On the Workbench -- Juno-106

My Juno-106 has had a dead voice since last summer.  And yes, it's the infamous 800017A IC.  The usual symptoms of a failed 800017A are a voice that is crackly, intermittent, or sounds like it has a bad connection somewhere.  Eventually it will go almost totally silent, with perhaps some cracking and very intermittent periods of good operation if you turn the VCF cutoff frequency and resonance all the way up.  Using the test mode (from Daredevil's Jupiter and Juno Page, a very good resource for calibration procedures, schematics, and factory patches), I determined that the dead voice was voice 5 (which I kind of thought already, but the test procedure confirmed it).  

So off it went to the workbench:

You access the interior of the 106 by removing the two screws recessed inside the handgrips in each of the end blocks.  Then, the panel opens up as you see here.  The long green board on the left is the module board, which is where all of the sound generation circuitry is.  (The smaller green board in the center is the CPU board, and the yellow board on the right is the power supply.)  There are five or six connectors to unplug, and six brass screws to unscrew, and the board comes out.  

The board is clearly marked as to which of the single-row ICs go with which voices, but I checked with the schematic to make sure.  Each voice has its own 800017A, which contains most of the VCF and VCA circuitry for that voice.  In between each pair of 800017As is another IC that provides the DCO for two voices.  The DCO ICs rarely fail.  IC5 is the 800017A for voice 5.  Most of the instructions I've seen for replacing these recommend cutting the pins from the failed IC in order to remove it, and then using a soldering iron to extract the remains of the pins.  This is to prevent any possibility of lifting traces off of the 25-year-old circuit board.  However, space was too tight for me to get my cutters in.  I solved the problem by just worrying the IC back and forth until metal fatigue snapped all of the pins.  Then, I used a soldering iron to heat each pin stub, pulling it out with the cutters, and using a solder bulb to clear the remaining solder.  The result:

You can see the space where IC5 was.  

There is some controversy as to whether certain lots of the 80017A are more likely to fail than others.  The Wikipedia entry for the 106 claims that lot #41 units are most likely to fail, but at least one poster at Matrixsynth has disputed that, saying that all lots are equally likely to fail.  The ones in mine are: voices 1-3 are lot 43B, voices 4-5 (including the dead one) are lot 42A, and voice 6 (which was replaced in 1991) is lot 61A.

To be continued...