CYBERYOGI =CO=Windler

If keys show a systematic fault (e.g. every 8th key does not work), then one of the keyboard matrix lines may have been interrupted (e.g. broken foil cable or a corroded PCB trace by battery leak acid or a bad solder joint). The bad connection needs to be found and resoldered.

I am no expert in CDP-100, but the symptome looks like a power supply issue (or dead batteries). The instrument will keep resetting (likely also changes preset sound and rhythm to default if it has that) by voltage drop during loud notes played through its internal amplifier because this draws higher current. Do you have connected it to a too weak (less mA than rated on the instrument, e.g. 300mA instead of 1A) non-original AC-adapter? Also a broken AC-adapter (e.g. bad capacitors) may cause this.

Check the headphone jack. There is a switch contact inside that disconnects the speaker when headphone is plugged in. If there is a bad solder joint at its input (by mechanical stress), neither that speaker nor headphone side will work. Also a badly oxidized contact inside the jack can make the speaker fail (happens often in 6.3mm jacks of 1980th Bontempi/Farfisa keyboards), although typically the headphone would work on both channels.

The plastic hook that keeps the key down has obviously cracked off (happens easily when a keyboard got carried roughly by lifting at the keys). If you can find the plastic fragment inside the instrument, reglue it with superglue. Else you may try to make a hook of sufficiently firm wire (e.g. stiff (non-wick) copper cable) and hotglue it to the key bottom.

I suspect that the hardware reason why Casio made no combined XW-P1/G1 model is that they reused a special CPU they had created already in 1990th, which sounds good by highly optimized handmade machine code (so they can't easily replace it), but supports a too small address space to access enough ROM (actually flash memory) to handle the features of both instruments at once. Of course they could have used bank switching, but there may have been a deadline for quickly finishing the chip or such stuff. That is to say, compared with average modern single chip hardware (e.g. Raspberry Pi) the system is surprisingly tiny, which can not be explained with RAM/ROM price anymore, but only by the oldfashioned virtue of "programming on the metal". An in depth technical analysis can be found in this thread: Pooched XW-P1

Are Casio XW-G1 now dying during firmware updates too? This absolutely sucks. :-( The main problem is an atrocious design flaw that completely erases the internal flash memory (including bootloader) before downloading the update into RAM from PC. Thus switching the instrument or power off after starting firmware update is absolutely deadly to the operating system and makes the mainboard unbootable. I.e. once firmware update has begun, *never ever* disconnect power to the Casio keyboard (no matter if your laptop crashes or windows reboots or you need to rent or buy a new laptop in between because yours refuses to finish that pesky procedure) before new firmware has successfully loaded into the instrument. An in depth technical analysis can be found in this thread: Pooched XW-P1 That bricked mainboard disease could have been easily avoided by not overwriting the flash boot loader during update. In computer engineering study I learned that building a system design like this is almost a sacrilege. But other companies built such trash too; e.g. plenty of Samsung laptops died of an UEFI bug by attempt of installing Linux. The XW-P1 and XW-G1 mainboards even seem to be identical (other than software), so I hope that hardware hackers will eventually learn how to refurbish those boards using an eprommer to rewrite the flash memory chips on them with the original firmware.

I see, the Casiotone 403 is one of the faux woodgrain models, which has an internal power supply and is mechanically different. It is likely mechanically similar like my Casiotone 401 (even same accompaniment chip). The 401 is quite heavy (about 10kg?) because despite its case looks almost perfectly like wood, it is mainly made of plastic coated sheet steel. (Casiotone 201 and 202 were of genuine wood.) In spite of this it is not as robust as it appears, because at the corners of the side pieces (made of pressboard?) the plastic woodgrain tends to peel off. I had to hotglue mine back into place, but this sheet plastic stuff also cracks off easily when accidentally folded to hard. The high quality speaker sits in an own thick plastic compartment of that even the cable hole was sealed with glue to turn it into a perfectly closed box. It is driven by a large hybrid amplifier module that is screwed to a sheet metal heatsink welded to the case bottom. This makes it very unpleasant to remove the mainboard because you have to unscrew the module and mess around with (possibly poisonous) old heat conductive paste. The front panel is held by screws (those in the 403 seem to be at the sides). I don't know how yours is built, but (if like 401) be careful with the 5 metal lashes with screw holes those hold the rear part of the control panel from inside. With case opened, these lashes are sharp as razor blades and easily damage the panel and PCB traces; with mine a preset sound LED failed by an accidentally cut trace after working on the opened keyboard. After soldering the trace I glued ribbon adhesive tape over the sheet metal lashes to make them less dangerous. After opening, a lot of broken small black rubber rings fell out of the case - likely they belonged under the piano keys somewhere for damping (keys feel a little loose), but turned brittle by oily room air or ozone. I don't have a service manual of my 401, but got a photocopy of Casiotone 403, which was of great help to understand the general hardware architecture, because it has the same accompaniment hardware combined with the main voice section of Casio MT-60 (first 403 models had CPU version D776G with external bugfix circuitry). They are a good didactic example how Casio in early instruments made several different CPUs cooperate and combine their keyboard matrices rather than the centralized master-slave approach found in later hardware, and how despite almost self-contained CPUs often minor functions like clock rate conversion or preventing wrong key press order during preset sound selection used a crazy amount of "glue logics" ICs cluttering up the mainboard.

