CWDRSSTC / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

CWDRSSTC

first previous 2 3 4 5 next last

Move Thread

LAN_403

Goodchild

Sun Nov 13 2011, 07:44AM

Registered Member #2292 Joined: Fri Aug 14 2009, 05:33PM
Location: The Wild West AKA Arizona
Posts: 795

Marko,

I simulated the control method you are talking about for the QCW using the comparator, but after it's all said and done with loading and such you get very bad ripple in the primary current even with delta sigma modulation of the primary current. Really the only way this control method could possible work is with a system that had a crazy high tank Z like say > 60ohms. Here is a result of one of my sims with a tank Z of around 25ohm:

6187175372 465688232d Z

45 ohm Z:

The reason for this is because at the start of the burst when the current is slow you have to skip so many cycles that it leaves a large ripple in the current. This ripple would be at a nasty low freq as well (<5KHz). The only way to solve this is with a really low coupling and a high Z to slow the rise and fall of the current and damp out the ripple effect. Although the bus modulation is by far the way to go, but it doesn't necessary have to be a buck! It could be as simple as half rectified mains or a level shift circuit like what a VTTC uses.

Pure CW is not my thing, I like making the biggest sparks with the least power.

Everyone has there own likes I guess. hehe

On the ZVS topic, I simulated that as well. It came out rather well in simulation and I even came close to building a full bridge ZVS drive for my QCW. I never ended up doing it though because ZVS doesn't work all that well with IGBTs because of there tail currents. If you were to build it with some large MOSFETs however a phase shift ZVS full bridge would be a rather fancy nice approach for a pure CW coil.

All the modulation could be accomplished through phase shifting the two half bridges.

Dr. Dark Current

Sun Nov 13 2011, 08:16AM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

Marko, the ZVS Tesla coil was a fail.
The first version was the standard push-pull circuit, with active feedback. Even though it did oscillate and I was getting some hundred volts per turn, I couldn't get it to make a longer discharge than 1cm. While increasing the tank frequency by tuning the tap, the frequency suddenly jumped from below secondary resonance to above secondary resonance. This was probably the result of high coupling, but I don't think a lower coupling would help.

The second version was with a self-oscillating ZVS fullbridge I designed. It oscillated, again producing some hundred volts per turn with the 4 sq.mm primary wire getting hot from the kVAR's, but the behavior was exactly the same as before. I just couldn't tune it.

PS. I just looked at your schematic again, and you have a Zener diode connected directly across the 74HC14 input. Zener diodes have high capacitance and are slow, that might be part of a reason why you got such a weak feedback.

Steve Conner

Sun Nov 13 2011, 11:13AM

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

The secret of the sigma delta control method is to turn all of the IGBTs off when the current goes above the setpoint, and let the tank energy regenerate back into the DC bus, the bridge running as a rectifier. That way, the envelope must fall as fast as it rose: the energy transfer is the same magnitude, just the opposite direction.

The back-and-forth of energy stresses the diodes more, but I believe the extra losses in the diodes are no worse than the losses in the high-level modulators you use just now.

I've been doing it for years, that's how the OCD on my coils works, and when doing demos in undersized cages they'll happily bounce off it all day for weeks on end. I just haven't tried substituting the constant setpoint for a ramp.

My PLL driver was supposed to work with CW SSTCs and induction heaters too. It is designed with two inputs, so you can drive the PLL from the secondary base, and the current limiter from the primary.

In practice I found trouble getting the PLL to lock when first applying the DC bus on my induction heater. With half wave rectified supply, it probably needs a keep-alive current.

Marko

Sun Nov 13 2011, 02:20PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Hi guys

Kilovolt:

Interesting that you couldn't get the current fed inverter to work, as I planned on something similar (for very high frequency TC) any data on what you did would be valuable.

Goodchild: My sight is too poor to make out much of your schematic (higher resolution pic please?) but what I intended to do is basically this -

the most of the circuit is Steve Ward's original DRSSTC controller, except there's a comparator in place of interrupter. Current signal (shared with OCD comparator, which now acts as auxiliary protection) goes into the - input, and to + input I'll feed signal from audio output of my PC (It will be a micro tabletop coil for start).

There'll be no hysteresis on the comparator so after the igbt's go off, the only "memory" is provided by the output synchro-flip-flop: every next cycle IGBT's get a chance to get turned on back again if the current has dropped low enough. As far as I know, this is what Steve Conner used as his OCD.

