0
\$\begingroup\$

Any references on how to go about modelling an AC/DC mains isolated H bridge buck converter in s or z domain? All the current resources online point to modelling buck converter with a single switch and in non isolated configuration.

The current process that I use for compensator design is that I look out for the frequency region where the LC low filter cutoff happens and at that region, the phase response drops almost to -180 degrees. Then I would design a phase lead lag compensator with some phase being added in the frequency region where the phase response drops to -180 degrees. Then I use the same compensator in the SMPS design without taking into account other components present. Wanted to know how much off I would be if I make this assumption that other components in the converter doesn't add much to phase response?

\$\endgroup\$
0
\$\begingroup\$

There are several options to get the control-to-output transfer function you want:

  • use an averaged model, a half-or full-bridge converter is a buck-derived topology and can easily be simulated with the PWM switch model in voltage- or current-mode control.
  • use a simulator which extracts the ac response from a cycle-by-cycle simulation: Simplis or PSIM do that and their demo version will allow you to simulate this simple structure in open loop.
  • use a small-signal model and apply the FACTs.

For a CCM full-bridge operated in voltage mode, the dc gain is that of a forward converter: \$H_0=\frac{NV_{in}}{V_p}\$ with \$N\$ the transformer turns ratio and \$V_p\$ the sawtooth peak voltage. The rest of the transfer function (poles and zero) is similar to that of the buck. You can look at seminars taught at APEC regarding small-signal modeling for both VM and CM.

I have captured a quick sim in Simplis (the demo version, Elements) and if you simulate the below 5-V/10-A circuit:

enter image description here

you get the transient waveforms and the dynamic response of the power stage. You see a gain equal to \$\approx\$23 dB which what the above formula gives (24-V input source with a turns ratio of 0.6). Updating this circuit into a CM mode would not be difficult either.

\$\endgroup\$
  • \$\begingroup\$ Great. Thanks. It cleared my doubt. I am planning to control the converter digitally using MCU. Can I replace sawtooth voltage with something else ?. Double poles exists at the LC filter resonant frequency ?. \$\endgroup\$ – renganathan b.s Feb 16 '18 at 8:58
  • \$\begingroup\$ The given example is a voltage-mode (VM) converter. The sawtooth is part of the pulse width modulator or PWM. You can digitally build the PWM pattern or consider a current-mode (CM) converter in which this sawtooth is replaced by an image of the inductive current. In the VM converter, yes, the \$LC\$ network peaks and compensation needs a type 3 circuit. In CM, the low-frequency response is of 1st order and compensation is simpler (type 2). However, subharmonic poles at \$\frac{F_{sw}}{2}\$ need to be damped. \$\endgroup\$ – Verbal Kint Feb 16 '18 at 13:07
0
\$\begingroup\$

You can model it in a simulator and, you don't need to model it as a switching converter because a linear model will work. If your switching frequency is (say) 100 kHz you could make an argument that modelling a delay of 10 us is useful to add.

Modelling it as a linear regulator has the main advantage that you can run an AC analysis and produce a bode plot.

May I recommend Micro-cap 11 student edition or if you want to go down to the bargain basement try LTSpice.

In Micro-cap you can model in the s-plane directly using formulae. I'm not sure if it will model in the z-plane because I've never tried but if you do go this route please make an answer to your own question. It will be useful to others I'm sure.

\$\endgroup\$
  • \$\begingroup\$ Thanks I will take a look. Suggestions looks very simple but yet effective assumption for first level design. I haven't tried z plane myself. Will update if I find something. \$\endgroup\$ – renganathan b.s Feb 16 '18 at 9:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.