Before we get to the Korg chips,
here are a few things I'd be happy if you could help me with:


Maybe you have the card or have downloaded a card image of Roland FANXRUP1 or FANXUP1 that was once floating around the web? If you have the card but don't know how to create an exact backup of it, I'll help you with it - it's easy.

Although unlikely, maybe you have Roland VP-9000 demo disk demo song files?

Or maybe copied floppy files Casio WK-1800 Accessory Disk (demo disk that came with it)? It doesn't seem to be for download on any casio support and fan sites. You might like the main reason I'm lookig for it - it is clearly a fun attack vector to non-destructively dump its maskrom (in Hitachi H8) since it actually allows to execute arbitrary code! While dumping can be done even by blinking a led chdk-style, in this case there's already initialized and working uart for that. Further, if all is well, waveroms (8Mbyte total) can also be dumped this way, since H8 can quickly access them, and procedure to do this for studying will be in the maskrom. On the other hand, I'll very likely at some point just desolder and dump waveroms anyways - but it could be more epic to dump it via diskette and midi.

I'm looking for an Alesis XT:C rom dump and also Alesis XT-Reverb rom dump, Here's a MidiVerb dump for you in exchange. In fact, you can find tons of my dumped roms on this great webpage.

Also - maybe you have an Alesis Wedge rom dump? Or, at least, have made some photos of its inside?

It seems that Yamaha CS6x/CS6R supplied accessory CD (with old original software and preload audio clips) is not amongst the available support downloads -  maybe you have it and can dump its contents?

It seems there that earliest version (A00) of the Yamaha FB01 firmware rom is somewhat rare, and I esimate, based on the units that I have, that at most 10%-20% of the units in the wild do have it. All usual ones have the most-likely-final C00, that I have desoldered, dumped and shared in the far away ancient times. However, for the sake of history and disassembly fun, having A00 dumped is a must as well, though I don't realy have too much of spare time to desolder and dump it, especially since maybe someone has already dumped this version (given the popularity and abundance of these modules). So if you have it, let me know. At some point I might desolder it and dump it, if it turns out to be realy rare, but I'd better spend that time desoldering and dumping something else.
Also, for the curious, initial hardware version (with A00 rom) only differs with (besides missing ground wire from the pcb inside) that it does not have a metal sub-chasis under the pcb, so you can tell that from the outside by looking from the bottom: early ones have two holes, but no screws in them, and instead has white plastic pegs in other holes right besides front panel. Serial numbers for the A00 version seem to be below 02000, while 07000 and up definitely has version C00. No idea about if there was a B00 version in betwen 02000 and 07000.
(I will likely make a separate article on FB-01 at some point)

Any of the ART MultiVerb, ProVerb or similar reverbs/effect rom dumps are also of interest for me, for example I'm looking for ART Proverb 200 - here's an ART Multiverb LT dump for you in exchange.

It seems that there exists a version 2.02 firmware dump of Lexicon MPX-500 - maybe you have it?

Speaking of Lexicons, maybe you have a dump of Lexicon LXP-1? Here's LXP-5 for you in exchange.

Judging by the immense popularity of PSS-51 in eastern europe, aisa and amongst russians, it is quite hard to believe that no one have dumped its roms. Maybe you have dumps of Yamaha PSS-51 program, abc and/or waveroms? Or maybe you know one for local pickup somewhere within Baltic states?

In regards to multieffect pedals - maybe you have a rom dump of Roland/BOSS GT-10B? There in this great webpage you can find my GT-3,5,6,8 dumps.

And maybe someone has already desoldered and dumped PSR-6000 waveroms? Those are three pieces of DIP42 chips, each 2Mbyte (besides two more chips with styles and then two DIP40 with program and demo songs). I am somewhat tempted to grab one from the local trash market to do this, yet not sure if I realy have to. And I don't have a clue when I could actually get to desoldering it, even if I' ll grab one now. Still GEW9 stuff seems rather interesting and not that common to come by in the keyboards.

If you do, please email me!




You've just stepped into
an unfinished page

So it is full of dummy placeholders and copied blocks of text. Don't take it too seriously.


