circuit bending details

• analogue VCF (with resonance and sustain preset trimmer)
• analogue percussion section (with decay trimmers)
• in the middle: main voice sound generator (sound chip D931C) and clock oscillator (below)
• accompaniment CPU D930G (which also polls the key matrix)

Most key matrix pins are at the top, and 2 at the left side of the D930G chip. The sound output and drum trigger lines are at the bottom side. Bass, arpeggio and percussion employ a capacitor decay envelope.

important: All PCB locations in this text are viewed from the solder side with keys facing down. Regard that all PCB writings at capacitors only tell their capacitance; they are not part numbers and thus ambiguous.

For correct APO (auto-power-off) control voltage, adjust trimmer VR1 until 5V is measured between left and right pin of the trimmer. See below how to disable APO.

keyboard matrix

All unknown function names and in/ out numbers in this chart were chosen by me. The input lines are active- low, i.e. react on GND, thus any functions are triggered by a switch in series to a diode from one "in" to one "out" pin.
 

legend:
"o"	= keyboard key
underlined	= function needs locking switch (i.e. stays active only so long the switch is closed)
R.	= rhythm
C.	= chord
O.	= orchestra (main voice sound)
orange background	= easteregg (unconnected feature)

• main voice sounds ("tone")

There are 10 main voice sounds selected through only 5 individual button inputs in combination with a bank switch input that needs to be pressed simultaneously (e.g. using dual contact push button switches). Although the sound select switches of the CT-410V are locking, the CPU memorizes the last pressed button and thus would also work with normal buttons. A locking "select" switch additionally doubles the available sounds to 20. (CT-410V uses all of them, but other keyboards don't.)

The main voice sounds are generated by the D931C sound IC that gets all data from the D930G CPU. Unlike various older Casio instruments, during trilled notes it does not occupy one new sound channel per repeated note, but re- uses only 1 sound channel per pressed key. This helps to save polyphony, but also makes the sound more boring because it prevents the typical phasing and volume increase effect known from older polyphonic Casios. I am still not sure why the main voice polyphony is reduced from 8 to 4 notes when chord/ bass or accompaniment are active, despite as well the accompaniment CPU as the main voice soundchip contain their own sets of totally independent tone generators. I can only imagine that by a design flaw the keyboard matrix decoder in the accompaniment CPU contains too little memory to handle more than 8 simultaneously held keys, which makes it cut down polyphony.

• envelope select

The Casio MT-65/ MT-68 has 2 locking switches to select 4 main voice envelope envelope combinations, those can be also added here. The O. envelope1 function changes the crossfading of the 2 suboscillator envelopes of a main voice sound, while the O. envelope2 mainly changes the attack rate of both suboscillators. What both switches exactly change depends much on the currently selected preset sound (e.g. with the "mandolin" they also reduce the ringing speed).
 
• rhythms

There are 6 rhythms selected by 6 OBS button inputs. Although the rhythm select switches of the CT-410V are locking, the CPU memorizes the last pressed button and thus would also work with normal buttons. A locking "select" switch selects between 2 sets of each 6 rhythms, which makes in total 12 different ones. (The CT-410V uses all of them, but smaller keyboards like the MT-45 don't).
 
• chord, bass & arpeggio variations

The 4 chord variations are selected by a locking 4 step switch which switches 2 independent chord select inputs through all 4 possible combinations. The 4 step bass variation switch works the same way with different inputs, and there are also 2 inputs for a 4 arpeggio variation switch, those are not used in the CT-410V by default (see below). Many other Casio keyboards also use only a small subset of the this way possible 64 variations.
 
• manual bass

My Casio MT-45 has a "manual bass" switch that turns the accompaniment section of the keyboard into a monophonic organ bass. Unlike the normal bass of the CT-410V (chord mode with muted "chord" volume), this bass is not limited to the scope of one octave, and it can be also played by hand together with rhythm and no accompaniment (the normal bass is chopped by accompaniment when rhythm on). Unfortunately in manual bass mode the main voice polyphony is still reduced from 8 to 4 voices despite the bass needs only 1.
 
• octave down

A switch here transposes the main voice 1 octave down (works only in chord mode).
 
• chord memory

With the Casio MT-45 setting this switch on makes organ chords (without rhythm) or accompaniment chords (with rhythm) hold when their keys are released; when off, chord or accompaniment both stop when all keys in accompaniment section are released, which permits more flexible play. With CT-410V instead organ chords simply hold never and accompaniment chords hold always. This behaviour is controlled by a fixed diode soldered into the key matrix.
 

chord memory (KI7->KC7)	chord memory mode (KI8->KC2)	rhythm on	chord holds
0	0	0	no
0	0	1	yes
0	1	0	no
0	1	1	no
1	0	0	no
1	0	1	yes
1	1	0	yes
1	1	1	yes

CT-410V and MT-65 have a fixed diode at KI7->KC7, while the MT-45 has at this location the "chord memory" switch and apparently the diode instead at KI8->KC2. Thus in CT-410V or MT-65 solder a switch in series to the diode KI7->KC7 and add a fixed diode at KI8->KC2.
 

• key hold

This odd feature simulates that all main voice keys stay held down/ get stuck. Only after the instrument runs out of polyphony, the first pressed keys get freed again with every new key press. Although this was likely intended for a pedal, also a switch here can be musically interesting to experiment with. E.g. with decaying sounds this simply makes a longer sustain, and with continuous tones for tekkno trance etc. you get your hands free to play with the VCF, stereo chorus or accompaniment etc. while the main voice stays constant.
 
• sustain + reverb.

Despite the soundchip does support sustain + reverb effect simultaneously, the slide switch in the CT-410V only offers them separately. The switch functions by a plastic handle with a leaf spring that presses down a conductive plastic foil through openings in an insulator foil onto PCB trace contacts. Thus I simply cut out the plastic bridge in the insulator foil between "sustain" and "reverb." and bent the spring a little flatter down. This way the switch will connect to both contacts simultaneously when switched into an intermediate position. Unfortunately this caused note mess during polyphonic play by shorting the key matrix lines KI5 with KI8 by the lack of diodes in between. Therefore I placed a germanium diode into the line from the switch's "reverb." output to fix this. (A silicon diode didn't work here because there are already other diodes in series those caused a too high voltage drop.)
 
