Garage IC building is cheap, so it can be done in industrial mass production scale in africa or asia in poor countries cheaply, no expensive machines are needed, if 2 micron feature size is reached that is enough for many analog and digital components. So third world countries can be in semiconductor business also, not just China, altough those 2 micron simple ICs no way can compete with modern ICs, not even close, but they have own niche market. ViralG is P2P system that promises to stop 99% illegal P2P traffic, it can be used in free internet / zeronet. In Hackaday netpage in prize entries section is Multiwind wind controller, Analog synth but in cello form, CEM 3340 module, and PolyMod microcontroller synth, similar cheap designs are suitable for third world. If J2 CPU costs 0,03 dollar, one cheap CPU can be used in cheap playable picture synth, one CPU uses software emulation to emulate several old CPUs of synths from 1980s. Synth music producing circuits can be roll printed or cheap silicon IC like CPU. Cheap phone can be docked to cheap keyboard controller, phone has softsynths. Similar docking is made for iPad, but cheap Android phone docked with cheap keyboard controller is needed in third world market. Ubuntu phone / mobile was project to make desktop PC / phone hybrid with similar docking of phone to keyboard and screen, phone is “desktop PC”. RDGAudio is award winning firm from India that makes music software. One playable picture can include several 1980s synths (Yamaha DX series in one cheap plastic sheet, Casio digital synths from middle of 1980s in another playable picture, Kurzweil series in third, Roland series in fourth etc). Also romplers, samplers, sequencers and trackers from 1980s are simply to make now, in roll printed plastic with electronics. Roll printed plastic sheet is simply ordinary roll printed plastic sheet of picture of synth etc, roll printed electronics that are miniature size is roll printed separately in fine-quality roll printing machine and then put on this very cheap plastic picture of synth that is printed in ordinary coarse roll printing machine. Cheap microcontroller and silicon cheap memory can also be inside plastic picture. Cheap microcontroller can even in softsynth format include 1980s FM synths and 1980s sampler / rompler / wavetable synth functionality, if small size roll printed electronics is not used. Softsynths are also in phones today. But those who have no internet connection or smartphone, cheap music softsynth platform with memory card slot (even plastic picture can have memory card slot) that has cheap 4-5 dollar ARM or x64 processor and cheap memory for softsynths, is ideal musical instrument for third world market. No internet connection, softsynths are loaded using memory card or directly to RAM and ROM memory, in RAM memory synths can be changed removing old ones and uploading new ones. This instrument can use smartphone / tablet PC (Android, and old Linux desktop also) softsynths without net connection. Its like tablet PC for music& other things like educational programs / video films without net connection. There is also cheap microcontrllers like ZF Micro ZFx86 and RDC Semiconductors x86, Intel Quark and expensive Ryzen / Epyc embedded CPU etc, in India chip foundries are starting so even cheaper than chinese manufacturing is possible. IBM Power RISC based microcontroller can use old Apple PC programs from 1994- 2006, old Apple iOS is open source now, or Darwin OS or emulator can be used. Microcontroller + FPGA combined, for example flash based FPGA, increases processing speed. Earliest FPGAs are coming to public domain so cheap microcontroller with some small FPGA circuit can make efficient and cheap processing for example in playable picture if roll printed circuits are not used. Cheap FPGA based synth is also possible if FPGA costs only few dollars. Even old softsynths from 1990s can be played with modern x86 microcontroller. In netpages EEWeb extremecircuits, Circuit-finder, and bowdenshobbycircuits are super simple VCO and VCF designs, and “Multi-sound for guitars” 2013. EU funded projects like Webinos, CORDIS projects RIFE, WiSFUL, UNIFY and iKaaS were for more efficient internet, where RIFE was “Internet for everybody”. Could those lead to free internet / zeronet, using for example blockchain (Moeda) based payment for advertisers like some music streaming services use blockchain? More about that in “Free internet for third world” post in RH forums in “latest” section. Because cell phone operators charge extra for virus / malware protection in their lines, free internet for third world can for example be free line cost but 2 dollar / month for virus protection. Because in India cheapest line costs are 2 dollar / month about for 4G or 3G with limited data plan, that means actually normal line cost with virus protection in “cost free internet”. Those who do not want pay for that extra virus protection that telephone operators charge, can have line for free, the same line that operators offer to those who do buy virus protection, but without virus protection. Because perhaps large part of those who use “cost free internet” choose to pay that extra virus protection charge, teleoperator gets actually normal payment for “free internet”. Another cost free possibility is that internet is cost free zeronet, cheaper slimmed down restricted version of internet. There is also celebrity bond - style payment schemes, for example to pay USA university student costs. Principle is that money is available immediately and that is paid later back. Similar model for third world cost free intrenet / zeronet can be done, like Spotify, has not made profit but is expected to make profit in future. “Netify” service that offers cost free internet in third world, not perhaps profitable now but in future, costs of operation are transferred to future using celebrity bond or other payment method. Abdul K. M. Saifullah: NGOs capacity in development management, 2001. Zeronet / free internet also has no net neutrality, if that makes free internet in third world possible. So in ordinary internet for paying customers there is net neutrality, but not in free zeronet, at least in some time, 5 -10 years without net neutrality from zeronet beginning, the situation can be changed if zeronet with net neutrality comes commercially possible.
EU projects Webinos, CORDIS projects RIFE, WiSFUL, UNIFY, and then iKaaS were for more efficinet internet, and RIFE was “internet for everybody”. Knowhow of those projects could be put to reality and make cost free intrenet in third world possible. Because teleoperators charge extra for virus protection, and in India cheapest 3G or 4G phone line costs are about 2 dollar / month per customer with data restictions, free internet in third world can be odinary phone net connection but without that 2 dollar / month virus protection. Also net content available in free version can be slimmed to zeronet, restricted and slimmed down version of that internet that is offered to paying customers. There is celebrity bond -style paying schemes for example to pay university student costs in USA. Celebrity bond paying scheme where money is available now but paying happens later can be used in zeronet, in “netify” system (like Spotify, Spotify has not made profit but is expected to became profitable in future, and has been great success despite always having negative cash flow). In cost free internet to end user, netify service, limited zeronet service, advertisement paid, is offered in third world, and altough not perhaps commecially profiatable when it begins, it becomes profitable in future like Spotify and costs of operation are transferred to future using celebriy payment type scheme or other, or like Spotify just waiting that zeronet becomes commercial success after few years of negative cash flow. Every year advertisement money for internet is increasing, in third world also, even faster than in western world. Data compression to maximum can be used in text, sound and video, and sparse sampling etc. Cheap musical instrument in third world market can be simple sample player, tracker type (sample player + sequencer, possible arpeggiator etc.). Made with roll printed plastic sheet, inside is cheapest possible memory, in ROM form because manufacturing scale is millions of units (or billion). People in third world have no money to buy expensive gear, so samples from most expensive different synths are used and collected to one unit. If chapest DRAM memory is below 0.4 dollar / gigabyte (slowest and biggest physical circuit size memory), lots of samples fit in very cheap device. 30khz sample rate with 13,7khz useful sound is enough, because human ear has sensitivity peaks in 6khz and 13-13,5khz, and no musical information can ear hear over 15khz, altough 16-18khz is highest heard frequency. High frequency can be made using High Frequency Replication / Reproduction. Bitwidth can be 16 bit or less, 14, 13, or 12 bits in linear PCM because cheap electronics used does not reach 16 bit / 96dB. Using ADPCM 4 bits is possible, like Yamaha SMAF. Delta sigma modulation with 1 bit but 8-16 bit multibit or multirate DSM internally, or pseudo-parallel DSM, makes 1 bit to 8-16 bit accuracy? Dither and noise shaping can be used, in DSM experimental noise shaping techniques reach about 10 bit / 60dB extra noise suppression range. If floating point is used 3 X 9 = 27 bits, + 5 bit exponent makes 32 bit value where three 9 bit (10 bits with hidden bit included) share one 5 bit exponent. NICAM style audio system where instead of white noise dither is used to improve accuracy can be used. Old Ensoniq keyboards used 13 bit samples from 16 bit, similar near-ADPCM like 12 bit can be used if sound changes suddenly, otherwise 4 bit ADPCM can be used, from 16 bit source. Data compression can be used to compress bits needed, for example Finite State Entropy compression. Ordinary MP3, Ogg Opus or AAC will do also, AAC and Opus have High band replication (SBR). If when one sample is played in keyboard it sounds unnatural when scale and pitch changes, 6-8 samples of same thing is needed. Limiting keyboard keys to 25, 32 