Discrete buck regulator

[Khron]

Hello, and Happy New Year to everyone!

In case the thread title isn't explicit enough, here's the rundown:

I've had some self-oscillating Class-D amps on the drawing board for some time now. The designs i've set my mind on, employ IR(S)2110/3 gate-drivers for the output FETs. Now, these chips need an extra 10-15v supply, referenced to the low-side source pin of the FETs.

Linear regulation isn't much of an option at higher voltages, and i've yet to find any IC buck regulators that can take more than 100v input (the final design aims to be a 130v single-supply full-bridge amp).

Sometime, a year or two ago, i googled for discrete buck designs, and the simplest i came across (from among very few) was this design.

Now, i've fiddled around with this design in LTspice, tweaking resistor values and whatnot, and at least for now, i've ended up with what you can see in the attached picture. Output caps will be Chemicon KZE 1200u/16v (i included the ESR in the simulation), and i've just ordered some NEC/Tokin 220u/0.33R inductors. 13.5V output, R5 provides a ~50mA load, and i slapped on that extra LC filter to knock down the ripple. As it stands, it's (supposedly) switching at ~152kHz, and the PFET burns about 3.8W.

If any of you brighter minds might have any tips on how to increase efficiency (even) further, i'd gladly listen Currently, LTspice says that for ~650mW load, it pulls 4.9W from the supply.

[Khron]

Re: Discrete buck regulator

Well, since it's been so quiet here today, i decided to google some more after discrete buck circuits, and lo and behold, i stumbled upon this site

Numbers look muuuuuuuuuuuuuuuch better now Pass transistor burns up a measly 93mW, for the same 50mA load

Haven't fiddled with component values too much (it's 3am here right now), but i've attached another screenshot, with a zoomed-in view of the output voltage. Switching frequency's around 125kHz, with about 5mV ripple on the output

Tomorrow i'll try to replace that zener with a TL431 and two resistors, and see how it behaves at various programmed voltages - maybe i can end up with a useful building-block for a mini-bench-supply

[ben7]

Re: Discrete buck regulator

Well, since it's been so quiet here today, i decided to google some more after discrete buck circuits, and lo and behold, i stumbled upon this site

Numbers look muuuuuuuuuuuuuuuch better now Pass transistor burns up a measly 93mW, for the same 50mA load

Haven't fiddled with component values too much (it's 3am here right now), but i've attached another screenshot, with a zoomed-in view of the output voltage. Switching frequency's around 125kHz, with about 5mV ripple on the output

Tomorrow i'll try to replace that zener with a TL431 and two resistors, and see how it behaves at various programmed voltages - maybe i can end up with a useful building-block for a mini-bench-supply

I am playing around with this in LTspice right now too.

[Khron]

Re: Discrete buck regulator

Well, today i got around to trying it out with a TL431. Also, for increased efficiency, i went ahead and slapped in an IRF9640 P-MOSFET

Took more than a few iterations, but the results, for 130V input, are as follows:

45V out, 2.25A load (that's about 101W), Fsw ~21kHz, ~90% efficiency And a mere 70mV ripple on the output (that's a measley 0.15% ), but that's with "just" those two 270uF, which are run well beyond their ripple ratings but that's not the point

With the values in the attached picture, it goes into some sort of "hiccup mode" if the output voltage is set below about 17v (R5<28k). But keep in mind, this is with 130V in

I see two further steps i'll wanna take:
1) replacing that NPN with a small N-MOSFET (no base current = lower losses ), and
2) trying to "mirror" the whole circuit, to see if it could be used to regulate a negative voltage as well

[Khron]

Re: Discrete buck regulator

Just as a "proof of concept" sort of thing, i gave challenge no. 2) a shot, and the result is...

It works!

Well, in LTspice, at least, for now

The N-channel FET burns up an extra 600mW compared to the P-channel one, but... I can live with 90% efficiency

Here's the schematic, for whoever's interested. Hope it helps ya

[Evil Lurker]

Re: Discrete buck regulator

Your way ahead of me... I'm still piddling with the LM2575 and MC34063 IC's for my buck regulators.

[ben7]

Re: Discrete buck regulator

He may be ahead of you somewhat, but not by much.

Khron666, have you done any tests for stability, at small load, and high load? Have you made sure that the parts i LTspice are accurate (I.E. the inductor also has a series resistance, and parallel capacitance)? Also, is there any overshoot in the output, when the regulator is starting up?