• arpeggio

My Casio MT-45 has arpeggio, and an MT-65/ MT-68 includes even 4 selectable variations of them. I really love this rhythmical chinking accompaniment effect, which IMO is one of the most typical and interesting style elements of classic electronic keyboards, thus it is a pity that with later Casio and Yamaha keyboards it was omitted because manufacturers apparently considered it out of fashion.

Also the CT-410V is missing this feature, despite its D930G even would have supported 4 variations of it. Fortunately it can be upgraded with it, although this is a little tricky. The easiest part is to add the C. arpeggio on switch and optionally a 4 step slide switch at KC19 with diodes to select the 4 arpeggio variations by binary counting through all combinations of {C. arpeggio1, C. arpeggio2}, i.e. 1= open, 2= KI7->KC19, 3= KI8->KC19, 4= both. (2 separate switches would work also.) Unfortunately this yet won't produce any sound, because the electronics also needs to be upgraded with a small  analogue circuit to generate the decay envelope for the arpeggio. I copied this circuit from my Casio MT-45, without that I certainly never would have found out how this works.

The squarewave tone for it is outputted at pin 6 of the D930G, and the envelope control signals are pin 79 and pin 12. Due to this is an SMD IC, it needs a fine soldering iron, a calm hand and/ or good luck to solder wires to the pins. Theoretically pin 6 would be sufficient to get an arpeggio, but it would consist of simple continuous organ tones, those keep tooting forever when arpeggio is stopped. To get the classic Casio arpeggio sound, the following components need to be added:

Connect an 1µF electrolytic capacitor from pin 12 to GND and connect a 100 kOhm resistor from pin 12 to pin 6. Connect a diode from pin 6 through a 22 kOhm resistor to pin 79. Connect a 22 kOhm resistor to pin 6 and to its other end a 8.2nF cap against GND and a 10nF capacitor, which open end will become the arpeggio sound output. It is a good idea to mount these components on a small wire wrapped daughter board, that can be hotglued somewhere below the D930G on the main PCB. For arpeggio volume control, connect the arpeggio sound output of the daughterboard through a shielded cable with the right end of a 100 kOhm log potentiometer, its left end with GND and its wiper pin will now become the volume controlled arpeggio output. (I mounted the pot between "main volume" and "rhythm volume" slider.) The output of the arpeggio volume pot could be theoretically connected with many places to get the sound audible. I added here a 3 position slide switch to route the sound either directly through the output amp ("normal"), or through the main voice ("tone" - i.e. gets processed by "stereo chorus" + timbre filter + VCF when they modify the main voice) or through the VCF:

• The switch output in the the "normal" position connects with the white wire of the shielded chorus cable, which is the left one of 2 such cables down in the middle of the left half of the main PCB, which goes to the output amp.
• The output in "tone" (main voice) position connects with the rightmost pin of the black hybrid module S19Z2052R in the middle of the PCB, which is the DAC for the main voice soundchip D931C. This modifies the arpeggio by everything that actually modifies the main voice (including its muffling capacitors).
• The output in "filter" position connects through a 2.2 kOhm resistor with a strange looking divided solder pad below the VCF resonance preset trimmer pot at the left side of the main PCB. This solder pad is round and consists of 2 semi- circular halves with a small bridge in the middle; likely it was intended as a cuttable jumper for test purposes.

• APO disable

According to Robin Whittle, a diode here (you may install a switch) disables the auto-power-off (also seen in Casiotone 405 schematics).

VCF and chorus input jacks

To send a signal only through the VCF (in any mode), connect a cinch jack through a 100nF cap in series to a 2.2 kOhm resistor with the strange looking divided solder pad below the VCF resonance preset trimmer pot at the left side of the main PCB. This solder pad is round and consists of 2 semi- circular halves with a small bridge in the middle; likely it was intended as a cuttable jumper for test purposes. The external signal this way will be sent always through the VCF, no matter if the VCF processes internal signals at the moment. (So far nothing is currently triggering the filter envelope, the "sustain level" and "cut off" controls must be set high enough to make the external signal audible.) These jacks can be likely also used as separate outputs.

main voice volume control

Here I added 2 potentiometers for main voice ("tone") and arpeggio volume controls.

timbre change & distortion

I soldered a 4 channel alternating switch into the control lines to the 4066 to make the 4066 inputs (pin 5, 6, 12,13) switchable between their original soundchip outputs and 4 external controls. Instead of 4 simple switches I connected here the wipers of 4 10 kOhm potentiometers to permit analogue control over these inputs (use shielded cables against hum). The 4066 in this instrument switches its load fully "off" at voltages below about 2.8V and fully "on" at about 2.87V, thus the pots need about 2.8V at the left and 2.87V at the right end. To generate these voltages, I built 2 voltage dividers those each consist of [a chain made from a 330 Ohm resistor, a 100 Ohm trimmer and a 220 Ohm resistor in series] from GND to +5V. The trimmer wiper of the 1st voltage divider is connected with the "left" pins of the 4 pots, and the trimmer wiper of the 2nd with the "right" pins. The trimmers have to be adjusted until the potentiometers block all sounds when fully turned left, and let it pass undistorted when turned a bit less than fully right.
 

preset sound:	4.7nF	33nF	5.6nF	470pF	IO-1	IO-2	IO-3
organ			X	X			H
pipeorgan			X	(X)
piano	X	X	X	X	H	H	H
e.piano			X	X			H
harpsi			X
vibraphone	X	X	X	X	H	H	H
flute	X		X	X	H		H
clarinet			X	X			H
oboe	X		X	X	H		H
accordion			X	(X)
trumpet	X	X	X	(X)	H	H
horn	X	X	X	X	H	H	H
violin		X	X	(X)		H
cello	X	X	X	(X)	H	H
celesta			X
harp	X		X	X	H		H
mandolin		X	X	X		H
e-guitar			X	X			H
funny			X
cosmic tone			X	(X)

(The left values were old observations, the IO-# are taken from MT-65 service manual.)
 