This can be everything. May be a crack in the PCB (torn traces) or a switch or jack has ripped out of it. If a special chip cracked, it can not be repaired, but likely parts need to be soldered back in. If it fell on a plugged in plug, contacts inside a (e.g. power supply) jack may be bent and thus prevent it from turning on.

Can a digital preset really "burn out" from being used too often? This type of bitrot sounds quite unlikely in modern CPU controlled hardware. I know that flash memory can wear out when the same memory cells (addresses) are written (not read) over and over again. That's why SSD harddrives use the infamous "wear levelling" feature (that instead wears down the entire chip evenly and make you basically loose all data instead of only one file by a voltage surge in the wrong moment). Unless the keyboard rewrites every accessed preset into flash memory, this makes no sense. But a dead or empty battery would be a plausible cause for very strange and crazy bugs. E.g. my much older Yamaha MK-100 makes bugs those can be absurdly confusing if you don't know what's going on. That is to say, the MK-100 stores all settings in battery backed up RAM; with no batteries inserted the RAM is backed up by a large electrolytic capacitor for a few days(?). When the cap runs empty, this messes up the data badly and even causes things to subtly malfunction those normally were expected not to be RAM dependant. E.g. sometimes particular preset sounds plays too silent or certain parts of them refuse to be editable, or their LEDs show mess or sustain doesn't work or even the chord volume slide switch refuses to change volume at some positions (e.g. only 2 of the 5 positions have different volume). These flaws can drive you crazy and make you take the entire thing apart for hours to successlessly search for dirty switches etc. etc. and even in the manual I downloaded from Yamaha there is no reset procedure for this keyboard explained to prevent this. (In manuals of later Yamaha PortaSounds with battery backed up RAM stands usually that certain simultaneous button presses reset the thing.) Replacing a soldered button cell very likely needs no Casio keyboard specialist at all, but only a sufficiently skilled ordinary TV repairman. The MZ-2000 service manual is at Elektrotanya.

By specification, many potentiometers survive less than 1000 motion cycles - in real life much more, but the fate of DJ console crossfaders (and some lousy slide pots in old Bontempis) prove that they breakdown by constant use. It is not abuse unless a Casio is smashed and burned on stage or such demolition show crap. But why this mechanical solution if you have a microcontroller!? This is not a Hammond Novachord (1st full polyphonic tube synth, with freakish complicated mechanical timbre switch mechs). You may simply simulate a variable resistor e.g. by an optoisolator (photo transistor + LED, or LDR + incan if you want smooth (but slow) transition) or simply controlling the input through an 4066 analogue switch IC and feed it with a current produced by the Arduino. Very likely the pitchbend potentiometer is polled like a homecomputer paddle port, i.e. a charged capacitor is polled by a digital input that counts the time until voltage drops below a threshold. (The duration is used as the detune value.) Then a current recharges the cap and everything repeats. You may easily simulate that by testing for the recharge pulse and then pull the voltage down (or up, depends on circuit) for an adjustable time through a resistor to simulate the potentiometer position. To keep the pitchbender functional, you may also wire your input parallel with the pot through a resistor. Voltage values may need to be tweaked (possibly through an op-amp, but likely a trimmer is enough).

I seriously doubt that a 2500$ synth (which may be sold in quantities of some 100 to maximum few 1000) would be considered by Casio worth to start a production run. Casio tends to make rather several 100000 copies of their plastic products than deciding to go that low. (Exception was the "Symphonytron 8000" modular stage organ system, of those only 100 were made - likely because it sold so badly.) I doubt they will add 8GB samples anywhere in their hardware. E.g. the XW-G1 contains only 32MB flash memory and 512KB RAM, operated by a COB (black blob) CPU that likely runs on something like 48MHz. I suspect that they perfected that special CPU in 1990th (thus it has limited address space) and only slightly improve it among keyboard generations. They have learned to write a nicely complex and versatile synth engine software (may be in highly optimized assembly language) on it, but can not easily port that to modern hardware and so keep making inexpensive instruments with low memory and strange limitations. in this thread is my brief analysis: Pooched XW-P1 Casio better should keep making their normal instruments and iron out technical flaws those deter professionals (like missing midi-sync in XW-PD1). Also offering e.g. their Privia hardware to furniture companies to built digital grand pianos (those don't sell in large enough quantities to be shipped by Casio from Japan) would make sense, because reasonably cheap piano-shaped objects by noname companies revealed to contain toy grade hardware (like those silicone rollup pianos) those weren't good for anything than room decoration. Casio is not that low-end as some snobbish musicians claim; nowadays Chinese no-name tablehooters often were much more horrible than Casio home instruments. (With Yongmei-like trash e.g. polyphonic play failed by omitted matrix diodes, flimsy plastic resembled yoghurt cups and tiny switching power supply PCB on flimsy wires looked like designed to set the room on fire once the brittle single screw plastic post falls off.) Casio should focus on making a new equivalent of the VL-1, i.e. inexpensive hardware with editable sound parameter. I would love to see e.g. a remake of the classic SA-series (SA-20 etc.) with LCD display, USB and a fully editable version of its unique "PCM" (first softsynth on a chip) sound engine, made for about 50 to 100€. It should be useable with or without computer and be capable to store sequencer data and sound patches on USB stick or SD card.

