About those resistors  you mention, as far as I see:

R19  is 14.3K. Are you sure the values written there in the schematics are not 28.9K and 14.3K  

R21 is 14K. I think you misread 196 on what is actually 1% and the new value is written below, hardly legible, but it could be 14.0K.

You are right, the 196 is in fact 1%

Lets think about the complete circuit and the effect of the trimmers:
We have the 7Bit D/A around the 4096 with the resistor cascade. VR3 determines the impact of the most significant bit.
After IC8 we should be able to read the outcome of this D/A.
VR2 determines the current, which runs from the D/A to IC7.
VR1 determines an offset current, which runs also into IC7.

If everything is linear below the E then mostly the MSB line (which includes VR3) becomes suspicious. Why From F (with transpose set to "down") to E (with transpose set to "up") there are 60 steps. We know from Tubbutec, that the firmware of the keyboard scanner is based on the Polysix (whose keyboard starts at C). From C to E we have 64 steps => the first 6Bits determine the notes between C and Eb. When the most significant bit (whichs output is controlled by VR3) goes high, then E and up are given.

So to me it looks like simply the description the first table in the adjustment instructions is wrong: VR3 is responsible for the high tuning. VR2 is for the V/Oct setting of the D/A and VR1 is for the offset (total tune).

In consequence I suggest to adjust the D/A measuring all voltages at the CV-Out socket like that:
1. set tune point to closed
2. set transpose to normal
3. adjust 1.000V per octave using VR2 while playing octaves jumps from lowest F to highest F. Do not obey the absolute voltage; only check, whether jumping one octave makes a voltage change of exactly 1.000V
4. when this is done, play always the highest E and toggle the transpose switch between normal and up; adjust VR3, that his octave jump reflects a change of 1.000V relatively (again: do not care for the absolute value)
5. when this is done, open "tune point", and then adjust VR1 to read straight 1.000V/ 2.000V/... at the C's,

If the adjustment of VR3 fails, then it might be necessary to decrease R52 to 1k.


What do you think about it

Florian