5

I am looking for an op amp to use as a unity gain buffer with a disable feature. The OPAx890 looks like a good candidate because it is unity gain stable.

I have noticed that some op amps labeled as unity gain stable have worse stability than others (e.g. OPA690 claims unity gain stability, but its stability is poor. The G=+1 small-signal frequency response shows >3dB of peaking in the datasheet).

The OPA2890 has a very nice looking small-signal frequency response in a unity gain configuration: enter image description here

I do not need a dual op amp. I was glad to find that there is also the OPA890 which looks like a single version of the OPA2890. It looks like the OPA890 has worse stability with some peaking at G=+1:

enter image description here

Does this make any sense? I would think the same op amp would have the same frequency response, or for the single to have better properties than the dual. Of course I can just get the dual and use 1, but is this an actual difference between op amps or just an error in the datasheet?

DavidG25
  • 1,292
  • 8
  • 21
  • 1
    How can there be stability issues when you use it as a follower? What does the word _stability_ mean to you? Under which frequencies? – Harry Svensson Aug 10 '17 at 17:29
  • Both are unity gain stable but one has a phase margin smaller than the other. – Andy aka Aug 10 '17 at 17:35
  • @HarrySvensson by stability issues I mean that when you hit the follower with a step response it will overshoot and ring. By stability I mean phase margin and the more the better. – DavidG25 Aug 10 '17 at 17:36
  • Woaw, that's a new definition of stability. I don't like that. Over to your problem at hand, isn't that something that can be solved with a small capacitor to alter the phase margin? – Harry Svensson Aug 10 '17 at 17:38
  • @HarrySvensson I'm not sure that's a new definition for stability. Typically I've heard that phase margin > 45 deg is stable and phase margin < 45 deg is unstable, but it's not really binary like that and the more phase margin the more stable. – DavidG25 Aug 10 '17 at 17:40
  • Having said what I said above they have the same graphs in each data sheet so TI are telling porkies somewhere. (http://www.urbandictionary.com/define.php?term=telling%20porkies) – Andy aka Aug 10 '17 at 17:42
  • 1
    Frequency response setup differs for these two graphs: one was done with 0.1V input, the other with 0.5V input. Could slew rate be involved with these different results? – glen_geek Aug 10 '17 at 17:45
  • @glen_geek nice catch, but the 890 has a smaller output swing and more peaking. Would have made sense if the peaking was caused by a larger output swing. – DavidG25 Aug 10 '17 at 17:48
  • You gonna use a GND plane, right? – analogsystemsrf Aug 11 '17 at 04:45
  • @analogsystemsrf ya decoupling caps, short traces – DavidG25 Aug 11 '17 at 06:01

2 Answers2

6

Does this make any sense?

Yes it does, from what I've seen they manufactures can use different silicon sometimes for a dual vs a single or quad part. But without actually seeing the die layout there is no way to really know (you could get some nitric acid and microscope -- be careful-- and find out for sure).

I would think the same op amp would have the same frequency response, or for the single to have better properties than the dual. Of course I can just get the dual and use 1, but is this an actual difference between op amps or just an error in the datasheet?

The manufactures datasheets should reflect actual testing of the parts, and purchased parts should reflect the test data seen in the datasheets.

The difference probably a result parasitics, maybe the die wires are shorter in the OPA2890 since they might have put two separate dies in the same package. Or it could be due to more capacitance in the OPA890. Its all in the parasitics of the package or testing, or both. I'd trust the datasheets unless you want to get a part and do the testing on your own, the datasheets should reflect actual testing done by the manufacturers engineers.

Voltage Spike
  • 75,799
  • 36
  • 80
  • 208
  • This makes a lot of sense. Same design but with different parasitics would give different results. I wish I could choose both your and Null's answer, but I'll just give you +1 for the good info. – DavidG25 Aug 10 '17 at 17:44
5

It looks like there is an actual difference between the OPA890 and the OPA2890. For example, the parametric table on the OPA890's description page shows that the OPA890 has a slightly higher GBW (260 MHz vs. 250 MHz), higher slew rate (500 V/µs vs. 400 V/µs), lower offset drift over temperature (15 µV/C vs. 35 µV/C), etc. The electrical tables in the two datasheets show more specific differences as well. The OPA890 is a bit faster than the OPA2890 so it looks like the two figures in the datasheets are correct.

Null
  • 7,448
  • 17
  • 36
  • 48
  • Okay so they're just different amplifiers and I was thrown off by the part number and the little table at the top that shows "related op amp products". – DavidG25 Aug 10 '17 at 17:38
  • @DavidG25 Well, they *are* related and probably share mostly the same design. But it looks like for some reason they are specified slightly differently. – Null Aug 10 '17 at 17:40