2
\$\begingroup\$

I have the following simple op-amp circuit, which is meant to drive a MOSFET (M1) gate and work as a voltage regulator. It is discrete for a mix of reasons (necessary bandwidth, using existing BOM items, avoidance of proprietary parts, cost).

schematic

simulate this circuit – Schematic created using CircuitLab

If there is neither RE nor CC present, the circuit can become unstable if the MOSFET is rather beefy (gate capacitance of about 500 pF or more) and if the output load does not damp the resonances well enough.

I found that there are two effective ways to stabilize the circuit: lowering open-loop gain via RE or lowering GBW product via CC. I wonder which is a better way and why?

Intuitively, I am drawn towards RE, because that does not increase the quiescent current to meet target GBW of ~20 MHz. But I guess CC is popular for some good reasons...

\$\endgroup\$
8
  • \$\begingroup\$ Why not simulate all the different conditions that your circuit may face. I'm sure there is no single answer that suits all scenarios so, happy simulation fun is the best method. \$\endgroup\$
    – Andy aka
    Commented Oct 14, 2021 at 8:45
  • \$\begingroup\$ @Andyaka I did simulate, and the result is that I favor RE. However, I wonder if this conclusion is "wrong" and I am overlooking something because CC seems to be the de-facto standard to stabilize op-amps driving capacitive loads. Potentially, the preference for CC comes from a desire for very high open-loop gain (which I don't need), so I am looking for other reasons in favor of CC. \$\endgroup\$
    – tobalt
    Commented Oct 14, 2021 at 8:52
  • 1
    \$\begingroup\$ Well, trust your simulator and the amount of scenarios you have covered is my first advice. 2nd advice is this: you say you need "necessary" (high?) BW but, if your MOSFET gate capacitance is 10 nF, the 2k2 gate (and emitter) resistor (R4) will create a poor circuit for dealing with load transients. When Q4 turns off, you have an RC time constant that implies a bandwidth of only 7 kHz. I'd go push-pull driving the MOSFET on that basis. But, with no details of load scenarios, power input ripple and transients, you are still facing simulation after simulation and no great advice from this site. \$\endgroup\$
    – Andy aka
    Commented Oct 14, 2021 at 9:04
  • \$\begingroup\$ It boils down to you providing more information about the load scenarios and the noise on your supply and the expected ripple/transients on the output. Without that information, you are effectively asking for opinions or asking too much of a potential answerer. \$\endgroup\$
    – Andy aka
    Commented Oct 14, 2021 at 9:06
  • 1
    \$\begingroup\$ @tobalt, IMO, While you have 10uF at load side, push-pull does not help. I have simulated. Though, moving CC from Q3/R3 node to Q4/R4 node would help, still need larger cap. \$\endgroup\$
    – jay
    Commented Oct 14, 2021 at 17:33

1 Answer 1

Reset to default
0
\$\begingroup\$

For the regulator to regulate accurately to a voltage of Vref the regulator needs to have a very high dc open loop gain.

Resistive compensation will reduce the open loop gain at all frequencies including dc which will reduce regulation accuracy.

Capacititive regulation (CC) provides a frequency dependent roll off of the open loop gain. So the open loop gain is very high at dc and low frequency, where it's most needed, giving regulation accuracy but the open loop gain reduces as frequency increases getting the loop gain down below unity well before the loop phase reaches -360 degrees giving adequate phase margin and ensuring stability.

\$\endgroup\$
1
  • \$\begingroup\$ The way you describe it, makes it sound like RE and CC are basically equivalent to roll off gain at higher frequencies, but RE has the additional downside that it also reduces gain at low frequencies, whereas CC does not. Indeed I would agree, but my simulations show different behavior: CC reduces the available GBW product. RE reduces only the low frequency gain but leaves the GBW product identical. However, now that I formulate it I don't even understand why RE then helps with stability at all, since it leaves high frequency gain actually unchanged. \$\endgroup\$
    – tobalt
    Commented Oct 14, 2021 at 10:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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