Snares and cymbals in classic drum machines with low sample rate are (unlike other drums) often output unfiltered, i.e. the samples will not sound dull, but the overtones you hear are fake (stair steps of the waveform). I haven't checked if this is done in RZ-1 (service manual can be found online). The same percussion IC however is also used in Casio CZ-230S. This is what I wrote about it. The percussion generator "NEC D934G" (64 pin SMD, pins count anticlockwise) is a percussion sound IC with external sample memory, that was used in some Casio keyboards from mid of 1980th and (2 of them) in the RZ-1 drum computer. Percussion samples can be in ROM or in RAM for a sampler function, but sampling has to be done by the controlling CPU (e.g. " NEC 7811G-120" in RZ-1). The samples seem to be only 8 bit, but can be magnified by 2 additional volume control bits (pin DB6, DB7) from the CPU for accents. Percussion sound is output at 20 kHz through an external 10 bit DAC. Its analogue waveform then has to be routed through 4 sample & hold circuits to separate the 4 multiplexed analogue percussion channels (synched by the 80 kHz CLOCK output). The D934G can output 2 percussion sounds on each channel (named 'alpha' and 'beta'), those apparently can be separated further by 2 trigger signals from the CPU. The CZ-230S uses more pins than RZ-1 - likely due to larger address space.

A faulty key group (or every n-th key is dead) is often a bad solder joint in the keyboard matrix. However if factory reset fixed this, then it is likely a software-related problem, which may be caused by power surges, static electricity, EMP (lightnings) or accidental unplugging while powered on. battery problem? Does the XW-G1 store user data in SRAM or flash memory? While the latter may make it prone to firmware corruption if implemented unsafely (i.e. no write protected address space), SRAM will freak out when battery runs empty. This may be even a soldered tiny rechargeable lithium button cell hidden on mainboard, that drains over time (possibly many months) when main batteries are not inserted or empty. E.g. my much older Yamaha MK-100 makes bugs those can be absurdly confusing if you don't know what's going on. That is to say, the MK-100 stores all settings in battery backed up RAM; with no batteries inserted the RAM is backed up by a large electrolytic capacitor for a few days(?). When the cap runs empty, this messes up the data badly and even causes things to subtly malfunction those normally were expected not to be RAM dependant. E.g. sometimes particular preset sounds plays too silent or certain parts of them refuse to be editable, or their LEDs show mess or sustain doesn't work or even the chord volume slide switch refuses to change volume at some positions (e.g. only 2 of the 5 positions have different volume). These flaws can drive you crazy and make you take the entire thing apart for hours to successlessly search for dirty switches etc. etc. and even in the manual I downloaded from Yamaha there is no reset procedure for this keyboard explained to prevent this. (In manuals of later Yamaha PortaSounds with battery backed up RAM stands usually that certain simultaneous button presses reset the thing.)

It is a bit strange that the modern XW keyboards have so little memory. My Casio MT-540 of 1987 had 768KB (mask) rom, which was quite a lot upon that time. (For comparison, the entire RAM of an Amiga 500 homecomputer of the same year was only 512KB and its rom 256KB.) The SRAM IC9 "R1LV0414DSB" on the XW mainboard photo is only 512KB (organized 256kWord*16bit) and thus way too small to hold the complete firmware during update. But if the hardware or firmware (by what reason ever) can only execute "chip erase" to format the entire ROM (i.e. nothing left to boot from), it for sure still would have been possible to load the boot code (flash loader) from PC into SRAM before starting erasure and then immediately write this part back into flash rom without depending on a still properly working PC program and USB connection. This way the keyboard would never become unbootable after power-off (unless a freak accident or extreme stupidity causes a power fail or powers-off just between formatting and reinstalling the boot loader) and could still always boot into the flash loader, no matter what happens to the PC or connection. Has anybody tried to load the BIN files from the update program on a small SD card (old hardware often supports maximum 2GB) or even USB stick and tried to start the keyboard firmware update mode with it inserted? If the CPU contains mask rom (it very likely does), it would be plausible to include such a fallback method if the normal method went wrong. It may be that special file names or other strange tricks are needed to make it work (possibly another button combination or even connecting 2 special pins on the mainboard), but before discarding the mainboard I would try to boot directly from SD or USB media. (What is the procedure with Privia or older Casios with updateable firmware? Is there something similar?) E.g. the CTK-2080 service manual mentions an easteregg, namely holding buttons "0"+"1"+"2" together during power-on enters a diagnostic program (aka test mode or service mode) where various functions can be checked. Items are selected with the cipher buttons. Button "9" displays rom version and model number. 0 = return to main menu 1 = button check (press all keys and buttons in a certain order) 2 = LCD check 3 = AC adaptor check 9 = rom version check If USB works correctly, it will show in Windows device manager an item "USB Audio Devices" under "Sound, Video and Game Controllers". Power off to exit. (But I guess this works only with intact firmware and not when it is already bricked.)

