I have a requirement for current sourcing about 30A and need to use a power amplifier for other reasons. Is it possible to club two OPA541 (each with 10A max) devices in parallel to achieve this? What considerations should be taken? Any gotchas?

The output of the two amplifiers is going to be a 50Hz sine wave. Inputs for both come from the same source. What's the best design to achieve this?

  • 1
    \$\begingroup\$ There are off-the-shelf high current op amps. I think, APEX do the job. \$\endgroup\$ – Gregory Kornblum Sep 28 '15 at 15:55
  • 1
    \$\begingroup\$ Yes, except all the hybrids (like PA03 etc.) tend to have rather shocking sticker prices. Whenever this comes into discussion I like to compare it with buying into mainframe technology, which if you know how the MCMs are made, it's not too far fetched of a comparison... \$\endgroup\$ – Fizz Sep 28 '15 at 18:06
  • 1
    \$\begingroup\$ Then use emitter follower for power amplification \$\endgroup\$ – Gregory Kornblum Sep 28 '15 at 18:45
  • 1
    \$\begingroup\$ Personally i have strong allergy to opamps in parallel. Power amplifier is a classic circuit, i bet will do the job \$\endgroup\$ – Gregory Kornblum Sep 28 '15 at 18:46
  • 2
    \$\begingroup\$ If your previous question on this similar subject (electronics.stackexchange.com/questions/191588/…) was answered to your satisfaction then please mark the answer as "accepted" or explain where you didn't understand the answer. On this question 10A + 10A does not = 30A. \$\endgroup\$ – Andy aka Sep 28 '15 at 20:25

Yes you can and I've done with lower power OPAs (in order to preserve low board height etc.) The main gotchas are:

  • You need a way (e.g. series resistors) to limit Voffset induced currents circulating between amps. Series resistors will blow away some power. Their value needs to be evaluated carefully. The higher they are, the more protection the amps have, but the more power you blow away needlessly. As @pjc50 correctly notes below the 10-ohm values shown in the circuit from the OP551 datasheet below are rather unsuitable for higher-power amps (OPA551 only goes up to 0.25A in "normal" use so loses in those resistors are about 0.6W each in that case. For 10A at with 10-ohm resistor you'd blow away a [theoretical] 1kW!! (In reality, the amp's output will likely saturate [hit rail] well before that as few amps can output 100V). Apex Microtechnology, which sells mostly higher-powered amps, recommends in their appnote (see link at end) only values in the 0.1 to 1 ohm for these series resistors (corresponding to power losses of 10W to 100W in the 10A scenario.) It so happens that the OPA541 actually has such a circuit in its datasheet (with a rather strange label, and it is not really being discussed in the text--hopefully they revise that) with a master/slave config and using 0.1 ohm series-resistors.

enter image description here enter image description here

(The capacitors in the latter circuit are there for stability improvement, almost certainly, see below. The odd looking connections from the output to what might be misinterpreted as a power rail of each opamp are actually going to the current sense pin[s].)

One more slight variation on this theme is found in the OPA549 datasheet. In this (also high-power, 2x10A) case they also use only 0.1ohm series resistors but they close the master's feedback loop locally, probably because the small values of the series resistors don't cause a significant voltage drop. enter image description here

Other issues:

  • The slew rate will go down (at least if you use the cascade config like all the above circuits do) but in your case [50Hz] you probably don't care about this in the slightest. You'll obviously need two slaves to get 30A in your case. The slaves don't need to be cascaded relative to each other. Regarding topology, some Intersil engineers say in their AN-1111 you're better off with a non-cascaded topology to get better slew rate. You'd need a few more passive parts for that and in my case I didn't need the improved slew rate, and you surely don't either.

  • Stability is worse if you change the feedback loop to be global, as done above with OPA551's. However for higher values of output resistor that's preferable because the global loop automatically deals with (for gain-setting purposes etc.) the voltage drop across the series resistor[s]. I've had to add some compensation capacitors that I didn't need with a single amp in order to deal with some extreme input cases that caused oscillation in the parallel config. But my inputs were 100kHz in some cases (and probably had higher freq noise on them too).

  • You may also need to worry about which opamp is going to limit first (temp and/or current) since you're planning to max them out (I wasn't).

Other resources worth reading on the topic are this TI blog post and the APEX note AN-26; the latter details more advanced issues like stability/compensation.

Good luck!

  • \$\begingroup\$ Those Rs seem quite high compared to (say) an 8 ohm speaker - I would have thought 1R would be more reasonable? \$\endgroup\$ – pjc50 Sep 28 '15 at 15:14
  • \$\begingroup\$ @pjc50: OPA551 has max output current of 250mA ("normal" use) and 380mA short, so it's not really intended to power speakers. For other power opamps (and low impedance loads), the series R's would indeed have to be re-evaluated. The amps blow power on these, of course. \$\endgroup\$ – Fizz Sep 28 '15 at 15:22
  • \$\begingroup\$ Ah, I'd missed that the question talked about OPA541 and 30A but the diagram showed OPA551. \$\endgroup\$ – pjc50 Sep 28 '15 at 16:13

In general the first consideration is what happens if one power amp has a (slightly) different output level than the other one. Will there be a current flow from one output into the other one? If yes, will this damage one or both components?

It is usually better to just use one op-amp and use a discreet FET, Transistor or IGBT for the power part.


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.