Watermelon Season (SMPS)

[NateS]

I've been looking at various schemes for dropping B+, and after I a while I realized that I've reinvented the SMPS a couple times. It occurs to me that 1) the old timers did this with mechanical vibrators, 2) most of the math and a large portion of the modes of self destruction all revolve around energy storage. My latest ruminations are:

1) Commercial 12V SMPS abound for cheap. This gives you a) heater supply (running your 6.3v heaters in series for the big bottle stuff), b) isolation and filtering (the stuff going back to the line), c) regulation. It's within 5% on supply voltage - so your tubes will have enough to start flinging electrons, and maybe actually increase tube life.

2) Run B+ off an unregulated open loop forward converter, powered off the 12V line. Input is regulated. Since we're transforming rather than storing, the requirements relax a bit, and I mostly care about finding something in the right turns ratio and core size.

I keep trying to fit variability into this, just a simple duty cycle reduction, but I think that probably complicates things too much (and output filtering starts sliding from simple filtering into real energy storage - and reintroduces the need for a closed loop system).

Thoughts?

[jazbo8]

I use a combo of laptop power adapter (12V) and a small DC-DC converter (12V-300V) to do what you described, it worked pretty well. Then I tried to scale the HV (a la powerscaling or vvr) by varying the FB voltage, that resulted in some nasty squealing or noise depending on the amount of FB, which I think is from the beat frequency between the power adpter and the converter. I am not sure what the laptop adapter switches at, but the converter switches at 40k.

Another experiment that I have not tried yet, is to vary the input voltage to the converter with a simple voltage divider, so the converter's duty cycle remains constant, I think that should work better... The money and weight saving by eliminating the PT are pretty significant, especially for higher power amps, so I am quite interested to learn what other configuration you can come up with.

Jaz

[NateS]

Another experiment that I have not tried yet, is to vary the input voltage to the converter with a simple voltage divider, so the converter's duty cycle remains constant, I think that should work better... The money and weight saving by eliminating the PT are pretty significant, especially for higher power amps, so I am quite interested to learn what other configuration you can come up with.
Jaz

I was hoping that I misunderstood - but I can't parse it that way. Feeding the power input to the dc-dc converter via voltage divider won't work - you're talking about trying to waste gobs of power as heat (and at these kind of power ratings, light too). If it did, you wouldn't need a switcher to accomplish power scaling. You'd need to either modulate duty cycle or frequency, and I believe frequency modulation only works when you're running as [pick topology] switchmode, rather than dc-transformer mode. (More specifically, the charge time would have to be less than half the period I think.)

I was talking about modulating drive PWM off a programming voltage. No feedback. Regulation would be input regulation at max duty cycle (50% in a non-storing transformer) and sag from there as you reduce on time. Output filtering would be limited to essentially smoothing ripple at the lower duty cycle limit. It's a filter, not an offline UPS.

[Steve Conner]

A voltage divider won't work too well for the reasons just stated. Although, some old designs of high voltage SMPS used a linear regulator powering a DC-DC converter with fixed 100% duty.

The modern way is just to modulate the duty cycle of the converter. But beware, if a converter was designed to run only at 100% duty (like the ones in cheap 12V to 120V inverters) then it will lack an output filter inductor, which is needed for it to respond to PWM properly.

To get rid of beat frequency noise, I've locked multiple SMPS together to a reference oscillator. However, it's quite possible for a single SMPS to go unstable and produce audible whistles, squeals and frying bacon sounds by itself. I have a couple of 1kW Xantrex benchtop supplies that do this as standard from the factory.

If you want to decrease the output voltage of an existing SMPS, don't do it by changing the feedback divider network, because that also increases the loop gain, which is asking for instability. The smart way is to vary the reference voltage going to the other terminal of the error amp, or inject a current into the feedback node through another resistor. I've used both techniques with success.

[jazbo8]

I think you misunderstood... All I am suggesting is to drop the incoming 12V to say 6V and let the DC converter run at the max 50% PWM duty cycle, no FB, so the HV could be scaled to 150V. May be the simple divider is not the most efficient choice but it is no worse than the dropping resistors used in conventional PS. Or am I still on the wrong track?