Regarding the initial current ripple, why is it so troublesome after all? It doesn't look like anything that would trouble the inverter too much.

One must note that the whole model is probably not too ideal as any secondary will have a latency time before it breaks out (even with breakout point!) during which it will have much higher Q and load the primary system more heavily. I'm thinking this might actually help rectify the initial current ripple.

And after all I intend to play with my waveform shape instead of just keeping it a plain ramp (If the ripple appears, I could even adjust the waveform with "anti ripple" to actively counter it). In any case, before the start of the ramp I intended to keep a steady DC signal for some time until all transients level out, and only then starting to ramp it up slowly.

Ofcourse this puts my PC at danger, probably not too much with a mini coil but for larger ones I'd probably want to have an isolation amplifier in between or use an MP3 player or something to produce the signal.

Still the original design was intended as normal DRSSTC, hence the tank impedance is low (10ohms or so) and I'd like to have it higher for QCW operation just for the sake of inverter losses. But for this I'd need about 3 CDE caps of 47nF or so and2-3kV rating, anyone happen to have some? :(

Anyway, to get back to the true CWDRSSTC's and not just part-time ones... I'm really serious about those 50kW (or 40-ish, what I think might be required for 1 meter CW arc, my initial goal), just not sure where to get them... bringing the coil to college would be fun but impractical, so my old school might be the next best option... it has it's own transformer station nearby (actually just few meters from the lab)... wonder what kind of feeds do they have there... Large arc flashes anyone? :O

Marko

Goodchild

Sun Nov 13 2011, 06:20PM

Registered Member #2292 Joined: Fri Aug 14 2009, 05:33PM
Location: The Wild West AKA Arizona
Posts: 795

First off I have to say this thread has become very interesting indeed

Don't see a lot of this anymore on 4hv.

Steve, I have tried what you are talking about with shutting off all the IGBTs and re triggering them on the next cycle. (this is what's happening in that simulation) Now under normal operation of a low impedance DR tank this works, because the current in most cases only takes a couple cycles to recycle back into the tank. But in the land of QCW with it's high impedance tank, it can take 10's of cycles to get the current to drop down. This makes for a nasty low frequency ripple at low power because it has to skip many cycles to lower the current enough to turn back on again.

Marko, Sorry I don't have a bigger pic, It's rather simple simulation, just a standard DRSSTC high impedance tank driven with a basic DRSSTC driver. The changed part was the addition of the integrator before the comparator and another flip flop to make the delta sigma modulator. It's basically what you are talking about doing only it has an integrator with the comparator. That hysteresis is very necessary in my mind. If the hysteresis wasn't there you would retrigger on the next cycle only to have the tank ring right back up again, some times higher than the point that you stopped it at the first time! This could lead to even more nasty oscillations.

Unlike cycle by cycle limiting in say a boost converter or other none resonant SMPS the current is sinusoidal, making cycle by cycle limiting kinda impossible without the addition of hysteresis. This is because with a boost you can hard switch and turn off right at the current you want to stop at. Where with the DR we have to turn on and off at the zero crossings. So we have to remember what the current was and add some delay so that doesn't keep re triggering.

The ripple can be bad for a couple of reasons, first off at 5KHz it would be audible! Also as Steve ward pointed out to me a while back, a ripple of that size will most likely effect spark growth. I very well could be wrong though, I still love the idea of only using the bridge for modulation, but practically I have yet to find a solution...

I hope my rambling wasn't to bad

Dr. Kilovolt, may I ask how the full bridge ZVS was implemented? The way I would do it is with a small amount of extra C in parallel with each switch and a small amount of delay in each gate drive. this makes the switch turn on when the current in the primary LC is at say 5% of it's peak value. That current will be flowing through the C on each switch and the switches internal C as well. So when the switch is turning on and off the voltage across each switch is 0. spice is a great way to get ZVS tuned right before building the real thing. Because you are right with an unturned ZVS is can be even worse than just regular ZCS!

Eric

Marko

Sun Nov 13 2011, 06:49PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Hi goodchild,

Well, it seems that the forum is being taken up by a new generation of newbies, I feel like a very old outsider here now!

Not sure if I was clear about that, but the "hysteresis" is provided by the 74HC74 flip flop (or whatever) that is normally used for soft turn-off in DRSSTC's, and the delay circuit which can be tuned to prevent turn on in one or more following cycles, so you may not even need any hysteresis on the comparator (I remember asking Steve Conner about exactly the same thing of not using any hysteresis at all for his OCD scheme, which is very similar concept from what I can tell).