I haven't documented my pot wiring order well, but filter capacitor #1 (5.6nF?) seems to be a pop noise protection or the like, because when its pot is turned fully closed (left), the main voice sound stays muted, independent from the other pot settings. Cap #2 and #3 simply modify the timbre. With these pots are in an intermediate position (between left and right) the sounds turn gradually quieter and distorts in interesting ways. Cap #4 also changes the timbre, but it behaves special, because when its pot is in an intermediate position, there is much static, hum, and depending on the other pot settings the amp even tends to toot loudly by self- oscillation. There is also a popping noise while pot 4 is turned (unless the sound is muted with pot 1).  

To the right you see the new timbre potentiometers with bypass switch for normal mode.  Left next to them are the switches for envelope control, key hold, octave and transpose. In the upper left corner you can see the added intermediate setting "sustain + reverb" for the effect switch.

note: When the main voice is routed through the VCF, these pots do nothing since the main voice apparently bypasses the normal timbre filter caps for this.

remove dullness

Unfortunately the label "C102" is ambiguous (only means "1nF") and PCB layouts may differ among versions. So for identification, temporary hold another slightly bigger capacitor (about 10nF) across its pins (wired parallel) while you play high notes. At the correct "C102" this will turn the main voice in all preset sounds (test this) even much duller and quieter, so you can safely remove it.

extended tempo range

VCF resonance improvement

Attention: Someone e-mailed me that this mod does not work with Casio MT-400V due to different hardware. I am not sure if both keyboards are indeed technically different or if there is even an error in my description. This  is what he wrote:

A couple of notes. On the 400v there's the resonance pot, right next to it is the cutoff pot. By tweaking these two it's possible to get the filter sounding pretty good. Also the wiring's different so I couldn't get the mod to work without having a drastic drop in volume on the filter output. 

On the arpegiator, the main voice in is the same, but the dry and filter ins are differerent. Particularily, I couldn't find the filter-in. The best I could do is get it routed through the bas/chord input. It does strange things to the amp and envelope of the bass, but it sounds pretty cool. The chorus distortion is a little bit different as well. If you want me to send you sound samples or pictures of the boards then let me know.

Thanks for being so intrepid with the circuit, the arpegiator adds wonders to the sound and the chorus distrotion is great fun to play with.

bass circuit

Interesting is that the Casio MT-45 produces a much softer, duller and more pressureful bass which timbre resembles a triangular wave or Roland TB303 (without resonance), while the CT-410V bass is a more sonorous droning squarewave tone. The MT-45 has instead of the 1µF envelope cap a 0.47µF one, and behind the 1µF output cap follows a 220 kOhm resistor against GND, a 120 kOhm resistor against +5V and a 47 kOhm resistor to the amplifier, but I doubt that this causes the difference. The CT-410V has after the 1µF output cap a black cable that leads to one input of the bass volume potentiometer, and its wiper (?) output is connected with a 22nF muffling cap against GND (located 3cm below the bass volume pot cable near the corner at the top of the middle of the main PCB) and resistors lead the signal to the amplifier. Bridging the muffling cap with a bigger one makes the bass indeed duller, but in spite of this it won't really sound as smooth and pressureful as the MT-45; a too big additional cap here only reduces the bass volume too much. I guess that the MT-45 circuitry has a much higher inner resistance and thus doesn't damp the bass as much as the bass volume pot stuff of the CT-410V does. (When I experimented with adding the bass potentiometer to the Testron I discovered similar behaviour.) Increasing the envelope cap value makes the bass decay slower (and thus sound less dry), but this also doesn't imitate the MT-45 bass sound. (The MT-45 has even a smaller instead of a larger cap, but this is likely rather caused by the higher inner resistance of the rest of its circuit, because it doesn't seem to fade the bass silent much faster. Certainly potentiometers can be added here to make the envelope and muffling capacitor values adjustable etc., but I didn't modify this.)

Possibly the CT-410V has even intentionally a less dull bass with more overtones to sound stronger when processed by the VCF. Which bass sounds "better" is a matter of situation or personal taste and there is no objective answer for this.

stereo chorus modification

In CZ-5000 service manual I found hints how the high-end version of a Casio stereo chorus works. In that circuit the mono audio signal first goes through a low-pass filter to remove components above 20KHz and then through a compressor, that amplifies small signals to reduce noise (similar like Dolby B). Then it is routed parallel through 3 lines (left, right, center), each made from a 256 stage bucket brigade device (MN3209) and an expander (to undo the compression). 2 three phase LFOs (0.54Hz and 6.1Hz) produce 3 out-of-phase control voltages those each wobble the clock frequency VCO of a bucket brigade device (between 33.3 and 55.6kHz) to vary its delay time, which produces the rotary speaker effect. At the end the center channel is partly mixed into the left and right channel audio by resistors. This kind of 3 line stereo chorus also exists in Casio CT-6000 (seen in service manual); the 0.66Hz triangular wave of its panning LFO additionally can be modulated with a 2nd faster triangular wave (8.33Hz) to produce the "ensemble" effect.

A bucket brigade device delays a signal by passing it through a chain of analogue RAM cells with a certain clock rate. It can be imagined as a crude sampler that records and playbacks simultaneously and so mimics the behaviour of a tape echo. So the effect is basically like modulating the tape speed of 3 tape echoes out of phase, shifting their pitch up and down.

However in CT-410V the stereo chorus circuit is much simpler, using only one MN3209 line without compressor and expander stuff. I expect that as stereo channels it simply forms the sum and difference between the MN3209 input and output signal.

• intensity control

At the "BA4558" IC (PCB label "4558DD-1") there is a 82 kOhm resistor at pin 7, which connects through a 47 kOhm resistor with pin 6. To make the effect intensity controllable, remove the 82 kOhm resistor and replace it with a 100 kOhm potentiometer (left end at pin 7, wiper at the trace to the 47 kOhm resistor).
 