The flash updater apparently first erases the entire firmware and so the only thing that keeps it capable to communicate through USB lives in RAM, i.e. power-off is deadly. The fact that the CPU has a software number (at least in CTK-2080) suggests that there is a mask rom built-in. But that it can not communicate by itself with the updater is a massive design flaw. Isn't there a way to save firmware (possibly those 2 .BIN files) on an SD card and press a button combination to make it boot from there? On the rear edge of the XW mainboard I saw (on photo) an unused 9 pin connector "CN9". May it be that this is used in factory to load firmware first time, or do they really need to flash the roms before soldering them in? The IC1 and IC11 have marks of 2 blue or 3 yellow dots, those may get added after successful flashing. But the handwritten model number on PCB back and the empty "MX-[______]" and "[NO._____] fields looks like when the mainboards are generic (made for several models) and only receive their firmware when they are actually installed (or sold as an expensive spare part). The flash ROM type "MXIC MX29GL128ETL2I-90G" can be found in the datasheet of "MX29GL128E" (56 pin TSOP) by Macronix. It has "128Mb x8/x16" architecture, i.e. it contains 128Mbit that can be accessed as bytes or 16bit words. (I.e. the instrument has 32 megabyte rom in total, which is not much in an age where you can buy several GB flash memory per Euro or US$. Nowadays it would be cheaply possible to put several CDs full of hifi samples into such an instrument, which makes me suspect that possibly the employed special CPU is too old to address more memory. It might even be that this was the technical reason to make a separate XW-P1 and G1 with each different omitted features, although I guess this was rather a marketing decision, because bank switching through external small parts is nothing new.) The MX29GL128E chip contains various security stuff including a "Solid Protection Mode" and a "Password Protection Mode". (These are likely functions needed for SD-cards.) It is unknown if the XW hardware makes use of them. Also individual sectors can be write protected. There is a "chip erase" command to auto-erase (format) itself without need of complex software interaction, which may be what the updater does.

This was typed in user mode page view (i.e. CSS disabled but Javascript enabled). Without page formatting the input field is tiny and horrible layout, but there is an input field (about 7x2cm) that can be typed text in. The input field seems to use use these classes: CLASS: "ipsComposeArea_editor" CLASS: "ipsType_normal ipsType_richText ipsType_break" CLASS: "ipsContained" CLASS: "ipsContained" Apparently there is a CSS bug. Here is some error console output about it:

I depend on Opera 12.02 because my PC (made from finest DOS-age hardware with AMD K6-3+@550MHz,768MB RAM on Win98SE) can not run newer browsers. When I try to reply in this forum, the editor window has turned into a solid grey rectangle with no text input field. So I can only edit with javascript turned off, which shows a plain textbox without style buttons that is awkward to handle, because I have to remember BB codes by trial and error and so e.g. can not upload photos or properly quote with name anymore. Please fix this!

Thanks for info. Its sad that no externally made solo synth patches can be loaded. But most FM home keyboards were preset too. By the way, the XW-PD1 somehow blantantly resembles my "Coleco Zodiac" astrological computer from 1979. Maybe a good hacker should now interconnect both to unleash some unheard noise that shakes the cosmos vigorously. (Remember what was done with a Casio MT-210 in the "Masters of the Universe" movie. ;-) ) Coleco - Zodiac - Astrological Computer 1979 - YouTube Coleco Zodiac - Handheld Games Museum

