Saturday, December 2, 2023

SUZUKI PK-37


Shortly after finishing the NTSH and NT05, I decided to start this project. I was looking for something to do that I could just slowly chip away at over the summer. I had been doing some reverse-engineering of some of my toy keyboards for fun. I started with a Casio MT-40, then moved on to the Casio SK-1 using Kicad to draw the schematics cleanly. After those, I needed a bigger challenge. I had been hanging on to this pair of Suzuki PK-37's for several years. The first one I found at the Goodwill bins probably ten or fifteen years ago. It was dirt cheap and was pretty busted up. The main-board inside was badly corroded, but I was able to get it to turn on and make some sounds. Aside from all of the hiss and noise that keyboards of this era tend to have, I really like the warm lo-fi quality of the sound and the classic home-organ features. It has some nice punchy analogue drums and a cheesy arpeggio in the accompaniment section. It also has a lead-chord mode that turns the lead keys into chords relative to the accompaniment key when it is on. I remember that function on another old home organ I had years ago that I accidentally fried and was sad about. The keyboard was in terrible condition, and some of the keys were broken, so I just added it to the pile of future projects. Years later, I found another one that was in perfect condition. Well almost.. I can't remember if I bought it on Ebay, or I found it at goodwill again. I am pretty sure I paid more for it though. Despite being cosmetically perfect, I was never able to get anything but the drums to play. I figured it could be repaired someday, and threw it in with the others. When I finally got around to opening them up and checking them out, the first one I found was a rotten mess, and barely turned on. The other seemed to be fried... 


The PK-37 is built around the M112 polyphonic sound generator chip, and is programmed by a M3872 microcontroller. Upon closer inspection of the non-working unit, I noticed that there was no polarity protection diode from the power supply. However, there was one in the rotten unit that still made sound. I imagine the keyboard was fried early on in its life, and that is why it stayed in such good condition. All of the 5 volt IC's were protected by a linear regulator, but sadly the M112 was toast. I later noticed that the keyboards had slightly different logos printed on them. The one that was fried just had SUZUKI on the front, whereas the rotten working one had SUZUKI with UNITED STATES printed underneath. Both were made in Italy by Siel. In fact, there were several variations of this keyboard by different brands. I think they were all made in Italy though. 

Original PK-37 main board

My schematic

I spent about three days reverse-engineering the PK-37 probing traces and components and drawing up schematics in Kicad. the circuit board was double sided which would have been expensive for when it was made. The through-hole vias were actual springs that were soldered in by hand. It was pretty interesting to imagine what the designers were thinking when they made this. It's not hard to see why there is so much hiss and hum in the circuit. I am no expert in circuit design, but it was pretty easy to spot some modern day faux pas. After confidently reverse-engineering the keyboard and drawing up a pretty nice schematic, I did a search on the M112 datasheet. I found the datasheet as well as the schematic for the Siel MK370. Aside from a couple different component values here and there, it was exactly the same as the Suzuki PK-37, and was actually cleaner than my schematic.. I actually ended up using it as a reference throughout the project instead of the schematic I drew.
Over the course of this project, there were a lot of ideas that I thought were good, but had to abandon. One of the first ideas I had was to incorporate a Hagiwo style drum sequencer to trigger the analogue drum sounds in sync with the accompaniment section. I was able to get it working, but it ended up being too slow and glitchy. Instead, I created a switch matrix so that the drum sounds could be triggered by any of the five drum pattern channels. I will have to save that Hagiwo sequencer for something else. The first circuit board I had made was basically just a fundamental-ish board for the M112 and microcontroller, just to get the keys connected and some voices coming out. Basically just a stripped down version of the PK-37 without any voice filtering or audio signal mixing. I replaced the 4Mhz crystal with a HC4046 VCO that can bend one octave. I also added a better vibrato circuit with speed and depth controls. 

M112 control main board

Top side assembled

Bottom side assembled

I designed the circuit board using Kicad and had it manufactured by JLCPCB. I designed the first faceplate for the left-most section of the keyboard to have all the rhythm and accompaniment controls, as well as the pitch and vibrato. The switches, pots and other hardware are mounted to a separate circuit board that I engraved myself with my CNC router. The hardware board is connected to the faceplate. In order to mount the hardware boards to the faceplate, I laser-cut cardboard templates for where the holes and other cavities should be cut so that the parts could pass through the body of the keyboard. I really wanted to keep the keyboard enclosure as whole as possible to ensure structural integrity. Rather than just cutting out huge cavities, I drilled and cut individual holes where I could with a dremel and a carbide endmill. I cut temporary faceplates out of cardboard to use while prototyping the circuits. This way I don't have to worry about damaging them, and I can easily remove or modify them while I'm working. 