• effect sounds

Connect the wiper of the intensity pot with the one of another 100 kOhm potentiometer. Connect its right end with one end of the transistor T6, and its left end through a capacitor with the right end. This way in middle position the sound will stay clear, while by turning right, the continuous panning will gradually turn into short blips and finally produce a helicopter- like throbbing. When turned left, the same effect will be milder and instead turn into a sort of squeaking, which resembles Hammond organ distortion. The timbre of the squeaking depends on the capacitor value; 47nF sounds most interesting. With the throbbing blip effects a lot of mind- machine like psychedelic sound effects for meditation music or tekkno trance can be produced, and this potentiometer also changes the VCF envelope in "wah wah" mode. The stereo chorus can be modified in a lot of other ways and e.g. turns into stuff like an American police siren or makes strange phasing effects when shorted with resistors or capacitors at certain places. But the siren effects are independent from the input signal, which is rather boring thus I didn't install them.

semi-analogue percussion

Different Casio keyboards with the same D930G can contain different percussion; my MT-45 has e.g. 2 drums (congas) less than the CT-410V.

This is what Casio considers the official decay settings (fall time down to 10% of initial level, found in MT-65 service manual).
 

percussion	trimmer	duration	circuit abbreviation
base drum	BD VR2 (MT-65: VR3)	15ms	B.D
snare	SD VR3 (MT-65: VR2)	7ms	SD.N & S.D
low conga	LC VR4 (MT-65: VR1)	27ms	L.C
high conga	HC VR5 (MT-65: VR4)	16ms	H.C
claves	CL VR6 (MT-65: VR5)	3.9ms	CL
cymbal	-		CYN
hihat	-		H.H

In that manual each "normal" drum and 'claves' employs only 1 transistor to produce a decaying sine wave (CR oscillation circuit). 'Hihat' and 'cymbal' each employ 2 transistors and the white noise output from D930G. Most complex is the 'snare', which mixes the outputs of 2 circuits (each 1 transistor) for drum and said hiss component. The trimmer names seem to differ among models; CT-410V does not match the MT-65 description. The MT-400V service manual does not even mention them.

pitchbend

shitshot

pinout D930G

The "D930G 011" can also be controlled by an external CPU (e.g. D7801G used in Casiotone 7000 or the Symphonytron 8000 accompaniment unit RC-1 - both control also D931C from there instead of connecting it with D930G). D930G xxx versions with other software numbers differ in accompaniments, preset sounds and have additional features (e.g. supporting percussion ICs) those typically reduce the count of accompaniment variations (likely due to limited internal ROM space).
 

software number	hardware class	notes & features
011	MT-65 (CT-7000, RC-1)	original version with 3x4 accompaniment variations
013	CT-620	61 keys support (changed matrix layout), arpeggio & bass volume pins swapped
017	MT-800 (old)	ROM-Pack support, different preset sounds & rhythms
018	MT-85, MT-800 (new)	bugfixed 017?
019	MT-210	uses percussion IC M6202-19, some preset sounds changed
020	MT-52	Super Drums support, different preset sounds & rhythms
022	MT-500	Super Drums support, uses percussion IC D934G, different preset sounds & rhythms

This pinout was based on the service manuals of Casio CT-405/MT-65/MT-68/MT-100, CT-430, CT-620, MT-52, MT-500 and MT-800. Apparently all "O-#" pins are outputs, "I-#" are inputs. "DB-#" are data bus. The naming convention in service manuals is ambiguous, so in many places "O-#" is named "O#", and even "KC#" pins in some places are refered as "O#" with same number. In later schematics the "-" was omitted.

Software number 019 behaves almost exactly like 011 (same keyboard matrix layout), but controls a percussion IC and has some changed preset sounds. 

The Technical Guide For Casiotone from 1986 mentions software number 018 also for MT-800 (mine has 017) thus it may be a bugfix release.
 

pin name purpose 1 O-34 chord wave out 2 O-35 chord wave out 3 O-36 chord wave out 4 O-37 chord wave out (only 7th) | MT-800: obligato wave out 5 O-38 bass wave out 6 O-39 arpeggio wave out | MT-800: obligato wave out 7 O-40 chord envelope out 8 O-41 chord envelope out 9 O-42 chord envelope out 10 O-43 chord envelope out | MT-800: obligato envelope out 11 O-44 bass envelope out 12 O-45 arpeggio envelope out | MT-500: bass trigger 13 O-46 base drum trigger out | MT-500: chord trigger | MT-210: not used 14 O-47 snare drum trigger out | MT-500: chord trigger | MT-210: not used 15 GND2 ground 0V (wired to GND) | MT-800: power down detection 16 O-48 low conga trigger out | MT-500: chord trigger | MT-210: not used 17 O-49 high conga trigger out | MT-500: chord trigger | MT-210: not used 18 O-50 claves trigger out | MT-52: rimshot trigger | MT-210: percussion ic reset 19 O-51 hihat trigger out | MT-52: cymbal & hihat trigger 20 O-52 cymbal trigger out 21 O-53 (not used) 22 O-54 white noise1 out (for cymbal, hihat, snare) 23 O-55 white noise2 out (externally mixed with O-54) 24 / (not used) 25 I-3 (wired to ground 0V) 26 O-32 MT-500: drum pad latch clock T1 out | MT-800: ROM-Pack clock out 27 O-33 MT-800: ROM-Pack clock out 28 I-1 /reset 29 O-25 /APO auto-power-off out 30 OSC (not used) 31 I-7 model select in? | CT-7000, RC-1: wired hi (through 1k to VDD) 32 I-8 clock in 1.6MHz (from D931C pin CLK3) 33 VDD supply voltage +5V 34 O-26 main voice strobe B (to D931C pin I-2) 35 O-27 main voice strobe A (to D931C pin I-1) 36 O-28 main voice data out (to D931C pin DB-1) 37 O-29 main voice data out (to D931C pin DB-2) 38 O-30 main voice data out (to D931C pin DB-3)  39 O-31 main voice data out (to D931C pin DB-4) 40 I-6 (wired to VDD)

