circuit bending details

Nick Santos sent me an e-mail with the following hardware tips about the Playskool PS-635, which apparently refers to the 40 pin version:

"Although I am still in the experimentation and documentation stage of this bend I have found some promising bends.  As you may have noticed, pins 10-26 seem to deal with they keys and buttons of this device.  I saw that pins 10 and 12 were being used but pin 11 was conspicuously unused, I used some wire and probes and proceeded to connect pin 11 to the other key control pins with pleasing results pin 11 connected to pin 15, 16 17 or 18 will produce the kick, cowbell, loud snare and soft snare respectively.  When pin 11 is connected to pins 21-26 it will produce lower pitch notes that were not included in the design.  You mentioned that you would like to use the pre programmed rhythms and cancel out the main instrument voice, I have discovered how to do this as well. Connect the pin nearest the headphone jack of the capacitor nearest the headphone jack, to pin 1 of the instruments main chip, this will completely cut out the main voice and leave the rhythm (if you attach to the wrong pole of the capacitor a very faint main voice will remain) to do the opposite and strip away the rhythm from the songs connect the same pin of the capacitor to pins 5 or 6 of the main chip (it will also boost the volume).
...
I hope this helps you and helps other people who wish to bend this instrument as well."

multipulse squarewave tone generator

What I identified as 'guitar' corresponds to 'piano' in MS-210B (see there) so it may be considered piano by the manufacturer. But IMO it sounds much more like an acoustic guitar than like the piano of EK-001 or an actual piano.

The multipulse bit loops look like this:
 

preset sound	multipulse pattern	comment
music box	1100	plain squarewave (1 octave higher)
violin	1010100000010001.
flute, banjo	11110000	plain squarewave
organ	1110100010101000	2 similar halves make this roar so bassy
guitar	1111111100001011.	(same like 'piano' in MS-210B)
horn	1111111100000011.	squarewave 3:5 (1 octave lower)
piano	1010001000000111.

On the PCB is a tuning trimmer that adjusts the clock rate.

By a strange glitch the accents of individual beats in rhythm patterns seem to change with tempo setting, i.e. individual percussions sound louder or quieter, which is not proportional with the selected tempo speed. Because (unlike EK-001) each percussion generator channel here is shared by 2 sounds, it may be that at certain speeds the accent bit is read or cleared in wrong order and so affects wrong percussion sounds. It also may be related with binary interactions of internal counters, in a way that e.g. the percussion envelope decay counter at certain tempo settings gets decremented already before the first waveform bit is output, making that step in the rhythm pattern sound quieter.

keyboard matrix

Because I do not own the 40 pin CPU version, row C4 is a guess based on that schematic and what Traktor told by e-mail. Theoretically also in the 32 pin version it may be possible to simulate the missing C4 signal by using the rising edge of the previous matrix out lines (likely C3) to turn a flipflop on and the falling edge of the next line (likely C5) to turn it off. But I haven't tested this by myself, so there may be timing issues or other problems to solve to make it work.

For adding preset sound switches, it is important to disconnect the line (cut or desolder the bridge) of the switch to CPU pin C6, because the slide switch has no intermediate open position and thus always shorts nearby pins.
 

14	13	12	11	10	9	8				PWH635 pin
26	25	24	-	23	22	21			635-21 pin
C1	C2	C3	C4	C5	C6	C7		pin name
out 1	out 2	out 3	out 4	out 5	out 6	out 7	in / out
o C#5	o C#4	o C#3	P. base	tempo +	O. guitar	R. pop	in 1	R1	27	15
o D5	o D4	o D3	P. cowbell	tempo -	O. flute	R. disco	in 2	R2	28	16
o D#5	o D#4	o D#3	P. snare	R. start	O. violin	R. march	in 3	R3	29	17
o E5	o E4	o E3	P. hihat	R. start/stop	O. organ	R. rhumba	in 4	R4	30	18
o F5	o F4	o F3	-	R. stop	O. horn	R. swing	in 5	R5	31	19
o F#5	o F#4	o F#3	-	demo	O. banjo	R. tango	in 6	R6	32	20
o G5	o G4	o G3 (1)	o G2	(vibrato)	O. musicbox	R. ballad	in 7	R7	1	21
o G#5	o G#4	o G#3	o G#2	sustain	O. piano	R. waltz	in 8	R8	2	22
o A5	o A4	o A3	o A2	S. clear	O. guitar	R. pop	in 9	R9	3	23
o A#5	o A#4	o A#3	o A#2	S. play	O. flute	R. disco	in 10	R10	4	24
o B5	o B4	o B3	o B2	S. one key 1	O. violin	R. march	in 11	R11	5	25
o C6	o C5	o C4	o C3	S. one key 2	O. organ	R. rhumba	in 12	R12	6	26

The input lines are active-low, i.e. react on GND. Any functions can be triggered by a non- locking switch in series to a diode from one "in" to one "out" pin.
 

legend:
"o"	= keyboard key
R.	= preset rhythm
O.	= preset sound ('orchestra')
P.	= drumpad
S.	= sequencer
orange background	= easteregg (unconnected feature)
grey background	= unconnected doublet

important: Because pinouts differ among CPU variants, in the following I refer the keyboard matrix pins by their "C" ("columns" = out pins) and "R" ("rows" = in pins) numbers found in MS-210B schematics. These are CPU pin names, not capacitors and resistors!