or 44 makes possible that “real” sample is in every third / fourth key, except in highest or lowest notes, 6 samples cover 25 note keyboard with one sample in evey third key, and 6 or 8 samples 32 notes (6 samples in every 4th key and 4 keys to edges of keyboard distance to nearest sample, if 8 samples are used then sample in each 3th key and 4 key distance to edges) to edges, 8 samples with new sample for every fourth key covers 44 key keyboard if distance to edges are 6 keys long to nearest sample. Highest and lowest notes in keyboard have perhaps one key or more distance to nearest sample than avarage sdistance between samples, but those far notes are not so often used. 8 samples is enough for 54 key keyboard if every 5th key holds “real” sample and 7 key distance to edges of keyboard to nearest sample are used. So now sound is natural and does not sound too much like sample played with wrong pitch. Divide-down electric organ style system / Polymoog can be used also with samples, but top octave generators needed are not manufactured anymore. If sample is taken from memory, sampled at 40/20 khz and then artificially increased to 2 mhz that top octave generator requires, that 20 khz sample upgraded to 2 mhz is the top octave generator. It needs 100 X frequency expansion, if sample is 24 khz (48 khz sampling rate) 80 X expansion is needed, if “artificial top octave generator” is created from sample that is taken from memory and then someway expanded to 80 X or more frequency. EML Polybox style system with samples can also be used if that makes polyphony from just one sample possible. Arturia Minilab has 6000 sound samples and requires 4 gigabytes of memory. Heavily data compressed or sampling rate / bitwidth limited samples make possible to perhaps 100 000 sample memory in only few gigabytes. Sparse sampling like Bonavolonta 2014 “Sparse sampling for sensing nodes” that make possible sample recovery when only 2% of bits of sample remain (50 X compression ratio) can be used. If lowest cost memory is used price is about one dollar for memory only, but because device is made with cheapest possible manufacturing it costs just about few dollars, 2-3 USD or less, even 1 dollar or less if smaller amount of memory is used (memory is most expensive thing in this design). For third world market. Keys of roll printed device are just touch responsive pictures of keyboard keys, no aftertouch etc., but perhaps possible if outside keyboard controller is used. Perhaps light indicators that shows what keys hold “real” samples in keyboard can be used, different samples use different 6-8 places in 25-54 key keyboard. When other keys than those holding “real” samples is pressed, sound is typical note-shifted sound from nearby key that holds “real” sample. Those 6-8 samples are from same sound source, but recorded from sound source in 6-8 different note positions. This is for acoustic samples, if electronic sounds from synth that is sampled uses just one sound shifted across several notes/keys, multiple samples for one sound are not needed but just one like in original source material. In PolyWaves VST samples are used like VCOs (“sample reader / sampliser”) to produce polyphonic sampling, that may be solution to polyphonic sound generation from just one sample. Old polyphonic synths in 1970s used one voice per octave, perhaps one sample per octave is enough for (limited) polyphony, if PolyWaves method is not used. Martin Vicanek has made several improvements in digital audio. 90% or more of cheap digital roll printed sound synth price is just memory, and 100 000 samples are in sound palette that has samples from best hardware- and softsynths (most expensive and best synths) in memory. For example Opus has 9 kbits/sec minimum music sampling, so 112 megabytes is enough for 100 000 low quality one second samples. Using TwinVQ (or Additive Quantization AQ, in sound sampling) or ADPCM like recursively indexed ADPCM RIQ-ADPCM is possible. Takis Zourntos one bit sampling is one bit ADPCM- like, so perhaps RIQ-ADPCM principle or AQ can be used with it. This simple sampler has ROM memory for 100 000 or less samples and some RAM memory for user samples, and inbuild sequencer so it is a tracker. It is sample player, sampler and sequencer. Price can be even below 1 dollar, if memory is cheap enough and roll printing used in other parts than memory. In netpage Sequencer is german language “Konzept : 1 X complex oscillator und 1 X sampler”. This tracker is hardware, not software like usual trackers. If memory must be saved old style “samples based on wavetables” principle like Fairlight can be used. If real synths are manufactured cheaply using roll printing, they can still have huge preset memory that is cheap, for example Dexed / DX7 has about 200 000 free presets, so internal 100 000 preset memory is possible in ROM, and tens of thousands empty RAM space for user s own presets. Price of synth is still minimal. Anadigm FPAAs can be used in more expensive (but still cheap) analog synth devices. FPAAs have noise and other problems that prevent them to become mainstream in high tech synths, but aim is to make cheap as possible musical instruments, so noise etc. is not aproblem then, and those Anadigm FPAA based analog instruments can be manufcatured at scale of millions cheaply. One FPAA can hold 128 filters (onepole?), and oscillators also can be constructed. “Anadigm demo industry s first analog synthesizer fully integrated with software controls” 2004. Oscillators, filters and amplifires can be Anadigm- based in (cheap) analog tech. ALD EPAD, Microsemi SmartFusion, Triadsemi VCA are other FPAAs. Making 8 VCOs per voice like old Oberheim and 4 analog filters per voice like John Bowen Solaris is possible when using FPAAs, cumulative noise is not a problem because those are cheap synths for masses. If 128 voices (or more) is made using FPAAs only solution for user interface is VST style software GUI, because no amount of knobs is enough for 128 or more voices analog synth. Instruments can be marketed as “toy keyboard” for children altough they are professional synths, for eaxmple cheap digital synths or samplers / sample players / romplers that cost few dollars or less can be marketed that way in third world countries (those ultra cheap devices perhaps never appear in any rich country). Simple VCO and VCF designs that are in netpages EEWeb extremecircuits, Circuit-finder and bowdenshobbycircuits can be used, in programming FPAAs. Also “simple wavefolder” design that is possible to build also can be combined in analog or digital form in those synths, in discrete circuit or in integrated circuit form, now both “east coast” and “west coast” synthesis methods can be used in same device. If cheap synth is just cheap microcontroller or cheap PC processor wavefolder can be in software. US patent 8487653B2 “SDOC with FPHA and FPXC:” by Tang System, 2009. Building cheap analog synth can bypass FPAA manufacturing stage and jump directly to analog ASIC manufacturing (Anadigm has also ASIC production in its FreezeFrame line), then analog synth is completely in single analog IC chip, polyphony and all, or modular synth can be build with several analog ASICs. Cheapest way for modular is digital Lunetta-style, and chips with similar functions can be roll printed now cheaply, and mixed with analog CMOS series silicon chips if needed. Way upward is digital modular synth using cheap microcontrollers, for example “Digital Formant synth (D-Formant) 130374” in Elektor magazine 2013. Real analog modular can be build cheaply using components from another Elektor magazie article “Analog synthesizer” 2016 but not in its “minimoog” form but modular monophonic. Similar cheap ASIC circuits can be used for cheap modular. Serge, CGS, MFOS, Yusynth, AE modular, Aries, Paia (1970s), Polyfusion (1970s, in monophonic not 8-voice form), can be used as model for super cheap monophonic modular analog synth, even using just single ASIC chip. Inside chip must be wiring for different modules, and outside of chip wires (cables) that connect modules, for example like AE modular does. Curtis and SSM ICs can used as building blocks if needed, because they are again available (they were used already in Digisound 80 long time ago). Using those, or transistor / opamp arrays and other components in multi-chip module, or special ASIC, or FPAA can be used. ASIC and FPAAs make possible polyphonic cheap synths, not modular, but semi-modular (Rhodes Chroma, Oberheim Matrix 1000), or just ordinary prepatched. Analog ASICs are made even 24 micron old manufacturing tech from 1960s up to today, and 24 micron manufacturing should make high quality analog circuits. Sequencers that cheap sample player and other synths use can be very complex, because exremely complex sequencers are now available in Dhalang MG, Ornament&Crime (using ARM microcontroller), Kirnu Cream, and VSTs Modulys, PolyWaves, Subconscious VST, Seeq One, Photon VST, Pandemonium VST, However if sample player is very simple, extremely complex sequencer that needs lots of processing power is not suitable, simple old tracker style can be used. Because Anadigm FPAAs are so cheap there can be several oscillators per one voice and one filter, for example 4 osc or more per voice/filter, and still 128 voices (or more) available, if multiple FPAAs are used. Sound synthesis using MPEG4 structured audio (Bonneville CPS), MPEG4 AABIFS (MPEG7 audio also) is that standardised synth is in hardware IC or software, so one standard programming is used in hardware synths or softsynths. This MPEG standard synth can be in cheap IC. Le Sound AudioTexture and “A new paradim for sound synthesis” are another new methods. If modern synthesis is not used, simply collecting old 1980s Yamaha DX7, Tx81 and OPN2, Casio CZ, Roland MT-32, Korg M1 and Kurzweil K250 to one cheap IC can be standard IC for cheap synths, Yamaha and Casio use almost similar FM method and Korg / Kurzweil / Roland wavetables and rompler, so circuits perhaps can be someway partially unified into one. Roll printing can be used. Fairlight and Synclavier were expensive in 1980s, but now their circuits can be roll printed at