CPU speed comparison Interesting to discover is that the COB CPU "UPD800468P-012" in CTK-2080 (chip likely identical with XW-series, last 3 digits likely indicate software number, seen in schematics) has a clock rate of only 48 MHz. With 48 note internal polyphony this may mean that the sound engine cycles through all channels with an update frequency of 1 MHz. Modern high performance CPUs (tablets, PC, digicams, Raspberry Pi) are clocked at least 10 times higher, which suggests that Casio keyboards use specialized dedicated (or just old) hardware and not a highspeed modern microcontroller to run state of the art DSP functions in software. (If the XW chip is specialized, this may be a reason why it can not simply play existing patches from CZ or VZ series, despite phase distortion in software emulation would not need much CPU power.) For comparison, the SA-series 4 bit RISC CPUs of 1989 was already clocked at (at that time quite fast) 21.725 MHz for 8 note internal polyphony, so clock rate hasn't increased that much, but instead parallel processing capacity was slightly increased. The 4 bit CPU would need to run at hypothetical 130.35MHz to compute 48 channels. To need 48 MHz for this job means it is only about 2.716 times faster than a chip made in 1989. For the 64 note polyphony of XW-G1 it would need either a 3.62 times faster CPU or run at 64 MHz clock rate. (I don't know how high it is actually clocked.) The MT-540 of 1988 even had 10 notes at 21.725 MHz, so it would need to be only 2.897 times faster to handle this. Please correct my if my maths are wrong. This may be a bit speculative and possibly unfair because the XW synth engine may process more complex synthesis and DSP algorithms per channel and so internally use a much faster CPU. But it gives an idea whether Casio designed really something completely new or only slightly improved a creation perfected in late 1980 to 1990th. The fact that XW-G1 has this strange chiptune appeal that tends to sound a bit grainy with less parameter resolution than modern softsynths and partly coarse timbre transitions would make it plausible, that it indeed feeds "good old" 1990th algorithms with only bigger samples and increased polyphony through otherwise very small hardware. This is not a complaint (Casio synth fans have learned to like that sound) but only a conclusion.

I yet own no XW-PD1, but I am interested because it takes way less space than an XW-G1 and has strongly dropped in price. But I read that PD1 is mainly a preset instrument, so I have some questions. - Can "solo synth" patches not be edited at all, or is this possible through PC software? (My CZ-230S is also preset based but can edit patches through midi.) - Can it load custom "solo synth" sounds made e.g. on XW-G1 or XW-P1? - Can it play the internal sounds through a keyboard when connected through USB-midi? - Do the drumpads really lack velocity? This is very uncommon with modern non-toy drum machines. - Does it contain internal wifi or bluetooth? To me this would be a definite exclusion criterium not to buy it, because these technologies emit brain destroying microwave radiation and I strictly boycott anything equipped with it.