• additional keys

The 32 pin version has 7 lower & 1 higher note keys addable. The 40 pin version even supports 13 lower note keys.

• drumpads

In the 40 pin version you can wire 4 drumpad buttons through each a dioded from CPU pins R1..R4 to C4.

• 8 preset sounds

For enabling all 8 preset sounds, disconnect the preset sound slide switch from CPU pin C6 (this switch has no intermediate open positions and thus makes trouble) and instead wire individual buttons through each a diode from pins R1..R8 to C6. (The original slide switch may be rewired for something else, e.g. 'sequencer record off'.)

waveform arpeggiator synth

If you want to enable the waveform arpeggiator synth (recommended, see EK-001) install here 8 locking switches instead of push buttons or an additional DIP switch block. Due to different matrix timing, it may even be that the doublets at pins R9..R12 produce additional sound variations, so 12 switches may be better.

• 8 preset rhythms

For enabling all 8 preset rhythms wire individual buttons through each a diode from CPU pins R1..R8 to C7. To use them, you have to move the rhythm slide switch into an intermediate open position or disconnect it.

rhythm combinations

Unlike EK-001, this hardware supports rhythm combinations. To enable lots of additional rhythms (recommended), install here locking switches or an additional DIP switch block instead of buttons.

(Traktor told me that their behaviour is not like the linear combinations of classic analogue drum machines, but apparently uses some binary logics to select or mix many different rhythms, and it first may need to switch the CPU into a different mode. I haven't examined this further. In my MS-210B only combinations of 'waltz' + 1 other rhythm button change patterns.)

• sustain

A sustain button can be wired through a diode from CPU pin R8 to C5. 

• vibrato

In the 40 pin version a vibrato button can be wired  through a diode from CPU pin R7 to C5. To make it work, you also need to install from the VIB pin to GND a 330K resistor in series with a 100nF capacitor; from the connection between these wire a 810K resistor to the clock pin OSC2 (seen in MS-210B schematics). If you only have the 32 pin version and want to add vibrato, you of course may install an external LFO oscillator and connect it the same way like the missing VIB pin. But Traktor told me that his vibrato is too slow (4Hz), so installing an external LFO may be a good idea anyway.

Like I did in my EK-001 you may install a potentiomater (e.g. 4.7 M) into the VIB line to control vibrato depth.

• one key play

To step through the sequencer, 2 'one key play' buttons can be wired through each a diode from CPU pins R11, R12 to C5.

• sequencer record off 

The sequencer is of little use, because it is always recording and so appends all played notes until memory is full, which is particularly annoying since it even keeps memory contents when powered off. Fortunately this behaviour can be changed.

The sequencer memory write enable pin MEMO is connected to GND by a transistor (base pin through 10K resistor on BUSY pin) when the instrument is on. The transistor is there because MEMO apparently has to be set hi (or open) before power off, else memory may be overwritten with data mess. When open, played note will not be automatically recorded and also the clear button does nothing. Install a switch into the MEMO line (pin 7 of 32 pin CPU). So (with batteries inserted) you can keep a sequence in memory as long as you want.

Traktor told me, that the sequencer even can be edited in a limited way. When the added 'record' switch is on (i.e. MEMO pulled lo), the sequencer normally appends new notes after the last manually played note. But when you step with 'one key play' through the recorded pattern, with 'record' on it will start overwriting the stored sequence beginning from that position. So you can overwrite individual notes and keep recording from there. But apparently you can neither insert nor delete individual notes, nor overwrite the 1st note of the recorded pattern. When playing the sequence with one key play, you can play the last note manually to skip the empty gap at its end.
 

pinout PWH635, 635-21, 210, HT3421A

The LSI "210" (32 pin COB) is the CPU of Yongmei MS-210B, which is a rom variant of it. Only preset sounds, preset rhythms and demo songs were changed, but behaviour stays the same. I later found the datasheet of Holtek HT3421A ("41 Matrix Key Multi-Function Piano with a Memory") of 1996-06-28, which is obviously a 40 pin DIL version of the "210" and makes me conclude that this CPU was originally created by Holtek. Their naming convention suggests that an alternative name of the "PHW635" (different rom) may have been "HT3421" or had another letter suffix. The Kiddesign - Barbie Music Center BE-630 (seen on website, I don't own this) has drumpads and contains a different 34 pin(?) COB module (shape resembling 32 pin DIL with 2(?) additional pins at its short ends).
 

