Here’s something cute but probably pointless. The module is intended to cover a range of 88 notes, which is 7.25 octaves, so it maps 0 to 4095 counts to 0 to 7.25 V. But suppose you decided a few of those notes at the top and bottom are useless, and you only care about 6.825 octaves, so you modify it to map 0 to 4095 counts to 0 to 6.825 V. Then for a perfect DAC (no nonlinearities, no gain errors) the differences between actual and correct output voltages looks like this:
Yes, that’s the correct plot. The arithmetic happens to work out that the resolution error is zero for every MIDI note! The reason has to do with the fact that 4095/12 is an exact multiple of 6.825 (in fact it’s 50 x 6.825), which turns out to mean 1 LSB corresponds to exactly 2 cents frequency change.
The resolution error is already small enough that it’s not really worth throwing away 0.425 octaves to reduce it to zero, but I thought it was amusing.