I'm trying to measure the feedback loop on a isolated flyback SMPS. My goal is to determine the phase margin and 0 dB cross-over frequency.

The most common circuit I have seen is the one recommended in e.g. this article, figure 10:

Measurement circuit for loop response

An isolation transformer is used to apply a small offset voltage to the feedback path. The gain and phase is then measured as the ratio between sine waves at points A and B. This seems to work well for frequencies above the loop cross-over frequency.

However at lower frequencies, the feedback action itself is causing channel B signal to vanish.

Before applying power to the converter, I see a sine wave on point B while point A is kept at 0 volts by the output capacitors:

Signal when power off

As soon as I turn on the power, the feedback loop acts to eliminate any wave on point B, while point A now shows the expected nearly-180° phase shift to the wave from previous image:

Signal when power on

Question: What am I doing wrong? Should I somehow deactivate the feedback loop before performing the measurement? Should I compare point B with power off against point A with power on?

  • \$\begingroup\$ Power off measurements are not relevant. Where is the plot of CHa/CHb? It'll be fun to see what CHa/CHb looks like when channel B is momentarily 0 volts. That article I note was written by someone in marketing (just saying). \$\endgroup\$
    – Andy aka
    Commented Oct 11, 2023 at 13:37
  • \$\begingroup\$ @Andyaka Yeah, I guess it is not literally A/B, but rather as the division of complex numbers / phasors representing the sine waves. But as you'd expect, even that has rather random values when B is very small. In this siglent appnote they seem to have reasonable results by adjusting the stimulus amplitude, but I don't seem to get reasonable signal at channel B no matter what I do. \$\endgroup\$
    – jpa
    Commented Oct 11, 2023 at 13:50
  • \$\begingroup\$ The trouble is... does anyone know what it is and does everybody have that Sigilent scope? My method is simpler; a sudden load discontinuity and look for ringing and the length of time ringing occurs for and, compare with a 2nd order filter response to estimate phase margin. \$\endgroup\$
    – Andy aka
    Commented Oct 11, 2023 at 14:18

1 Answer 1


Turns out this was a combination of two factors:

  1. My loop response was way out from what I was expecting. The loop gain at the frequencies I thought were near the 0 dB crossing was actually more than 40 dB. And what I though was the 0 dB frequency (because A and B signals had equal amplitude) was actually the frequency where phase shift got to 0. I'm not sure why it didn't start oscillating wildly.

  2. In this measurement, the larger amplitude is limited by the transformer output voltage. The smaller amplitude is then determined by the loop gain. So with a 100 mVp-p signal on point A and 40 dB loop gain, I could expect a 1 mVp-p signal on point B. It simply gets lost under the noise.

Raising the stimulus much above 100 mV starts to affect circuit behavior too much, and commonly signals below 10 mV vanish behind the SMPS noise. This gives applicable measurement range +- 20 dB for loop gain. That's usually the most interesting range for stability.

For values out of range, you can determine:

  • Point A signal vanishes: Loop gain below -20 dB
  • Point B signal vanishes: Loop gain above +20 dB

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.