almost no cost. Hydi VST and Metatron 2 VST. If analog synth is made cheaply, it uses digital modulation, sequencer, ADSR, envelopes and LFOs (but oscillators except LFOs and filters are analogue, so signal path is analog, virtual analog LFO is absolute similar to real analog LFO if wanted, so no difference in sound). Because modulation and sequencer are digital and can be very complex (and LFOs also), cheap PC CPU chip like Atom or ARM with 4-5 dollar price can handle digital tasks, and because sequencers and modulation etc are in software different sequencer and modulation software can be improved and changed like softsynhs do, so that makes analog synth very flexible (but still very cheap). Result is analog softsynth in hardware form (real analog oscillators and filters). Different “softsynths” can be used in this analog hardware for modulation, ADSR, LFOs, and sequencer, because it has x86 or ARM processor for those duties. Digital software can also use digital oscillators and filters and bypass real analog circuits, creating digital-analog hybrid synth from real analog synth, making synth even more flexible. Bucket brigade ICs can be used for delay/chorus/flanger. Those are available at cheap price from China, but cheap BBDs have noise problems. But in cheap synth noise is not a problem, and every voice can have its own cheap BBD delay/chorus/flanger (even in 128 voices), digitally controlled like everything in this synth. Combining digital control from CPU and FPAAs will make very flexible synth. If only cheap digital softsynth platform is used for making cheap music device (no hardware synth circuit other than CPU), Intel Quark is microcontroller (and other x86 microcontrollers exist), but perhaps it has computing power of early 1990s / 2000s x86 chips, so softsynths of that era can be used in Quark microcontroller (in modified form), in very cheap device, and new softsynths specially made for cheap microcontrollers. Armonyx was softsynth for PC already in 1985, and used very little computing power (compared to modern designs), similar can be used in Quark at ease. Audion VST by Transonic uses new music description language called “Fast”. If music device is just sampler/sample player, granular synthesis can be used to turn sampler to synth (Borderlands iOS app etc.). Timeline based modulation (Progress audio Kinisis) can be used in device because it is simple. TC-11 by Bit Shape and Concentric Rhytm iOS apps are sophisticated sequencers. Zen preset browser and VST Plugin Preset Match can be used to find presets that work together in different softsynths. Discrete summation formulae synthesis (DSF, Moppelsynth) creates many harmonics easily, like FM modulation (vector phaseshaping and phase offset modulation). GPU instead of CPU can be used in synth (additive-, granular-, physical modelling-, spectral modelling-, digital waveguide-, Bezier curve-, raytracing-, modal-, formant-, DSF-, vector-, convolution-, wavetable and polygon wave sound synthesis, and version of FM synthesis also). Subtractive synthesis is not suitable for GPU. Fourier transforms and filters can also be done in GPU, and “vector oscillator” by Vicanek perhaps also, “The Geometric oscillator: sound synthesis with cyclic shapes” 2017, scanned oscillator, granular oscillator, Kronos (Kronos PWGL) is music description language that is specially suitable for vectors, altough not designed for GPU computing, Hasan Hujairi: “Parametric composition” as possble approach for non-western art music. If microcontrollers are programmed for softsynths, Forth based “Formula”, Ample and APL based “Dyalog”, TCL and Q/Pure are compact programming languages. Dyalog is for vectors also. Other compact music programming languages are Sporth, “The Smoke”, Sonic Pi, Mondrian, Python Snack (sound processing enviroment). GPU programing languages are Futhark, Harlan and LambdaCube3D (rare ones). Microcontroller based platforms are Zynthian, FreDSP and others. Real cheap, about one dollar or less plastic sheet has ordinary printed picture of synth with touch sensing parts for keys and knobs and switches. Inside this ordinary plastic picture is either roll printed electronics made by fine-accuracy printing machine and put on ordinary plastic picture, or microcontroller, or both, and then cheapest possible silicon memory. “softsynths” that use microcontroller have no GUI, they use primitive controls of playable picture. There is 4 kilobyte art, 4 kilobyte softsynths, tinyapps netpage have similar minimal programs. Dexed and other efficient programs work in microcontrollers also (needing much more than 4 kbytes). Instead of 4 kilobytes perhaps 4 megabyte class for softsynths, GUI memory not counted in 4 MB because one standard GUI is added to softsynth like Blue adds GUI to CSound. More important than memory space is how much softsynth needs from CPU processor resources. Because fastest consumer CPU is now 1 teraflops and GPU about 30 teraflops softsynth that uses 0,1% of CPU power class from CPU and GPU can be created, so that is 1 gigaflop from CPU, and 30 GFLOPS in GPU if softsynth is designed to use GPU not CPU computing. Older CPU s and GPU s need more than 0.01% but this can be marketed as 0,1% processing power softsynth if only speed of fastest processors is used when testing these softsynths. Simpler like monophonic softsynths can have even smaller than 4MB class. Softsynths are almost all made with SynthEdit or Flowstone (SyntHMaker). Becuse there are (propably) over 10 000 free softsynths (or VST format or Linux / standalone, and there are just few bulding blocks for these most of these (some are not made with Synth Edit or Flowstone, but those exceptions often have other standard “building blocks”). To save memory space, if cheap PC or other device for third world market has inside SynthEdit and Flowstone modules, and only those parameters that programmers use when building softsynths from those building blocks are in softsynth itself, that saves from copying those modules over 10 000 times, and makes memory saving of almost 10 000 X possible. If several softsynths are used in same PC same time, several copies of those building blocks must be in PC memory, now memory saving is not 10 000 X, and GUI must be also (but it is build with standard building blocks often also). When softsynth is loaded from internet or memory card or DVD-ROM, only those parameters that make individual softsynth from those building blocks are transferred, and GUI also. It perhaps takes little time when PC “builds” softsynth from those parameters it has received, but end result is same softsynth who use building blocks that come every time when softsynth is loaded, altough they are just same building blocks every time (and owner of free softsynths now has thousands of copies of just one program, SynthEdit or Flowstone, if he has collected thousands of softsynths). But free softsynth building programs are rare, using Csound as base like Cabbage and Silence, or PD in PdVst is possible, but most of free Synthedit style development enviroments are often incomplete and very rare (AudioKit and SonicBirth are for iOS only, Cuence Construct, Transonic Audion, Open Sound World, Soundpipe/Sporth, Dplug, Bonneville CPS, Audio Plugin Generator are rare, and Jsyn, Libmodsynth, SndObj library, Sqnc music editor, Axiom synth). But few standard building block systems for softsynths that are free is needed so memory saving softsynth platform(s) in third world and everywhere can be used, free building blocks in PC memory and “softsynth” is just parameters to those blocks. Cabbage and PdVst is something similar. To save memory space, and packed in DVD-ROM ecodiscs and then sold from example 100 free softsynths and effects in 1 dollar DVD-ROM. Now free softsynth developers get money, and weak or nonexistent internet connections in third world are not needed for loading content, if they use internet, such amount of data when loaded through internet will propably cost in third world more than that 1 dollar. Ecodisc DVD-ROM can be multi-layer, and two-sided disc so more gigabytes can be in one disc. DVD-ROM because price of disc / software can be printed in the sleeve of disc, so those who attempt to sell higher price wil be revealed. Memory card packages can also have price printed in package, cheap small plastic bag wth price printed in bag and info, and inside is memory card with software, but copying memory card data is easy, so DVD-ROM is better. Actually chinese CBHD disc is better, it has 15 GB capacity and can be made in standard DVD factory. CBHD standard does not use western audio or video coding (no high license fees) and three layer HD DVD that CBHD is based has 3 X 17 GB capacity, so dual side triple layer Ecodisc ROM disc has 102 GB capacity. This CBHD - ROM disc can also be used as cheap mass memory in very cheap PCs in third world, in partially replacing tasks of hard disc. Manufacturing costs of ecodisc CBHD is minimal in third world or chinese factory. Ordinary ecodisc DVD -ROM is even cheaper, and can store 4,7 Gb single layer, or in 80mm diameter 2 GB and 60 mm disc 1 GB. Audio and video can use DVD audio system, no CD standard needed, audio disc is DVD without video but audio only. Using 20 bit 48khz sound to avoid copying to CD piratism. I counted 80 and 60 mm disc capacity for available area of discs compared to 120 mm, altough oficial capacity of 80mm DVD is less han 2 GB. 60 mm disc has over 70 minutes recording time in 20 bit 48 khz even counting the fact that DVD audio track cannot use full (video) capacity of DVD in DVD-audio standard that is not used in west anymore. Using 20 bit 36 khz 120 minutes is possible. So 60 mm mini-DVD audio ecodisc is minimal price music distribution system, and small size of discs make transportation and storage cheap. Also plastic adapter that plugs small discs to 120 mm size or using manufacturing of 120 mm plastic disc that has only 60 mm reflecting surface can be used. In third world. Price is minimal. Instead of HD video like CBHD or Blu ray simple 3 X DVD standard that offers HD video in DVD can be used in third world. So CBHD can basically be used just as mass memory RAM / ROM