20KW DRSSTC / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

20KW DRSSTC

first previous 8 9 10 11 next

Move Thread

LAN_403

Steve Conner

Fri Apr 27 2007, 01:02PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Well, I have been designing microcontroller and DSP based instruments for a day job for several years now, and I'm not about to try it. The reason being that it's a real time job that requires very fast response, whereas anything controlled by a micro has latency. (aka lag, all too familiar to any player of FPS.)

Latency is unavoidable because the micro has to execute instructions to transform sensor input into control output, and each instruction takes time. Even a 100MHz uC could only do about 10 instructions in the turn-off delay time of a TO-247 IGBT, and that doesn't count the time taken to fetch a result from an ADC, etc.

So to me, this means that any control algorithm running on a micro would have to make its decisions based on what it observed on previous cycles, not the current one. Therefore, I believe it would just end up as a fancy digital PLL.

I could be completely wrong, but even if I was right, there could be scope for hybrid controllers that used hardware for the time critical stuff. One possibility might be a Steve Ward controller preceded by a phase lead block whose amount of lead was adjustable by a microcontroller. The micro would run a kind of digital PLL algorithm that continuously tweaked the phase lead to get the best zero current switching. It could also do overcurrent detection, duty cycle limiting, and the like.

HV Enthusiast

Fri Apr 27 2007, 02:26PM

Registered Member #15 Joined: Thu Feb 02 2006, 01:11PM
Location:
Posts: 3068

Steve Ward wrote ...

Sooo... who will be the first to implement the brains of a DRSSTC driver on a micro-controller?

I've already done this quite successfully with my integrated DRSSTC commercial unit which is packaged in a small brick enclosure. In fact, i'm designing several digitally uProc controlled current mode control DC-DC converters for a client of mine as we speak. When I say digitally controlled, I mean the entire feedback system etc... is digital, no analog amplifiers or compensation or anything in the voltage or current mode loops.

wrote ...

So to me, this means that any control algorithm running on a micro would have to make its decisions based on what it observed on previous cycles, not the current one. Therefore, I believe it would just end up as a fancy digital PLL.

I could be completely wrong, but even if I was right, there could be scope for hybrid controllers that used hardware for the time critical stuff. One possibility might be a Steve Ward controller preceded by a phase lead block whose amount of lead was adjustable by a microcontroller. The micro would run a kind of digital PLL algorithm that continuously tweaked the phase lead to get the best zero current switching. It could also do overcurrent detection, duty cycle limiting, and the like.

Not true. In fact, a large part of the power supply industry is moving into this direction. For a client now, I'm actually doing a supply now utilizing real-time digital control using 100MHz Lattice CPLD for the controller. I can get 1Mhz bandwidth with pulse-by-pulse realtime voltage / current feedback and the CPLD can control pulsewidth with resolution less than 200ps. This all in a phase-shifted full-bridge topology. Pretty neat stuff.

Tom540

Fri Apr 27 2007, 04:52PM

Banned on 3/17/2009.
Registered Member #487 Joined: Sun Jul 09 2006, 01:22AM
Location:
Posts: 617

I was thinking maybe using an programmable logic xilinx chip. Just using the schematic type programming and cramming all the logic into it. Of course you'd still need some analog so there would still be the comparator for current limit etc.

HV Enthusiast

Fri Apr 27 2007, 06:09PM

Registered Member #15 Joined: Thu Feb 02 2006, 01:11PM
Location:
Posts: 3068

You can do all the current limiting, etc... digitally. The TI competitor to Lattice has onboard A/Ds. With the Lattice, you need to use an external A/D.

thedatastream

Fri Apr 27 2007, 08:01PM

Registered Member #505 Joined: Sun Nov 19 2006, 06:42PM
Location: Yorkshire!
Posts: 329

I've been looking at the Atmel AT90PWMx series of micros

Which look interesting for this kind of application

James

Finn Hammer

Sat Apr 28 2007, 07:14AM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

Steve Conner wrote ...

Also, I've not gone through the math, but I think the Steve Ward style feedback driver can produce a drive waveform that actually contains both mode frequencies at once, whereas mine is limited to one frequency at a time. I don't know what the implication of that is, though.

Steve,

I know you have officially declined to try to explain this any more, but when you say that a signal can contain both frequencies at one time, I have to ask a couple of questions.

