Opto-couplers in smps...

Sureshot way : Use a pulse width modulator to transfer the amplitude information across the optocoupler; average it out on the other side or better still, use the time domain information. That way the optocoupler is always operating either at full power or zero. And hence highly INsensetive to varying gains. Or you could use a Hall effect DC current sensor (also called a DC Current Transformer / DCCT) to Ohmically isolate the two sections and still transmit the required data.

--- Anirban

Simply specify an optocoupler with a more tightly spec'd coupling ratio. They exist for just this reason. Paul Mathews

It changes with time.

Best regards, Spehro Pefhany

Way to go, at least when riding close to the cliff in terms of loop stability.

Another simple method is to use a little RF transformer instead of the opto coupler and transfer a carrier where the amplitude is proportional to the input signal. I often use 13.56MHz for this kind of stuff, less worries at the EMC lab :-)

Hide quoted text -

Of course, I agree that you have to design around the non-ideal characteristics of optocouplers, but they are out there in the world, functioning reliably in the billions of units. The drift due to aging is toward less coupling, and is one of many factors that the designer must consider when deciding how much margin to require. Transformer isolation is a good alternative, and it is often a requirement in space and avionics electronics, with optos not allowed. In addition to using various pulse width feedback schemes, there are plenty of examples of amplitude feedback thru transformers, including one in the TI/Unitrode Seminar Notes, IIRC. Paul Mathews

rd"

com

f=2Ecom-Hide quoted text -

Does anyone of you have an example of how to implement a pwm feedback using opto? How to integrate the pulses to achieve a correct "copy" of the original error signal. This must require a carrier frequency more than 10 times the switching frequency of main converter?

I appreciate any advice.