Wednesday, January 9, 2008

On the Workbench: Juno-106, part 2

Calibrated the 106 last night.  First, some photos from the end of the repair job.
The module board was too wide to fit on the Panavise the conventional way.  I had to stand it on its end:



The new 80017A ready to go.  Note that the lot number is actually older than the one it is replacing!


The new part installed and ready to rock:



Next up: calibration.  The last time I calibrated the 106 was four years ago, and I was curious to see how far off things were.  I already knew the VCF on voice 4 was a bit sharp, and when I played the synth after powering it back up, the one on voice 5 was about a whole tone flat -- not surprising since it had just been replaced.  There are 10 steps in the calibration procedure (from Daredevil's page; there are two files.  Scroll to down near the bottom.)   The first is to check the power supply voltages and adjust as needed.  The procedure wants the voltages accurate to within 0.01V.  That's just within the precision of my DMM.  I measured the -15 supply at 15.06, so that was very close.  Careful tweaking on the trim pot got it to -15 on the nose.  Probing the power supply board:




The next step is adjusting the DCO control voltage offset.  In the test mode, you press the MIDI CH button and it is supposed to go to exactly 0V.  I was a bit surprised to find that mine was dead on (sorry about the photo orientation):




Step 3, VCA bias.  It was within range.  Step 4, VCA offset.  This adjustment minimizes control voltage feedthrough, which causes thumps or pops in the audio output when the control voltage changes abruptly.  This is what control voltage feedthrough looks like on the scope:




This was from voice 5.  Voice 2 had a bit of offset, not this bad.  The other four voices were already as flat as I could make them.
The next steps get into the VCF adjustments, and are the trickiest part of the procedure.  Step 5 calibrates the VCA resonance for each voice.  A test mode puts the VCA sets the resonance high enough to cause self-oscillation, which generates a sine wave.  Which has to be visually trimmed to 4.8V peak-peak on the scope.  (I don't have one of those fancy scopes that can measure it for you.)  With the reticle illumination cranked way up, this is what I was looking at:




One thing that helped was using the delayed sweep on the scope to position the peaks exactly on the reticle line intersections (and at the left edge, as you see above), which made the position of the peaks relative to the reticle easier to see.  Voices 4 and 5 were a bit high; the others were close enough not to mess with. 
Step 6 adjusts the VCA gain, using the VCF-generated sine wave.  All voices were on spec except voice 6 which was a bit high.  
Step 7 adjusted the VCF cutoff frequences.  You're supposed to use a frequency counter to do this.  But my scope doesn't have a frequency counter; the DMM does but it's not accurate enough.  And anyway, I don't like doing it the way that the procedure recommends -- they want you to tune all of the VCFs a semitone flat for some reason.  So I do it by ear.  In the test mode, you can still use most of the patch editing controls.  So I checked the VCF self-oscillation against the DCO by turning the pulse wave on and off, and listening for a pitch change.  I tweaked the PWM control to get a similar volume to the VCF self-oscillation.  Holding down the middle C, I toggled the pulse wave on and off and listened.  
Much to my surprise, the voice 4 VCF (which I had thought was sharp) was dead on; all the others were flat.  Most of them were flat about a quarter tone.  The trim pots have a large range and you need a delicate touch to bring them those last few cents.  Voice 5 of course required a big change.  I recall that the last time I calibrated the 106, the VCF on voice 5 was way sharp and I had to crank it a lot to bring it in.  I seem to recall then when the 800017A on voice 6 failed in 1991, it was sharp just before it failed.  I'm starting to wonder if a VCF that has drifted sharp is the advance warning of an 800017A about to fail.
Step 8 adjusts the VCF scaling.  The scaling interacts with the cutoff frequency calibration, and the last time I calibrated the 106, I was caught in a loop between these two things for a while.  But this time, the scaling needed very little tweaking, mercifully.  I think I just touched voice 3, not enough to effect the cutoff frequency tuning.  
Step 9 adjusts the noise level.  I was surprised to find that it was waaaaay hot, 6V on the "average peak to peak" method shown by the procedure, as opposed to the spec of 4V.  Last week, when I was working on Solar Flare, I recall thinking that the noise on the 106 was rather prominent in a patch that I was using.  Maybe I should go back and mix that again to see how it sounds.  
Step 10 calibrates the PWM adjustment in the DCO.  You trim the circuit to get a 50% pulse, and then again to get a 95% pulse.  This was on spec and I didn't have to touch it.  A photo of the scope display:


Note the bit of ringing on the leading edges, and the slight downward (towards zero) slope on the tops of the pulses.  The pulse wave on my 106 has always looked like this, and it's the same on every voice.  I've always wondered if this contributed some to the character of the pulse wave on this synth.  But I don't know if other 106es exhibit this behavior.
The last step is supposed to adjust a DC bias in the chorus circuit, to minimize clipping in the chorus.  It's kind of a pain because you have to connect a signal generator to a test point on the audio board, scope the synth's output, and then adjust some pots on the jack board that are kind of hard to get to unless you put the synth right at the edge of the workbench where you can open the panel up all the way.  (As it was, I put a block underneath the back of the synth to allow the panel to open as much as shown in the photos.)
It works as good as new now, and the sound is still that great 106 sound.  One more photo, of the completed and calibrated synth just before I buttoned it up:



I'm going to put up an expanded version of this post on the Web page eventually.  It will contain the complete calibration procedure with my notes, photos, and illustrations, along with some troubleshooting tips I've compiled in my years with this synth.

1 comment:

Lau said...

Read your post on the Juno, very informative! I've acquired a Juno 106 myself, but it's silent, completely, not even a crackle. From what I've measured the powersupply seems to not be outputting +15/-15VDC, suspect the power transistors may have died.