[sdiy] Hand-matching capacitors for filter stages

[thx1138]

On 2/9/10 11:49 AM, "Antti Huovilainen" <ajhuovil at cc.hut.fi> wrote:

> On Tue, 9 Feb 2010, thx1138 wrote:
> 
>> I define good engineering as the ability to not have to use matched
>> components is projects whenever possible.
> 
> Which in this case "obviously" means not using analog at all but instead a
> DSP with its much lower distortion and accuracy.
> If the specs call for, say, "Moog filter", then we're pretty much stuck
> with the circuit topology and can only improve on it by matching and/or
> trimming.
> 
>> In some cases I find that it is easier to use 1% resistors to tighten up
>> a circuit.
> 
> If the caps have 5-10% tolerance, matching resistors to tighter than 1%
> (the basic accuracy I buy mine in for peanuts) won't make any difference.
> The accuracy is dominated by the capacitors.
> 
>> Transistors are a different case as I like to use transistor arrays like
>> the CA3046/CA3086 but these are getting harder to find in stock around
>> town.
> 
> More importantly, IMO, matching transistors directly affects CV
> feedthrough, while capacitors have no effect on that.
> 
> Now for my 0.02e:
> Matching capacitors to reasonable accuracy (1-2%?) should be fairly
> trivial if you have a DMM with capacitance measurement. The absolute
> accuracy won't matter at all as long as you match only same value
> capacitors.
> 
> For material, I don't see how there would be any audible difference
> between identical value poly* capacitors. As long as the tempco is within
> your specs, any material should work fine (mica and np0 ceramic too).
> 
> In a 4 pole filter the effect of nonidentical capacitors would be slight
> shift in resonant frequency and resonance amount. This is unlikely to be
> audible except near self oscillation. The exact effect is easy enough to
> calculate when you remember that 1) resonant frequency is the frequency
> where filter phase shift is 180 degrees and 2) self oscillation happens
> when filter + feedback gain exceeds 1.0 at resonant freq.
> 
> Antti
> 
> "No boom today. Boom tomorrow. There's always a boom tomorrow"
>  -- Lt. Cmdr. Ivanova
Personally I find writing DSP code boring after doing it for over 25 years.
The smell of rosin based solder is much better and easier to debug.

I love .1 inch centers for prototyping old analog and hate building adaptors
for surface mount, but PCB design is much easier than the old tape method
from the past.

Matching caps is a best interesting as it they tend to drift over time and
age anyway. 

Harold Bode once told me the secret to analog is to use what is on hand and
not depend on parts that are unavailable when you need them. It seems
prudent to trust that you can make things work in a pinch and trust your
ears as well.

Your comments are correct and I have not been able to detect much difference
in audible artifacts.

My filters were designed to track pretty well when put into oscillation.
Mind you "Tracking" is over a very limited range but well enough to
calibrate on each module and give some useable service.

It got better on the SSM 2040 and 2044/45 devices. Too bad these are getting
so hard to find. Ron Dow built some really good chips back in a day.

Regards,

Terry