OPA277: OPA188 oscillation issue

Well Brian, why would C27 be hit and broke? In any case, Kai is right that creates an unstable circuit. Without C27 you essentially have a postive feedback configuration in the SKF where C27 normally shunts that feedback path off at higher F. If C27 can get broke, you should expect this circuit to oscillate. 

Also, 

60deg is a very conservative target, close to a Butterworth response in amplifier stages. I recall that 65.5deg phase margin is exact Butterworth. 

45deg was actually the nominal target for a lot of national op amps to get a bandwidth extension for marketing purposes. 

I use, 

>30deg, probably ok 

20 to 30deg, a little concerning - if easy to raise above 30deg do so

10 to 20deg, absolutely need to work on it to raise above 20deg min, 30deg preferrable. 

<10deg, you should expect oscillation in production and over temp. 

I think I put some of that in this article, and then #6 is for VFA specifically. 

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

Hi Michael,

30° phase margin can result in an overshot of 40% in the step response as you show in your nice article. Many time domain applications cannot tolerate this. Even the 23% overshot with a phase margin of 45° can be too much.

Also, a higher phase margin in the simulation compensates for unexpected board stray capacitances (especially helpful in high ohmic feedback loops), load capacitances and imperfections of the Spice model itself. Think only of the many models which do not properly model the complex output impedance of OPAmp, which you also discuss in your article 

Kai

All correct Kai, 

As I said in the article what is acceptable is 1st of all in the context of the application. Where I ran into this a lot was in the context of MFB filter using FDA's 

The newer FDA's are RRout and very reactive Zout - you can implement a very nice LP MFB where the overall loop is maybe 20deg phase margin, and that will be just fine as that is way out beyond the passband, 

Just one of many examples of this is a redesign I did (for myself kindof) for the THS4561 front page MFB filter - the original one I did oscillated, but here I fixed that, 

THS4561 front page MFB updated ckt Aug2018.docx

Hi Kai,

Thank you very much for your detailed explanation.

And I still have two other questions:

1. I saw the phase margin is not enough, but the input is DC signal.

    Is that means if the phase margin is not big enough, no matter what the input is, the circuit will always oscillate?

2. My customer use the right circuit in their application. And only in very few cases the C27 can be hit and broke, for example when there is a violent impact.

    So we want to choose an opamp that when C27 is broke, the circuit can still operates normally.

   Would you please help me to find which spec is the most important ?(Which spec makes OPA140 has a 20 more phase margin than OPA277?)

Thanks,

Brian

Hi Michael,

In fact, this is an automotive application, there are often crashes.

And my customer find that in some crashes, the C27 will be hit and broke. And there will be oscillation.

So I use this EVM board to do these experiments.

And I also find if I removed C29 at first. No matter C27 broke or not, the circuit will always operate as normal.

Would you please help me to analysis this phenomenon?

Thanks,

Brian 

Hi Brian,

Brian Li said:

1. I saw the phase margin is not enough, but the input is DC signal.

    Is that means if the phase margin is not big enough, no matter what the input is, the circuit will always oscillate?

Yes, it will oscillate even if you only want to handle a DC signal. When the circuit is instable because of a too small phase margin even the least noise in the circuit coming from the feedback resistances and the OPAmp itself is enough to result in oscillation. This noise is broadband noise and contains all frequencies, including the frequency at which phase margin is too small and the condition for maintained oscillation is fullfilled. The step signal from the power-up ramp of supply voltage also contains all frequencies, including the oscillation frequency, which can make the instable OPAmp immediately oscillate after power-up.

Brian Li said:

2. My customer use the right circuit in their application. And only in very few cases the C27 can be hit and broke, for example when there is a violent impact.

    So we want to choose an opamp that when C27 is broke, the circuit can still operates normally.

First, I would try to find out why the cap gets damaged at all. I say this, because if the cap is in danger to be destroyed then all other components are in danger to be destroyed as well. Most SMD components get damaged because of board flexure caused by vibration or violent impacts. Then encapsulating the whole board by a suited potting compound can help. Another remedy is the use of a ceramic cap providing soft-termination:

https://www.tdk-electronics.tdk.com/en/374108/tech-library/articles/products-technologies/products-technologies/automotive-grade-soft-termination-and-megacap-type-mlccs-with-c0g-characteristics/1952460

Internal cracks can also be avoided by taking a leaded capacitor (through-hole) or by taking a plastic foil capacitor. Mounting several caps in parallel (three 3.3nF caps, e.g.) instead of using one single cap (10nF) can help to avoid instability even in the case that one or two caps gets damaged. Another remedy is to avoid the active low pass filter topology using this "critical" positive feedback at all and to only use passive low pass filtering.

But in any case the customer should do all he can to prevent the cap and other components from being destroyed at all.

You see that I don't want to recommend an OPAmp for the circuit with damaged cap because I think that the phase margin will always be too small equally which OPAmp he takes. Also, I don't think that the customer will be satisfied with the change in the frequency response of his low pass filter when the cap gets damaged:

Kai

Hi Kai,

Thank you very much.

This is an automotive application, there are often unavoidable crashes. 

And I have another question, I have seen the videos on TI precision labs, they taught me to simulate like this:

And I saw your circuit is:

Would you please teach me why simulate like this?

Thanks,

Brian

Hi Brian,

this method is similar to what is shown in the TI's training video series on stability:

https://training.ti.com/node/1138805

"My" approach is originally from the TI's former employee Tim Green.

I introduce a stimulus directly to the input pins of OPAmp and look what comes back via the feedback loop. From the frequency response and phase response I can directly determine the phase margin.

Because the feedback loop is opened at the input pins, the feedback loop no longer sees the input capacitances of OPAmp ("C4", "C5" and "C6"). Because of that the input capacitances have to be "mounted" externally.

"L2" and "L3" close the feedback loop for DC and allow the OPAmp inputs to be properly DC biased.

"C3" provides an AC coupling of stimulus without ruining the DC biasing of OPAmp inputs. And because the AC coupling shall be invisible even for the lowest frequencies "L2", "L3" and "C3" are chosen to be "infinitely" high.

The advantage of this a bit more opulent method is that the complex output impedance of OPAmp is also taken into calculation. See Michael's link.

Kai

Dear Kai,

Thank you very much for your detailed explanation.

It really helps me a lot.

Brian

Hi Kai,

I find a new question about phase margin.

When I simulate the circuit with OPA365 and without C2.

The AC transfer characteristic is as below:

When the Gain=0dB, the phase margin is 34deg.

But the phase margin can be 1deg when the Gain is 19dB.

I wonder can this 34deg means the circuit is stable?

Hi Brian,

sorry, I have this answered here:

https://e2e.ti.com/support/amplifiers-group/amplifiers/f/amplifiers-forum/1170725/phase-margin-issue

"I would say that the circuit is not stable.

Because of manufacturing tolerances fo OPamp and imperfections of the model, the phase margin should not only be >45° exactly at the point of 0dB gain (122kHz), but also in the immediate surroundings. And because the phase margin goes all the way down to 0° at 100kHz, I would say that the circuit is not stable.

Can you post your TSC-file and tell more about your application?"

Kai