I am currently designing a flyback converter and I am interested in measuring the stability of the converter (with a Bode plot). I will try to connect an isolation transformer on the feedback loop that inject a small AC signal and measure the transfer function with an FRA.

However, I'd like to compare my measurements to a simulation model. In order to do that, the idea is to make a small-signal model of my converter on LTSpice.

I have seen on internet different models of a flyback converter, but My questions related to this subject are:

  • Is it possible to model the whole flyback converter, including the IC controller? Does one has to do each model of each controller or is there a pole/zero/delay/... equivalent?
  • Are stability models created in the industry of the IC controller usually does the job (integrated compensation, protections,...)?
  • A TL431 is often used for FB + optocoupler. I have seen equivalent models with poles, how about the model of a chip with integrated feedback like the Innoswitch family?
  • The book of Christophe Basso seems to cover this subject, has anyone successfully made a model based on his book?

Thanks a lot for your advices. Regards.


2 Answers 2


The answer given by Antonio51 is good and uses the auto-toggling CCM-DCM model built on the PWM switch in 2005. This is great to determine the small-signal response of many switching circuits and not only a flyback converter of course. What is cool is that the subcircuit determines the operating mode itself and works for ac but also dc and TRAN large-signal analyses. The model has been ported to LTspice in many different flavors and I posted a few on my webpage.

Let me add another option for determining this transfer function. In the example documented by Antonio51, this is an averaged model in which the switching component has purposely been neglected as we are interested by the average component values. So if you want to analyze a circuit, you first need to identify the averaged model configuration and then run a separated ac analyses with confirmation that all bias points between the switching and the averaged version are identical (or very close) to confirm the circuit and validate the ac results. Depending on the circuit, the exercise can be long - but nothing insurmountable though - and some structures such as resonant converters don't have an averaged model. This is because energy is conveyed by the switching fundamental and its harmonic that we purposely ignored in an averaged model.

In this case, one cool thing is to resort to a piece-wise linear (PWL) simulation engine such as PSIM or SIMPLIS. These programs let you extract the ac response from a switching circuit of any kind, without resorting to an averaged model. For your information, LTspice also does offer this possibility with a built-in frequency-response analyzer (FRA) but I don't think SPICE lends itself well to doing this type of exercise swiftly. Look at the below isolated current-mode flyback converter implementing a TL431:

enter image description here

This a circuit excerpted from the free 60+ ready-to-simulate simulation templates I posted with the release of my last book covering the determination of transfer functions for switching converters. These platforms are described here, from my webpage. And what is interesting is that most of these templates work with the free SIMPLIS demo version. If you press the start button, then you'll have the switching operating point and the ac responses in few seconds:

enter image description here

You will need to model your PWM controller internals to reproduce the feedback path and account for division, filters etc. Here, nothing complicated beside a divide-by-three block and a clamp. You ignore all protections as they are silent in normal operation. You can also add delays if you deal with fast dc-dc converters but for a flyback converter, as you limit crossover because of the RHP zero, propagation delays can be ignored safely for loop analysis I would say.


microcap v12 has some example of these models.

Flyback VM model.

enter image description here


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.