If phones, tablet PCs or other devices should as cheap as possible, cheapest solutions must be used. Near field communication (NFC) is in some mobile phones. A version of it that uses sound as data transfer method and does not need special NFC electronics, because only microphone and loudspeaker is needed, and that phones have anyway. Tagattitude Near Sound Data Transfer (TagPay) has been in use since 2005, and similar techniques Narette Zoosh (using ultrasound) and NearBytes (from Brazil) are in use. Encrypted voice data transfer can be used as payments etc. similar way NFC uses radio waves. Voice data transfer is ideal for cheap devices because it does need extra electronics, as cheap substitide for radio frequency NFC. NearBytes has 10 cm transfer distance and works even in noisy enviroment. Directional audio loudspeakers (Panphonics) propably makes transfer distance even longer, like acoustic version of Bluetooth. Cheap techniques like acoustic data transfer as substitute for radiowave NFC is needed in super cheap devices that are given free to users of free internet connection (advertisement paid connection by commercial sponsors, limited access to internet and communication compared to customer pays- model). Also what kind of cheap entertaiment free internet connection will have? Public domain movies are either very old, or then very bad c-class movies, and their amount is limited. But Bollywood movies, old and new, are offered in dozens or hundread netpages for free, and amount of them for free download is staggering, over 6000 Bollywood movies is in one netpage (Nyootv.com) alone. All are downloadable for free. In Youtube in "Cinemascope Movies" is over 400 indian films at least, probaply many more. Netpages like apnaview.com, oldmovietime.com, bharat-movies.com are full of Bollywood movies, old and new. So if cost free to end user internet becomes reality Bollywood movies (perhaps mostly old Bollywood movies, the cheapest option) are the most significant entertaiment. Because content in cost free internet to end user must be cost free too, so public domain or other minimal copyright cost material must be used. One way to make super cheap devices to development countries market is musicmaking software and hardware. Free softsynths are vailable at Android and Linux platforms. Making Linux run Android Apps is possible (several conversion solutions are for free), Windows music softsynths can be run on Linux or Android using Wine and Qtractor, Ardour etc. Even iOS apps can be run using Cider APK (Cycada) or iEMU APK, in Windows iPadian is similar solution. For poor countries free apps like free soft synths for musicmaking are the most important free goods. If very cheap device can run free music apps, not only Android and Linux together in same device, but Windows and IOS too, that would be ultimate solution. Windows is offered for discount price to cheap laptops, about 25 - 30 dollars. Hardware memory is restricted to 4GB / 32 GB and also the size of screen is limited, and processor must be cheap (Intel Atom, or Via, for example). Cheap "music computer", mini PC style small box, small display (made using e-ink or other cheap), DVD-ROM is needed (altough it is nowadays forbidden in cheap Windows requirements), and perhaps slight increase in memory requirements (to 6 GB / 54 GB for example) because some sample-based softsynths have several dozen gigabytes of samples. The music computer can run Linux, Android apps, and it can have cheap Windows also, perhaps in compresssed OS form like some laptops have. Main OS is Linux or Android but it can switch to Windows if needed. Also iOS apps can be used through Cider APK, iEMU APK in Android or Linux mode. Windows increases the price 25-30 dollars compared to without it using only Wine, but now softsynth platform is better and stable. Outside typing keyboard and large screen (sold as additional extras, not in standard configuration) is needed also. Price must be as low as possible. Altough it is possible to run Android programs aimed for ARM processor to x86 processor using conversion software, and different operating system apps and programs will work in Android or Linux (both Windows programs and iOS iPhone and iPad apps), conversion with processor hardware base from ARM to x86 and vice versa makes this difficult. iPadian is (?) an conversion program that makes (?) iOS apps made for ARM processor to work in x86 Windows (?). If that is true, however conversion to operating system to another is diffiult enough, and conversion to one processor architechture to another even more difficult (if at the same time operating system must be changed to another one) so perhaps this cheap "music computer" needs two processors, other is ARM, another cheap x86 (Atom or Via) so that Linux desktop PC, Windows desktop PC, and