for PCs in third world, for example storing softsynths, but as video disc also. If audio compression is needed in audio or video disc chinese DRA or Ogg Opus or other free codec can be used. Because four core ARM SoC is available for tablet PC / phone for 4 dollars (or less) and four core Intel Atom for tablet PC for 5 dollars or less, really sophisticated but still cheap softsynth platform muiscal insrument is possible to make cheaply. No internet connection, softsynths are transferred using memory card or DVD-ROM Ecodisc (then DVD-ROM / CBHD player is needed in device, it can be used as media player / small TV also not just as softsynth platform). One dollar price DVD ROMs can be sold in third world and include large number of softsynths in one disc (those that are free in internet), now free softsynth developers get money. Cheap chinese elecronic toy pianos cost 1,5 dollar cheapest. When in this toy piano are added few knobs and switches and modulation wheel, and perhaps DVD player made from cheapest portable chinese CD player componets with DVD (DVD-ROM) playing capacity altough it is not used in portable CD players, and then less than 4 dollar ARM quadcore or less than 5 dolar Intel Atom quadcore or compatible chinese x86 or other x86 cheap processor, result is super cheap professional synth that has modern synth features but it is build using cheap chinese toy piano as manufacturing model (in same production line even). Price (selling price) can be 10 -15 dollar, at least version with just memory card. I compare that price for cheapest chinese Android smartphones that have factory price of 6-7 dollars or so in large orders. But this is not smartphone but synth in toy modified piano manufacturing (using same cheap ARM SoC than phone). In ARM SoC also GPU can be used in sound processing (ARM Mali). Price is still very cheap even with DVD-ROM player, and real proper keys and knobs in (toylike) keyboard. If cheapest screen (touchscreen? Or electric ink black&white or other cheap type) is added, similar like in cheap phones, softsynth controls are not just hardware knobs anymore, and this thing can be used as PC also and as DVD / CD player. Ultra cheap (chinese) video glasses and data gloves are another option. 15 -20 dollar price when cheapest components and manufacturing (China, India) are used? Simple music Linux distro (like m-dist) can be used, or free version for Widows for cheap devices, or use Windows VSTs through emulator. That four core ARM processor is open source (used in Pine64). Simple small loudspeaker is needed (toy pianos have those), like roll printed electrostatic loudspeaker that is perhaps even cheaper than used in toy pianos. Old hungarian Muzyx81 (Muzix81) programming language used Sinclair Spectrum PC as platform 1982, and before 1985 was capable of sampling and Fourier transform in additive synthesis like Fairlight used. All in Sinclair Spectrum CPU from 1984-5. So Muzix81 was the world s first softsynth, before anything similar was done in west, like Armonyx 1985 and Tecmar softsynth 1986. If Muzix81 was so capable using 1983 era Sinclair Spectrum, even ARM 0+ series CPU microcontroller, that is most primitive today, should be capable of using enormous performance “supersynth”, because ARM 0 microcontroller has enermously more processing power than Sinclair Spectrum from 1983. So sound synthesis functions can be used economically, and sample rate 30khz and bitrate 10 bit integer, because softsynths often have just 60 dB (10 bits) dynamic range, altough they use 32 bit floating point processing. So very capable softsynths should be possile to be made even in most primitive microcontrollers. Bela Platform / board (not primitive but efficient softsynths platform, Pine64 is another example). But even microcontrollers should work in music platform. “PCH2CSD: an application for converting Nord modular G2 patches” is for converting G2 patches to Csound. “Multi-channel simultaneus data acquisition through a compressive sampling-based approach” 2014 that only 1/250th of bandwidth is needed to reconstruct signal, leading 250 X compression ratio. Xampling is analog sampling below Nyqyist rate, how it can be done in analog sound synthesis or not I don t know. Walsh-Hadamart transform can be done in analog music synth however, instead of Fourier transform. Feeback amplitude modulation, vector phaseshaping and allpass filter chain modulation are new sound synthesis methods. DocNashSynths has made many new sound synthesizing techniques. Joaquin Aldunate: “Modular composition enviroment”. IRCAM Orchids. “Macleyver command language” was for controlling analog synth with digital programming language. Linux synth-a-modeler compiler. Michael Norris Spectral Magick VSTs. Moselle Synthesizers VST. Firebird VST. Zero Vector VST White Noise Audio. If roll printing as used in electronic components inexact computing / approximate computing can be used and “bespoke processor”. “Alias-free nonlinear audio processing (ALINA)”. FFT transforms can be replaced, with sparse FFT or Walsh-Hadamart, or in Google goups forums “Why sine wave in Fourier series?” 10.8 1999, and then answer 12.8 1999 by Timo Tossavainen, Karhunen-Loeve transform is best (but works in analog only), wavelet, packed wavelet, Slant transform for saw waves, Hartley transform, permutation matrix for DFT. Because analog tech is now manufactured in 16nm Karhunen-Loeve transform can use it and be faster than digital signal processing. Mcleyvier command language was for controlling analog synth components digitally, special programming language for analog signal processing. Perhaps similar is needed now? Programming language for KLT and other analog processing in 16nm circuits, like anamorphic strecth transform (AST and PST) that has functions for analog circuit behaviour in programming language. If softsynths have polyphony that reaches thousands of voices, large screen is needed, for example 80 inch TV screen that has separate touchscreen adaptor screen that is sold nowadays to addition to TV set screens for those who need large touchscreen. Now virtual knobs can be turned in touchscreen by touching picture screen, not keyboard controller, and because screen is large thousands of knobs and sliders fit in it, if it does not fit zooming of the screen can be used. So thousands of voices softsynth or hardware synth that is controlled straight from touchscreen can be used. This is for sophisticated western softsynth only, not third world version. Lowering sample rate to 30 khz and uisng 16 bit integer or 16 or 8 bit floating point, and alias free waveforms makes thousands of voices softsynth possible. High Frequency Replication makes possible to add frequency from about 13,5 khz sound if integers are used (1,5 khz goes to quantization noise). Or 27,2 khz sampling with 16 bit FP, no quantization noise as much. Even 12 khz sampling rate is possible, 6 khz sound with 16 bit FP, up to 12 khz using HFR and then that to 18 khz using aural exciter. Integer accuracy can be increased using G.711 telephone standard type 16 bit to 13 bit or NICAM style 14 bit to 10 bit least significand bits dropped. 28 bit companded to 20 bit and then that 20 to 16 bit using Sony Super Bit Mapping that makes quiet noise shaping without affecting much signal waveshape / sound (SBM claimed 22 bit accuracy from 16 bit sound so from 20 to 16 bits without LSB dropped is realistic because it is already dropped if signal uses companding from 28 to 20 bits before SBM). Waveform is processed as 16 bit integer and when played back expanded to 28 bits again. For softsynth waveforms.
For enterntainment there is tens of thousands PC and other computer games (for Android for example) that are free. There are also dozens of free game engines. So if cheap PC, tablet PC or phone has game engine or several game engines inside, every time game is downloaded game can use game engines inside PC or phone, so downloaded data would be smaller. Hundreads of games can use same game engine so bitrate and memory saving is very efficient. This is for third world computer games. Also web games from browser can be downlaoded and saved as offline version, like some gaming netpages offering free web games are doing already. PC games are very large, from 40 gigabytes to 100 or more GB. In netpage Gameslay is compressed games, 300 MB, 200MB, 100MB, 50MB and 20 MB classes. These can be standard PC game size classes for third world market for free games, because highly compressed versions of popular games fits in 300 MB. Then could be 1,44 MB size minigames for PC (old floppy disks had 1,44 MB capacity). For mass storage DVD-ROM or CBHD-ROM (Chinese version of HD-DVD) can be used. DVD and CBHD can use either chinese audio and video codec or free audio and video codecs. MP3 is now free. Those optical discs can be used as cheap PC memory storage, in ROM and RAM form if rewriteable discs are used. Three layer 51 gigabyte HD DVD was proposed, doubble sided three layer CBHD disc then has 102 gigabytes capacity, and manufacturing can use ordinary DVD ecodisc factories in China or third world countries so manufacturing cost is minimal. Audio can be stored in soundtrack of DVD disc, making cheap video / audio media player possible without CD playing capacity. DVD needs always video track for system balancing, but there is no minimum bitrate for video. 20 kilobits or even 1 kilobit / sec video bitsream is possible in DVD. So if 20 bit / 48 KHZ is used for audio and 20 kilobits for video that shows for example TV test picture while music is played, that is 99% bitsream for audio and 1% for video. Discs can smaller than 12 cm, 8 CM mini- DVD and 6 - 6,5 cm “business card” DVD size. Single layer 8 cm disc has 2 GB capacity and 6,5 cm 1,27 GB if reflecting area surface is compared to 12 cm disc. In small 6,5 cm ecodisc about 90 minutes of music at 20 bit / 48 khz fits and 20 bit / 36 khz is about 120 minutes. So super cheap music distribution is possible using very light and small discs that mule or camel can carry to bazars around the world. 