OPA552: External Compensation for Stability Below 5dB gain

Hello Shant,

I see the issue you mention regarding the datasheet statement, "For the values shown in Figure 31, the f–3dB is approximately 956 kHz. This frequency is less than that predicted by simply dividing the GBP by NG1." And certainly, if the OPA552 GBP is 12 MHz as listed in the Electrical Characteristics table and NG1 = 2, the frequency we arrive it is 6 MHz which isn't close in any way to 956 kHz.

I searched the archives and found the original OPA552 datasheet from 20 years ago and it provides the same information regarding compensating the OPA552 as today's section 8.2.2.3. Unfortunately, this particular form of noise gain compensation doesn't appear to be documented elsewhere for the OPA552. About the only assumption that can be made is the GBP divided by NG1 might be a typo that has carried through revisions.

The OPA552 typical GBP is listed as 12 MHz. If the original intended relation for the bandwidth was instead GBP divided by NG2, then 12 MHz/10 V/V = 1.2 MHz, which is closer to 956 kHz but still different enough to raise questions about its origin. I suspect that since 956 kHz is a very specific number that it is the result f an actual bench measurement of the bandwidth.

Usually noise gain compensation involves a resistor in series with a capacitor from the op amp inverting input to ground so this form of noise gain compensation is different than what I have seen before. I went ahead and set up the Figure 31, Compensation of the OPA552 for G = 1, circuit in TI's TINA-TI Spice simulator to test the phase margin. The phase margin analysis is based on the information in TI's Precision Labs - Op amps series. Do note that the OPA552 simulation model is dated and less sophisticated compared to today's modern models so the results aren't exact, but the results should be close to what can be expected from the actual op amp circuit. The open-loop test circuit and Bode plots are shown here:

The simulation indicates that the phase margin is about 56 degrees for a resistive load. This result verifies that the OPA552 noise gain compensation is working well and stabilizes the circuit when the op amp is operated in a noise gain of 2 V/V. When the circuit is simulated closed loop the cutoff frequency is 1.14 MHz, which is very close to a GBP/NG2 value of 1.2 MHz. That tends to support that this may have been the original intended relationship.

I considered your gain case of -4 dB (NG1 = 1.63 V/V, in v G = -0.63 V/V) and set up the original OPA552 circuit using the assumptions in section 8.2.2.3, except NG2 was assumed to be 1.2 MHz. I ratioed the feedback resistor for NG1 = 1.63 V/V and applied the same noise gain pole created by RF and CF, NG (pole) = 1/(2 pi RF CF), which was 765 kHz in the original example. The feedback resistor was set to 630 Ohms and feedback capacitor to 210 pF. Since the NG1 is 1.63 V/V, and it was decided to set NG2 to 5xNG1 (like the original example) or in this case 8.15 V/V. Multiplying CF by 8.15 results in a Cs of 1.7 nF. A TINA-TI simulation was performed and a phase margin of 47 degrees was obtained. Anything above 45 degrees is quite optimal.

Since we don't have a complete picture regarding this particular datasheet noise gain technique used for the OPA552, you may want to consider a compensation method that is more fully documented. Here is a link to a TI AAJ article that goes deeper into noise gain compensation:

http://www.ti.com/lit/an/slyt174/slyt174.pdf

Regards, Thomas

Precision Amplifiers Applications Engineering

Thomas,

Thank you for taking the time and running these simulations. I had a feeling this was a typo. 

I would like to ask for confirmation on the capacitor values I calculated for the following gains for sufficient compensation at this 1.2MHz pole (210 pF).

2 times voltage gain in inverting mode (Rg= 1 k, Rf= 2 k): Cs- 1.6 nF 

3 times voltage gain in inverting mode (Rg= 1 k, Rf= 3 k): Cs- 1.4 nF

4 times voltage gain in inverting mode (Rg= 1 k, Rf= 4 k): Cs- 1.3 nF

Questions:

1- What was the reason for adding a series feedback inductor "L1" in the simulation?

2- Was there any particular reason for changing the Cf value from 208 pF to 210 pF between the simulations you performed?

3- Is this formula of compensation applicable to both inverting AND non-inverting configurations?

4- We will also be running the OPA552 at gains up to 10 times V/V. Do you anticipate any issues by operating in inverting mode at these levels? Id like to know if one mode will be more beneficial than the other at these higher gains?

Regards,

Shant K.

Finished some other things, and came back to this to try a LG sim how I would do it. 

This approach traces the signal around the loop and adds Ccm+Cdiff at the summing node where LG is measured. Here, I get a good sim but cannot map LG=0dB crossover frequency and phase margin to closed loop BW very well. That approach in Insight #5 applies to a 2nd order LG analysis - I would say the OPA552 model is not that simple by any means so the 2.1MHz LG=0dB is confusing vs. closed loop 1.55Mhz F-3dB, but the phase margin is encouraging at 52deg. 

And here is this file, 

OPA552 inverting comp LG.TSC

Hi Shant,

the reason for adding "L1" in the simulation is explained in these TI's "stability" training videos:

Kai

Hi Michael,

Thanks much for jumping in and providing all of the relevant information regarding this compensation technique you developed for decompensated op amps.

Regards, Thomas

Precision Amplifiers Applications Engineering

You bet Tom, 

So I was publishing and presenting all that circa 1997 inside BurrBrown, but no one bothered to ask my inputs for that section of the OPA552 circa 1999 - oh well, probably more than they wanted to get into - I don't think I developed the simplified discussion until early 2000s

Hey Kai, incidentally ---

I tend not to send out this stability links currently in this TI series - while there is a lot of good information in these generally, and this one in particular, they have kind of institutionalized an obvious error in their LG setup circuit shown here, 

This approach might work a lot of the time, but not all of the time. It is ironic with all the recent effort put into modelling the open loop output impedance that this approach isolates that element from the feedback network. I actually showed how far off this approach can be in a relatively simple OPA837 circuit in this article, starting in the section titled "Simulation approaches to estimating nominal phase margin". Really too bad they printed up all those "Analog Engineers Pocket Reference" booklets with this error in there. 

https://www.planetanalog.com/stability-issues-for-high-speed-amplifiers-introductory-background-and-improved-analysis-insight-5/

Hi Michael,

I'm very happy to read all this! I have never understood why isolating Zo from the feedback loop components should give correct results in the phase stability analysis. I always that I'm too dumb to fully understand what is going on with all this magic TINA-TI stuff :-)

Thank you very much for helping me to fix this painful gap. You are a really brilliant head!

I also enjoyed this article very much because it demonstrates how to handle biasing components connected to the +input of OPAmp in the phase stability analysis:

By the way, what do you think about paralleling input bias current cancelling resistors with caps? You could have both then, a big enough resistor to fully cancel the input bias currents and a low impedant path to signal ground for higher frequencies to keep the noise low and to improve stability. I'm only talking about standard VFAs here. Is this a good idea or have I overlooked something which could result in stability issues?

I'm looking forward to your answer :-)

Kai