While with older videogame consoles soft- & hardware piracy is common (TV-Boy = Atari 2600, Famiclone = Nintendo NES, Wannawii = Sega Megadrive in Wii-like case etc.), yet I never saw any genuinely "pirated" cheap home keyboard clones. Obfuscating by "potting" a known microcontroller chip would only make sense when the actual sound engine was purely software based (SA-series was, but laters likely added hardware). And do not think that chip manufacturers laboratories have no means to decap ICs and dismantle them under an electron microscope. A Casio keyboard is not a banking card, so (unlike homebanking) extracting a master key from it would not mean to destroy the entire system or business model. It takes much more than a bit of code to imitate a keyboard hardware mass-produced cheap enough to compete, and they are big enough to be detected on trade fairs when counterfeit for profit. Modern softsynths on PC (accessing gigabytes of samples through a fast CPU) tend to have higher timbre quality than Casio hardware, so pirating modern models for profit in software would not be that reasonable nor destructive to Casio, those instrument benefit from being self-contained with real haptic controls those no mouse or touchscreen can replace. Nobody buys a Privia only as a midi sound source when he does not want to play on its keys. Multi-gigabyte sample software pianos on PC can do higher quality cheaper (and take less space) if that was the only purpose. I expect that it would be relatively easy for a laboratory to desolder and read the flash rom from an XW mainboard to extract code and samples. It would not surprise me when even the official firmware upgrade files have no strong encryption at all but only a checksum. E.g. of my Samsung WB210 digicam the downloaded firmware file could be easily read with a hex editor and revealed plenty of setup commands in plain ASCII text. I haven't dug deeper, but I am quite sure it is nothing encrypted. Here is what I found out: Samsung WB210 review (big touchscreen,wide angle,non-wifi|Hacking?) In music keyboards I did find clones of simple Yamaha FM chips like YM2413B (OPLL) and the cheapish squarewave chip YM2163 (DSG - which was not used in any known Yamaha keyboard but only sold to other companies) remade in Chinese no-name instruments (e.g. Medeli, even pinout was camouflaged by bigger IC package). But the main CPUs always differ and particularly I never found any well known digital high quality sound engine pirated. Keyboards with imitated case shape (sometimes Casio knock-offs) instead typically contain chips of the Holtek HT36 series. E.g. the Holtek "Ad-Lib Micro" was the grainy sounding heart of "My Music Center" toy keyboards. In late 1990th this microcontroller series became extremely popular in cheap no-name single-chip keyboards and sound toys, and likely also caused the almost demise of quality brand (Casio, Yamaha, Kawai) toy- and mini keyboard sales. The impact can be almost compared with the videogame market crash of 1983. Despite the timbre quality of the flood of Chinese toy tablehooters was lower than in-house developed chips of the classic quality brands (and many had other flaws), they were simply very much cheaper. So finally even Casio tried to sell such a rebranded piece of cheap awful sounding junk as MA-150/MA-170, and today Bontempi still offers a mix of rebranded Chinese toy tablehooters and own higher quality models. Also Holtek later made reasonable sounding digital keyboard CPUs based on sufficiently high resolution samples, and nowadays of course also other companies make them; but what it began with was the Ad-lib Micro. With multitimbrale polyphonic melody voices, samples, accompaniment and demos it was the first cheap single-chip microcontroller intended to compete e.g. with Casio SA-series, despite Casio's 4 bit softsynth CPU of 1989 was much more powerful (21.725 MHz, about 55.7KB DRAM, 63.5KB sample ROM+14KB program ROM according to patents). Unfortunately nobody spent time to program similarly complex sounds on the Ad-lib Micro, which makes it a bit boring. But among toy-grade keyboards it had the last genuinely novel sound synthesis engine, which timbre was still majorly defined by hardware constraints and not only by programmer's choice of sampled waveforms. Several versions of this Taiwanese lo-fi synth chip exists, those mainly differ in polyphony and amount of memory, but the general sound engine with "ETS" (Electronical Tone Synthesizer) 7 bit static waveform ADSR synth and dedicated PCM voice (percussion) channels of compressed 5 bit samples stays the same. The sound is output through an 8 bit time slice DAC with characteristic glassy timbre. In many instruments for each melody voice 2 detuned ETS channels are layered in software as a (often stereo) chorus effect, which halves polyphony. The datasheets looks detailed enough to emulate this thing if anybody finds a method to dump the rom. ETS channels have 8 static waveforms of 7 bit with 64 steps and ADSR envelope with 4 bit parameter resolution, sustain & vibrato. PCM channels (intended for percussion and effect noises, not main voice) use 5 bit samples with 8 to 5 µ-law compression, but sample rate can be only set for all channels together with coarse frequency resolution. The 8 bit CPU has only 96 byte RAM. The intended clock rate is 3.58 MHz (but can be overclocked higher - it is unknown if also keyboard manufacturers did this). Known generic versions are e.g. HT3630, HT3650, HT3670 and HT3690. The highest model HT3690 has 8 ETS channels and 4 PCM channels (32K*5 bit voice ROM, 128*5 bit per segment, but only 13 sample rates with max. 13.98 kHz) with 8 step volume control. The DAC has 2 melody outputs (statically assigning each 4 ETS channels) + 1 separate output for percussion (PCM channels). It has 4K*15 bit program ROM. Versions with less polyphony have higher sample rate. Holtek published some datasheets of in-house example versions (e.g. as advertisement to bait keyboard manufacturers), but in their component list many other ICs are marked "restricted" (i.e. confidential) because 3rd party manufacturers can order custom software variants of ICs and keep datasheets secret as advantage over competitors. (However upon a time the original "My Music Center" CPU was such widespread that it was obviously offered to many different keyboard companies.) Such modern Holtek successors are named "8-Bit Music Synthesizer MCU" and have naming scheme "HT36+letter+cipher" (e.g. HT36A2, websearch the datasheet). Some of them support midi, USB etc. and are used in fullsize Yongmei keyboards those at least visually compete with Casio and Yamaha and are offered under plenty of fantasy brand names and -prices (from 40 to sometimes fraudulent thousands of € - typically 100 or 200€) on eBay. In the official Holtek forum I read that of the HT36 chip series exists no OTP (prom) version, so normal ones contain mask rom, which will be good news for long term stability and safer circuit-bending, since mask rom does not suffer of bitrot and can not be accidentally overwritten by messing with wrong pins. For hardware developers apparently there is a development board "HT36DB", and of the (newer) "Music MCU" HT36B0 exists a flash version HT36P03 that can be written with a "HT-C300 Flash Writer". The caption mentioned "Music Studio HT36", which may be the name of a development software. But this is all I know. I am not their developer and haven't seen any of this. Holtek earlier also made other awesome sounding unknown toy keyboad LSI chips with strange blip percussion (including the lovely multipulse squarewave based "Hing Hon EK-001" and Creatoy keyboard) those are technically similar like Casio VL-1 but not copies. Sorry if this lead too far off-topic, but this is the typical stuff found in "clone" keyboards; unlike game consoles I never found actual pirated modern sound engines of a name brand keyboard in cheap tablehooters. That is to say, the (now outdated) SA-series patent text US5319151 is detailed enough to reimplement its well decribed softsynth on any modern hardware. (May be it's me who will eventually do this. ;-) ) Despite the SA-series IMO sounds better and more interesting than modern toy grade keyboards, Chinese companies don't seem to care but prefer making their own stuff. (Samples with simple envelope are obviously easier to adapt to changing market demand than a sophisticated real softsynth engine.) By the way, on ElektroTanya I found a Casio CTK-2080 service manual of 2012. The depicted PCB "M822-MDA1" is quite similar like XW-G1 (but has only one flash rom). Unfortunately it also tells nothing about repairing misflash accidents. But the schematics may be useful to understand general principles.