6,5 cm ecodisc DVD packed in paper sleeve. Ecodisc is half of ordinary DVD weight, and 6,5 cm diameter and paper sleeve makes disc even lighter and smaller to transport or store. Or use 12 cm plastic disc that has 6,5 cm reflecting surface, manufacturing price is still minimal. Small mini PC with DVD / CBHD player and memory card slot and cheap display, or simple portable media player without PC processor. And tablet PC size device without optical disc player is another version, with solar panel electric charger that is separate from tablet PC. Or use tablet PC bottom for small DVD / CBHD player like small portable “discman” player, so result is tablet PC and discman player in same “dual side” device. Ogg Opus for sound or CBHD chinese audio / video codec can be used. For video also free codecs like Google s, or Libde265 etc. can be used in hardware. 65mm ecodisc would be almost same size as old 64mm Minidisc so jewelcase Minidisc packages can be used as 65mm DVD video film packages etc, or simple paper or cardboard packages that are almost same size as disc, for music and videos. For music production toy piano modified to tablet PC with processor SoC and additional knobs and sliders. Problem is compared to computer gaming that there is no much free softsynth “building blocks” that are in use except Synthedit and Flowstone. Similar free exits but are rare. If inside every PC is free “audio plugin generator” only parameter changes to those building blocks are needed to download and hundreads of softsynths could use just one Synthedit style building blocks inside PC. No GUI needed because program inside PC automatically adds standard GUI to softsynth, like Csound and Pure Data uses GUI layout separate from softsynth. So Csound and Pure Data already are those free standard softsynths, but something like free VST style system could be done. MPEG4 SA and MPEG4 AABIFS were hardware designations for standard hardware synths (not software) inside every PC, so like old soundcard standards with hardware synth. But no hardware using those MPEG standards was made. MPEG 7 and MPEG21 are also. If MPEG hardware standard ever comes to reality, it can use most computationally effective synth methods (subtractive synthesis without filters, LP-BLIT bandlimited impulse train etc.) , aliasfree waveforms (ALINA project) and low sample rates. Also analog sound circuits in FPAA structure or other can be used. Analog electronics is made now at 16nm so hybrid digital analog or all analog that has 32 voices and uses KLT transform or Anamorphic stretch transform can be cheaply mass produced, and can be in phone like old ringtone sound cicuits etc. and uses low amount of electric power. Bonneville CPS. In previous post was listed netpages for simple VCO and filter designs. If inside for example 180nm manufactured analog synth cicuit is FPAA structure variable 32 to 128 voice structure can be used, 128 voice being very simple circuits. There is “simple transtor organ” and “simple electronic piano” circuits in intrenet, so cheap and simple transistor electric organ and analog electronic piano can be inside FPAA structure analog chip also, offering real analog electric organ and elecric piano sounds instead of samples, altough sound quality of those simple circuits may not be so good. Digital side of MPEG hardware synth can be 128 voce like Yamaha SMAF, but much modernized using most modern and effective synth methods. So result is 32 - 128 analog plus 128 voice digital circuit. Things like ADSR and sequencer can be in software, and LFO also if it is used. Most modern digital synth methods do not even use filters, so perhaps not LFOs also. Things like complex FM sound synthesis and other that are difficult to do in software can be in standardized MPEG chip that is cheap to manufacture, and can be inside every phone and PC like MP3 player. If real softsynths are used there can be classes of 10 MB, 4 MB, 1,44 MB and lastly 400 kilobyte minisynths. These refer to “four kilobyte art”, but are for sound not visuals. If sound generator blocks are in PC memory and GUI is automatically added by PC actual synth is just changing parameters to building blocks that are already inside PC. Only GUI blocks parameters that can be changed is perhaps colour, 4 bits for 16 different colours for differnt GUI blocks that softsynth can decide, all other comes automatically from PC. If low CPU and GPU GFLOPs usage is in softsynth hundreads can be in VST host simulataneusly. There can be class of 1% or 0,1% CPU / GPU usage softsynths. If 1 teraflps is maximum CPU speed today and about 30 teraflops in GPU, synth that uses 30% of its processing in GPU and 70% in CPU has then 9,7 gigaflops in use if it uses 0,1% of computer resources. Effective and low power requirement PC can be build using Kalray CPU. It needs special Linux and can use Windows only through software emulation, but it uses very small amount of electricity and has teraflops computing power. So cheap but fast PC can be build. Altough GPU is most powerful part of PC today ratio of CPU versus GPU computing power can be changed if Kalray CPU is used, altough Kalray nay have difficulties (normal programs PC programs do not work properly). But if cheap, powerful and low power requirement PC is needed in third world Kalray processor can be an option, altough it is more DSP than CPU processor. There is 80 free game engines in Wikipedia game engines list, and that is not perhaps all free game engines available in the world, so there is lots to choose for example putting several free game engines inside every PC so actual games are not so memory intensive if every game when downloaded have game engine inside, now this can be inside PC. But for softsynths there has few free “patcher audio programming languages” meaning visual audio programming languages to use like free VST. There is Csound and Pure Data and other less known like PWGL, Kronos and Aura (Dannenberg). Dannenberg: “Is music audio processing embarrassingly parallel?” If GPU computing can be used in softsynth it saves CPU a lot. Monolog X EcoSYSTEM, MobMuPlat, and “Mrs Watson” are examples towards free VST-style systems. Cabbage, Csound/Blue and perhaps Common Music / Grace, IRCAM Najo are suitable for VST style free sound synth library from which only some outside parameters are needed to build softsynth. Actual building blocks are inside PC. There is “music DSP archive” netpage. Together with softsynths can be standardised cheap hardware synth chip. Divided to analog and digital parts and both can be manufactured as separate die at 180nm (or smaller nm) to save cost (analog and digital separately made) but are packed together and work together. Digital synth that uses wavetables or samples or digital waveguide (Yamaha YMF 724) has then lots of software based control, and ADSR and sequencing in software also. So result is mixture software/hardware synth. “Omnisphere 2.5 hardware synth integration” as one example. Difficult synth methods like complex FM can be in hardware. Analog part can be FPAA based flexible. Using simple few transitor based VCO and filter structures voices can maximum 128 from standard 32 voices. Simple wavefolders and simple electric organ circuit and electronic piano (analog transistor piano sound, not electric mechanical piano like Rhodes) can be in FPAA, altough due to simplicity of circuits electric organ and electronic piano analog sounds are perhaps “toylike”. Samples can be used also, but that is not analog sound. “Simple electronic piano using 555 timer”. “Simple electric organ circuit” (many different in internet). Analog part can have bucket brigade flanger / chorus / delay in every voice or just one BBD circuit shared between all voices if needed because it is FPAA structure, and filter can be Mutronics Mutator based. Control can be “timeline based modulation” or “one knob” style if simplicity is needed. If softsynth has thousands of voices it not only need 70 / 80 inch touchscreen, or virtual user interface with data gloves and virtual reality glasses so that thousands of knobs can be used, and sound of it must use 360 degree soundfield so that thousands of voices at least somehow separate at each other. HRTF sound like old Q sound is needed if only two loudspeaker is in use or in headphones panning synth sound not from left to right but left, right, front and back and using HRTF functions to make sound coming from 360 degrees is needed. Synth waveforms if they use smaples or wavetables can be made using 32 bit floating point, then put to synth memory using DAC that turns 32 bit FP to 24 bit integer and then that to 16 bit integer. Now perhaps 16 bit integer has few additional bits of accuracy compared to plain 16 bit integer? Or use 32 or 64 bit integer and that to 16 bit. This 16 bit integer is taken from memory and used in waveform processing of synth, and when sound is played back through 32 bit FP DAC or 32 or 64 bit integer DAC that adds “header” to 16 bit sound, making it 32 or 64 bit, then 16 bit sound has again those few extra bits of accuracy? Floating point or integer 32 bit can perhaps be truncated to 8 bit for memory and processing, and played back as 32 bit sound with few extra bits of accuracy compared to just 8 bit sound? “DC-accurate, 32 bit DAC achieves 32 bit resolution” 2008. Audio bitrate / sampling rates other than 16 bit / 44,1khz must be used to counter pirate CD sales. 36khz / 20 bit, or 40khz / 20 bit or 16 bit can be used.
