The calculations...

[Ingo Debus]

Hi Roel,

several comments...

If I got you right you are trying to make your MIDI-to-CV as accurate as
the 16-Bit DAC is. I think this is not necessary. An error of 1 cent (=
1/100 semitone) of the pitch of an attached VCO would be excellent, two
or three cents perhaps still very good. Assuming your output range would
cover 128 semitones (full MIDI note range, more than 10 octaves!) this
would allow for an error of 1/12800 of the full output range. This is
much more than a 1/2 LSB (16 Bit) error.
Hey, do you want to outperform today's digital synths? Your next project
is a VCO with that accuracy, right? ;-)

In your calculations you are dealing with leakage currents in the nA or
even pA range, but you're taking into account only the leakage currents
of MOS switches and opamp input currents. Are you sure that there aren't
other leakage currents of that magnitude, like parasitic current on the
PCB or capacitor leakage?

And don't worry about a few wires on your PCB. Yes, in theory you should
do everything right at the first attempt. But look at commercial
electronic devices, you'll often find wires on the PCBs. And these are
from series production, while yours is the first prototype.

Ingo