I wonder about the "quality difference" between using an secondary side reference + optocoupler to provide voltage feedback in an isolated flyback design and using the third winding of the transformer.

As far es I understand things, "better" regulation is possible with first variant because the tolerances of voltage reference and optocoupler can be much tighter than the tolerance of the transformer winding ratio from secondary side to the third winding.

Now assume that i am happy with an output voltage regulation of +/- 10% , are there other benefits than better output voltage regulation when using the optocoupler variant? I heard that load regulation should be better but don't know why this should be the case.

I also saw that in the variant with third winding, the voltage is more or less (rectification) fed back to the controller FB pin whereas with optocoupler feedback you can directly "bypass the controller's error amplifier" because the optocoupler's feedback is a current signal that can be fed into the output of the controller's error amplifier(i.e. the compensation network). I don' t know if this is somehow a benefit.

NOTE: All the text abobe assumes an isolated flyback design with current mode regulation and a constant frequency controller. Also, I assume to only have one output, so please no explanations concerning cross regulation topics.


1 Answer 1


Load regulation will tend to be much worse with transformer feedback because the windings are not perfectly coupled - there is leakage inductance in a real transformer. This can be minimized by keeping the feedback and secondary windings as close as possible but there are limits to that since galvanic isolation depends on the insulation between the two windings. To put it bluntly, if the wires short, people could die.

The voltage across the output rectifier diode(s) will also increase with increasing load current and the average voltage across the filter caps will decrease a bit, and there will be drop due to non-zero output winding resistance. None of these changes occur in the feedback winding so they will not be automatically compensated.

You can see these effects in DC-DC converters where only one output is regulated, an loading the regulated output causes the other outputs to increase in voltage. With minimum loads applied this can be tolerable, but it is very noticeable. It really is a cross regulation thing- the feedback winding is another (lightly loaded) output that happens to be tied to the primary side.

It may be possible to (partially) compensate for these factors by fiddling the feedback a bit with increasing primary current, and I believe some controllers adopt this strategy to avoid the opto. Another approach is to follow the poorly regulated output with an LDO linear regulator.

  • \$\begingroup\$ Thanks! I understand that the contributing errors are:. Diode voltage (because not constant), transfomer inductance tolerance between second and third winding (because (fixed) error of feedback voltage), secondary side transformer ESR (because non constant voltage drop). Concerning the diode voltage: it is of course a function of current, but looking at diode curves, shouldn't the diode conduct "fully" (more or less)? In this region the change in voltage over current should be rather negligible considering an desirable 20% output voltage regulation accuracy? I will do a simulation at home ;-) \$\endgroup\$
    – Junius
    Commented Oct 29, 2016 at 11:51
  • \$\begingroup\$ Diode drop is a more important effect with a 3.3V output than a 12V output. \$\endgroup\$ Commented Oct 29, 2016 at 11:52
  • \$\begingroup\$ Yeah, i plan to use a 12v output. But a very quick simulation few days ago showed good performance when non-isolated, but extremely bad ( essentially no regulation) when done via third winding. Is it ok, if i send you a LTspice simulation once i have refined it? Another thing I am not sure about: with optocoupler feedback, the information about output voltage is present of the controller feedback pin all the time. With third winding feedback, the information is only available when energy is transferred to the output. I guess this is ok because only in this interval the voltage is of interest? \$\endgroup\$
    – Junius
    Commented Oct 29, 2016 at 12:01
  • \$\begingroup\$ Add a question here and someone will help. You don't want to pay my retainer and consulting fees, \$\endgroup\$ Commented Oct 29, 2016 at 12:04
  • \$\begingroup\$ The nature of smps feedback is that there is a dead time of some part of a cycle. Nothing you can do about that for a given frequency and topology unless you are feeding back output current, which is rare. \$\endgroup\$ Commented Oct 29, 2016 at 12:11

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.