How do isolated DC/DC converters get feedback from output?

Many put the error amplifier and reference that the output is compared to on the output side of the isolation and use an opto coupler between the error amplifier's output and the control circuitry on the primary side. The opto has a big tolerance, but this approach puts that tolerance in with all the other tolerances of conditions and components inside the error amplifier's control loop, not between the reference and the output voltage in the error judgment part of the process.

Some designs put the controller on the output side, with transformers to control the driver on the input side and a driver that will self- oscillate before the driver comes on-line.

The 2004 ARRL Handbook has such a design for a 12V, 100W, off-line converter.

Take a look at "FOD2741" opto isolated error amplifier.

Mike

Hmm I had thought about doing it that way but decided that the power on condition would be annoyingly complex to deal with. You don't happen to have a schematic handy, do you? It would be interesting to see how others dealt with that.

Also - what exactly does "off-line" mean in this context? I occasionally run into that term but am not familiar with it.

Thanks,

-Michael

Three ways:

1. Use a seperate winding. This is a typical solution with flyback converters like the 3842. Since the flyback pulse's peak voltage is proportional no matter what winding you're on, it works pretty well. Leakage inductance needs to be low enough, which isn't usually a problem in the 100kHz range commonly used. In typical use, you might see a UC3842, MOSFET and transformer (in that order) in your computer monitor's power supply.

1. Put the oscillator and error amplifier at the output. Use a coupling transformer to drive the switches. This begs the question, what starts it? Sometimes the switching circuit is arranged so it kick-starts itself, others use a seperate free-running oscillator. You often see three ferrite transformers in computer supplies for this reason.

2. As mentioned, optoisolators running in the linear range. One can be used, but two are better, making a local feedback loop that compensates for its own error. This preserves the value, instead of having to account for all conditions of the opto, which changes with manufacture and age. I've seen a flyback type chip with integrated MOSFET, I think it was a FANxxxx something, with feedback via opto.

Tim

-- Deep Fryer: A very philosophical monk. Website @

I don't have a schematic in electronic form -- that's why I referenced the Handbook.

"Off-line" in this context means that the supply is designed to rectify the AC from the wall and generate isolated regulated DC straight from that, with no intervening linear supplies. It's what's in your PC and all the newer wall-warts.