[tom66]

Re: Discrete buck regulator

Also the inductor is the major efficiency loss in a buck converter due to magnetic hysteresis. Real world efficiency will be close to 80% but who cares anyway? It's for a 13V, 100mA supply. That will all be lost in the gate drive of the FETs as resitive heating from the capacitive charge/discharge.

I'm working on a small low power DC-operated flyback based on a two-transistor circuit.

[ben7]

Re: Discrete buck regulator

Also the inductor is the major efficiency loss in a buck converter due to magnetic hysteresis.

^^^
Even he knows more about this than I do!

[tom66]

Re: Discrete buck regulator

Magnetic circuits, dude. They're awesome.

[ben7]

Re: Discrete buck regulator

Magnetic circuits, dude. They're awesome.

Agreed.

Awesome when used properly

[tom66]

Re: Discrete buck regulator

Logic gates, anyone?
https://www.youtube.com/watch?v=p7SkE5pERtA

[Khron]

Re: Discrete buck regulator

I looked for discrete designs precisely 'cause i needed a greater input voltage than most chip designs can take

Yes, i *have* included the series resistance in the big 220u inductor (as per the datasheet), but not the parallel capacitance. Not yet, anyway. But don't worry, i'll take the practical / physical testing one step at a time - i will *NOT* start out with 130Vin

I got the IRF9640 model from the Vishay website, the MPSA42 model from the Fairchild datasheet (if i recall correctly), and i also included the ESR for the output caps.

I launched a sim last night, with 100mA load, and let it run for a whole second (of simmed time). There DOES seem to be the tiniest amount of overshoot (think 70uV, that's micro-volts), but nothing to worry about, apparently. But, of course, this is all in virtual-land, for now

[Khron]

Re: Discrete buck regulator

It's been a while since the last post but i have news!

After loads more fiddling in LTspice, i ended up including one of these 2-transistor buck converters on the PCB for a class-D amp using a TAS5630 chip. This needs a 12v(-ish) rail for the gate drivers inside it, in addition to the main power rail (20-50v).

Since i don't need all that much power out of it (150-200mA max - about 80-100mA for the chip and two opamps, plus the 100mA cooling fan that gets triggered by the 125C "overtemp" pin of the IC), sims indicated i could get away with a SOT23-packaged PNP switching transistor.

I finally got around to actually ordering a batch of PCB's from China, and they arrived today So i promptly got to populating the buck circuitry.

Did an initial batch of load tests (10-200mA in 10mA steps), with promising results. In the 30-200mA range, efficiency's hovering between 70-80% with a 24v input Also had my scope hooked up on the output - with a single 10uF tantalum cap, i didn't see ripple higher than 10-12mv. The board has pads for 3 of those in parallel, as well as further LC-filtering

Gonna continue a bit later with 30v, 36v, 42v, and maybe 48v if i feel like it

For those more curious, i can attach a spreadsheet with the numbers when i'm all done

For now, attached you can find the final schematic.

I went with the PBSS9110 PNP because i needed something similar for some other project, so when i put together my Mouser order, i figured what the hell :P

[Khron]

Re: Discrete buck regulator

Well, i might've spoken a tiny little bit too soon...

LTspice was a bit too optimistic with the efficiency figures, as it turns out.

With 24v input, all is fine and dandy up to 150-200mA load, as mentioned in the post above.
With 30v or more though, efficiency figures drop considerably:
24v in, 200mA load @ 11.66v, i'm burning away about 640mW (difference between Pin and Pout)
30v in, 160mA load @ 11.81v, 900mW loss
36v in, 110mA load @ 11.88v, 961mW loss

That's... suboptimal, as you might imagine

[Khron]

Re: Discrete buck regulator

Mystery solved, pretty much - it seems i was a liiiittle bit too optimistic about the capabilities of the 1210-sized inductor i was planning on using. There's more than a good chance that it was saturating, and as such, screwing up the whole shebang.

For testing purposes, i tacked on a beefy radial Panasonic ELC11D 180uH inductor (good for 3A DC or so), and totally different figures, i'm glad to report

48v in, 56mA draw @ 11.51v / 150mA load, 961mW losses / 64.2%
48v in, 40mA draw @ 11.49v / 100mA load, 771mW losses / 59.8%
36v in, 46mA draw @ 11.45v / 100mA load, 511mW losses / 69.1%
36v in, 64mA draw @ 11.44v / 150mA load, 588mW losses / 74.5%

Now that i'm a lot happier with Guess i'll have to order another batch of higher-rated inductors...