Jaz

[Steve Conner]

No tube equipment PS ever used a dropping resistor to throw away half of the voltage for the main plate supply, it would have been too big and given off too much heat. Also, the DC-DC converter may be unstable with a lot of resistance in series with its input.

[jazbo8]

If you want to decrease the output voltage of an existing SMPS, don't do it by changing the feedback divider network, because that also increases the loop gain, which is asking for instability. The smart way is to vary the reference voltage going to the other terminal of the error amp, or inject a current into the feedback node through another resistor. I've used both techniques with success.

Yes, I think that's what I did wrong - by changing the feedback divider network. Will try changing the reference voltage next. Thanks! BTW, the converter is based on the TI494 reference design, do you have some experience with it?

Jaz

[jazbo8]

I see what you are getting at, yes, it would be dumb to throw away half of the voltage for the plate supply using dropping resistors....but in this case, we are merely dropping 6V at say 100mA, so my thinking was the resistor would be similar in size to the ones used in the conventional supply... Not trying to be argumentative or anything, just trying to learn and clarify my own thinking... so feel free to correct me

Could you please elaborate on the second point, why would a smaller input voltage cause instability for the converter if we require no input or output regulation (as an example)?

Jaz

[Steve Conner]

I assumed you wanted to run an actual tube power amp, and those tend to consume tens of watts at a minimum. For a tube preamp I suppose you could get away with it.

As for the second point, I assumed that (like most people) you want to avoid actually building a SMPS, so you took a DC-DC converter module off the shelf and want to mess around with its input voltage. The controller chips in these things have undervoltage lockout and so on. Below a certain input voltage they'll quit working, and around the lower limit they may pulse on and off.

If you took care of a separate supply for the controller chip, then I guess it would work fine, since without feedback there is nothing to get unstable.

I've worked on a lot of SMPS based on the TL494. It really is well overdue for retirement. Nowadays you can buy a little step-up converter board made by some hobbyist (I forget who) intended for powering Nixie tubes.

[jazbo8]

Yes, even though I mentioned higher power amp before, so far my experiments were limited to pretty low power amp. I now see how I would not be able to "scale up" with my simple topology, and thanks for taking the time to answer my questions.

Cheers,
Jaz

[NateS]

Yes, even though I mentioned higher power amp before, so far my experiments were limited to pretty low power amp. I now see how I would not be able to "scale up" with my simple topology, and thanks for taking the time to answer my questions.

Cheers,
Jaz

Well, I started the thread, and yeah I was shooting for a two bottle EL-34 monster. I was also talking about rolling the DC-DC converter design myself, and planning for duty cycle modulation in the output filtering. I was hoping for something like 200mA from 70-450v. I think making it variable - particularly with a duty cycle range on the order of 15-50% pretty much pushes me into closing the loop - but that doesn't require giving up the forward converter/transformer topology - and I think that still simplifies things a lot - from the tweaking/practical side even more so than design work. I think in the interest of actually getting rolling on something I'm willing to consider anything from there down to fixed 12v*whatever winding I can find between 250-400v @ 150mA.

I also had hopes for finding an off the shelf transformer - I've seen that they do exist - but I'm not finding retail vendors so far, just 'call our sales droid, and he'll take you out to a steak dinner with a $100 bottle of wine and we can discuss your application and potential sales figures."

[Steve Conner]

SMPS transformers are much easier to wind yourself than the ordinary iron-cored type. For instance, say you were to use the commonly available ETD39 core set (easily good for 100W continuous throughput) your high voltage winding will end up with only about 75 turns. You just wind the bobbin and clip the core halves onto it.

If you have to have a ready-made one, I think one of Coilcraft's capacitor charging transformers might do it. They're designed to run in flyback mode, but that's the simplest approach, since as I already pointed out, a forward converter needs an output filter inductor if you want to apply PWM to it, but a flyback converter doesn't.

I can't be bothered figuring out which one for you right now. Coilcraft Capacitor Charging Transformers - menu

[NateS]

Thanks - actually, after reconsidering following the input on transformer winding I started to consider flyback again. And subsequently discovered LT3751, which in fact does have a bunch of $10 OTS transformers available, After looking the data sheet and a few articles over, it certainly looks promising. Time to go reread Billing's flyback chapters again.