Hole cutting template

Temporary cardboard faceplate

The second circuit board I had made was the rhythm voice board. Like the M112 board, I redesigned all of the circuitry using surface mount components, with the exception of the capacitors for the drum voice oscillators. I used red poly caps, but I think I could have just used SMD MLCC with no problems. The drum sounds are mostly MFB/Twin-T style oscillators with some white noise mixed in the snare and the cymbal voices. Some time in the future, I plan to add pitch knobs for some of the voices. I kind of regret not including them in the design, but it is a super easy mod, and I will probably end up modding this thing a lot as time goes on. I also added a basic filter to the mixed output of the rhythm mix. Originally, I had just a cutoff knob and a HP/LP select switch, but the high pass and lowpass filters were only 6db, and they left a lot to be desired with the minimal drum sounds, so I went in and changed it to a 12db lowpass filter and used the switch to turn the resonance up. I called it "GAIN" on the faceplate graphics. Now the filter gives the rhythm mix a little more punchy compression sound. 

Rhythm section PCB

Rhythm PCB mounted to hardware board

Rhythm and accompaniment control section installed

The rhythm clock oscillator is unique on the PK-37. Rather than just having a tempo control pot to the microcontroller, there is an external R/C oscillator clocking the drum patterns and accompaniment. This would make it possible to not only sync to the PK-37, but also sync the PK-37 to other clock sources. The clock speed is pretty high.This makes it possible for the PK-37 to have more dynamic rhythm patterns, but in order to use more conventional external clock sources, I had to add a clock multiplier to the sync input. The clock multiplier is applied to the internal clock and the clock sync output as well. It can be set to x2, x4, x8 or x16.
 
Chord Bass Rhythm temporary faceplate installed

Once the left-most section was working, I moved on to the accompaniment section. The original accompaniment section consisted of a mono bass and a chord section with three separate footages. Both voices had a VCA and envelope generator that were gated by the microcontroller when the accompaniment was playing. The accompaniment patterns are super cheesy and not nearly as cool as some older Italian accompaniment machines. Like the drum voices, I added select switches to the bass and chord voices. They can be triggered by one of any rhythm, bass or chord pattern channel. The mono bass voice was pretty basic. Just a square-wave with a weak enveloped VCA. I got rid of it and added a voltage controlled PWM filter, followed by a diode VCA, followed by a 12db lowpass filter with resonance. I added a proper attack/release generator to modulate either the VCA or VCF, depending on the mode switch. I added the same attack/release generator to the chord voice, but without a depth knob. The chord voice had three separate octaves coming from the M112 chip that were mixed together. I took them apart and added individual VCA's to each one. I wanted to be able to modulate the levels separately with LFO's. The chord voice seemed kind of basic. I wanted to give it some life, so I added a PT2399 based chorus circuit to it. I had originally planned to use the "Little Angel" circuit, but it was too subtle for the mix. In the end I had to modify the circuit to modulate the delay-time pin, and get rid of all of the other Little Angel trickery, as it was not necessary. The chorus circuit has a rate and depth knob. The effect sounds really good, but only really in contrast to it not being there at all. It really sounds its best when the rate is being modulated slowly. 

Hole cutting template for Chord, Bass, and Rhythm section faceplate

Holes cut for hardware to pass through keyboard enclosure

Once the accompaniment section was prototyped and working properly, I had the third circuit board made. Along with the chord and bass voices, I included the lead voice to the circuit board. I was already anticipating trouble with the next section to come, so I wanted to have all the easy stuff on one board that I knew would work, or at least be easy to fix, if need be. The lead voice is an 8-note polyphonic oscillator. The eight separate voices are mixed together externally, but they cannot really be separated in any useful kind of way. The voices are not separated note-by-note, but by footage and key groups. It was kind of fun to destroy the waveforms by mixing the eight voices in different ways, but it was hardly the sound I was going for, so I left them mixed as intended. The PK-37 had eight different standard home-organ timbres. There was a kind of interesting filter matrix using a CD4051 multiplexer that I had never seen done before. Behind all of the filtering and modulation, there were really only six different waveforms. I got rid of the filter matrix and added a 12db BP/LP filter with resonance and a nice attack/release envelope to modulate it. The M112 chip has some volume envelope controls, but they were not very consistent. The "attack" control pin also seemed to effect the envelope depth, which caused my filter envelope to trigger at different times, so I only added a knob for the decay/release. Each of the six waveforms have slightly different envelope characteristics. some have really short release, and some are long. The "Envelope" control knob acts more like a piano sustain, and can be set to almost continuous. It is a nice feature, but it doesn't play too well with the filter's envelope generator either. In order to trigger the VCF envelope generator, I built eight separate gate-to-trigger circuits to trigger the VCF envelope every time a key is pressed. The PK-37 has eight capacitors connected to the M112 for envelope timing. The gate-to trigger circuits are connected to them. If the "Envelope" control knob is set too long, the volume envelope can have trouble re-triggering the VCF envelope generator. It is not a perfect solution, but it is pretty great to be able to re-trigger the VCF envelope and get a nice paraphonic filter effect.

