Roland's first polyphonic. I was surprised to see that this uses the standard comparator ADC circuit! They did a nice job making it sound smooth..and maybe it was just an experiment to see if they could make the technology work reliably! Because since it after all doesn't even recognize slider movements while it's not in 'manual' mode....I see no point in it. They could have just put in switching between slider voltage and DAC playbacks to S+H cells but...they didn't. It's really quite similar to most of the other synths with memory except they didn't go so far as to make it so it has a patch buffer so that you can change an aspect of a stored sound. Bummer. ...or is it? Maybe somehow it sounds better because of that limitation. I doubt it but think rather it's probably just the way the signal paths were set up and fiter selections/locations along the way. The envelopes on this machine are actually pulsed capacitors it appears. Again they did a good job making it smooth sounding. Love this machine's arpeggiator in particular also. I fished for one quite a while before finally grabbing one at around $210. One of the synths you'd have a hard time prying out of my cold dead hands :-).
Other than cleaning issues haven't seen much wrong with mine. :-) I delved more into it when installing the IO modification. The main problem encountered was with the noise volume control. In the original mod the designer used 30usec pulsing it appears to modulate the gate voltage of the FET switch Q23. This results in a distortion of the white noise spectrum until it reaches peak level where a 9volt or so dc voltage is present and all pulsing is turned off. AS OPPOSED to the 15V that would be found there if the line were not hooked up when noise is on. Turning noise up to where it will mix well resulted in the horrible outcome of noise bleeding through all the time and not a good sounding noise but a clippy poppy noise. Tech Peter Lakowitz got his working adequately by changing the 56K R101 to about 2.5K with a trimmer in place of that resistor of 10K value. This of course changes the front end bias of the FET and no doubt compresses the noise dynamics but it does allow for more volume headroom on the noise. I've proposed a solution to the inventor of the mod where we use instead a digital pot in place of the 33K feedback resistor in the inverting amplifier circuit where VR11 sets the noise gain.
Meanwhile, Here's a mail from Dan Wilson on a repair he did of the ensemble effect:
May I take this opportunity to share with you a particularly weird fault which I encountered on a JP-4 last week as a thank you for all your help over the years...
I came back from Studio Goldfrapp with Will’s JP-4 which had developed a crazy fault with its chorus ensemble. Rather bizarrely, it was taking longer and longer for the ensemble to kick in having hit the ensemble button and was becoming more and more noisy once running. He demonstrated this to me and sure enough enough several seconds after hitting the button a cacophony swept BBDL clocking noise came through the monitors!
On closer listening what was so bizarre about the fault was that not only could you hear clock noise but the chorus was warbling just like a cassette player with a bad pinch wheel!
This seemingly simple fault took me several hours to trace last night and turned out to be quite a little gremlin!
I quickly established that both MN3004s and their driver ICs were working fine and that the supply rails didn’t look too shabby. I could see a healthy enable line coming into the chorus assembly and was driving the signal path enable FETs Q7 and Q15 just fine. Some of the caps looked a little tired so I recapped the pcb with no effect. There was some of the dreaded brown cement on some of them which is known to go conductive with age so I removed that too – no effect.
On closer inspection there is a clock disable feature on this ensemble. After much poking around it transpired that there simply wasn’t enough negative voltage swing on standoff resistors R93 & R92 to the defeat transistors Q20 & Q21 in each swept clock generator. Tracing backwards I found that although the correct voltages were being applied to the transistor that drives the defeat transistors (Q22) the device was not saturating and thus not swinging sufficiently negative to take the defeat transistors out of conduction. I replaced the transistor and sure enough it all started to work correctly. My theory is that what was happening was the transistor had gone leaky and was sitting in its linear region on chorus enable. This was causing the defeat transistors to partially conduct with any noise that might be tied up in the supply rails about Q22 modulating the clocks. Since everything was running in the linear region it would fully explain why a change in temperature in the transistors would eventually cause the circuit to enable hence the slow engagement.
Contact Sound Doctorin'