0
\$\begingroup\$

Referring to the capacitor between the output and the inverting pin in an op-amp, in parallel with the feedback resistor --- should I use ceramic (NP0), or thin film?

We're talking bandwidth in the order of tens or low-hundreds of MHz. Capacitor value in the order of a few pF, maybe up to 10pF or 20pF. Ceramics seem to offer higher frequency response; however, their non-linear characteristics presumably make them unsuitable to be part of an analog signal path, and I'm thinking the feedback capacitor qualifies as "part of the signal path", right?

What about other types? Anything other than ceramic or thin film that is recommended for this?

\$\endgroup\$
3
\$\begingroup\$

There is a lot of talk going on about capacitors and their uses these days.

Unfortunately half of that is over-generalised.

For the purpose of below nF capacitance an NP0/C0G capacitor can be tiny and affordable without much problem and will likely out-perform most other types of capacitors over frequency, voltage and temperature, not to mention life.

It is true that "lower grade" ceramics, which yield a higher K, are known for large influences of DC voltage, AC voltage and temperature to their capacitance and may have large pieco-electric effects associated with them, which may also not be great in feedback scenarios, depending on the application.

But the ceramics marked C0G/NP0 are very rugged and experience nearly no aging or expansion effects, nor are they compressible enough to generate measureable voltages unless you smash them hard enough to break.

Just as an example, look at the graphs for voltage biasses and temperature of this 68pF capacitor in a 0201 (metric: 0603) package (note that the graphs are 1000's of MHz, not single MHz). http://psearch.en.murata.com/capacitor/product/GRM0335C1H680JA01%23.html These cost virtually nothing to such an extend that I usually just buy the whole reel for my lab&mini-fab.

As an aside, as you get into higher capacitor values, NP0/C0G will become somewhat huge, which brings other issues such as parasitic resistances and inductance into play, but certainly not when we're still using pF to denote their value. Up to 10's of nF NP0/C0G should be reasonably easy to still specify without breaking the bank or your board-space.

((I used the double name everywhere to avoid confusion, add completeness to the answer))

| improve this answer | |
\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.