2
\$\begingroup\$

When using a level 2 UniversalOpamp in a unity gain capacitive drive follower application, I find that it is quite stable.

I decided to compare the results to an actual opamp model of the LM7321. I matched the open-loop gain, the GBW product, the output current, the phase margin. Surprisingly, the actual model shows much inferior stability. The ESR of the cap is 1 Ohm.

Is the UniversalOpamp a bad model for such stability analysis ? Or would you infact trust the UniversalOpamp more than the models which often have strange stuff going on.

enter image description here

enter image description here

\$\endgroup\$
9
  • \$\begingroup\$ In LTSpice, if the part isn't tied to an actual circuit model of a real opamp (same goes for FETS, BJT and diodes) then it's an ideal model. Meaning, don't trust it to model real life behavior. \$\endgroup\$
    – Aaron
    Oct 14, 2021 at 13:47
  • \$\begingroup\$ Actually the same goes for RLC too, but they have a lot fewer issues, and are easy to add parasitics to. \$\endgroup\$
    – Aaron
    Oct 14, 2021 at 13:49
  • \$\begingroup\$ Have you considered that the LM7322 is driving much too much capacitance and that this sort of thing is bound to happen. The DS implies no more than about 1 nF. \$\endgroup\$
    – Andy aka
    Oct 14, 2021 at 14:03
  • \$\begingroup\$ @Andyaka How so ? The datasheet says 'unlimited calacitive drive' in many places. \$\endgroup\$
    – tobalt
    Oct 14, 2021 at 16:31
  • 1
    \$\begingroup\$ @Aaron Agree for Diodes, transistors and RLC. But for more complex parts my experience actually is that the part models do a lot of strange (overcomplicating?) things and that real-world seems to be more benign. That's why I indeed consider that the nom-ideal (!!) UniversalOpamp could be closer to the truth than the manufacturer lib. \$\endgroup\$
    – tobalt
    Oct 14, 2021 at 16:37

1 Answer 1

2
\$\begingroup\$

If you'll look at the examples/Educational/UniversalOpamp2.asc you'll see that there are 4 levels for the UniversalOpamp2:

  • level1 is a basic VCCS and an R||C output
  • level2 is a single pole opamp
  • level3a is a two pole opamp and programmable phase margin
  • level3b has an additional dominant pole and delay

You used level2, so a simple opamp with a single pole, compared to the model for the 7322 -- which I don't know how it is (haven't looked), but I'd wager it's more than a single pole. Therefore if you want more meaningful results you should use level3a or level3b. And, yes, the UniversalOpamp2 can stand its own against many opamps, when properly tweaked. In fact, personally (and there are others who agree) I'd recommend it over many models of opamps.

Of course, everything is SPICE, therefore an approximation, even the 7322, so take all the results with some grains of salt.

\$\endgroup\$
2
  • 1
    \$\begingroup\$ Thanks for the suggestion of level 3. I was wondering why the phase margin property of the UniversalOpamp never seemed to have an effect. After trying a bit, I can sort of reproduce the stability behavior of the LM7322 with a level.3 op-amp; however, with quite inferior values for the phase margin than those given in the datasheet. The Aol und GBW are worse, too. So my hypothesis is that the op-amp model might present some worst case specs (which isn't bad for design), but explains why a naively set-up UniversalOpamp outperforms the LM7322 model. \$\endgroup\$
    – tobalt
    Oct 14, 2021 at 20:05
  • \$\begingroup\$ @tobalt Don't forget that it's meant to be universal, so it will (probably) never be able to precisely match every extremities you'll find, but it may come close. It will outperform many opamps in .TRAN, though, since it has only 7 internal nodes. Not lastly, check the 7322 against the datasheet, as well, because too often opamp models don't bother to implement every bells and whistles under the hood, and that may come at the expense of some other bells and whistles. E.g. the more acute peaking may be due to higher order poles (or ladder approximation, still haven't looked). \$\endgroup\$ Oct 15, 2021 at 7:00

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.