Korg DSP Chips

This section was overdue for more than a decade, sorry. Despite a popular belief that I'm a yamaha chip and rom zealot, I am quite a harcore korg chip nerd as well. So there it goes!

Korg Poly800 can actually be considered a CASIO. Yep, you red that right - tone generator in it is a lowest-grade single-chip primetive OKI (MSM5232), unable even to bend pitches for individual voices (so no pitch envelopes and only a cheesy transistor organ single common LFO vibrato) that pretty much prefectly explains how that piece of crap sounds...

Sorry for the mess - take these notes as they are, as it's quite unlikely I will fix this page anytime soon.

MB = Fujitsu
 

Korg MDE MB87405
Korg MDE MB87405 DRAM 20bit x64kw
  • Korg MDE
    MB87405

    This chip is just for effects processing - approprietaly called "Multi Digital Effects" - and stuff on it does sound very good - mostly very dark, pixely, muddy and vintage.

    DRAM for it is comprised of five chips (nec D41464V-10), totalling to 20bit x64kwords.
Korg TGI, TGII, VDF
  • Korg TGI, TGII, VDF chipset
    TGI = MB87402
    TGII = MB87403A
    VDF = MB87404

    Used in M1/M1R, M3R, T1/T2/T3 and such. Sounds very lo-fi, but fun and booleg-house-y to the core. Usually this chipset is complemented with MDE for effects processing.
Korg TG88, DF88, WS89 inside O1Wpro
  • Korg TG88, DF88, WS89 chipset
    TG88 = MB87726
    DF88 = MB87727

    Used in Wavestation, O1W and such. Sounds fantastic in dark and mysterious ways. WS89 was an experimental rom-lookup based waveshaper. Usually this chipset is complemented with MDE for effects processing.
Korg TGL MBCS35104
  • Korg TGL
    MBCS35104

    Mostly cheap-ass, dark and muddy sound.
    Used in a lot of stuffn with simple non-resonant filters. Can be paired together for doubling the polyphony, since they can share wavebus, using it only half the time.
Korg TGL2 MB87A104A
  • Korg TGL2
    MB87A104A

    Mostly cheap-ass, dark and muddy sound.
    Used in a lot of stuffn with simple non-resonant filters. Can be paired together for doubling the polyphony, since they can share wavebus, using it only half the time.
Korg TG92, DF93 in Trinity Rack
  • Korg TG92, DF93

    TG92, TG92B, MB87103, MB87D103, YEN-KYO
    DF93, DF93A, MB87104, MB87D104, KATYO-HEN

    It might likely be that Fujitsu gate array base for both of these chips is called MBCE31751, with the customization numbers 021 for TG92 and 030 for DF93 respectively.

    Sucessor chipset to the one in wavestation and family. Amazing, detailed and spacy sound, substantial upgrade over the old ones. Sad, though, is that it has devastatingly low polyphony.

    Trinity! 48kHz samplerate.

    TG92 does great quality 32 voices
    DF93 does resonant filters for each.
Korg TGL96 MB87F1710 in Triton Rack
  • Korg TGL96
    MB87F1710

    Sucessor chip to the Trinity's TG93&DF94 chipset. Amazing sound. While polyphony has been doubled, it is still tragically low, especially since these chips cannot be paired together to share wavebus.
    Tritons and MicroX! (5V I/O and 3.3V core)

    Does great quality 60 voices with resonant filters and a bunch of simultaneous effects, all in true 48kHz samplerate.

    Effects processor's capabilites are about exactly that of two Panasonic MN19412A - so could likely have 512-step WCS.

    By the way, all the tritons actually are fully capable of 120-voice polyphony... but only 60 voices are at your disposal. How so, you might ask? Well, there are two TGL96 in every Triton (actual ones, so not counting Le, TR61, TR88, etc), with the second one having its 60-voices unused. Reason for this is that waveroms are connected to only one of the chips, and cannot be shared between them. Only Triton that has this solved to a degree is the Ex(treme) model, which has first 60 voices as usual for normal triton sounds and your loaded samples, and the other 60 voices (second TGL96) connected to separate waverom bus with some expansion waves (and therefore cannot play user samples or "normal" patches).

