CCPS (Capacitor Charging Power Supply) / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

CCPS (Capacitor Charging Power Supply)

1 2 3 4 next last

Move Thread

LAN_403

Finn Hammer

Thu Aug 23 2007, 09:41PM

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

All,
I am building a medium/large Tesla coil together with Daniel Uhrenholt, a fellow coiler and toolmaker here in Denmark.
During a period, where progress was slow, due to me having other commitments, Daniel submitted: I am going to have a go at a CCPS for the coil, I`ve seen Marco`s and Ward has done some work on it also.
I thought:"Yeah, right" and forgot about it, but then some promising videos showed up on youtube

And so I thought Hey, this looks good, I`d better pay attention.
So during the next weekend, we got a microsim model set up.

I had some of the big cores lying around, waiting for an opportunity like this, and we started to make some measurements on them.
By applying 40kHz to them, turning up the voltage and watching the triangular current curve, we looked for the point where the curve got an upward slope. This was also the point where they got warm, indicating beginning saturation.
This happened at 50volts/turn.
We also made mock up windings to measure the leakage inductance of the transformer, and it turned out to be 58uH.
Plugging the numbers into microsim, with a series capacitor of 68nF, a resonant frequency of 80kHz revealed some nice numbers:
At the start of the charging cycle, 19A at the end 39A for 15Arms
Drawing this kind of current from a 560V supply gives 8400VA and with 2 cores available, a 16KVA supply looked feasible.
After this we concluded that there was only one thing to do: Built the damn thing.

I had a pcb for a plasma sonic leftover from another life, and we decided that it would be a good platform for the development stage, until we had a working design. I severed it in half,

dumping the H-bridge section, and modified it to start and stop in syc. with the oscillator.

IGBT``s are not cheap, and sifting trough the maze of 20USD+ devices, I stumbled upon this little beauty:

And at 2USD I decided to build 2 bridges with 8 devices in each.
Since I figured not to change one defective IGBT in the case of a bridge failure, I went for a compact design.

A HP214A was gutted and stuffed with quality parts. 3phase inlet, resistive softstart, timer bypassed. dual 3300uF smoothing caps and bleeders. The transformers and bridges. After some shuffling around, this is the current stage:

It does work!

and this is still with the secondary shorted securely.

I know what you are thinking! You are sitting at your little monitors thinking:
Show us your gate waveforms! Show them to us, so that we can tell you that you are a schmuck!

I won``t disappoint you of course.
With a 22ohm gate resistor, there is no overshoot at the zero transition, and the gate charge gets transferred in a little over 300Sn

As soon as voltage is aplied, things get a bit more ragged, with a nasty 20mHz ripple.
I am starting to run into measurement difficulties, such as different output from a Pearson 2100 and a Tek P6021/134. The latter follows the gate signal with greater fidelity than the pearson.

With 15Arms flowing trough the bridges, and a voltage drop across each device of 1.9V, one would assume 3.8 x 15 = 57W dissiption, and the finns get hot in about 30sec, so I guess a big fan has to be installed.

This weekend, Daniel and I are going to work on a voltage probe, so that we can set the desired output voltage.
We also intend to pass the trigger signal from the trigger generator trough this supply, so that we can derive a delay, to kep it disabled during the coils ringdown, and so that it starts charging the cap after the BRISG has quenched.

And perhaps try it out on a jacobs ladder.

We invite you all to comment, and in particular help us add protective circuitry, to keep this beauty running.

Cheers, Finn and Daniel

Steve Ward

Fri Aug 24 2007, 05:15AM

Registered Member #146 Joined: Sun Feb 12 2006, 04:21AM
Location: Austin Tx
Posts: 1055

Nice work so far guys.

For snubbing the GDT ringing, i found that around 5-10 ohms on the *primary* side of the GDT was all that was needed (no secondary side gate resistance). This produces very clean waveforms.

My CCPS currently produces a very healthy arc and about 20KV peak output voltage from a voltage doubler. It works great for charging up my beastly 5kJ 12kV pulse cap.

It should run a TC great as well, just be sure to stick some sort of output filter on the thing. I had some strange issues when running a spark gap coil. I think a SISG will be a little nicer than a spark gap system anyway.

Also, you might still want to watch for core saturation. The way i did it was just put another few turns of wire on the core and scope it. As you bring up the voltage, you should get a good idea of the odd voltage waveform. If you begin to saturate, you will likely see the voltage drop off to zero. I think the inverter is pretty well protected against saturation effects since the magnetizing inductance is typically shorted out anyway (which is basically what happens for saturation). Only thing is that the core gets very hot!

Keep us posted!

Marko

Fri Aug 24 2007, 11:59AM

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

Hi Finn

Awesome project guys. You always find those beautiful metal cases for these things.

Any pics of the monster transformer for me?

For 16kW although I would be worried about those heatsink size.. those two look really wimpy even for like 99% efficiency, make sure to do something in order not to overheat the IGBT's.