Android and iOS mobile apps work in same computer. Main point is that because Windows (and Linux) softsynths are offered for free (thousands of them), and Android musicmaking apps also, and to lesser extent iOS apps (for free), and people in the third world countries does not have money to buy a hardware synth, so offering to them at cheap price (or for free) softsynths and other musicmaking software would be great help. So (super)cheap platform device ("music computer") that can handle four operating systems (Linux, Android, Windows desktop and iOS mobile apps) at the same time and for free (except Widows desktop that costs 25-30 dollars at cheap devices) is needed. It would be little box like mini PC, and desktop PC monitor screen and typewriting keyboard would be extra features not sold with the box itself. But using two cheap processors (ARM and x86) perhaps makes performance not so good, so selling two different devices (other with ARM processor, other with x86) is also possible. Specs should be such that discount price Windos desktop OS (25 - 30 dollars) can be used, or then using Wine emulator instead (which can make softsynth / VST operating unstable and difficult). Selling softsynths to this device is perhaps with DVD-ROMs that have softsynths etc. software. Ecodisc single layer DVDs are cheap to manufacture. Softsynths can be downloaded for free, but net connections in development countries are not the best, so selling DVD-ROMs directly to users, price printed on DVD disc and package, 1 Dollar price collecton discs with music software. Music software like softsynth can be compared to music album. Streaming services like Spotify pay 0,0084 - 0,006 dollar per stream, 10 downloads is an album. But only 0,0016 USD or less (0,0011 USD or less is avarage in Europe) goes to artists, record company takes the largest amount. If music album is compared to softsynth then 60 softsynths make 1 dollar DVD-ROM, 4,7 gigabits or 2 X 4,7GB if double sided single layer disc is used. Softsynths here are those that are otherwise offered for free in internet. Now some amount of money (like spotify payment, but from product that exists on DVD-ROM or memory card) goes to free softsynth developer. Same method can be applied to selling music directly to users in development countries, that music that is nowadays offered for free download to internet and without record company support (indie artists giving their music for free in the internet). If 60 million DVD audio records or DVD-ROMs are sold one music album or softsynth will bring one million dollar money to its maker. Beacuse mobile devices does not have DVD-ROM inside (phones and tablets), SD memory card distribution of content is also possible. Internet connections in third world countries are weak and slow, so downloading gigabytes of softsynths would take ages and broken down link connection propably corrupts the data files, so direct sale of content to customers using memory cards and DVDsis one option. People in poor countries have no money to buy iPhone or iPad, so conversion program that translates iPhone and iPad free musicmaking apps to platform that is cheap is needed. If those free apps cannot be downloaded at Apple store because devices are not Apple manufactured, other distribution channel for those iOS mobile apps that are free or few dollar price is needed. Direct sale using memory cards and DVD-Roms is one option. Also softsynths are "too sophisticated" nowadays, 32 or 24 bit sound with 96khz sampling rate is too much and useless, and same goes to high sample rate sampling instruments also. Lower bitrate 36 khz sampling rate is enough for human hearing system, or even 32 khz (higher sound than 18 khz humans does not hear, and musical notes over 15khz humans does not hear anyway). Simply adding zeros (silence) to this 16 or 18 khz sound will make 22 or 24 khz sound if needed in the output stage. Softsynths have poor dynamic range anyway, about 60 - 70 decibels, so 12- 16 bits is enough. Some softsynths in mobile phones etc. perhaps have 50 db, or less than 48db "dynamic power" (dynamic range) and that is even below to 8 bits chiptune music quality. So not wasting bits low sample rate and bitrate is needed. Some sampling collections for softsynths have dozens of gigabytes high quality (too high quality) samples. If floating point processor is used why not use differential method like DPCM makes 16 bit linear PCM 8 bit DPCM sound, from 32 bit or 16 bit floating point sound 11bit or 10bit floating point sound, but with differential ("DFP") form would have same quality but lesser bits. 10/11 bits is used in OpenGL graphics. If integer processor is in use 4 bit DDPCM (Dynamic DPCM, bitsnbites.eu- netpage) is enough for softsynths, and using simple dither 1-bit signal can have 4 bit precision, so 1-bit