[sdiy] Query about best cap type for '2164 stability network

[Richie Burnett]

> You've given several reasons that C0Gs are good in general, and that's
why I tend to go for C0G for small value caps if I don't have a good
reason not to. But in the spirit of trying to learn about this stuff
properly, I wondered if there are any reasons particular to the 2164
application that would call for C0G over something else.

Ok, I'd go for C0G/NP0 because of low microphony (I don't want PCB vibration 
getting converted to an AC input current right at the sensitive input to a 
VCA,) and also because this dielectric has a very low loss at high 
frequencies.  The last thing I would want to find is that my VCA breaks into 
oscillation at a few MHz, and I've seen this happed with some VCAs if the 
correct compensation components aren't right.  Stability is what these 
components are about.  (In addition I would benefit from the commonly known 
advantages for C0G, like low voltage coefficient, low temp coefficient, and 
tight tolerance, for very little extra cost/space given the relatively small 
capacitance value here.)

> It doesn't seem to me that the capacitance change with voltage or
temperature are essential in this application.

Possibly true, but I don't like the thought of any capacitance change with 
applied voltage as this can imply intermodulation distortion in audio 
circuits.  So, I'd always use caps with low voltage coefficient in the audio 
path.  (Although, granted their shouldn't be much audio-frequency voltage 
across this cap in this particular location.)

> The high frequency performance may be. Is it?

A 560pF cap in series with 500R resistor connected from the summing node of 
an op-amp to ground puts a zero in the noise-gain transfer function of the 
amplifier at approximately 3.5MHz.  Whilst this isn't a particularly high 
frequency, I would say that it is high enough that *I personally* would be 
concerned about the effect of using a poor dielectric here that was lossy. 
The cap is in series with a 500R resistor, so intuition tells me that if I 
get a big amount of ESR from my capacitor at this frequency, it will add to 
this reasonably low resistance, and the impedance of this little RC series 
network isn't going to be what it's meant to be any longer.  (I guess you 
could fit a lossy cap and decrease the resistor value accordingly though if 
you felt comfortable with this process.)

> I realise that from an actual building standpoint this is a somewhat
academic question - we can just use C0G and stop worrying - but I like
worrying / learning.

Yes, just fit it and don't worry ;-)  But it's good to understand why.

As for what this R-C network is doing...  The short answer is that it is a 
required part of the amplifier circuit to make it stable.  They just either 
couldn't fit these bits on the chip or decided not to because the values 
change subtly depending on the details of the application.  The network has 
nothing to do with filtering your audio that passes through the VCA, and has 
everything to do with keeping the VCA from misbehaving at MHz.  For more 
technical details read about "Phase Lead Compensation" and things like 
"Lead/Lag compensation" in analogue electronics text books like Neil Johnson 
mentioned.  The R-C network modifies the loop gain *around* the amplifier so 
that the closed-loop crossover point occurs at a suitably high frequency for 
a good transient response but also with sufficient gain and phase margin to 
ensure acceptable stability.

-Richie,