Each of the band pass filters were adjusted to have ~12 volts(p-p) outputs. It alone made audible improvement to the overall output.
Just today I read from some forum, that this(Paia) vocoder, has an inherent property in form of deviation of the bpf outputs. When I was measuring the compander's output levels, it wasn't so dramatic, so that could the reason, why the designer of this device(Craig Anderton) did let them be, as they are.
But now it's been taken care of, and the previously mentioned compressor for speech signals may not be necessary at this point.
Optocoupler pair, as a module, and built one(1 of 4):
There were two odd things, that I stumbled upon. First one was, that I had to swap mic/instrument connections for the led-ldr pairs 2-4-6-8. Otherwise they wouldn't give any output. Probably some logic error in the bus boards. I didn't want to spend too much time to find them(errors), instead it was easy to fix the problem, when they were on the breadboard.
Second one was utterly strange. When I built the above led-ldr pair as a module, there was usual testings, to see if it works, as expected. Procedure was this: feed a low frequency square wave(1Hz) to the led inputs, and higher one(audio range) freq. signal to ldr's. That way I expected to see, how the slowly flashing led turns on and off the signal, that goes through the ldr.
No such thing. The signal went through the ldr, and the flashing led did affect it only barely visible amount, when looking at the scope. I used signal generator for the signals, and dialed the levels up and down, but no visible reaction.
There I was, thinking: "bugger! this doesn't work!". Then I did the same tests with 'previously aurally tested and working' -setup, that was on the breadboard. Exactly similar results.
My conclusion was, based on level measurements, that it shouldn't work, but when connected to the rest of the circuitry, and listening to the output, it works, for some reason.
There must be some 'plan 9 from outer phase' -audiochemical processes going on.
But seriously, the way they work, isn't obvious(at least for me). I would most probably have skipped the whole concept, if I was counting on my skills to find out, what's happening, just by looking at the numbers or curves.
In the end, no off-topic ramble this time.
Makes one wonder, why the optoisolators aren't used on diy-vocoders(as far, as I'm aware of).