LLC with an explicit discrete Lm

A multi-output forward doesn't need inductors and has decent cross-regulation, but you can't PWM regulate. How about a buck switcher ahead of the forward?

Or no regulation ahead but low-drop linear or switching regs after, as needed. I've done that a few times.

Power supply design is often a giant puzzle.

This reminds me the buck-pre-regulated push-pull and indeed is a viable option. An unregulated Baxandall driving the trafos through a magnetic amplifier is another.

Too complex overall. Curently I have been considering the TV-like approach: make the transmitting side overly complex for the beneft of simplified receivers. Can have a big net gain, given the number of secondaries.

Surprisingly, it is the auxiliary PSU that drives me nuts. The main power train turned out to be pretty straightforward, even if pretty advanced: a totem-pole PFC powering a high-voltage LLC converter.

Best regards, Piotr

Small switchers, like TPS54302, with a cheap stock inductor, are easy. So it can make sense to keep the power parts, especially the transformer, simple and regulate on the secondary. Really small stuff can use linear regs.

If your main converter has to make a lot of AC anyway, maybe you can couple that into some small transformers to get the low power stuff.

On a sunny day (Sun, 24 Nov 2019 18:22:03 +0100) it happened Piotr Wyderski wrote in :

For the boost converter I have this:

I like this part, but it solves only the easier part of the problem.

The isolated auxiliaries need to be energized even when there is no AC power, using battery backup. So it is simpler to power the smaller converter from the secondary end of the main LLC.

Best regards, Piotr

Consider planar -- the resonant thing I showed earlier has pretty good cross regulation. There's about 10nH between the secondaries.

Toroids are fine too, just a pain to wind. If you're farming them out then do it however you like. ;D

To get comparable leakage, you will need short winding lengths, and preferably use twisted pairs or even quads for each winding to keep the impedance low. Also, TLT techniques don't mind what the core is, it just needs to up the magnetizing impedance.

Also, don't mind adding LDOs to clean up [cross]regulation, it's just one more part, not a huge burden.

Also also, you can save some diodes if you use CT windings, which would be a pain for a hand wound transformer, but is almost free in planar. Just a PITA to design them...

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website:

Hi Tim

Can you post a link to the planar thingie?

Cheers

Klaus

Tim

That's really impressive, but I don't see how a planar could host that many windings.

Not a problem, there is a company that could wind them in quantity. Now it is for a proto, so the winding complexity is not a big issue.

Times the number of secondaries, 15 parts then (not all voltages need to be regulated).

The plan for this week: grab an LM5041, extract the full glory of it in the cascaded configuration and see if it suits my needs. Unfortunately, the built-in drivers are not particularly strong, so MOSFET selection will be a PITA. It looks promising and I can have SR drive for free (well, at the cost of even more windings...).

Best regards, Piotr

Apparently it does make sense, but I wasn't the first who invented exactly this combo:

LOL! :-)

Best regards, Piotr

out

speaking of synchronous rectification, this is a neat idea,

em out

tml

Nice, could not find a price.

English datasheet:

ood

them out

st

to

in

y,

.html

On a sunny day (Wed, 27 Nov 2019 15:09:36 -0800 (PST)) it happened Klaus Kragelund wrote in :

Interesting, always a RC network needed across it?

This is a typical RC snubber, often found in synchronously rectified flybacks in order to make the life of the MOSFET bearable.

Best regards, Pior