pin name purpose 41 I-5 (wired to VDD) 42 I-4 (wired to VDD) | MT-500: drumpad interrupt /INT in | MT-800: APO release in 43 I-2 tempo clock in 44 KI8 key matrix in 45 KI7 key matrix in 46 KI6 key matrix in 47 KI5 key matrix in 48 KI4 key matrix in 49 KI3 key matrix in 50 KI2 key matrix in 51 KI1 key matrix in 52 KC1 key matrix out, address bus A0 out 53 KC2 key matrix out, address bus A1 out 54 KC3 key matrix out, address bus A2 out 55 KC4 key matrix out, address bus A3 out 56 KC5 key matrix out, address bus A4 out 57 KC6 key matrix out, address bus A5 out 58 KC7 key matrix out, address bus A6 out 59 KC8 key matrix out, address bus A7 out 60 KC9 key matrix out, address bus A8 out 61 KC10 key matrix out, address bus A9 out 62 O-11 MT-500, MT-800: rom CS out | MT-800: also ram CS out 63 O-12 MT-500, MT-800: rom CS out | MT-800: also ram CS out 64 O-24 MT-500, MT-800: ram WE | MT-500, MT-210: also percussion ic CS 65 DB-4 data bus (ram, rom, percussion ic, drumpads) 66 DB-3 data bus (ram, rom, percussion ic, drumpads) 67 DB-2 data bus (ram, rom, percussion ic, drumpads) 68 DB-1 data bus (ram, rom, percussion ic, drumpads) 69 O-23 MT-800: ROM-Pack chip select  70 O-22 KC20 key matrix out | MT-800: data/address select out 71 O-21 KC19 key matrix out | MT-800:  tempo led out 72 GND ground 0V 73 O-20 tempo led out, address bus A13 out | MT-800: ROM-Pack chip select 74 O-19 address bus A12 out | CT-620: KC21 key matrix out | MT-800: also ROM-Pack address 75 O-18 address bus A11 out | MT-800: also  ROM-Pack address 76 O-17 address bus A10 out | MT-800: also  ROM-Pack address 77 O-16 MT-500: chord filter switch out (H=dull, L=short release) | MT-800: latch 2 clock out 78 O-15 MT-500: chord release out (H=long, L=short) | MT-800: latch 1 clock out 79 O-14 arpeggio volume out | CT-620: bass volume | MT-52: /hihat cymbal select | MT-500: tempo led | MT-800: latch 4 clock out 80 O-13 bass volume out | CT-620: arpeggio volume | MT-500: auto bd led | MT-800: latch 3 clock out

Many unused pins of D930G version "011" (e.g. data bus) have a meaning in other software numbers. E.g. MT-800 ("017") supports a ROM-Pack slot and MT-500 ("022") a sample percussion IC; both also have additional RAM. In MT-800 many things get demuxed from address lines by latches clocked by pins 77..80. In MT-500 4 drumpads are buffered in a latch that is clocked by pin 26 to output its 4 signals to the data bus (pin 65..68). According to its service manual, a 4-input NOR triggers an interrupt at pin 42 when multiple (or any?) drumpads are hit. Pin 73 selects the active nibble of the 8 bit percussion ROM to be read through the 4-bit data bus. In MT-210 its early percussion IC is controlled through pins KC#, 18, 64 but no data bus. The CT-7000 has no sound IC wired to its D930G because it feeds its 2x D931C directly from an additional CPU that controls the D930G apparently by simulating key presses. 

In CT-7000 the undocumented pin 31 is wired through an 1k resistor to pin 33 (+5V), so it may inform the D930G about the absent sound IC or even change the data format for being controlled by an external CPU. This looks like a model select feature. The 1k resistor hints that pin 31 can also act as output. The same is done in RC-1 (Symphonytron accompaniment unit), which pin 33 is even wired through a 100 ohm resistor to +5V and so may get influenced if pin 31 can pull lo.

Apparently a direct predecessor of this hardware class was the accompaniment CPU D910G with main voice sound IC D990G, those were used in Casio MT-60.

pinout D931C

Additionally it can select (in CT-410V and others) 8 timbres through an external analogue filter circuit with 3 digital switch inputs. (The maximum possible number is unknown, but there seem to be up to 7 CPU controllable IO pins while the filters are actually set as a 4 bit value which suggests 4 lines for up to 16 timbres.) In MT-800 IO-4 mutes sound as click noise prevention while filters are switched. Beside filter, the sound engine seems to be multitimbrale.

As a pure sound IC the D931C contains no preset sound definitions. Only envelope durations are mostly predefined and so can be switched only through few bits. E.g. the mandolin ring is hardsynced to a fixed 12.7Hz LFO of the vibrato circuit and gets retriggered only after envelope has finished section 1 and 2. Also tone scale and octaves are predefined, while transpose is handled by the main CPU. During keyboard play the CPU only sends note-on and -off codes to the sound IC. So it is unknown what later ROM revisions were supposed to do (none are known); perhaps they should permit additional increment widths (as the partly redundant 4 bit resolution value suggests - the 2^n limitation was apparently chosen to permit fast multiplication by bit shifting - see below) for finer waveform definition. The pinout is really a mess, which suggests that it may be based on a generic microcontroller running a special software in its internal ROM, but the high mixing frequency makes this unlikely.

The IC is often labelled "NEC D931C 011", which suggests that 011 is the software number of internal ROM, but no other versions have been found, so it may have been named after the matching "NEC D930G 011" CPU it first got paired with.