    Single TGL96: TR61, TR88, MicroX, X50, Le
    Dual TGL96 (one only for fx): Triton (all the classic, pro, rack...)
    Dual TGL96 (second one with its own waveroms): Triton Extreme
Korg FXD03 MB87R1420
  • Korg FXD03
    MB87R1420

    Multi-fx chip that has an excellent algo library for it.

    It also has a built-in DIT for spdif digital out that would be so great to make use of in most korg synths and pedals that miss it, however you need to configure and set it up through registers, so it unfortunately is not active with stock firmware.

    Have done with it some serious tinkering, and will someday make a page about it, and how to boot, program and control it via its serial interface. You can make a nice development rig with it from a dirt cheap AX guitar pedal.

I am going to grow the lists gradually, whenever I will find some spare time for doing it.

  has amt.

Off-the-shelf DSP chips in korgs

This section was overdue for more than a decade, sorry. Despite a popular belief that I'm a yamaha chip and rom zealot, I am quite a harcore korg chip nerd as well. So there it goes!

Sorry for the mess - take these notes as they are, as it's quite unlikely I will fix this page anytime soon.
 

Panasonic MN19412A in Korg Trinity
  • Panasonic MN19412A

    Used as an evolution of the MDE, 48kHz samplerate.
    Used in Trinity and Z1. Fantastic sound!

    Korg really did an amazing masterwork on this chip porting best algos they had at that time, and developing new ones.

    Also can be found in otherwise absolutely unremarkable CD player Denon DN-C680 and similarly (though less) unremarkable Kawai K5000 (which is, thanks to witty ebay sellers ~ "youtube reviewers", is completely submersed in bullshit about it and its "grand unique vinage amazing rare" features; who would otherwise call it a flat sounding 4Mbyte samplerom rompler-GM-workstation - that is unless they need to sell one, or brag about having), where effects are nowhere even comparable to korg's and also are rather uneditable (almost no parameters to tweak at all).

    Currently I only have a pinout of this chip for you - no datasheets found so far, since it seems that all in-depth info about this thing is only in the printed literature from 90s. All that can be found is a scanned index of some panasonic book with page numbers referring to this chip, and page 48 of this book just gives some basic specs.

    These basic specs seem to suggest 256-step WCS with 16bit instructions and 24bit paired immed coeffs or registers. Also there's some rom, which could likely be some sine and/or log lookups.

    Finding its datasheet - especially its assembly & machine code reference - will definitely make disassembly and figuring out the dsp microcode much easier.
Korg TMS320M10NL inside DVP-1
  • TI TMS320M10NL with cusomized bootrom

    Used in DVP-1 vocal harmony processor (and no it's not an actual vocoder by any means). A stock DSP with custom bootrom, that loads the signal processing code after startup.

    And again, no, DVP-1 is not a vocoder, except in the sense of voice harmony processor terminology - in which a "vocoder mode" is a feature of harmony processors that can force input voice to an unrelated arbitrary pitch (fixed, selected, pressed via keys, etc). Holding or shifting formants is done via pitch-synchronous splicing (PSOLA). In case of DVP-1 - a very low quality one at that.
Korg Kaossilator TMS320VC5501
  • TI TMS320VC5501

    Used in kaossilators, kaosspads and such - reverbs are horrid, but ported va-synthesis stuff is fine.
Korg Radias TMS320VC5502 FXD03
  • TI TMS320VC5502

    Used in all sorts of good stuff, but best if together with Korg FXD03 effects dsp, since reverb and such effects written (or poorly ported?) on these TIs in korgs are an abomination.
Korg Prophecy/Solo-Tri TMS57002
  • TI TMS57002, TMS57070FFT

    Used in a lot of stuff, but good are only for VA and Karplus-Strong kind of synthesis things on them. Effects on those are very poor, besides delays which are trivial.

    Here's some incomplete datasheet on TMS57002
Korg something with DSP56362
  • Motorola DSP56362 (I think I've seen 364 and 367 in korgs as well, but maybe not)

    Used in MicroKorg, MS2000, Electribe ER1 and such. Chip is cool, but effects algos written by korg on it are not good at all. Virtual-analog stuff is in range from meh to poor as well, and yes, I know you very likely and easily influenced by seller bullshit legends about MS2000 and microkorg "THE SOUND" but it is nowhere near good, nor virtually analog. It is ok, though. But I sometimes enjoy it in a sort of guilty pleasure way, since there are some fun aspects sometimes.