dithered signal with 4 bit DDPCM quality is enough for soft synths. Sony Super Bit Mapping had 22 bit quality using 16 bits, and no audible alteretions in sound, so SBM method in 1 bit sound would have 6 bits precision and signal is still not different than without SBM because changes are below audible range, and this signal can be treated and processed as ordinary 6 bit signal without noise shaping changes to sound (or as 1 bit signal with 6 bits of accuracy) , but it is only 1 bit signal (with 6 bits of accuracy). This "differential floating point" and 1 bit integer that can be dithered/noise shaped to 4 bit, 6bit, or 8 bit DPCM (ADPCM / DDPCM) accuracy can be applied to sampled sounds also not only sound synthesis algorithms, and to general audio (music and speech, any sound) codec also. If 8 bits precision is needed from 1 bit sound noise shaping goes to audible, so if this 1 bit noise shaped to 8 bit DPCM file is played without similar noise shaper codec, difference in sound is noticeable. But 1 bit to 4 bit simple dither or 1 bit to 6 bit Super Bit Mapping sound can be played even without precise codec because difference in sound is unnoticeable, changes of sound will stay below audible level (altough noise shaping is used). Also DTS sound encodind that uses "coherent audio" coding that encodes only differences between channels, so in its maximum (HD form) several thousand sound channels can be compressed and encoded. If same methods are applied to soft synths, different oscillators and filters are used but only difference between them is computed and similarities are copied from polyphonic voice to another, perhaps then softsynths with thousands of voices would be possible. But modern softsynths propably use these methods anyway already. Phase shift keying, frequency shift keying, amplitude shift keying etc. methods use low bitrate, some even are1-bit for default, and applying them in "coherent audio coding" style methods perhaps make low bitrate softsynths, those keying methods use sine waves, square waves and sawtooth waves already, like oscillators of sound uses. Standard waveshapes that are modulated with those phase shift keying methods (some of them are only 1 bit by default) are already used in softsynths. Other bitrate savings using sparse sound synthesis, numerical sound synthesis, "Audio synthesis by bitwise logical modulation", "Even more errant sound synthesis", "GLITCH: noisy fusion of math and music", XOR synthesis, "Evolving sinusoidal oscillators using genetic algorithm", Vector Phaseshaping Synthesis, "CORDIC based digital sound synthesizer", "CORDIC and Taylor based FPGA music synthesizer", "Design of FFTs using CORDIC and parabolic synthesis as an alternative", Polynomial Transition Regions PTR (yofiel.com has many examples of these), Feedback Amplitude Modulation FBAM etc. Pine64 is cheap (15 dollars) PC platform, Unuiga S905 (25 dollars) and "Kodi box", cheap chinese set top box PC (30 dollars) are examples of cheap tech. Instead of set top box / mini PC style music computer (with cheap e-ink display, even black and white display, in the device, because Windows needs it according to cheap desktop OS requirements) perhaps Android TV stick / Linux stick style small device that ihas small display that is required and uses memory card only as storage device (DVD-ROm is missing), so in appareance this device looks like Plugiator softsynth platform but instead of DSP it has ARM and x86 processor inside. Cheap large display like desktop PC, typewriting keyboard and musical keyboard / MIDI controller is sold as extra features, not in standard configuration (that has just "TV stick" music computer). Cheapest Android / Linux TV sticks are being sold below 10 dollar price in China, and because they are computers with Wifi, memory etc, they are like cheapest Android phones that are being sold at 10 dollar price range in India and China. Instead of ARM or x86 cheap processor (or processors) perhaps Kalray manycore processor will do, it is for embedded devices and it is reasonably cheap, has good computing power, but it is neither ARM nor x86, so programs written for ARM or x86 processors work equal badly in Kalray. But because conversion programs from one OS to another or one processor architechture to another are emulator programs anyway, processor that runs emulations as OS like Kalray is perhaps suitable for many operating system "music computer". Any device that can handle four OS (Linux, Android, Windows desktop for softsynths, and iOS mobile apps) is needed. Because in rich countries there is no need for this device it will be sold only in third world countries. For sound (sound codec, sound synthesis, compression, sampling and processing) perhaps vector compression like Additive Quantization (Martinez), also called extreme vector