Old Minidisc sized DVD / DVD audio discs that are manufactured like ecodisc CDs would be suitable when digital distribution of video and audio content (films and music albums) are made in third world. Minidisc package had 105 X 86 mm dimensions. If disc is 65 mm size it fits inside that. Or simple cardboard or paper sleeve as package. 65 mm ecodisc has minimal manufacturing cost, but capacity as DVD disc is enough. Chinese HD DVD or other efficient but cheap / free video coding can be used. However video content is not in HD format, it can have same pixel count as PAL or NTSC TV or even smaller, and not HD quality. New experimental systems are “Nanoscale track - follow performance for flexible tape” (for tape), “Nano- optical long-data memory”, Hyper-CD ROM (Storex) and “Digital Multilayer Disc”. Those are said to be cheap alternatives to ordinary DVD. As audio disc 65 mm diameter DVD has about 130 minutes of storing capacity. Using lossless compression like FLAC it is about 230 minutes. Lossy compression uses 16 bit bitdepth, altough cheap media players and smartphones, even some expensive smartphones have only about 70 decibel dynamic range and noise ratio (SNR ratio). Some portable CD players / radios with CD have only 60 decibel dynamic range and noise ratio. So if smartphones / feature phones or portable media players are medium of digital content, audio of 16 bits is too much. When internet connection is through wire to desktop PC 16 bit is suitable, but not for wireless communication. Almost all lossy compression uses 16 bit bitdepth. If goal is to compress audio as small as possible it makes no sense use 16 bits in 70-60 decibel device. 12 bits (72 decibel) or even 10 bit (60 decibel) linear PCM is enough. And lossy compression makes audio signal bad quality anyway, so bitdepth drop is not so important. However there is no option in audio compression to use lower than 16 bit bitdepth. MP3 file has about 40 decibel noise (SNR) ratio, which is 7 bits, and maximum noise ratio of MP3 is 64 decibels (in 320 kbps?) I read. So 16 bit audio is not needed. Even 8 bit audio with lossy compression can be used, old C- cassettes with Dolby had 46 decibel max dynamic range but sounded just fine, better than MP3 files. Lossy compression like MP3 or Ogg Opus or AAC are not using noise shaping / dithering either, altough it would bring efficiency also? Also lossless compression like FLAC are using dithered / noise shaped audio in special cases, for example when 32 bit audio is converted to 24 bit dithered and that is then FLAC- compressed. Using dithering and noise shaping in lossless compression standard can then make compression more effective? 8 bit noise shaped and dithered sound can have 14 bit dynamic range or even more. “Hidden” noise shaping like Sony super bit mapping adds 4 bits of accuracy in sound but because it is hidden sound can be used as standard linear PCM soundfile. FLAC have options like “ReplayGain” and “stereo side - mid side” adjustment that makes FLAC partially lossy if needed. FLAC is also fast to decode. So suitable for simple media player. If high frequency replication is added to FLAC it makes FLAC even more lossy. There are new lossless compression methods like “achieving lossless compression of audio by encoding its constituted components (LCAEC)” 2018 Uttar Mondal, “Lossless compression catalyst based on binary allocation via modular arithmetic”, patent “Audio lossless compression and frequency decoding method” CN103280221A. Using sampling rates other than 48 khz, like 44,1 khz, 40 khz, 36 khz, 32 khz, 24 khz, 16 khz and 12 khz and then high frequency replication is possible. if 12 bit linear PCM 32 khz (either lossless or lossy compressed) is sent to phone, it has about 14,4 khz audio range when quantization noise is removed, but high frequency replication can improve high frequencies, nobody would notice difference from normal 16 bit 44,1 khz soundfile. If lossless compression is not enough to store lengthy audio material in small 65 mm disc, in 40khz 16 bit, using compression rate that makes sound max 58% of size of original, and then if lossless compression is not enough to fit sound in disc some lossy form is used. Usually lossless compression makes 50 - 57% size compared to original, but sometimes up to 70% large. But 58% is enough in almost all cases, so downgrading from 70% to 58% does not affect sound so much. Only when 58% is not enough “real” lossy compression is used. ADPCM compression is not used in audio anymore, altough its low bitrate would be suitable for DVD film audio track. ADPCM has high frequency problems, but system like high frequency replication can improve it, HFR that uses high frequency ADPCM information as cue and then corrects ADPCM making it somehow better? New ADPCM schemes are “Subband/ADPCM audio coder using adaptive vector quantization”, “Low-delay vector-quantized subband ADPCM coding” 2015, “Ultralow latency audio coding based on DPCM and block companding”, “Cascaded prediction in ADPCM codec structures”. Some of those ADPCM codecs need only 2 to 2,5 bits per sample for near transparent quality of audio. “Designing a hybrid codec with the integer MDCT and to estimate the audio quality by of SPL and CR”, Hydrogenaudio “Good 1-bit ADPCM codecs” 2018. Apt-X is also ADPCM based and it has almost- lossless mode of 572 kbs with 48 khz 24 bit, that would make 320 kbs 16 bit 40 khz. “Bit rate reduction BRR” is another method. ADPCM can use perceptual compression and Huffman coding and other such methods but are not used in ADPCM coding as often as linear PCM lossy codecs. Making ADPCM version of nowdays linear PCM lossy audio codec would make bitrate smaller than linear PCM, if ADPCM in 4 bits has almost 16 bit quality (except in high frequences). Oversampling increases quality of ADPCM, so 96 khz 4 bit ADPCM is much better quality than 48 khz 8 bit linear PCM. Using 10-12 bit floating point or 8 bit posit instead of integer value would increase audio codec efficiency also. ADPCM with 5 bits is possible but not higher bit depths are suitale for ADPCM. Subband ADPCM like ITU telephone standards, or other like “ADPCM in multiple subbands”, “Delay free lossy coding using shelving ADPCM”, recursively indexed vector quantizer RIQ-ADPCM, RIVQ-ADPCM, noise feedback coding NFC-ADPCM, adaptive quantizer DPCM (DPCM-AQB), backwards adaptive DPCM, dynamical DPCM (DDPCM in “bitsnbites eu” netpage). Pyramid-, spherical-, cubic- and spot vector quantization can be used, Additive Quantization AQ, “Optimal noise shaping using least squares theory”, Fibonacci delta coding etc. Even 4 bit linear PCM can be used, 16 bit to 70 decibel quality which is about 11,5 bits, that truncated to 8 bits which is similar truncation used in NICAM sound (14 bits to 10), then that 8 bits to 4 bit noise shaped and dithered sound. NICAM used 14 bits truncated to 10 so its like between DPCM and linear PCM, but still considered as PCM sound. Similar system can be used if linear PCM must be really small. That 4 bit PCM can have almost 70 decibel range and 70 decibels is range of cheap phones and media players audio reproduction, so difference is not so noticeable if either 16 or 4 bit PCM is used, but 4 bits requires only 1/4th of bitwidth. MPEG-4 ALS and MPEG-4 SLS are lossless compression methods, altough not used much. In cheap device it is no matter of coding efficiency, because all lossless audio compression formats have about same compression ratio, but lossless audio coding that is simple and fast to decode is needed, and FLAC is both. Ogg Opus is lossy compression that uses much CPU, but making processing slower, or increasing number of sample rates (not just 48 khz, but adding 44,1 khz and 40 khz etc.) can make Ogg Opus more CPU friendly. If Ogg Opus is in hardware not software I don t know if there is any difference in complexity etc. compared to other codecs. Nothing to do with previous, but Kawai/Lowrey stopped making electric (electronic) organs 2018. Lowrey organs were last real analogue transistor organs available, Hammond has been fake analog digital for years. If real analog transistor organs like expensive deluxe models are not commercially possible anymore, cheap combo organs that were much cheaper and popular during 1960s / early 1970s can be made again, with real analog transistor organ sound, not digital fake analog simulation. Electric analog pianos are being made once again in Brazil and USA, and Clavinet (called Vibanet now). So combo organs can have marketplace also once again. They have become collectors items and are much sought after. If expensive or less expensive transistor organs would be made as rackmount modules with full Midi etc. they could be used with keyboard controllers and in rackmount units without keyboard their price would be cheaper, graphical user interface like softsynths. Those modernised transistor organs can have market place once again, and combo organs were simple and cheap already in 1960s. Analog synths have returned, electric organs / combo organs can be next analog trend, electric analog pianos are having already revival. XiFEO FLAC (XiVero) and MQA are lossless sound compression methods, improved FLAC. Optimfrog is alternative to FLAC. Chord Electronics (Rob Watts) has made upsampling DAC that makes 16 X upsampling from 16 bit sound. So 1 bit ADPCM, Takis Zourntos model 1 bit or 1 bit delta-sigma can be upsampled in receiving end (DAC) to 16 X sampling rate, so 1 bit sound without oversampling in source audio will have 16 X accuracy / 16 X oversampling when this DAC is used? From 1 bit 44,1 khz sound for example to 16 X sampling rate of 1 bit, not 16 X sampling rate of 16 bit sound like this DAC normally does, 1 bit X 16 is much simpler than 16 X 16 bit, requires less hardware. Lossless “compression” (actually upsampling, but same sound quality than 16 bit altough just 1 bit is used?). Chord Electronics also has noise shaper that uses extra 8 bits to 16 bit sound to achieve 301 decibel range. It means 50 bit dynamic range accuracy, 16 bit source + 34 bits exta. So linear 8 bit PCM + 8 bit noise shaping is then 8 + 34 bits = 42 bits accuracy. This from 16 bits. 1 bit sound + 8 bits is 9 bits, so 1 bit non- oversampling sound is 35 bits then when only 9 bits are used (1 + 34 bits accuracy)? Low bitrate audio codec, lossless that uses very small bitrate like MP3 but lossless codec, is possible if 1 bit X 16 expansion like Chord Electronics DAC is used. For internet streaming etc. Perhaps even PC software, not in hardware, is possible some kind of upsampling from 1 bit ADPCM/DSM 48 khz source to 16 X sampling rate 768 khz? If pseudo parallel delta sigma 1 bit or similar is used perhaps even below 1 bit is possible, like pseudo parallel 1/16 bit to 1 bit DSM and then that to 16 bit using upsampling etc, so 16 X 16 = 256 X “compression ratio”. Or 32 bit or 32 X multirate DSM (1 bit actually, or Takis Zourntos 1 bit / 1 bit adpcm with 32 bit / 32 X multirate internal processing) and 48 khz sampling