I haven't figured out what the integrator is for though, just generating a ramp from the input signal, or is it actually in feedback path?

I'm pretty definitely going to try something like this in next few months on a small DRSSTC I have in parts here, after I redo the control board to include a comparator.
I might add a bit of optional + feedback hysteresis too just in case, but otherwise I want to see if this uber simple scheme will work if I can have good control over input waveform. (hopefully without frying my PC)

Kilovolt: How did you implement your full bridge current-fed inverter? It's troublesome since it needs gate drive overlap instead of deadtime, so you'd need isolated gate drives controlled by optocouplers. But advantage would be zero voltage switching as well as no diode conduction and unity power factor which is attractive for HFSSTC's.

Marko

Dr. Dark Current

Sun Nov 13 2011, 07:01PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

All,
here is my CFPR fullbridge circuit. It oscillates as expected, but the IGBTs got hot at 200 kHz and I got just a small spark. Maybe it would help to use MOSFETs, but their higher voltage drop could be a problem for the circuit. Note: 21V is regulated from LM317.

Marko

Mon Nov 14 2011, 02:35AM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Jan, that circuit looks like a plain shock horror to me. Have you tried using a "normal" SSTC driver circuit instead (this is what I'm considering, but will likely have to do with isolated gate drivers in the end)

Marko

Goodchild

Mon Nov 14 2011, 04:21AM

Registered Member #2292 Joined: Fri Aug 14 2009, 05:33PM
Location: The Wild West AKA Arizona
Posts: 795

Marko, I also wasn't being totally clear now that I read my own post again. The hysteresis and integrator are one in the same. There is no hysteresis on the comparator it's self but rather hysteresis of the whole feedback loop, provided by the integrator. So yes the integrator is a big part of the feedback loop!

The feedback from my system goes though the integrating amplifier and is then squared up by the comparator. This signal along with the feedback clock is then fed to the flip flop latch that turns the bridge on and off at the zero crossings. This method of control is better known as delta sigma modulation.

Let me give an example why you need the hysteresis: So your tank just rang up to say 100A and your reference is set at 100A. With a normal OCD the controller would stop the bridge right there and wait for the next interrupter pulse to enable the flip flop again. But for the QCW we would want to re-enable the driver after the current has dropped to an acceptable level. So we uses the clock to re-enable the flip flop. But with no hysteresis in the feedback loop the system would turn back on again after only one cycle! This would result in ringing because the current in the primary LC is not going to ring down a whole lot in just one cycle. So the addition of the hysteresis in the feedback loop makes the system react a little bit more sluggish, avoiding the oscillations. This is at least how I have looked at it so far. hmmm maybe sluggish is not the right word....

I'm not sure if Steve McConner uses the same OCD configuration? I'm familiarized with Steve Ward's driver. Steve do you use the same OCD as Steve Ward? (wow lots of Steve's in there

)

Steve perhaps you could elaborate on how you would implement the delta sigma modulation in a little more detail? I'm curios to the difference if any bettwen the way I'm looking at it and how you would do it.

Dr. kilovolt, That driver looks a lot like a flyback ZVS driver

I agree with Marko in that you should isolate both the feedback and the gate drive. As I said in my other post IGBTs are going to suck hard with ZVS!!! This is because of the charge stored in the device being wasted in heat at turn off. This is probably where most of your heat is coming from! MOSFETs would be a better choice for a ZVS design because of there lack of a tail current.

I "had" a great paper on IGBTs and MOSFETs in ZVS and ZCS systems.... If I find it again I will be sure to link it.

Eric

[EDIT]

Well would you look at that I found that paper after all

Dr. Dark Current

Mon Nov 14 2011, 08:05AM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

Hi,
yes the circuit is the standard "zvs flyback driver" adapted to full bridge with smaller gate resistors to make it run at a higher frequency, also it adds bias to the gates, to make sure they turn off properly. This circuit would probably not work with FETs at higher power, if their drop rises to the point the other transitor which should be off starts turning on, something pops.

The idea of this was just to make something simple and foolproof. Of course I could use active feedback, in which case we could use MOSFETs. I'm just not into complex circuits, and I try to avoid as much control ICs as possible, as they are most often the source of unexpected behavior, are disturbed by interference etc...

first previous 2 3 4 5 next last

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