Protyping Chord, Bass, and Lead sections
  
Lead envelope VCF re-trigger circuit

Chord/Bass/Lead PCB

I engraved more hardware boards with the CNC engraver, and laser-cut out new cardboard faceplates and hole-cutting templates for the enclosure. The lead section would share a faceplate with the next section, so I cut a smaller cardboard prototype for the temporary lead section.

Temporary faceplates and mini faceplate for Lead section to the right

Underside view of hardware pass-through holes

Everything so far wired up to hardware boards

Aligning the panels

Small circuit board add-on's here and there to make things work better



Testing the four voice sections

Preparing for the effects section

Once the accompaniment and lead sections were all wired up and working, I began working on the next section that I knew would be the most difficult and frustrating. The "Effects" section was supposed to be pretty straight forward and as uncomplicated as possible since I have had such bad luck with all things digital... I have had a lot of fun over the years learning and experimenting with microcontrollers, but when it comes down to it, I can never really get them to do anything useful. And when it comes to DAC's, Codecs, DSP, sometimes I feel like I'm cursed... But actually I think my problem comes down to buying bad or fake parts from sketchy people online. After a couple of weeks trying to figure out why my effects section wouldn't work, that is the conclusion that I came to. 

Effects section PCB

The original plan was to have two effects channels that any of the four voice sections(chord, bass, rhythm, lead) could be routed to and mixed back with the dry signal. I chose the coolaudio(behringer) V1000 DSP chip because I have used it many times before in other projects, and I love the way it sounds when you modulate the sample rate with an external VCO instead of a typical crystal oscillator. In the past, however, I have always bought old cheap behringer delay pedals and modified them. This time I thought I would save some money and some extra work, and just order the chip along with the V4220 codec from Ebay. Of the five V1000's and five V4220's I bought for $5 each, one V1000 worked. The rest either made no sound or barely made any sound at all. The most frustrating thing was that I couldn't be sure if the chips that were working badly were just picking up interference from being on a breadboard. I decided to design a circuit board around the chips anyway, and ordered them hoping that the well designed PCB would solve the problems. Unfortunately it didn't, and I ended up scrapping all of the V4220 chips.

Breadboarding

Prototyping the effects section

I had a similar issue with the second of the two effects sections. I wanted to have an echo/delay section, but I wanted something with a little more quality than the PT2399. I opted for the HT8955A delay chip. It had a lot of advantages over the PT2399. The datasheet sounded great. I found a bunch for sale on Aliexpress along with the D-RAM chips needed for about $20. Unfortunately, only about three of the ten HT8955A's worked, and I fried one of them in a very important lesson about AMS1117 linear regulators from no-name companies. If you are reading this and didn't already know, many surface mount linear voltage regulators, namely the 1117 variants, are not built withstand high temps, transients, and even too much voltage. Read this. I tried to power all of my 5 volt effects chips with one AMS1117.5, and ended up cooking them all... Once they were all replaced, and a proper regulator was installed, the HT8955A was just kind of OK sounding. It was pretty noisy and needed a lot of fine tuning, but it is kind of nice to have separate controls for delay time and sample rate. The delay times are all stepped though, so the effect is very digital sounding, but also very dark and dirty. It's not really what I was after, but I like it anyway. I will not be using these chips in future designs though. 
 
Effects hardware board

Cut template

Holes cut

Cleaned up

Dry-fit

Aligning temporary faceplate

Everything on the effects circuit board worked except the V1000 circuit. I really didn't want to redesign the circuit board and have to move parts over to a new version. Especially if it didn't work again. Instead, I dug out these old effects modules I bought on Ebay a couple of years ago. They are very much like the V1000 circuit. In fact the V1000 is a knockoff of the alesis AL3201 chip. The pinout is almost exactly the same but the effects are slightly different. The AL3201 effects modules had a codec chip, regulator and all of the other essential circuitry to add DSP effects to something. I think they were originally intended for Mackie mixing consoles. The PCB was kind of bulky, but it was actually pretty easy to incorporate into the rest of the keyboard and effects section. I took out the AL3201 and replaced it with my working V1000 chip.

Effects PCB mounted to effects hardware board

The V1000 has 16 different internal DSP effects. Most of them are reverb. To switch between the 16 effects, there is a four bit binary input. Rather than using an expensive and hard to find rotary dip switch, I used an Arduino mini to convert an analogue voltage from a potentiometer, to one of 16 four bit words to select the effects. I used the rest of the Arduino to display the name of the effect on a tiny OLED screen. I asked chatGPT to write the code, and it actually worked! I have not been able to get chatGPT to write any working code since. HAHA!