This pinout and description is based on Casio DG-10/ DG-20 schematics (doesn't tell functions), service manuals of MT-65, MT-800 etc. and Robin Whittle's detailed analysis Casio-931-2006-06-17.txt in which he identified the data format to control the D931C in software. Apparently all "O#" pins are outputs, "I#" are inputs and "IO#" can be both. "DB#" are data bus. The pin naming convention is ambiguous; the service manuals in many places insert or omit a "-" before the # number.
 

pin	name	purpose
1	DB4	data bus D-3
2	DB3	data bus D-2
3	DB2	data bus D-1
4	DB1	data bus D-0
5	I3	/CS
6	I1	/strobe A
7	I2	/strobe B
8	OSC1	clock in (4.9MHz)
9	OSC2	clock out
10	CLK3	clock/3 out (1.6MHz to CPU)
11	I5 (M/S)	clock master /slave (wired to VDD)
12	IO8	not used
13	SCH1	/reset
14	IO4	filter switch out (not used) | MT-800: mute switch out
15	IO3	filter switch out
16	IO2	filter switch out
17	IO1	filter switch out
18	TEST	(wired to GND)
19	IO5	not used
20	I4	(wired to GND)
21	GND	ground 0V
22	I6	(wired to VDD)
23	IO7	sample & hold pulse | DG-20: filter switch out
24	IO6	not used
25	O17	dac bit out (LSB)
26	O16	dac bit out
27	O15	dac bit out
28	O14	dac bit out
29	O13	dac bit out
30	O12	dac bit out
31	O11	dac bit out
32	O10	dac bit out
33	O9	dac bit out
34	O8	dac bit out
35	O7	dac bit out
36	O6	dac bit out
37	O5	dac bit out
38	O4	dac bit out
39	O3	dac bit out
40	O2	dac bit out
41	O1	dac bit out (MSB)
42	VDD	supply voltage +5V

The 'clock master /slave' input is wired to VDD to generate its own clock. According to Whittle's description it should be pin 12, but in the DG-20 schematics and my own CT-410V it is pin 11. After audio bits change, pin 23 goes hi to trigger an external sample & hold circuit (according to service manuals in MT-52 and MT-500 only 25kHz) for a better DAC, which fails because bits have not stabilized yet when the pulse arrives. The 4 bit data bus is completely asynchronous (handshake controlled through 2 strobe lines) which makes it independent from the main CPU clock (which e.g. may be synced to midi or a display) and so permits analogue pitch control (tuning, pitch bend) through clock speed. The bus protocol has a simple infix structure that transmits triplets of 4 bit address, 4 bit data and 4 bit subaddress while 2 strobe lines indicate which nibble is currently sent. Interesting is that the serial data protocol defines envelopes by feeding a bit sequence (26 nibbles using only bit 1 and 4) through a long shift register chain, which to me (I am no expert) strongly resembles JTAG. The apparent reason for this is that the sound generation for the 8 polyphony channels is time multiplexed (summed at the end as a 17 bit DAC output value), so all register contents of the sound IC is anyway stored in circular multi-bit shift registers (see patent below), because for fast hardware multitasking, after processing each channel they cycle to the next entry.

It would be interesting to program a modern microcontroller with matrix LCD as a graphical sound editor with step sequencer and midi to control this sound IC.

Robin Whittle explained in his great bulletin "Modifying the Casiotone Instruments" from 1981 about the MT-65 hardware family (which he calls "Series V"):

"... these are really worth getting excited about! The sounds are brilliant by any standards - better in my opinion than the Series I [Casiotone 201 etc.] and the sound LSI puts out 16 [genuinely 17] bits, only 14 of which are normally converted. I have made a good 16 bit anti-glitch DAC and can tell you that the results are stunning. [...]

The MT-65 and CT-405 are carbon copies of each other - both use a single chip microcomputer to scan the keys and control switches, play chords, melody and bass and trigger drum sounds. This LSI [the "D930G 011"] talks to the upD931 sound LSI via 4 data lines and 2 strobe lines. Three nybbles are transferred in each burst. The turning on and off of notes and the setting of vibrato and sustain modes is done in 1, 2 or 3 bursts, but the voice specification takes about 98. The way in which waveforms and envelopes are specified is wierd and wonderful by any standards, and I have figured out 95% [...] If you are a serious computer hacker and would like to know more [...] then write to me."

Major parts of the D931C sound engine are disclosed in US patent 4590838 (with Japanese priority dates from 1978). However this does not describe a finished sound IC, but the circuitry of an obviously incomplete self-contained instrument prototype with 48 keys and layering capability ("multi-performance mode") in the manner of US patent 4387619 (see Casiotone 201). Unfortunately patent 4590838 is even more cryptic than the latter because the schematics even use proprietary miniaturized logic gate symbols to save space, and instead of using ROM, all preset sound parameters (like waveform and envelope) are vaguely referred as a bunch of "switches". (Particularly in "multi-performance mode" real switches would make no sense at all, because they permit no multiplexed selection of different sounds for each channel.) It even patents a unique synthesis mechanism that seems to be absent in the actual D931C.

In patent 4590838 the sound generation for the 8 polyphony channels is time multiplexed, thus like in most 1980th Casio keyboards, all register contents of the sound and envelope hardware is genuinely stored in 8 stage circular multi-bit shift registers. As a form of hardware multitasking, after processing each channel they cycle to the next entry every clock step.

The waveform is defined by increments. Any waveform has up to 256 clock steps (depending on pitch) divided into 16 blocks; each block is rectangular (like a stair step of a sample) and defined by a 3 bit increment ("differential coefficient" +/-{0,1,2,4}). Another bit selects which of 2 simultaneous envelopes {alpha, beta} shall affect it. Only block 0 stays always zero with fixed 10 step length (possibly due to unstable data during fetch cycle). Thus a waveform definition takes 4x 15 bit.