rate. 48 khz audio can reserve over 14-27 khz sampling to quantization noise if only 1 bit is used so over 7-13,5 khz sound is using High Frequency Replication. Sliding from 7 khz to 13,5 khz sound depending how much space quantization noise needs, ear has sensitivity peaks in 6 and 13-13,5 khz so sound sliding from 6,912 to 13,44 khz perhaps, easily divided by number 32 and 6,9 khz is 1/3,5 and 13,44 khz is 55% of 24 khz frequency range so room for quantization noise. Dithered noise shaping can be used with FLAC, it improves quality of low bitrate (8 bit or even 1 bit?) sound but adds only about 2% more to FLAC file. Same bitrate reduction that lossy codecs do can be used in 1 bit sound also, making bitrate even smaller. Making 16 bit 48 khz sound smaller if 0-16 khz is lossless, 16-36 khz lossy, and over 36 khz Sub Band Replication only (so sampling rate is actually 36 khz, highest sound 18 khz, some “cue” information over 36 khz SBR can use if needed), that will make good quality digital sound. Quantization noise makes highest sound 16,4 khz about, over it is just SBR to 20 khz. If small 60- 65 mm ecodisc is used to store audio or video and then put to plastic ring of 120mm diameter, or normal size 120mm disc used, cheap organic layer that modern CDr and Bluray discs use can be used in ecodisc production, very cheap. Ecodiscs are made with two layers now, so Super Audio CD in ecodisc form or ecodisc that has one layer of CD sound and other layer of DVD sound, DVD shows still picture when music is played, 24 bit 96 khz is possible, for cheap music albums (in third world) with hifi quality. Like SACD, CD/DVD hybrid disc has CD audio for cheap CD players and better quality DVD sound for DVD players with audio output. If playing time of disc allows, CD and DVD / SACD information can be together in one layer only, in separate tracks. 80-99 minutes is possible in CD/CDr, so if album is 40 minutes long 50-60% disc space is left for SACD or DVD audio with still picture.
It is possible that in cheap organic dye layer disc like CDr and modern Bluray discs, manufactured like Ecodisc, 120 mm diameter, one layer disc, has three different sound formats, in separate tracks, in one layer only. One is CD sound, another 24 bit / 96 khz (or 48 khz) DVD sound track that shows still picture in TV when audio is played, and third is SACD sound. Disc is super cheap to make but has three different audio formats, from normal CD sound to high quality, but only one layer. It depends length of music album is this possible. For third world market music distribution. If CD sound is not used and DVD audio only or that with SACD, even small 80 or 60 mm discs can be used, or reflective surface is 80 or 60 mm only and outside it is plastic ring so disc is 120 mm diameter (if needed). Small 60 or 80 mm discs can use if disc space permits and music album is not long 32 bit 48 or 96 khz sound FLAC compressed playable by computer and then 48 or 96 khz 24 bit DVD sound track also. 80 mm disc is enough for about 40 minutes of CD audio (36 minutes for regular CD, 43 minutes for CDr) if only CD sound is used. In hydrogenaudio forum “Good 1-bit audio codecs” 2018 is ADPCM that has more or less transparent quality in 7-9 X oversampling. That may be the base for low oversampling 1 bit audio codec, even non oversampling if from 48 khz sampling / 24 khz sound frequencies over 6,9 khz to 13,44 khz (sliding from 6,9 khz to 13,44 khz when sound quality permits) are for sound and rest for quantization noise. 6,9 khz is about one 3,5th of 24 khz frequency range, so room for quantization noise, actually 3,5 X oversampling 1 bit audio. Over 6,9 khz can be lossy audio 1 bit (if codec that uses 6,9 khz is lossless) or High Frequency Replication / Spectral Band Replication (not “sub band replication”, that was error of mine, I did not notice that SBR means spectral, not sub band) that uses some “cue” information from 6,9 - 15 khz to reproduce sound, so small amount of high frequency information is included also in this coded. When sound quality permits 1 bit frequency can slide up to 13,44 khz which is about 55% (56%) of 24 khz, so less than 2 X oversampling. XiFEO FLAC and MQA (FLAC) style compression can be used, MQA is actually lossy not lossless. Adding dithered noise shaping to FLAC file adds only about 2% to FLAC but can make for example 8 bit PCM to 16 bit dynamic range, similar to CD sound. Or use for example 10 bit floating point or 8 bit posit as sound format. SACD is 4 X oversampling of 16 bit PCM if bitrate is compared, and has about 20 bit quality, even if using 1 bit, not multibit, delta sigma DAC. So low oversampling ratio DSM or ADPCM or Takis Zourntos model 1 bit sound codec is possible, or using only 48 khz in 1 bit codec (or 4 or 5 bit ADPCM). Variable bit rate from 3 to 5 bit can be used. “Variable bit rate ADPCM-MQ” multi quantizer, “Variable bit rate modulo PCM VBR-MPCM”, “ADPCM explicit noise coding”, patent EPO369682 “Efficient coding method” Matsushita. Chord Electronics uses upsampling (“taps”). Upsampling 1 or 4 bit audio is simpler than 16 bit (requires 65 000 taps, 1 bit needs only 2 taps and 4 bit only 16). So Chord Electronics upsampling can be used perhaps in software, if upsampling has small amount of “taps”. Sony has quality improvement system also, perhaps different from Chord Elecronics system, called Sony DSEE HX. Those sound improvement systems can improve low bitrate audio. Also if better than 16 bit CD quality is needed, oversampled ADPCM offers greater improvement than linear PCM, for example 192 khz 4-5 bit avarage bitrate AptX must be highly superior to 48 khz 16 bit PCM, AptX has then 4 X samplerate, and ADPCM improves greatly if samplerate is increased. 1 bit non oversampling 48 khz sound or 3-5 bit ADPCM can be improved using “taps”, using 16 X, 20 X, 24 X or 32 X upsampling or taps for 1 - 5 bit sound. So non oversampling 1 bit sound can have CD quality altough it is very low quality 1 bit 48 khz in first place? Only about 13,44 (about 50 - 55% of frequency range, actually 56%) khz is needed for good quality sound and rest of 24 khz range used for quantization noise and noise shaping. Noise shaping / dither can perhaps be used in sound that also uses “taps” to improve sound quality even further. Over 13,44 khz sound can use lossy codec if below 13,44 khz codec is lossless, or use high frequency replication. MQA folds frequencies in 96 or 192 khz files and makes soundfiles smaller, similar MQA that folds over 13,44 khz, 15 khz or 16 khz to 48 khz frequency range of 48 khz sound, so bitrate above 13,44/26,88 or 15/30 or 16/32 khz is not needed (because information above those high frequencies are encoded in lower frequencies) can perhaps be done, altough perhaps encoding those frequencies needs more bits than 3 bits that is needed in MQA that encodes 96 khz 24 bit to 16 bit 44 khz, because those lower frequencies have more energy. XiFEO is another similar high frequencies to smaller bitrate method. Perhaps XiFEO and MQA can be used together, but then must be paid two license fees. And Optimfrog or other efficient instead of FLAC with them. Folding for example 4 X, 8 X, 16 X, 20 X, 24 X or 32 X oversampled 1 bit audio (or 3-5 bit ADPCM) is perhaps economical for MQA or XiFEO? More economical than oversampled linear PCM? So if MQA and XiFEO works with high oversampling ratio 1 bit sound or ADPCM, it can make really extremely small soundfiles of good quality sound? And Chord Electronics upsampling also. Perhaps MQA type downsampling can be used with Chord Electronics type upsampling, making super small soundfiles of high resolution sound, and even making normal 48 khz audio smaller size? With XiFEO and other type of dithered noise shaping. Chord Electronics has noise shaper that uses extra 8 bits to achieve 301 decibel dynamic range from 16 bit sound, so it is 50 bits of accuracy, 16 + 34. So using 8 extra bits it makes 34 bits of extra accuracy. If cheap 120mm ecodisc is having three sound formats in one layer in separate tracks, and SACD is supported only in some CD players, then three format one layer CD can have CD sound first, then 48 or 96 khz DVD soundtrack 24 bit, third can be 32 bit floating point 48 khz or 96 khz sound in DVD-ROM tracks, that is playable in computer only not in DVD player, but it can use FLAC or other compression. SACD has similar lossless compression, but perhaps not supported in hardware or free software? CDr / CD-ROM sound format can use FLAC also (or Optimfrog or other free codec), but FLAC and other codecs do not play in regular cheap CD player. So if purpose is to make disc that plays in every CD player then album length is the factor that defines is it possible to add 24 bit DVD / DVD-ROM tracks and 32 bit FLAC DVD-ROM to one layer ecodisc together with CD audio. Sampling rates like 44/88 khz or 40/80 khz or 36 khz (and 72 khz? Double sampling rate of 36 khz) can be used if it is possible in hardware DVD player or in computer software player. Etalon DAC that uses Super R2R topology is improvement of previous DAC structures, but is enormously expensive. However Super R2R can be better than other digital to analog conversion methods. Noise shaping can also be used with lossless coding, there are many free noise shaping plugins and programs, so FLAC or other compressed sound file that uses some of the best free noise shaping methods can make file smaller still. So in one layer disc fits 32 bit 48 khz or 96 khz sound losslessy compressed noise shaped / dithered together with normal CD sound. Or 16 bit floating point can be used. If sound is 192 khz 16 bit, 192 khz can be divided to 4 X 48 khz sections, sections 1, 3 and 4 are used, they have 4 X, 2 X and 1 X sampling rate of 48 khz sound. All three have sound, so first section have sound, and also ultrasonic frequencies over 48 khz are not “empty” but used to store sound, another track. At same time is high frequency spectral components to be heard, they are not ultrasonic anymore. But efficient and free noise removing programs exist. Sound also can be dithered using dither that is pleasant to human ear, so altough previously ultrasonic noise is heard, it can be removed or made quieter. So instead of only one sound track in 192 khz there is now three. One 48 khz is not used so empty gap 48 khz makes ultrasonic noise smaller, last 48 khz section can use only 36 or 32 khz, making distance to previous section 12 or 16 khz wider, making ultrasonic noise smaller (less energy, quieter). Last section has cumulative ultrasonic noise of all three previous sections, but more distance (herz) makes noise quieter and less energetic, it can be removed using computer. Perhaps in quiet places can be heard parts of previous three sections sound, but very quietly, or it is dithered to quiet humming that can be removed. So instead of just one music track in 192 khz, there is now three. 