If there is 2 frequencies present in the primary current, how does this affect the waveform of the signal?

What I am aiming at is this: the signal can only cross zero once per half cycle, and from that point of view it can only have one frequency at a time. But this frequency can change from cycle to cycle.

In audio, we look at a signal`s base frequency. This frequency is based on the time from zero crossing to zero crossing. If this signal contains contributions from other frequencies, this is called distortion.
The presence of odd order harmonic flatten the signal. even order making it more pointy

So is this what you mean? The current is not a perfect sine wave but distorted due to the other pole frequency mixed into it?

Isn`t it so, that it is a property inherent of 2 coupled LC circuits, that both pole frequencies are represented in the signal.

Cheers, Finn Hammer

Steve Conner

Sat Apr 28 2007, 01:21PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

EVR: A CPLD is neat, but it's a completely different thing to a microprocessor, so I stand by my original assertion.

Finn: The concept of "frequency" only really has a meaning as an average across several cycles. Sure, you can measure the period of one cycle, and say the frequency is the inverse of that. But that is only true if all other cycles of the waveform have the same period, and in order to find that out, you have to observe several cycles on a scope.

When there are two frequencies present at once in the primary current, you see ripples in the envelope of the waveform as they beat with each other. If they are both the same amplitude, you get notches, which are just very big ripples. You don't see them as waveform distortion, because they are not harmonically related.

The big question for me was always how to synthesize these two frequencies in a square wave drive that has a constant envelope that can't be made to ripple. I think it comes down to moving the zero crossings around in some special pattern, but what pattern? Does the output from a Steve Ward style driver follow this pattern? Always or only under certain conditions? Is it possible to synthesize two frequencies while maintaining zero current switching? Who knows?

HV Enthusiast

Sat Apr 28 2007, 01:53PM

Registered Member #15 Joined: Thu Feb 02 2006, 01:11PM
Location:
Posts: 3068

Finn,

The primary current does contain the two major frequency components. Depending on how the coil is tuned will determine the magnitudes of those components. If you hook a current transformer to a spectrum analyzer, you'll see these components quite easily.

hvguy

Wed May 02 2007, 07:37AM

Registered Member #289 Joined: Mon Mar 06 2006, 10:45AM
Location: Conroe, TX
Posts: 154

Updateâ€¦

I have about 55lbs (48 bricks) of IGBTs ready to go for the new driver

I plan on using three or four stacked H-bridge modules composed of these CM600s to drive a divided tank cap then the primary of the 16â€. Total tank capacitance will probably increase to .88uF for four modules or stay at .66uF for three. The DC rail voltage will be dropped to 700V to accommodate the 1200V rating of the CM600s. Each brick will have its own driver board based on the design I mentioned in my last post. Each bridge will also have is own CT and the system will have a modified OC circuit. Here is the revised schematic

The more I think about this stacked bridge concept the more I like it. It takes up quite a bit more physical space than a single bridge but the heat and high current are distributed which make the mechanical design considerably easier. I will probably machine some .5â€ AL water cooled heatsinks to save on space; four 2â€ thick finned heatsinks would make the final assembly almost 18â€ long. If I go water cooled I should be able to keep it under 12â€ and increase the duty cycle to 100% (if my neighbors can handle it)

I have some new controller ideas I am currently testing but no significant results yet. The idea of using a micro to control one of these things has been discussed for some time now. My big problem with them is reliability; Iâ€™m just not convinced they are better than an analog circuit. Although the fact that your using them commercially in a phase shifted bridge, Dan, does go a long way to convincing me. I have seen them used in buck/boost converters for some time but anything can do that. A phase shifted full bridge is a whole other thing though. More to comeâ€¦

Steve Conner

Wed May 02 2007, 08:03AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

I was never sure about divided tank caps. Sure, they force sharing of current at the fundamental frequency, but what about the harmonics? The impedance of a capacitor falls with frequency. If one bridge switches a little earlier or later than the others, switching edges are made of such high frequencies that the tank caps might as well be short circuits.

If I were paralleling bridges, I would use air-cored ballast chokes instead. Or better still some crazy RF transformer arrangement so the outputs end up in series. I've seen this done in broadcast transmitters.

I don't know, maybe the switching edges from large IGBTs will be slow enough that it'll work fine. But it's one more thing to think about.

first previous 8 9 10 11 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob