Planning ahead: compensation capacitor footprints

[cellularmitosis]

Howdy!

I'm going to be taking a swing at some projects involving op-amp feedback loops in the near future (e.g. a high-side current source).

Ideally I'd like to plan ahead and add some optional footprints to my PCB design so that I can add a compensation capacitor here or there as needed.

Where would you place such just-in-case footprints in a typical op-amp circuit?

For example, here's an Intersil appnote about compensation: http://www.intersil.com/content/dam/Intersil/documents/an94/an9415.pdf

They show a few examples of where you might want a compensation capacitor (see attachment)

[Kleinstein]

The more common capacitor at an OP is in direct feedback, thus output to the inverting input. However this only applies to voltage feedback OP - it does not work with current feedback OPs.

The other place that is sometimes used is between the inputs.

[T3sl4co1l]

Add an R+C.  This way you can always use the lone C (R = zero ohm jumper), or add a "shelf" or pole-zero compensation.

If this is more about signals than control loops, it may be desirable to add filtering or bypass at the signal input and/or output, away from the op-amp itself.

Consider,
Q: When is an op-amp not an op-amp?
A: Near GBW.

An active filter only works while the amp gain is high.  As amp gain drops to 1 (and below), the compensation capacitor ceases to compensate, and becomes a feed-forward capacitor instead.  Typically, for the circuit shown, the frequency response has a low-pass response, where gain falls from nominal (below fc) to maximum attenuation at some modest frequency, then rising again.  This is accompanied by a notch, where phase reversal occurs.  Above that frequency, gain is determined by the resistor ratios between input and output (including the op-amp's output impedance, which is nonzero, despite what the textbooks say).

So, you can't get unlimited, asymptotic response from a GBW limited amplifier.  But you can get close with a passive filter.  Keep this in mind when you need squeaky clean filters!

Also, filter RFI away from any wires that leave the board.  Only use the bandwidth you need, and filter away the rest.

Tim

[cellularmitosis]

Thanks guys!  That explanation is really helpful Tim (a short at high freq).

I was just toying around with a high-side current source in LTSpice and was able to fix some oscillation with 1pF across the feedback capacitor!

[cellularmitosis]

(oops, that should be 1pF on the cap, not 1pC)

[T3sl4co1l]

You should probably compensate the error amplifier U2 instead.  Note also that the limited speed of U2 (with or without compensation) will make the output characteristic capacitive at high frequencies (i.e., the output impedance is dropping with rising F).

Tim

[cellularmitosis]

Actually I tried that (a cap from U2 output to inverting input), but I had to make the cap ridiculously large to stop the oscillation (like 100uF)

[T3sl4co1l]

You didn't forget to put a resistor in series with U1's output, right..?

Also, a R+C gives you control over phase margin, so you can get much higher cutoff frequency than with just a C.

Tim

[cellularmitosis]

Ah, I did forget to do that, thanks

[Kleinstein]

In this circuit, there should be compensation at U2, and the circuit around U1 should be rather fast.

 To make the part around U1 fast, the resistor values should be considerably lower than 100 K, to reduce the effect of parasitic capacitance. The amplification around U1 is usually stable without an extra capacitor.