As expected, a volume envelope is multiplied with the waveform, but the implementation is really bizarre. The "waveform program designation section" (here obfuscated as a strange bunch of switches) outputs the differential coefficient of the actual block to an integer multiplier, that multiplies it with the output of the volume envelope counter that is stepped with the "control amount" for envelope steepness T. Like the differential coefficient, also the control amount (+/-{0, 1, 2, 4, 8, 32}) is always a power of 2. The reason for this is that this simplifies the multiplier implementation into a small gate network (only 34 gates) for hardwired bit shifting, that works within a single clock step and also forms complements for subtraction. (A general purpose multiplier would need to deal with carry and need many clock steps or a huge lookup table.) The envelope counter has 32 steps ("envelope coefficients" 0..31) and is wired through the "synchronizing set register" to step only during block 0 (to avoid glitches?). The result is summed by an adder + accumulator and (after the pitch correcting stage) finally output through a 13 bit DAC. The envelope register has 7 bit; 5 bits indicate the 32 envelope steps, 2 upper bits the envelope state {clear, attack, decay, release}. Some envelope shapes use the additional internal signals "high release" (rapid damping, truncates to zero when decay reaches 8) and "slow release".

The envelopes are defined by these parameters ("switches"): Decay clock DA (1 bit), DB (3 modes), Release clock RA (1 bit), RB (4 modes), RC (3 modes for slow release), organ envelope OA (1 bit), vibrato clock SA (1 bit), SB (1 bit). Aa, Da each select envelope addition value {1, 2, 4, 8, 32} for attack, decay, release, slow release, high release. T1 = tremolo flat, T2 = touch tremolo (only when key held and T1 is off), T3, T4, T5 control tremolo depth {1, 1/2, 1/4}, T6 enables mandolin ring. All T# parameters are implemented by gates affecting the envelope counter. The tremolo loops potions of the decay and release phase. Switch B enables vibrato.

While this ADR envelope generator itself is only linear and not that unusual, the really crazy part comes now, that is to say, there are 2 envelopes running simultaneously, and unlike normal synths, in the prototype these don't control different oscillators, but different sections in the same waveform. I.e. each of the 15 blocks (set by its beta bit) can be multiplied either by the alpha or the beta envelope to produce lots of freakish timbre changes. So the 1 bit parameters ("switches") S1, S3, S5 enable the attack, decay, release sections of the of alpha volume curve. S2, S4, S6 do the same for the beta volume curve. S10, S11, S12 (3 bits) indicate alpha and beta period modes to produce undertones (and buzzy bass range) by skipping every nth waveform cycle (set it 0) in alpha or beta {(1, 1), (1/2, 1/2), (1/4, 1/4), (1, 1/2), (1, 1/4), (1/2, 1/4), (1/2, 1/3), (1/3, 1/3 OR 1/2)}. The "1/3" mode is implemented by making the 8 step cycle counter skip steps 4 and 5. Parameter R1 inverts alpha/beta designation signal (beta bits from waveform definition). R2 is non-separation indication only with duet performance (2 layered sounds). R3 inverts all relation of alpha and beta (for each block). Depending on settings, either alpha and beta affect all lines memory (polyphony channels), or alternatingly alpha for even and beta for odd cycles.

In older Casio patents note pitches are produced by frequency dividers and binary counters. To correct the pitch of some notes, gate logics make counters stop at certain clock cycles. But here pitch correction works differently. The output is written into a kind of ring buffer ("shift memory") that holds the actual increment (10 bits) of the 8 polyphony channels, but unlike the known Casio 8 stage shift registers for polyphony multitasking, this one has additionally 3 address lines to write directly into different cells than the active polyphony channel (controlled by gate logics depending on selected note and octave) and so delay the signal differently for each of the 16 block in a waveform to correct pitch. While block 0 always has 10 steps, the period duration of others depend on note pitch and varies by 1 during the other 15 steps (e.g.  in highest octave the high C stays 8, but the low C toggles between 15 and 16). One basic clock step (here 272510Hz) is one step of the sound wave. Each of the 16 block of a waveform cycle has to be >=8 basic clock steps therefore each "clock" (block?) address has >=8 steps. The highest note pitch has 130 clock steps. (Ring buffers are commonly used for jitter compensation and delay adjustment in software based sound synthesis, so it is a bit surprising to find one here. The idea is to put values at varying positions onto a running conveyor belt (or moving a tape head during recording) and so stretching and squashing the output.)

But this reference implementation strongly differs from the actual D931C. As a self-contained keyboard instrument, an 48 stage shift register (one bit per key) memorizes which keys are already held, so only changes during key scan are sensed as note-on and note-off commands. The multi-performance mode layers polyphony channels by writing data to 2, 4 or 8 tone generators per key press (parameters W1=no, W2=duet, W3=quartet, W4=octet). A special flipflop holds notes for minimum 45ms. Interesting is that in this implementation the note pitch value of a released key is used to identify to which polyphony channel a note-off command should be sent. This may be the reason why with trilled notes older Casios occupied multiple polyphony channels (making nice volume increase and phasing) while newers don't. There are no analogue filters shown behind the DAC, but the patent claims mention that they may be added and that e.g. instead of switches, waveforms may be stored in ROM or magnetic cards, and that all those infamous 8 stage shift registers (of course) may be replaced by conventional RAM. It says that a note may be also made from layering different preset sounds or from layering the same waveform with small pitch difference (for chorus effect). However it does nowhere suggest implementation as a sound IC controlled by an external CPU.

About the data format also see below. For further technical details read Robin Whittle's analysis. He mentions various unidentified parameters in the D931C; it would be fascinating to figure out if the mysterious block envelope synthesis perhaps still slumbers unused inside as an easteregg. The patent text claimed about it:

"Thus by indicating two volume curves of alpha and beta, the system can simultaneously indicate waveforms alpha and beta. Therefore when waveforms are different, rises and falls of the volume curves may be changed so that a proper combination of them provides a great variety of a musical sound waveform synthesized. Accordingly, the time-variation of a harmonic structure of the waveform is remarkable to produce a musical sound wave with rich timbre. Accordingly, the musical instrument thus constructed according to the invention can produce a musical sound with features peculiar to the sound produced particularly by brasses and strings..."