LFO PCB

Once the effects section was finally done, it kind of felt like the keyboard was finished, because I knew the next section would be the easiest of all of them. The LFO was one of the first things I came up with in FALSTAD circuit simulator. The LFO is based on the KASSUTRONICS variable waveform LFO. I modified it slightly to have voltage controlled rate. I built two of them on a breadboard, and designed a circuit board before even testing them out on the keyboard. I was that confident they would work. Most of the modulation inputs on the rest of the keyboard were 0-12 volts, but there were a few that needed 0-5 volts, so I added a separate buffered output that was scaled down and diode protected for the 0-5 volt destinations. Both LFO's have 11 modulation destinations on a 12 position rotary switch. The 12th position is "off". Both LFO's can modulate Bass-PWM, Bass-VCF, Lead-VCF, and Chord-Low, Med, and Hi(volume). LFO1 can also modulate Rhythm-Cutoff, Rhythm-Volume, Lead-Volume, Bass-Volume, and Chord-Volume. LFO2 can also modulate Delay-Time, Delay-Sample rate, DSP-Sample rate, Vibrato-Speed, and Chord-Chorus speed. The LFO circuit board was so simple I didn't need to build a separate hardware board for the faceplate. Instead, I just mounted all of the LFO's hardware to the PCB itself. I had a little room left on the faceplate, so I put the master volume control knob in the corner. I also wanted to keep the original Suzuki logo in the corner, so added a little cut-away to the bottom of the faceplate so that the logo is not covered. 

LFO PCB and rotary switches to the left

Wiring up LFO rotary switches to modulation destinations

While going through the process of installing the LFO faceplate, I made a pretty big mistake. The master volume potentiometer was not part of the LFO PCB hardware, so I had to use a larger pot that bolted to the faceplate. The potentiometer was much bigger, and needed a bigger cavity in the keyboard enclosure than I had cut. It was late at night, and the LFO faceplate was already installed and working. I didn't want to go to the trouble of taking it all apart again, so I decided to just carefully widen the hole for the volume pot with my dremel from the underside. I drew a simple outline of the pot on the bottom of the keyboard enclosure, and proceeded to grind away the plastic, careful not to go too deep and hit the underside of the LFO faceplate. Unfortunately I was not paying attention to how far I was going out to the side, and when I was finished there was light shining through a big hole between the LFO and Effects section faceplates... It was pretty disappointing, and I should have known better. I always seem to make these kind of mistakes late at night when I am trying to do just one last thing. Thankfully I had a whole other enclosure that I could use. Unfortunately the backup was the more beat up one. I had destroyed the one in perfect condition. The next couple of days were spent disassembling the keyboard, drilling and cutting the holes and cavities in the new enclosure, and reinstalling everything to get it back to where it was before I screwed up. The new enclosure is not nearly as nice looking. It has a lot of scratches and chips, and there is some graphic text poking out below some of the faceplates that I hadn't noticed before. I do like that it says SUZUKI UNITED STATES on this one though. Not really sure why, it just seems unusual. I might go back someday and just paint the enclosure and cover up all of the scratches. I think it would look better if were red or baby blue. We'll see..

Bored...

With...

These...

Yet?

The faceplates are all laser-cut from a cream colored acrylic I bought from TAP plastics here in Portland. I love that store! The text is all lightly engraved with the laser, then filled with acrylic paint. Once the paint is cured, I sand the surfaces lightly with a 320 grit paper. It removes the excess paint and leaves a smooth matte finish on the faceplates. All of the knobs are plastic. I have been ordering them from Aliexpress for pretty cheap. I used to buy everything from Ebay or amazon, but I have found that Aliexpress is so much cheaper and usually has a lot more variety. It is pretty much all the same stuff I would find on Ebay anyway. When it comes to buying components like capacitors and IC's, it's probably best not to buy from Aliexpress or Ebay. I have been buying more from LCSC, and Mouser when I need to be sure I'm getting quality parts.




I finished the project off by building a small amplifier board for the output and a speaker power amplifier(LM386). The keyboard has a mono line output, a sync input and output jack, sync clock multiplier switch, a 12 volt adapter jack, and a power switch on the back panel. I didn't connect the battery compartment, because I don't have any intention of buying eight D batteries anytime soon.

Effects section with effects module add-on and power-amp board(silver)

Rear panel view of sync input multiplier switch and jacks



Gut-shot

Buttoning up

Tucking in

Closing up

4 comments:

  1. the truth is i love reading about the process as much or if not more as seeing and hearing the outcome of your builds!

    ReplyDelete
  2. Good to know. I will keep that in mind for future posts

    ReplyDelete
  3. I think that Berhinger needs to hire you, and make clones of your creations. I would buy that thing in a second, and I've only heard 30 seconds of it.

    ReplyDelete