8 X 44,1 khz can also be used, 8 sections for sound (some are not used). At same time high resolution sound and small bitrate. In 8 X 44,1 khz audio sections 1, 2, 4, 5, 7 and 8 include sound and are 44,1 khz wide, sections 3 and 6 empty and 44,1 khz wide, so that cumulative ultrasonic noise gets further away in spectrum. Secton 8 has cumulative noise of previous five sections that contain sound, but that ultrasonic noise that now becomes audible has little energy and it is far away in spectrum mostly, so perhaps most of of those high frequency spectral components that now are audible in section 8 (and audible in other sections also except section 1) can be filtered out or dithered or something. Very sophisticated noise removing software is available, and “cue” information how to remove noise / spectral ultrasonic components that become in audible spectrum when higher sections are used, can be added in bitstream. Small amount of noise or other sound can remain in audio even if noise removing is used, so in quiet passages in background can be heard this other channel s sound that was in ultrasonic range for them but now is heard in those sections that are higher sections in 4 X 48 khz or 8 X 44,1 khz range. So instead of one 16 bit with 8 X 44,1 khz sampling rate, there is six 16 bit sound channels, each one less quality than another until section 8 has worst quality. 16 bit must be floating point to minimize quantization noise. 192 khz 4 X sampling rate 48 khz sections must be 32 bit floating point or 16 bit FP for same reason. Instead of FP, posit, dual fixed-point (its almost logarithmic so suitable for ADPCM or delta-sigma perhaps), ELMA exact log-linear multiply add, dynamical fixed point DFXP, or flexpoint (16 + 5 bit), multiple description coding (delta sigma, ADPCM, Takis Zourntos model) can bew used. AptX uses 8 bits in low frequencies, so it can use posit or ELMA in 8 bit sections, radically increasing performance. Audio compression divides sound to small blocks and compress them, those blocks can have common exponent like flexpoint 5 bit and then 16 bit integer per herz. Similar principle like dividing sound to small blocks is used in fractal picture compression. Because modern cameras have huge megapixel count picture has very large amount of pixels. Instead of one picture this large pixel amount can have large amount of small pictures for example 16 X 16 matrix and then this one picture has 256 smaller pictures. This one picture that contain 256 small pictures is compressed using fractal compression, is compression ratio now better than normal MDCT- based compression? When picture is decoded those small 256 pictures are extracted and used as separate small pixel count pictures. “High definition vinyl” is modern high quality LP manufacturing that offers better quality and cheaper manufacturing for analog sound. For 78 rpm LPs LP diameter can be oversized, like 315-320 mm or 350 mm so playing time of 78 rpm LP is enough and 45 rpm also, and sound quality better in outer grooves than in 300 mm LP. 13 inch LPs (330 mm) were pressed in 1990s and perhaps even lately, so 319 mm 12,5 inch LP that fits in LP slots of record shops or 350 mm “14 inch” LP is not so crazy. But then 78 rpm would have long playing time and good quality sound for LPs. Also LP standard is 12 inch so 305 mm, altough in USA its 302 mm and in Europe 300mm. Making LPs 305 mm wide would offer 2-3% more playing time than 300 mm. For lightning, laser lamps that turn electricity to energy much better ratio than LED lamps etc. can be used in third world to save energy, in “Sam s laser FAQ - diode lasers” is “Diode laser light bulbs” and how much better they are than other lamps to save energy. Mass produced laser lamps (light bulb socket type and other) would be very energy efficient. In 8 cm mini-CD discs have official playing time only 24 minutes for standard CD and 34 minutes for CDr. However they have surface area for 36 minutes if 120 mm CD has 85 minutes playing time (actually longest CD is 83 minutes) and about CDr early 2000s was 120 minute CDr offered factory made (but not sold?). So 80 mm mini-CD with CDr standard can have 50 minute playing time. Those official short playing times of mini-CDs do not count outer edge “overburn” area perhaps, however “overburn” is matter of 120 mm CDs, so 80 mm mini CDs could be made easily that use all outer surface safely and have long playing time. CDr standard does not mean that disc must be CDr, it can be factory made music disc with content like standard CD, but playable only in those players that are CDr capable. Optical layer can be cheap organic dye like CDr and Bluray discs. But intent is to make cheap as possible music discs for third world and some cheapest CD players do not play CDr discs. Lasers used in cheap CD players perhaps use nowdays 650 nm lasers, not 780 nm, and those have perhaps some light frequency up to 600 nm, so from 600nm to 780 nm means 30% more space, and if standard mini-CD 8cm that has 36 minutes max playing time uses 30% more space it is then 46-47 minute standard CD-DA mini-CD. 650nm means 42 minute mini-CD. Using FLAC or other lossless or lossy coding, and FLAC + dithered noise shaping (noise shaping can offer 8 bits extra accuracy but adds only about 2% to FLAC file) and 80mm CD-XA (CDr standard) disc can be used, mini-CD capacity expands dramatically and can even surpass normal CD, but now mini-CD disc is further away from normal CD and does not play in cheapest CD players. Instead for going to high sample rates high resolution sound can be 64 bit 48 khz sampled, but for example in 24 khz sound only 6,9 khz is 64 bit, 6,9 -12 khz is 32 bit and 12-16 khz 16 bit and 16-20 khz SBR or lossy encoded if 16 khz and below is lossless. Also one bit sound (delta sigma, ADPCM, Takis Zourntos model) can use floating point system, not integer (altough delta sigma is logarithmic). Floating point DSM has been proposed. If floating point suits for DSM or any other one bit system then also posit, ELMA, or “Between fixed point and floating point by Dr. Gary Ray” article s “reversed Elias gamma exponent”. SAR ADC / DAC can be pseudo or parallel (perhaps even pseudo parallel) like DSM. Vinyl LPs are made of PVC, but also polystyrene was used. Vinyl LPs can have 30% of recycled PVC; if LPs are made of polystyrene like they once were (up to 1970s), recycled polystyrene can be used, perhaps more than 30%, or even near 100% recycled polystyrene for LP record. LPs are not ought to be so heavy, 180 grams nowdays. The weight of the LP has nothing to do with its sound, even thinnest and lightest LP has exactly same deep grooves and bass response as heavier LPs, 180 gram LPs are only marketing gimmick, they are not any better soundwise than lighter LPs. So thin light “featherweight” LP made of recycled polystyrene is eco-friendly. LP diameter can be bigger than 300mm for 78 rpm and 45rpm records, even 16 inch “transcription LPs” were made and there still are some modern LP players that can play them. Thin and lightweight oversized LPs can be made. “High definition vinyl” can increase LP playing time 30% for example in 78 rpm record. Music content can be said to be square waves derived to “peaks”. So music synth that uses square waves derived, can imitate almost any sound? Also analogue square wave generator when derived can imitate almost any musical content? So sampler / wavetable type analog synth is possible using only derived square waves? TIM (transient) distortion can appear in sound in amplifier, and make sound worse than measurements of distortion show. But it is possible to build TIM distortion noise shaper and make sound better. Ortospektra was surround sound system that used only three speakers, used mono signal plus two side signals. Making low bitrare audio system for home theater (DVD or internet TV transmission) that is cheap and low bitrate can be build using ortospektra. Only one speaker with full frequency and two small speakers with no low bass response needed and only about 3000 hz max high frequency. Stereo signal can be used (from DVD). So in third world very cheap surround system for home TV can be build. Also because ortospektra uses mono signal with little left - right side information, bitrate can be very low if only mono audio is used with that small amount of side information to 3000hz at sides, and that side information is just for ambience in sound, so using encoding like lossy audio codecs to store that small amount of side information and central mono signal also, bitrate can be very small. For internet TV broadcast with surround sound and small bitrate. Also perhaps headphones can use ortospektra system, nowdays MP3 files etc. use false stereo “joint stereo”, mono signal with fake stereo in low bitrates. If only mono signal + ortospektra side information to 3000 hz in left - right channels are used, and “enchanted stereo” or “surround” mode in headphones that some audio players have, that may offer better fake stereo at low bitates than usual joint stereo from mono to stereo system? LP records can use Trimicron pressing that tripled playing time, High definition vinyl and Trimicron grooves, at 78 rpm 12 inch diameter disc or larger. Trimicron grooves degrade sound quality but 78 rpm improves it, so 78 rpm discs with Trimicron 3 X playing time can be made. This 78 rpm disc should have better quality than 45 rpm disc, despite Trimicron grooves, and longer playing time than 45 rpm or 33 rpm disc.