I am going to grow the lists gradually, whenever I will find some spare time for doing it.

  has amt.

S ries

The20).

Rym.

D-7well).

Chipset
Features & capabilites
Timeline
Sound character
Synths / effects processors
Dual TMS320M10
4-voice pitch shifters with pitch-synchronous splicing; approx 22kHz
~1986
Low precision, glitchy and lo-fi harmony processing pitch shifters that use some primetive implementation of PSOLA algo.
DVP-1
TGI+TGII+VDF+MDE
"AI"
"Optinized Covolution" (sic)
16 voices; 2 fx; 31.25kHz
~1989
Dark, aliasy and pixely sound, but good.
M1, M1R, M3R, T1/T2/T3
TG88+DF88(+WS89)+MDE 32 voices; up to 6 fx; 31.25kHz
~1990..1994

TG88, DF88 developed in 1988;
WS89 developed in 1989
Considerably dark but exceptionally tasty.

If you like the Fifth Element and Sonic-CD soundtrack, this will absolutely be your cup of tea.
O1/W, Wavestation
TG92+DF93+3x PanasonicMDE
"ACCESS"
32 voices; a couple of resonant filters for each voice; a ton of insert and master fx; 48kHz
~1996..1998

TG92 developed in 1992;
DF93 finalized footprint 1993/12/24
Airy and smooth - fantastic evolution of wavestation's character.
Trinity, TR-Rack
Dual MDE
48kHz (37.1kHz for A2, A3)
~1991
Pixely, lo-fi-ishly oversaturated and exagerrated sound, yet strikingly enjoyable.
A1, A2, A3
TGL
"AI²"
32 voices; 2 fx; 31.25kHz

Muddy, low fidelity; kinda like it to a degree, yet more of a guilty pleasure
i1/i2/i3,
TGL2
"AI²"
32 voices; 2 fx; 31.25kHz
Muddy, low fidelity; kinda like it to a degree, yet more of a guilty pleasure
Dual TGL2
"AI²"
64 voices; 2 fx; 31.25kHz
actually capable of 4 fx, but korg didn't care to implement it in main firmware;

Muddy, low fidelity; kinda like it to a degree, yet more of a guilty pleasure X5D, X5DR
TGL96
"HI"

~1997
It's good. TR61, TR88, MicroX, X50, Triton Le, PA50
Dual TGL96
"HI"

~1997 It's good. Triton (all non-Le), Triton Extreme/Ex
TMS320VC5501


most if not all Kaossilator, Kaosspad
TMS320VC5502+FXD03
"MMT"
8 voices; 5 fx; 48kHz
~2007
Besides low polyphony, it's good. R3
Dual TMS320VC5502+FXD03
"MMT"
24 voices; 5 fx; 48kHz ~2005
It's good. Radias
FXD03



AX3, AX5 and a lot of other pedals
Dual TMS57002BPHA 48kHz


A4
Triple TMS57002BPHA

~1995

Prophecy, Solo-Tri
TMS57070FFT

~2000

Electribe EA1, EM1, ES1
Twelve TMS57070FFT + PanasonicMDE
"MOSS"
12 voices
~1998
It's good. Z1
DSP56362
too few voices, and a very crappy primetive fx;
~2002
Chip is great, but korg's sound on it is not. There are fun aspects to MS2000, don't get me wrong - I do have a bunch of MS2000 myself. But it is nothing to write home about. All the tales about legendary-ness of those are bullshite stories from ebay sellers parroted by wannabe reviewers.
MicroKorg, MS2000
NEC uPD77213
"MMT"

~2004
Incomparable sonic upgrade to the original cheesy electribes.
Aaah, if only EMX had an actual editing of sounds, instead of that basic single-function few knob abomination.
EMX-1, ESX-1






* - multiply

By the way, all byte multiplicators k and M here are Si - meaning x1000 and x1000'000 respectively, and in most cases are precise byte values, rounded to whatever looks good in the table.