[sdiy] Help, I'm Desperate! (Charge Injection with DG408)

David G Dixon dixon at mail.ubc.ca
Tue Dec 11 08:10:35 CET 2018

OK, so to continue the saga:

I rebuilt my circuit with a flash-ADC comparator chain, and it all worked
reasonably well, except that when the CV was stuck at 0V (right in the
middle, where my fader has a detent) one of the comparators was oscillating
like a mo-fo, and this was causing the DG408s to freak right out, putting
all sorts of noise onto the audio output.

So, I decided to put hysteresis on the comparators.  This required an input
resistor (10k) to the + input of each comparator, plus a large hysteresis
resistor (2.2M).  I also added 10k resistors to the - inputs, so added 21
resistors to the circuit overall.  Somehow, I managed to lay this all out
with just 0.4" of additional length.  I didn't want to rebuild the entire
board to test this, so I built a daughterboard with just the two LM339s and
the CD4532 on it, and I removed all of these components from the main board,
and wired up the daughterboard at seven connection points.  This worked
great, and the daughterboard floated above the main board on the stiffness
of the connecting wires.

Result?  Now it works ALMOST perfectly.  The noise has disappeared (as I
knew it would) and the comparators do not oscillate at all.  There is now
only one tiny problem: There is still an almost imperceptible click when the
CV crosses 0V (and the logic control voltage crosses 2.5V) in the positive
direction.  This is when the logic switches from 011 to 100 (i.e., all three
bits change).  Interestingly, I don't hear the click at all when the CV goes
in the other direction (100 to 011).  Hence, what I think I am hearing now
is just charge injection from the A2 bit, which only occurs in one direction
according to a document I found on the web about it.  Also, the click is
very faint, and could thus be about 20pC of charge as per the datasheet.

So, I believe that I have now eliminated all of the problems I had
introduced to this circuit with my naive approach to logic design.  I don't
know if I'll be able to get rid of this tiny click, but I'm going to try two

1) Adding 10nF caps to ground from the switch outputs, as suggested in the
datasheet and the document I found, and
2) Reducing the input impedance on the downstream circuitry from 100k to
10k, which the datasheet suggests should help (lower impedance is better).

According to what I've read, these things should reduce the effect of charge
injection.  (If it is charge injection, then I would have been no better off
with digital SPST switches.)  However, this is going to have to wait until
tomorrow afternoon.

More information about the Synth-diy mailing list