These ICs are a modernized variant of the archaic Hing Hon EK-001 CPU HT3371 and thus has a very similar sound engine. The main differences are that the main voice is only monophonic, and also the percussion generator was halved, so it has only 2 instead of 4 outputs and its envelope resolution was reduced to 3 bit (8 steps), resulting in shorter envelopes. Apparently percussions here share 2 generator channels, because unlike EK-001 they truncate each other, so 'base' + 'cowbell' or 'snare' + 'hihat' can not sound the same time (tested with MS-210B drumpads). The keys have become part of the 7 x 12 keyboard matrix, there is a simple sequencer with battery backed-up SRAM and auto-power-off with power on/off button. Also the main voice multipulses differ. This pinout was originally based on the schematics of Yongmei MS-210B, which CPU "210" is very similar. I later added missing pin names from the Holtek HT3421A datasheet, which describes a 40 pin DIL version which pinout blatantly corresponds to PWH635. Comparing these pinouts, you can nicely see how in COB ICs the position of pin 1 is not fix at all, but the order of pins (unused ones may be omitted) around the silicon chip (see die pad order) will stay the same.

PWH635, HT3421A DIL pin	HT3421A die pad	635-21 pin	200 pin	name	purpose
1	36	17	6	OUT	main voice audio out
2	37	18	7	ENV	envelope analogue mux (capacitor 4.7uF to VDD)
3	38			TEST4	tempo clock in/out (not used)
4	39			TEST3	sequencer disable in (not used)
5	40	19	8	S/H	percussion snare+hihat out
6	1	20	9	B/C	percussion base+cowbell out
7	2		10	VIB	vibrato LFO out
8	3	21	11	/C7	key matrix out
9	4	22	12	/C6	key matrix out
10	5	23	13	/C5	key matrix out
11	6		14	/C4	key matrix out
12	7	24	15	/C3	key matrix out
13	8	25	16	/C2	key matrix out
14	9	26	17	/C1	key matrix out
15	10	27	18	/R1	key matrix in
16	11	28	19	/R2	key matrix in
17	12	29	20	/R3	key matrix in
18	13	30	21	/R4	key matrix in
19	14	31	22	/R5	key matrix in
20	15	32	23	/R6	key matrix in
21	16	1	24	/R7	key matrix in
22	17	2	25	/R8	key matrix in
23	18	3	26	/R9	key matrix in
24	19	4	27	/R10	key matrix in
25	20	5	28	/R11	key matrix in
26	21	6	29	/R12	key matrix in
27	22			TEST5	test pin (not used)
28	23			TEST2	test mode pin (not used)
29	24	7		/MEMO	sequencer write enable in
30	25			TEST1	analogue reference voltage? (not used)
31	26	8	31	/RESET	reset
32	27	9	32	/OFF/ON	power on/off button
33	28	10		/BUSY	auto-power-off out (hi=on)
34	29	11	30	VSS	ground 0V
	?	12		TEST6	APO timer test pin (not used)
35	30			/MODE	sequencer record on/off button in (not used)
36	31	13	1	OSC1	oscillator in 512 kHz (48k to OSC2)
37	32	14	2	OSC2	oscillator out
38	33		3	MLED	memory led out (not used)
39	34		4	TLED	tempo led out
40	35	16	5	VDD	supply voltage +4.5V

Of the Playskool PS-635 I only own the 32 pin version "635-21", which has many pins omitted. So it lacks keyboard matrix out pin C4, VIB and both LED outs. My pin 15 has not even a trace to the chip and so is completely NC.

My pin 7 (MEMO) seems to be sequencer memory write enable; it is connected to GND by a transistor (base pin through 10K resistor on the BUSY pin 10) when the instrument is on. The transistor is there because pin 7 apparently has to be set hi (or open) before power off, else memory may be overwritten with data mess. When open, played note will not be automatically recorded and also the clear button does nothing. Possibly a keyboard with 3 step slide switch was planned, that on positions 3 ('off') disconnects power and write enable. On position 2 ('play') it connects power to the amplifier, and in position 1 ('record') it additionally grounds the write enable pin. The use of that mode switch would work like in Casio VL-1. In MS-210B schematics I don't see that pin, but pin order corresponds to my PS-635. I later found in HT3421A datasheet that lo='memory' and hi='play' mode (needs locking switch).

Pulling my unused pin 12 lo (tested through 1k resistor) switches power off after about 2..3 seconds. This is not by overload (I measured current), but likely an APO test pin that speeds up the internal timer counter for testing. Strange is that this pin (I name it TEST6) does not exist in any other version and is not even shown in the "HT3421A" die pinout (but there is enough space between pad 29 and 30 where I expect its position).