quantization, is suitable. Also "Maximum mutual information vector quantization" 1995 Wilcox, "Soft vector quantization and the EM algorithm", "Amplitude-adaptive vectro quantization system", "Musical sound analysis-synthesis using vector-quantized time-variyng spectra". Other: Waverazor synth, Tone2 impulse modeling synthesis IMS, decoy.iki.fi/dsound: "Sound synthesis", "Towards timbre synthesis and method for sound..." (Kantz), "Extenting the Macalay - Quatieri analysis of synthesis with limited number of oscillators" (Fritz). There are virtual analogue models that can reproduce exact analogue waveforms, but they need several megaherz bandwith and huge oversampling, Novation Peak uses 24 mhz sampling rate. Delta-sigma modulation (or Takis Zourntos nonlinear oversampling) uses oversampling, and in the Hatami distributed model has low bitrate (Hatami model has been improved to even better now), so virtual analogue synthesis using Hatami or other 1-bit model (or "A higher-order mismatch-shaping method for multi-bit sigma-delta modulators" A. Lavzin 2002) would have low bitrate. Same methods can be applied to general sound codec also. Also music description languages can have code density that has low bitrate, Forth- based (Formula, Ample, Moxie, MacMox, Neon, iMops, Sporth), Common Lisp Music (Scheme, Impromptu, Extempore, Racket), and even APL programming language and its many versions to music generating (altough no known examples of APL applied to music description language exists). Faust (and Faustina), and Futhark (GPU programming language), Lytescript, Bard etc. effective programming languages can be used. "A declarative metaprogramming language for digital signal..." (Kronos) is Common Lisp-based. Earlier music programming languages used only few kilobytes of data, because computers did not have memory and capacity, so early Csound (from 1980s) or earliest PureData versions (from 1996) propably has small memory footprint. Also very early sofsynths from 1980s like "Armonyx" or "Tecmar sound syth" for IBM PC (propably MS DOS programs) were very small programs to nowadays standards. Also "Fractional Fourier shift transform" (Rutanen 2016), "Algebraic integer quantization", "Sparse atomic decomposition of audio signals", Xampling, "Non-negative matrix factorization" (Koivunen 2016), "Recent advances in PWM techniques and multilevel inverters" 2014 (space vector PWM SVPWM), Differential Pulse Position Modulation (another modulation method like phase shift keying), "Differential spatial modulation", "Dynamic differential modulation of sub-carriers", "Angle differential modulation for odd-bit QAM", "Duobinary modulation for optical systems", "Dispersion modulation using allpass filter", "Abstract frequency modulation", "1bit quantization and oversampling at the receiver", multiple description delta coding, "Bit-interleaved coded differential space-time modulation". Also conversion of different operating system to another: Google Chromebit, and Intel Compute Stick are kind of cross operating system devices, CloudReady, and Igel Software are cloud based cross platform and OS systems. Google Chrome is only wide scale succesful thin client system that has been a success when Firefox OS was cancelled. Free cross platform solution that can handle music software (and several operating systems) is needed. Processor architechture where memory and processor are the same like iCube UPU exists, but it has not gained much use, and is like experimental thing and not gonna make widespread use. Other: "GPU driven killer synth software", "A ROM-less direct digital frequency synth based on fifth-degree bezier curves" 2013, dinisnoise.org, Delta sigma direct digital synthesizer (Orino), "A modulation matrix for complex parameter sets" 2011, "Bilevel sparse model polyphonic music transcription", "Neurogranular sampler" Mironda 2009, "Parallel evolutionary optimization of digital sound synthesis parameters" 2011, "SKMD: single kernel on multiple devices for CPU-GPU..." Lee 2015, "How to implement digital oscillator ?" 2011, "A survey of computational calculi used in musical applications" Sarria 2013, "Harmony: a system for musical composition", "Another fast, fixed-point sine approximation", "ARMO (Auto Regressive Moving Avarage filter)", "Extreme compressive sampling for covariance estimation" 2015, "Compressive phase-only filtering at extreme compression rates" , "Enchancement in DASS for low cost sensing nodes" Bonavolonta 2014, "Perceptual signal coding for more effective usage of bit codes", "Universal rate-efficient scalar quantization", "Quantization noise shaping on arbitrary frame expansions", "Sparse composite quantization", "Pairwise quantization", "Analog additive dithering", "Optimum quantization" James Bruce. OSCulator is Open Sound Control software, cross platform. In