New G1 synths are not made anymore, so if I find a dead XW-G1 cheap on eBay (unless smashed or otherwise mechanically broken) chances are high that the firmware is toast - isn't it? :-( That an interrupted or wrong firmware update can brick a device is a severe design flaw. At the uni in software engineering we learned that updateable firmware always must contain a fallback flashloader that will be never overwritten during flash process and resides in a write protected memory area, from where the device can be booted by a button combination (or at least jumper contact on PCB) to permit re-flashing if the main firmware got corrupted. (In Casio "Databank" watches 2 contacts need to be connected after battery change with corrupted SRAM contents to make it boot. But Casio instead wanted owners to mail their watch to a service center for each battery change, which sounds pretty much like a moneymaking scheme. I hope that such 2 pins exist here too, but I doubt.) But even PC manufacturers don't care, so their EFI Bios is often extremely vulnerable (e.g. Samsung laptops got bricked by installing Linux, because too long error message log files written into the same flash chip overwrote the Bios). In DOS-age IBM Thinkpads a full battery had to be present during BIOS upgrade, because the resetting mainboard chip turned off the power supply input control pin while flashing, thus an empty or broken (since long time not made anymore) battery pack could end in a disaster. (Flashing the XW with full batteries inserted may be a good idea too against mains brownouts.) Also my "Willem Pro4isp" eprommer (connected to parrallel port in Win98SE) software warns to turn off antivirus and CPU throtteling when accessing certain ICs (e.g. Intel MCS-48 microcontrollers). Nowadays where the IoT (Internet of Things) plague is taking over the mankind, necessarity of so-called "security" (or national trojan spying?) firmware updates in all kinds of cheap hardware will be likely the next obsolescence method to make perfectly intact household items unuseable. People those modified their "dBox" settopbox to run Linux (for PVR software etc.) also often ended with bricked hardware (I found a dead one on a fleamarket), but these were neither made for not encouraged by the manufacturer to flash such inofficial firmware upgrades (which included connecting wires to IC pins and such things). That original stock firmware flash can fail catastrophically is unacceptable. I hope someone finds a hardware method to re-flash dead XW mainboards. In past the presence of COB chips (except in watches) was a warning sign indicating overly cheap hardware, although nowadays speed of such uncooled CPUs certainly has become good enough to do something serious. Technically flash memory is such cheap now that an entire 2nd copy of unchanged original firmware can stay in it for emergency. It sucks if Casio really avoided additional memory capacity for a fallback flashloader to save a few (or a fraction of 1?) cents. Casio should have at least a cheap repair service for this and particularly allow to only mail them the mainboard for exchange instead of shipping the entire bulky instrument (imagine a fullsize wooden home piano - yikes!). Does a dead XW mainboard still mount this drive? Then it might be possible to copy contents from an intact to a dead one. Wouldn't it make technically more sense to wipe the old firmware after the reset? So the time without firmware (running only on RAM contents) would be as short as possible.

Thanks for the link. With "encapsulated" I thought of "potted". Here is what real potted hardware means. Look at this 1970th speech synthesizer and what an archeological excavation task it was to document the inside. http://www.kevtris.org/Projects/votraxml1/ another model: http://kevtris.org/Projects/votraxpss/unpot.html But my Atari 800XL power supply indeed was potted too (likely to improve heat distribution, but may be planned obsolescence), so in 1980th I had to build a new one from a kit when the overheating thing failed. I have not coded for Casio and yet own no XW synth (very little space left and mainly interested in older stuff, but may indeed buy an XW-G1 or the small XW-PD1 tekkno groovebox if it is sufficiently editable through a PC app). But I am interesting in technical similarities (e.g. CTK-1000 is obvious audible related to the classic SA-series sound engine). I am not solely into sound toys, but also own some big stuff (e.g. a Vermona SK86). The XW-G1 has that strange chiptune appeal (much more resembling C64 SID than e.g. Minimoog) that I like.

