SLR Driver With Feedback Regulation

[Tyler Christensen]

I usually post to 4hv these days, but figured I might as well put this one up here since it relates fairly well to fusors. I have designed and built an SLR driver that has regulation, so it shouldn't suffer from the no-current-ringing-voltage syndrome that most of these home built supplies seem to have. It always maintains zero current switching, so it should have the same efficiency as a typical SLR which is quite high.

It operates by comparing voltage and current output to references set by pots, and then running this enable/disable signal through a 7474 flip-flop to assure that actions are only taken on zero current moments.

Preliminary testing shows it to be working well, however I won't be able to test it on high voltage until I'm home for winter break (since I don't have my multiplier at MIT). I'll post more comprehensive test results on dummy transformers and loads over the next week or two.

Attached is the driver, obligatory beautiful-waveform picture, and schematics (one for a power supply since it requires 5 rails on the logic board, and the other being the logic and gate drive system). The gates are driven via a gate drive transformer. I used a big CM200 IGBT brick for a 10-20kW rating, but it should be transferable to any size. If you are using small transistors, there is no need for the P-N half bridges to drive the gate drive transformer (although you'd have to use UCC37321 instead of UCC37322 since this P-N topology is inverting).

[Chris Bradley]

I think you might find that the use of external components to shut off the switching signal is not necessary. I believe if you pull the Dead Time control high (pin 5, I think), then the dead time will effectively shut the signal down. [Or maybe it turns it fully on, whichever way your switching inverts the pulses. I'm quite sure you'll figure whether that works for you or not!]

You may also want to work out how to connect up the comparator pins so that it will follow the ac ripple you will surely get on the outputs, when running off of the mains.

I have been doing work on swithing circuits recently myself, and the 'nearly not operating' condition is definitely a tricky one to deal with. Worse still is when the drivers get confused by a very short PWM signal and end up flipping erratically - exactly NOT what you want when trying to control small outputs!! One other way I had considered is to fix the pulse length and use variable pulse separation, rather than width. You can implement this relatively easily with a 555.

Current control is very useful especially where you use it to balance out the net current to avoid flux walking, which you may get with a half-bridge and may burn out your bridge if the core walks its way into saturation. (I'd never heard of flux walking before until recently when I started experimenting with cores and experienced the problem first hand.)

I suspect that you are in for spending many many hours tuning the driver circuits to get a good output drive. I attribute many of the difficulties of driving high volts/high currents to the gate being pulled around by the capacitance in the FET/IGBT chip and the changing voltage on the drain/collector, and you need to think carefully about how to manage the various voltages attached to the chips. That view might be a bit simplistic (or 'not even wrong') but might be a reasonable 'mechanistic' explanation for why you might be facing some uphill work to make it work under loads. Handling kW power is non-trivial!

A few documents that might be useful to you. The specifics may not apply to your config, but the comments regarding IGBT function are in 'plain speak', you might find them useful.

http://www.infineon.com/search/en?q=an-2006-01

http://www.vishay.com/docs/81227/81227.pdf

Good luck!! I hope it all goes well for you.

[Carl Willis]

Nice idea, Tyler, and fairly straightforward to implement even as a retrofit to drivers with fixed, dead-reckoned timing. You've got my attention.

Two comments: First, looking at your 'scope trace, it appears that the switching happens uncomfortably far from the zero-current crossing in that particular trace. This is probably amended easily enough with the current threshold pot. Or, it may just be that you set this pot to provide safety at a higher power level, and now you're at a low power and a different load condition. My second comment is just a caveat that a lot of the non-ideal behavior I observe comes from the transformer. Yours looks to be low-turns-ratio. But at high turns ratio, a lot of that ringing comes up due to stray reactance.

Good work!
-Carl

[Tyler Christensen]

Here is a video of it operating with feedback. Works exactly as expected, I like how it's coming out. The top line trace is the output voltage (negative polarity, so lower is more voltage) and the current is shown. This test was done with full-wave rectification, a small filter cap, and only voltage feedback.

http://www.youtube.com/watch?v=eEwlHi5mUXE


Carl: The poor current appearance was due partially to a lack of effort in tuning it during this test, and also due to a crappy current measurement. It's much better now (as can be seen in the video). Also, I agree that leakage generally causes the ring-up. Although interestingly I have huge ring-up with even these few turns. Without feedback, it'll draw a half inch arc with this ~1:3 turns ratio test transformer with 50V on the bridge (no output rectification or filtering or anything). Pretty remarkable! Of course with feedback, this is completely eliminated.

Chris: The deadtime pin control probably would work, although it won't really save on component count. You still need the flip-flop so that you turn it off at a zero crossing. Not sure what you're saying about the short PWM signals, this is very much not a PWM based controller, since it has to maintain SLR operation. The pulses are fixed length, and basically does have variable pulse separation.

[Chris Bradley]

Tyler Christensen wrote:
> Chris: The deadtime pin control probably would work, although it won't really save on component count. You still need the flip-flop so that you turn it off at a zero crossing.
I think an adjustment of the dead-time control would have effect at the next dead time, not cut off a pulse mid-stream. Would this not be the same effect?

> Not sure what you're saying about the short PWM signals, this is very much not a PWM based controller, since it has to maintain SLR operation. The pulses are fixed length, and basically does have variable pulse separation.
OK, sorry, I took it that R2 adjustment would affect pulse width.

[Cliff S]

Just because I’ve been down this road 25 years ago, reviewing countless ways to control resonant converters…
The best way I have found, (without using FW logic devices), is to take a VCO, any VCO, (even a PWM using the Rt pin to control freq, and keeping PW at full), and send outputs ( 0 deg and 180 deg) to a mono stable multi-vibrator, “One-shot”, like the MM74C221. Gives you fixed on time, triggering off one edge of VCO’s. Set one-shot ON time >1/2 period. Best way to set ON time is to run into shorted output, where tank is completely reactive, period the longest, and set it for just slightly greater than the ½ period. (Remember, diodes across switching devices will conduct the reverse current automatically, so no need to extend ON time much past ½ period. Maximum frequency can also be set under this condition, for maximum safety. (Will result in some dead time under load, which can be compensated for by using voltage feedback to increase Fmax).

Feedback regulation is typically done with a PID controller or simple integrator. A “bang-bang” (Hysteretic) controller, can also be used, but will result in higher output ripple, due to multiple pulsing causing asymmetrical charging/discharging operation. It also “sounds” like all hell is breaking loose.

Sorry no schematics to show. If the verbiage doesn’t sink in, let me know, and I can dig up some schematics.

Nowadays, it’s all done in FW.

Best Regards,
Cliff Scapellati
Executive Vice President of Engineering
Spellman High Voltage Electronics
631-630-3110

[JohnCuthbert]

OMG!
TLA
SLR?

[Cliff S]

Oh, I left out one other obvious controller method: There are OTS Resonant Controller I.C.'s available on the market that have everything you need, built in: Fixed ON time, OFF time, variable frequency, Fmax/Fmin, Error amp, etc. Any major analog I.C. manufacturer has these in the Power Conversion control portfolio. We use these too. They work well, (especially when you don't have any onboard FW to do the job).

Best Regards,
Cliff Scapellati
Executive Vice President of Engineering
Spellman High Voltage Electronics
631-630-3110