Likely direct successor of the IC D931C was the Music LSI "NEC D932G" (64 pin zigzag DIL) that was used only in Casio CT-6000 and CPS-201. It communicates with the CPU through 8 input and 4 output pins, has 17 bit audio and additional 14 control output pins for external filters, mixing ratio, stereo chorus and the like. The velocity sensitive highend midi keyboard contains 3 of these unique sound ICs; each is wired to an own DAC with different fixed analogue filter (each 3 control lines). The outputs are mixed; in chord mode the 3rd sound IC is used for chord only, else all 3 are layered as main voice. I suspect that in classic Consonant-Vowel manner the mixing ratio between multiple differently filtered sound ICs changes with keyboard velocity. The existence of so many switchable filters seems to disprove the rumour that CT-6000 was a secret unofficial first phase distortion instrument. The high timbre quality rather results of 17 bit DAC resolution and complex analogue post-processing through filters and a costly 3 line stereo chorus. Even its unique percussion generator IC HD61701 (54 pin SMD) routes some outputs through 2 external filters to shape timbres.

---

keyboard matrix D930G-013

All unknown function names and in/ out numbers in this chart were chosen by me. The input lines are active- low, i.e. react on GND, thus any functions are triggered by a switch in series to a diode from one "in" to one "out" pin.
 

legend:
"o"	= keyboard key
underlined	= function needs locking switch (i.e. stays active only so long the switch is closed)
R.	= rhythm
C.	= chord
O.	= orchestra (main voice sound)

Re: [music-dsp] patents (the softsynth was patented 1997)

To: music-dsp from shoko calarts edu 
Subject: Re: [music-dsp] patents (the softsynth was patented 1997) 
From: Robin Whittle <rw from firstpr com au> 
Date: Sat, 14 Oct 2000 22:56:37 +1100 
Organization: First Principles 
References: <Pine.LNX.4.21.0010131840120.5467-100000 from shoko calarts edu> <00101412325300.00991 from linuxhost localdomain> <39E832AD.FA73C0CD from cableinet co uk>
Reply-To: music-dsp from shoko calarts edu 
Sender: owner-music-dsp from shoko calarts edu

Regarding digital musical instrument patents, around 1981 when I began modifying the Casio M10, I met a chap who was the agent here for Allen digital organs.  These were serious "pipe organ" like things for churches, full of Rockwell LSI chips with round metal chip covers, black plastic packages with quad inline staggered leads.

He assured me that Allen had licensed a patent, by one Ralph Deutsch (if I remember correctly, and I haven't thought of it for 18 years or so) which was for a musical instrument which worked by storing a digital representation of a waveform in its memory.  He showed me the patents - I may still have copies somewhere.

He was intrigued by the Casios - the first production digital keyboards.  He was convinced they used a curious form of synthesis, the name of Walsh Functions - some mathematical abstraction of little obvious importance - because he was sure that Casio would not want to be caught out on Allen's worldwide patents.

The Casio M10 chip and its siblings in the MT-30, MT31, CT-201 (the very first Casio - 4 octaves, full size, two chips with different waveforms) work by generating complex 16 step waves (I assume this from what I know about a later chip I describe below), eight notes at a time.  The waves are made of two component waveforms and there is crude envelope control over each.  The sum of the 8 waves appears as a 14 bit binary number at a sampling rate of about 500 kHz.  This means nice, crisp, non-aliased high tones!  The DAC was inverters driving a R-2R resistor array for 12 bits and a few resistors for the least significant 2 bits.  There was no sample and hold - just an op-amp - so timing anomalies in the bits from the chip and the inverter slew rates caused spikes when the wave went from 10000x to 01111x.

The waveforms were pretty crude and of course made of stair-steps.  The signal went through a switchable analogue LPF - but my mods bypass that.  I even made a super-low glitch tweaked dual 14 bit DAC mod board for the CT-202, using the standard Casio R-2R network and some extra resistors, trimpots and judiciously clocked HC174 latches so the rising and falling edges were symmetrical.

Later, in 1982, Casio produced a similar sounding MT-65 and its full-size equivalent.  The sound chip in this is a 42 pin custom LSI which has its note playing and waveforms loaded into it by an external CPU and software.  Around then, I figured out the protocol for talking to the chip and wrote a C program (BDS-C for Z80 CPM) to write waveform data to the chip via the parallel port of a Big Board I.

The reason I mention the Casio is that the chip (the NEC D931) did not actually receive, or apparently store, waveforms.  It received and stored *increments* - and used these steps to generate the waveform.  If your increments did not add up to 0 then all sorts of trouble occurred!

Let me look into my archives . .  In less than a minute I found the patent!

US Patent Office 2 June 1970  Patent 3,515,792

Ralph Deutsch, Sherman Oaks Calif. assignor to North American  Rockwell Corporation.

A digital electronic organ wherein a digital representation of an organ pipe wave shape is stored in a memory. A frequency synthesiser activated by a manual or pedal key produces a clock frequency at Nf, where f is the frequency of the note selected, and N is the number of sample points of the stored wave shape. The digitized wave shape is read out repetitiously at the generated clock frequency and converted to analog form to produce a musical note having a waveshape corresponding to that stored in memory. Circuitry is provided to sum digitally notes which are played simultaneously; to shape each note in attack and decay using digital operations; and to read out stored multiple wave shapes to implement harmonic and mutation stops. 

Using increments (albeit simple +/- 1, +/- 2 +/- 4 +/- 8) was probably harder than simply storing the waveform, and led to less flexibility than a stored waveform - but I figure that Casio did it so they didn't have to worry about the Allen patent.  So the dull force of patent law made a popular instrument more awkward, or at least more idiosyncratic.

I just found my doco file for the chip in the MT-65 - the D931.  All the guff is there on how to talk to it.  I have C-code as well - and 8 D931 chips, seven unused.  I was able to load novel waveforms into the D931 in my MT-65 it and then play it from the MT-65's CPU via its keyboard, or the MIDI interface I added.  I also made up a second D931 on an external board with an independent clock source so I could have two waveforms and detuning.

A web search for:

"Ralph Deutsch" and patent

http://www.allenorgan.com/book/jbook.htm

Honoring the Intellectual Property of Others
Ralph Deutsch and the Dark Side

- Robin

dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links
http://shoko.calarts.edu/musicdsp/