In 1986, Roland brought out a piano that sounded amazingly realistic for the time. The RD-1000, and it's rack module equivalent, the MKS-20, used something called "SAS" or Structured/Adaptive Synthesis to create their tones. There are various theories floating around the net about exactly what this was, some people claiming there were NO samples used. My impression is that there were, but they were harmonic samples used in a kind of additive re-synthesis, tailored for different keyboard zones and velocities. Anyway, it was pretty hi-tech for the day, and it paid off. By the end of the eighties, Elton John was playing an RD-1000 and most serious studios had an MKS-20 in their racks.
While the digital side of the instrument was impressive, I think the genius of it was how Roland combined it with analogue technology to create such a warm, lush sound.
A block diagram of the architecture shows two PCBs, CPU-A and CPU-B, create the monophonic audio from Roland's patented SAS method. This audio signal is then passed to an analogue "Effects" section comprising a parametric bandpass filter, then high and low pass shelving filters (controlled by Curtis CEM 3360 VCA chips). Then at this stage the signal is split into two channels for stereo, to a chorus circuit based around Panasonic MN 3007 BBDs (bucket brigade delay chips), then a (panning) tremolo circuit again based on Curtis VCAs. Lastly there is a CPU-controlled voice muting circuit that basically acts a noise gate when notes aren't sounding, probably to shut off the hiss from the chorus unit.
Taking a closer look at it, I wondered if it would be possible to gain external access to the analogue effects of the instrument. Well, it turned out to be pretty easy.
Above: with the cover off, CPU-B pcb is at the rear of the unit. The connector just to the left of the large black block (filter inductor) at the bottom left corner of the photo is where I accessed the audio output.
Above: detail from the schematic (copyright Roland Corp) showing the audio output from CPU-B at terminal 11.
Looking inside, the MKS-20 is basically divided up into a digital top deck, and an analogue bottom deck, and the connection between the two that I was interested in was the monophonic audio line. It leaves CPU-B board at connector #3 (marked CN3). By cutting that connection and bringing it out onto two switched jacks, I was able to take the raw mono audio from the digital boards out for external processing, say, by a saturator or guitar pedal, and then send this back into the effects section to be treated by the EQ, chorus and tremolo units, emerging at the regular outputs in "stereo".
Above: view rearwards of the cut audio wire as it leaves CN3, the yellow wires taking the signal to two switched jacks at the rear panel. Note the new resistor at the "piano out" jack.
Obviously, any audio could be processed just by plugging in to the effects side. However, there was one trap with this arrangement that baffled me at first - it turns out that unless the CPU opens the noise gate you won't hear any sound! It does this when it receives a midi note-on message, so you need to play a note every so often to keep the gate open - since the "piano" output has been interrupted you won't hear it. The gate closes with a gentle slope, and so can't really be used for creative triggered-gate type effects (damn!). There's an audio demo mp3 below.
The two-jack access allows you to bring the connections out to a typical normalled "send-return" type patchbay, so that when nothing is plugged in, the MKS-20 behaves as it was designed.
There is no external control of the analogue section other than Midi CC numbers 92 and 93 which switch the tremolo and chorus on and off, and volume control via CC numbers 1, 4 and 7. Considering how good it sounds this is a shame - I think you could get a nice Leslie simulation with smooth rate control of both the tremolo and the chorus. At present you can only control these, one parameter at any one time, by selecting one for editing and using the "alpha dial" encoder. One way around this might be to program different settings at different memory locations and use midi program change messages to switch between them, although the rate changes wouldn't be smooth. Also, this will shut any midi note off, so to keep the gate open another midi note-on needs to be sent as well.
Further possible mods could involve either external/MIDI control of the alpha dial (if that's possible), or perhaps more efficiently, direct control of the parameters on the effects board itself.
Here's an mp3 of a single chord held on a string preset from a Casio CZ-101, going through the effects section of the MKS-20. Now I know the musical input could have been a bit more interesting, but I'm trying to demonstrate the effect on the timbre and stereo field, so a static patch like this was ideal. First you hear the basic patch going thru the unit with all effects and EQ zeroed. Then I quickly zip thru some presets randomly (by sending midi PG messages, as well as note-on messages from a sequencer, on the MKS midi channel), each demonstrating a different eq, etc. Make sure you listen in stereo, at a fairly healthy volume because some of the changes are subtle, especially the chorus. This was knocked out quickly - I just used the existing programs without preparation. More musical applications might involve, say, incremental changes to effect parameters on adjacent program locations, for a smoother effect, or making the PG messages quantised with a rhythm track.