new devices perhaps something better than MIDI is needed, like OSC, TUIO, or Nonlinear Labs control method (TCD). MPEG4 had SAOL standard, that can directly control hardware synth (in phone or tablet PC, but those have no hardware synths, only software- ). If phones or tablet PCs will have inbuilt hardware synths, more modern control than 20 years old SAOL is perhaps needed, and those standardised hardware music synths must use most sophisticated modern synth methods. There is MPEG7 AABIFS etc. Even MP3 player circuit can be turned to hardware music synth, and GPU or graphics circuit used as softsynth platform etc. About modulation: Multiple Description Coding (delta / delta-sigma), and "Multi-Pulse Pulse-Position modulation", "Pulse Group Position Modulation", "Modulation by multiple pulse per group keying..." patent EP 1477001A 1, (Charles Hartmann 2002), Bit Angle Modulation and Digital Pulse Interval Modulation are different modulation methods. Tinyapps.org has very small music software (kilobytes), so several hundreads of these programs can be packed to small memory space, like old Symbian ringtone/music softsynths that had just few kilobyte memory requirement. Softsynths can have hundreads of voices, so but humans only 10 fingers, so system that uses some algorithm to expand those 10 notes that are played to dozens is needed, for example expanding 8 notes to 24 voices (3X8) so that 8 voices melody line goes on bottom and 8 voices top human played 8 notes. Unlike MIDI control that replicates notes played, algorithm or artificial intelligence program brings variations to those additional notes that are not played by human, so it sounds like theree persons with six hands would be playing variations of melody. for 10 notes (fingers), program would expand to 32 notes (voices of synth etc.) 11 notes played below and 11 notes on top of human played melody. 64 voices / notes version would expand 8 human played notes to 8X8 melody lines played by 8 different hardware or sof soft synths for example, not direct note for note copy like MIDI, but algorithmic /AI generated variations of basic human played melody line. 10 notes would expand to 10 + 6X9 note melody lines, one 10 note line played by human, AI plays 6 other 9 note sequences with different synths/ softsynths. Also sample player/ soundfond player that uses AI to find from samplebank suitable sounds to play with human played keyboard melody . "Automatic position loop for heterogenous systems" 2016 Baier, "DAINO: a higher-level framework for parallel and efficient AMR" 2016, efficient algorithm for longest common subsequence, generic inverted index, dynamic sparse matrix allocation, "DQMRES: a quasi-minimal residual algorithm", "Sound signals as RGBA textures" (Gallo). ADC DAC structures like "Multi-phase VCO-based ADC" Madhulika 2016 can be used as musical instruments, A/D converters have oscillators, ring oscillators etc, that can be used as musical instruments in "reverse mode". If polyphony reaches hundreads of voices, virtual reality user interface with video glasses and data gloves is one option for synth control, even in cheap devices (if video glasses and data gloves are cheap also). If synth sound frequency is limited to 7 khz (14 khz sampling rate) and higher frequencies are made using High Frequncy Replication and aural exciter, and synth patches designed accordingly, multiple voices are possible. 48 khz or 44,1 khz sampling rates can be divided for example integer 7, 48 khz would then have slightly less than 7 khz frequency and sampling rate of about 13,7 khz (48 khz divided to 3,5). 44,1 khz would have 6,3 khz sound and 12,6 khz sampling rate (44,1 divided to 7 / 3,5). This 7 khz or 6,3 khz sound can be expanded using High Frequency Replication in final output stage when all voices go through one mono channel or stereo channel, so final output voice will have about 14 - 16 khz frequency, even more if aural exciter is used. Aural exciter and HFR circuits are required in phone / tablet PC for regular sound codec (speech codec etc.) and music reproduction that uses low sampling rates . Aural exciter circuit themselves can be used as music synth (EXCTR synth by Jari Suominen, Spatial Aural Exciter). Unusual MIDI controllers like Samchillain or Udar can be used in mobile devices and soft synths designed accordingly. If human ear has two sensitivity peaks, one at about 6 khz and another at 13 - 13,5 khz (for young people, over 40 years old don t hear over 12 khz frequencies), and highest musical sound heard is 15 khz and highest sound about 18 khz, no higher sampling rates than about 24 - 27 khz is needed if sound is 12 -13,5 khz and everything over that have HFR or aural exciter, and nobody notices the difference. If integer processor is used quantization noise requires some extra bandwith.