The Playskool PS-635 of Traktor has the CPU "PWH635, 9136M" (40 pin DIL, seen on photo). "9136M" here may be a serial number, but because this is in the 9000th range, it also may be a chip type (e.g. an EK-001 CPU version is "9037" and the similar Superb Sound EK-210 CPU is "CIL-9038"). Traktor told me what his version's additional pins do. Regard that he is no electronics expert and has no oscilloscope, so his descriptions may be inaccurate.

His pin 3 (TEST4) outputs pulses synched with the selected tempo. The audio he recorded from it (mixed with rhythms through a 100k resistor to the amp section) seems to be the internal step resolution (1/16?) of the rhythm patterns. Rhythms stops while pulled hi. Also pulling it lo stops rhythm, but first outputs some drum noises and makes all sounds change. Also touching with test cable triggers the same strange percussion timbre changes like the tempo control accent glitch. Shorting it with a clock pin through a resistor (which lowered pitch by a semitone) turns rhythm into strange jungle grooves. This makes me conclude that it is a tempo clock (test?) input that is internally connected to the tempo clock oscillator output (which speed is set by tempo +/- buttons). Perhaps this CPU was even originally designed to support here an external tempo oscillator (for tempo knob) but has that line bonded to an internal tempo oscillator (or enables that by software) to cut cost. So it might be possible to feed pin 3 with an external oscillator for freely adjustable tempo. But it may even be time multiplexed as a serial input to trigger internal functions in a certain order (like a keyboard matrix input), and so produces sound garbage when pulled lo.

His pin 4 (TEST3) completely disables the sequencer (record and playback) while pulled lo. After release (pin open) playback works, but you have to press 'one key play' to make the record mode work again. The MLED led ignores pin 4, so it stays on also with sequencer disabled. (I conclude that this function was perhaps designed to disable the sequencer in CPU specimen with faulty memory to install them in keyboards without - much like Intel sold broken 4 core Pentiums as Dual Core etc.)

His pin 7 (VIB) can be wired like in MS-210B, but the resulting vibrato is quite slow. So it may be that the CPU was originally planned for a higher clock rate. The HT3421A datasheet shows that at 512kHz clock rate the LFO is 4Hz.

His pin 27 (TEST5) does nothing, but reverse diode test with DMM shows that it is internally connected.

His pin 28 (TEST2) seems to activate a test mode for all kinds of functions. It apparently makes the 'one key play'(?) button cycle through different demos, notes, sounds, rhythms and tempos and even played sequencer memory with different preset sounds and together with percussion.

When his pin 29 (MEMO) is pulled low, it starts recording into the sequencer. The last note played on the keys before enabling MEMO will become the 1st note of the recording. This can be prevented by pressing other sequencer buttons (e.g. 'one key play') before enabling it. HT3421A datasheet: "Connect this pin to VSS/VDD to work in either the memory or play mode." If hi or open, MODE switch toggles between play/memory.

Touching his pin 30 with a test cable causes a very high pitched thin tone (self-oscillation??), that also interacts with the ENV pin (pulls it lo?). (I think this may be an internal analogue reference voltage for DAC or VCA that picks up keyboard matrix out frequency when touched.) The pin normally seems to be at about half supply voltage and freezes the sound when connected to ground 0V or +Vs.

His pin 35 (MODE) is a memory write ('record' on/off) button input. It only works when pin 29 (MEMO) is not pulled low (e.g. open). So in the 40 pin version the locking MEMO switch is not needed to switch the sequencer on and off. After APO, this way recording is automatically turned off (with memory contents preserved). HT3421A datasheet: "Toggle select of play mode or memory mode. The system is in the play mode at power-on."

In Yongmei MS-210B schematics the depicted circle (between B/C and VIB pin) stands for the empty hole in the COB PCB that is closest to the blob. It may be that this was originally meant to mark pin 1, but I consider counting anticlockwise from the corner after the biggest gap (with the screw between trace bundles) more logical as pin 1, thus the pin labelled "210" is pin 32. The VSS pin 30 (ground 0V) is placed here differently (skipping 2 pins), but the schematics shows a 2nd pin in its real place (where in PS-635 MEMO would be), i.e. likely the silicon die has multiple ground 0V connections or it was simply shorted with MEMO.

I.e. in the "210" CPU unfortunately the MEMO pin was omitted (likely internally hardwired to ground), so its sequencer is always recording. Also the APO timer is still doing its job, despite by the missing /BUSY pin (how stupid is that?!) it can not turn off the amplifier. So there still is a slide switch and you additionally always have to press 'computer start' (aka 'on/off' button) to make the thing work. Possibly this chip was a pirated version of "PHW635" (with rom changed to enshroud this?), but the manufacturer did not know how to handle APO properly and so removed all pins he did not understand.