CPU controlled analog with dco's that are timer controlled pulses creating square and pulse waves and a sawtooth op amp generator that is digitally controlled. From there on the signal path is typical digital controlled analog.
First note that it is normal to have the 'countdown' at bootup as the machine aligns. Aftertouch tends to go bad. Here is a page describing the problem and a sometimes fix. The feedback resistor also should be modified often in the op amp gaining up the signal from the AT sensor. I seem to recall 3x the value being used. 150K I think. I finally got one in here 9/2012 and it cleaned up nice and I retrieved the bezel and polished it and goop'd it back in :-). The AT on this one though has initial value without pressing any key. Not a lot but enough to be annoying if you turn the sensetivity up. STRANGELY this appears to be normal though as I check the design. The AT sensetivity delivers 0 to 2.9V to the non-inverting input of the op amp. The inverting leg has a 15K feedback resistor and also attaches to a 470 ohm series resistance with the AT strip to ground. I get 8M ohms when there is no pressure and 9K when you press hard on this unit in 2012. 27 years old. SO it maybe has less than that when it's in good shape? Anyway the fact remains that with the high impedence on the strip, the output of the op amp will be essentially the same as the Non-inverting voltage. So when you turn the slider up you will have 2.9V there basically and that's enough to overcome the series diode and put voltage on the input of the ADC! So I'm frankly baffled by the design on this! Calling Roland... Michael said he wasn't aware of the issue but agreed with my analysis. And even consulting one of the guys who worked on it, we conclude it was probably just designed that way! If anyone wants to resolve the apparent 'fault' the obvious way is add an op amp. Design a common subtractor circuit that goes between the output of the op amp already there which is what U3 I think...the 4558 pin 1 anyway. Then ..wait is the other half of that even in use? If not go for it! Anyway subtract the voltage that is being provided by the slider (ie. the voltage on pin 3 of the 4558) from the voltage on pin 1. Thus 0 on the slider to 2.9V with NO movement of the sensor strip would still give zero output. At maximum sense pressing the strip would give a peak of possibly 2.9V less if it wasn't already crushing the headroom :-) So possibly installing a 22k in place of 15K in the feedback of first op amp would be in order. You may want to do that anyway....Look for occasional bad solder and lots of dirty switches!
Roland may still have some parts for these. And also there are some discrete components that I usually stock.
Contact Sound Doctorin'