dual-supply switcher

Dual_PS_Buck.JPG

That's brilliantly funny.

joe

Why does the LV psu feed to before the coil rather than after? Why the PSes in series - if parallel, only either one psu would need power to produce output.

NT

They do something sort of like that for ATX power supplies, the 3.3V rail:

Topology is full wave forward converter, with centertapped secondaries. One big long winding gets +/-12V, with taps for the +/-5V. Those just go straight into the rectifiers and filter (the main choke being a common winding dealie).

What to do about +3.3V? If you take one side of the 5V and half-wave rectify it (you need a buck diode from ground, and a forward diode from the winding), you get 2.5V, give or take losses in the rectifier and stuff. Which is pretty substantial considering the 3.3V rail has to be the most precise.

So they do that, they start with 2.5V crude. They add in the other 5V phase, which would give 5V at poor regulation, but they put a saturable reactor on it, so it's adjustable. A TL431 and transistor controls it directly, making it the best regulated rail on the supply.

It's not quite the same as stacking in series, but generally the same idea -- the regulator control has Vmax to Vmax/2 range, roughly speaking.

Tim

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

This might be useful for something, maybe:

I guess the L could be on either side of the diode.

I need to drive a 250 amp load, current-mode but only over a limited voltage range. Imagine a giant LED or something. Stacking a relatively low power switcher in series with a giant constant-voltage supply might work.

wouldn't that be a similar concept to either a QUAD current dumping amplifier, small class-a in parallel with big class-B

or CROWN current balanced amplifier, buck source in parallel with buck sink

-Lasse

OT, but

seen these?:

breakouts for sc70-6,so8,sot32-6 with adhesive backing

-Lasse

Interesting for kluging a PCB.

I like the Bellin adapters for breadboards. They are relatively cheap.

What did you use to glue the pcb "bus bars" to the large pcb blank?

Ed

Crazy glue, but it's messy, so I should maybe use epoxy. Or solder paste.

Right. I just assumed it would cost a lot on energy efficiency to have 2 supply circuits in series at high current, both with their rectifier V drops.

NT

}snip{

Right, but if one of the two were a big battery, that would be another story.

joe

Yup. I'm betting it's not though :)

NT

We'd buy a couple of commercial voltage-output switchers, like 5 and

10 kilowatts or something. The series stacking would improve the efficiency of the constant-current regulator that we'd build.

I wonder how much power a roughly 0.5 square foot PCB can dissipate. The hard part about high power design is always thermal.

I don't know. What measures are you willing to take ?

On the topic of things I don't know, could one avoid low side diode drop to some extent by using transistors with relatively high Veb and low Vcesat? I'm guessing such a combination doesn't exist.

NT

I was thinking about squashing the board against a water-cooled cold-plate, with a thin (1 mm) Bergquist gap-pad between them as the insulator. One Bergquist part is 5 W/m-K, which works out to about

0.005 degC/watt if I did the math right.

I'd use mosfets.

Yes. I'm partly thinking of a different situation where BJTs are plentiful but fets fairly rare, and the currents involved low. I've seldom looked for high Veb trs, but they seem very few. I wonder what other characteristics comes with the high Veb.

2SD2704K Veb=25v, Vcesat 50mA @ 30mA CMPT404A Veb=25v, Vcesat 0.15v & rising 2Sd2114 Veb = 12v, Vcesat 0.4v @ 0.5A I see they all have very high betas.

But these things are too rare in the wild (ie in scrap equipment) to design around them (for 3rd world apps).

NT