I found no such picture anywhere (not even Google), but I seriously doubt that Casio "encapsulates" chips to enshroud what is inside. The only consumer product I saw something encapsulated inside was my TechniSat cable TV receiver, which contains a rectangular block of black epoxy resin casted into a plastic frame around a PCB section to prevent non-destructive access to the pay TV decoder circuitry. And this rather suggests that in opposite the sealed block contains documented standard parts (e.g. standard microcontroller) they want to lock away from hackers. Pinouts of my 1980/90th Casio keyboard sound ICs and single-chip CPUs don't match anything commonly known and thus are very likely custom ICs anyway. What you likely mean is a COB IC, which tells not more about the chip (and potentially suspected security paranoia) than that it has low power consumption (needs no heatsink) and can be reliably mass produced in huge enough quantities to justify making it cheaper by directly bonding it to a PCB. The low power constraint however means that it certainly contains no cutting edge ultra high performance general purpose CPU. ("Raspberry Pi Zero" is a good example how fast such an uncooled chip may be. With specialized hardware it may need even less power to do audio processing.) A related Casio patent (about integrating a main CPU with several audio LSI cores into one chip) may be US5541360 of 1996 (with priority dates going back to 1992). On this page I found a small photo of the inside of the XW blob: http://www.casio-music.com/de/instrumente/technical-insights/xw-technologie/ Here is an excerpt from my WarrantyVoid FAQ: about COB ("black blob") ICs: Did you ever wonder why ICs in modern toys and consumer devices are often welded in a black blob of plastic directly to the PCB? It is a false myth that the blob of such COB (chip on board) ICs serves the purpose to prevent commercial espionage by hiding the circuit inside. In reality it is simply cheaper to mount the silicon die of a chip directly to a small PCB (so far it needs no heat sink) than to package it inside a rectangular plastic or ceramic case and then solder its legs to the PCB. The plastic blob protects the die against corrosion and mechanical damage, and the blob is not black to enshroud the interior against spectators but to shield it against light, which would otherwise generate electricity in the silicone (like with solar cells) that prevents proper working. (Despite this I even found white blobs in certain toy keyboards.) Particularly with some old COB ICs the blob was made instead from hard plastic resin from soft silicone rubber. Be very careful not to scratch or tear such coatings, because it may easily destroy the chip or its fragile bond wires. Do not use silicone oil near these; it may dissolve the rubber over time and so weaken the seal. Also be generally careful not to burn the blob with a soldering iron. In some old handheld electronic games I even found bare microchips without blob, those were only protected by a hollow hard plastic cap that was bolted to the PCB; by the lack of corrosion protection the life span of such ICs is very questionable. I own an LCD game which has a hole in that cap (production fault) and continuously crashes. about unlabelled & camouflage ICs: In music keyboards and other devices sometimes there are ICs without printed type number. Often it is sanded off but may be also painted over and sometimes additionally stamped over with a short internal name. Obvious purpose of this can be to make hardware piracy difficult (like with Wersi organ kit ICs). But especially in cheap noname hardware it can mean that it was build from cheap rejected name brand ICs (e.g. Yamaha sound chips in a Medeli tablehooter) those failed the quality check (e.g. function only at 900kHz instead of the rated 1MHz) and so were debranded by the IC manufacturer in the fear of spoiling their brand name with substandard parts. Another explanation is that these even may be re-branded scrap parts. I saw a TV docu that poor people in China and 3rd world countries recycle e-waste by desoldering electronic components (often causing heat-damaged by primitive methods like blowtorch or coal fire) to bulk-sell them untested. To make more money, criminals then often print fake new labels on and ship them to the industry as counterfeit new parts (often faulty or wrong types those can be a safety hazard). I don't know if also music keyboard ICs are affected. A sanded off type number often can be made readable by wiping a drop of water or silicone oil on the surface and shining on it with an LED lamp at a flat angle. A digicam helps to enlarge details. Particularly COB, but also other custom ICs were often released in several different package variants (e.g. to fit on different PCBs or to enshroud what they are based on). While the silicon die stays the same, packages can have unused pins omitted or for configuration internally bonded to nearby pins ("bonding options"). Also one of the supply voltage inputs (negative, 0V or positive) often covers the whole bottom of the die; so this pin may be placed anywhere (also as multiple pins among others). Packages that are too large for a particular die sometimes have many pins (often every n-th pin) internally not connected; such NC pins will show high resistance (also in diode test mode) in both directions against all others. Additional pins may be also internally connected to their neighbours. (E.g. FM sound ICs in Chinese keyboards had many additional NC pins to distract from that they were copies of much smaller Yamaha sound ICs. But even Yamaha himself proudly created a "16 pin" sound IC YM3427 with 8 of its pins NC.) To identify camouflage IC variants, it is most important to know that despite possibility of omitting or inserting blank or duplicate pins, the pin order will stay the same (at least unless the die was redesigned too, or a complicated adapter built into the package). Regard that the placement of pin 1 still may differ when the die was installed rotated or the numbering was changed. So it can be a good idea to compare the actual order of function pins around the chip rather than their absolute positions. (Of course this tip is only valid for classic ICs with pin rows along their rims, not modern packages (like BGA or Pentium CPUs) with hundreds of contacts covering their entire bottom.) With COB the pins typically count anticlockwise when looking at the blob side.

If you don't like modulation wheels, why not placing spring loaded potentiometers outside, connect a spool of ribbon and attach that to your clothing or a loop around the body. So you can change modulation by bending e.g. your trunk back and forward. This can also be used to avoid volume pedals (e.g. if as a yogi you play sitting in lotus seat on floor). I know similar mods for Indian table harmoniums, where the rear air bellow is moved by a belt around the body (instead of left hand) to permit dual handed play.