48VDC from 120VDC needs to be isolated

if so, you need to locate, and then read, the regs

in fact, and in actuality, you get a better converter if you use a transformer ...

in fact, and in actuality, you get a better converter if you use a transformer ..

Click to expand...

Thanks, yes, and in fact i think you mean eg as in the leakage inductor in a full bridge SMPS acting as a turn on snubber?
Also, the ZVS switching that you can get in a PSFB.
Do you think say a couple of paralleled PSFB's offers much advantage over a couple of paralleled full bridge's at 120vdc input and 48v, 100A output?

And these days, with quick data rates...i am actually wondering, about having a couple of paralleled converters with a single error amplifier looking at vout and feeding each PWM comparator.....but digitising the error voltage , and feeding it to each pwm comparator.....then DAC'ing it back to analog for input to the PWM comparators?

This seems preferable to having an analog error voltage getting bussed around the expanse of a couple of big power stages...with noise problems and what have you. I mean, the error voltage could be turned into PWM say, then easily converted back to analog, with say a filter? A high enough frequency PWM would only need light filtering, and thus not introduce too heavy (low frequency) pole into the feedback loop.

I like the notion of disaggregating the control section
from the switch*magnetics section. There are PWM
controller ICs with digital outputs, there are digital
isolators and isolated FET drivers, the feedback only
needs to be itself isolated without bothering the
control loop stability / performance (current sense
would be tougher but believe I've seen that too,
in isolated forms). Managine or combining N current
sense feedbacks, that could be "interesting" (we did
this in ICs with multiple sense comparators in a
cellular "power farm", but at board level different
"stuff" might apply).

Depending on your end market sourcing appropriately
rated and qualified magnetics for 100A at all, let alone
fitting allotted volume and mass, might be your
biggest problem. A good argument for parallelism
I expect.