Other thing could be bad power factor.. But it's natural outcome of need for smooth output.
I imagine it could only be practically helped by a humongous boost converter like Steve's.

Steve: I always wondered how much sense does it make to use all the diode-resistor mess in gate drive.
If my transformer is made well enough, I thought I can avoid gate Zeners after all, just by using clamp-to-rail Schottky diodes on driver outputs?

Still if I want dead time (which you guys have plenty of here) I need separate resistors and diodes for each gate. (Meh.)

Finn Hammer

Fri Aug 24 2007, 05:11PM

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

Steve Ward wrote ...

For snubbing the GDT ringing, i found that around 5-10 ohms on the *primary* side of the GDT was all that was needed (no secondary side gate resistance). This produces very clean waveforms.

I was eager to try this out, but it didn`t work with my setup.
I tried with 10, 5 and 3.3 ohms, about the same results. Shown here are the 5 ohms traces.
Here is a scope shot of the unloaded gate voltage

The miller plateau is not distinctly visible.

This shot shows the gate with voltage on the bridge

The yellow trace shows the corresponding current curve. The turn on is delayed relative to the other bridge.

Next pic. shows the current waveform, the yellow one.

To me, it says that the IGBT is not conducting fully. The leading current pulse is smaller in amplitude than the freewheling one.
And the bridge is heating more.

Marko wrote ...

Hi Finn

Any pics of the monster transformer for me?

For 16kW although I would be worried about those heatsink size.. those two look really wimpy even for like 99% efficiency, make sure to do something in order not to overheat the IGBT's.

To the left is a prototype of the homemade fiberglass former. This is covered with mylar and other dielectric from stock.
Primary wire is 14 turns 1.1 x 5 mm flat wire, over electrostatic shield. Sec is 0.75mm wire. We may rewind this for more clearance at the hot end.

The cooling finns, yeah, I wish I had thought about that before I made it. I just grabbed a sample, without making any calculations.
Silly Me.

Cheers, Finn Hammer

Marko

Fri Aug 24 2007, 05:56PM

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

Finn you are brilliant, and transformer with this flat wire is wicked cool.
I'm eager to see you progressing.

I wondered if there is actually a benefit from having the primary under the secondary.

Few days ago I was drawing arcs from similarly constructed transformer, except on a flyback core and hard-switched mosfet halfbridge.

It was curious how with no ballast I couldn't get more then some 200-300W (very roughly) into the arc, even after I ramped the frequency to 100kHz and V/turn quite up.

At first I used primary ballast what I thought as reasonable, but in the end without it and no gaps I couldn't get satisfactionary amount of power through the core, even though setup should be good for like kW of power at least. (mosfets got barely over room temperature)

It appears that leakage inductance is really a big limiting factor, what can be done for it?

Steve Ward

Fri Aug 24 2007, 06:54PM

Registered Member #146 Joined: Sun Feb 12 2006, 04:21AM
Location: Austin Tx
Posts: 1055

To me, it says that the IGBT is not conducting fully. The leading current pulse is smaller in amplitude than the freewheling one.
And the bridge is heating more.

Are you saying this was caused by using a resistor on the primary of the GDT? Well, surely this can change the switching speed of the IGBT, but they are definitely turning on completely for nearly all of the half-cycle. My CCPS also had one current peak smaller than the other. I never really figured out exactly why that happened, but my Pspice sim often produced that same result. BTW, i see the same non-uniformity in current peaks in your original scope shot from post #1.

One reason the primary resistance (vs secondary side gate resistors) might be somewhat desirable is that in the event of a IGBT failure and assuming it fails with all pins shorted, you end up killing the gate drive to the other devices. This is good since it prevents the other switch in the half-bridge from switching ON shorting out the DC bus. In this case, most of the gate drive voltage is dropped across the primary resistor, rather than only the secondary gate resistor. Of course, i dont think ive more than 1 failure since implementing the SLR topology, and that one failure was a single IGBT that had managed to arc through the sil-pad isolator to the heatsink (this happened while powering a SGTC and using no filter between supply and TC).

For 16kW although I would be worried about those heatsink size.. those two look really wimpy even for like 99% efficiency, make sure to do something in order not to overheat the IGBT's.

Well, first, its kVA (not kW), and this distinction is very important for the SLR CCPS because it inherently has a lot of reactive power and a lot less real power (much like the "ZVS flyback driver"). But i agree, the 57W figure is pretty optimistic.

Finn Hammer

Fri Aug 24 2007, 07:18PM

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

Marko wrote ...

It appears that leakage inductance is really a big limiting factor, what can be done for it?

About leakage inductance: If you resonate it with a series capacitor, and drive it at 1/2 the resonant frequency, with 50% duty cycle, it becomes your friend.

Read Marco Denicolai`s thesis for a full coverage of the CCPS topology.

Cheers, Finn Hammer

Steve Conner

Fri Aug 24 2007, 07:24PM

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

The second half-cycle is always smaller than the first because you're looking at the beginning of a ringdown.

Marko: the leakage inductance is part of the SLR converter circuit, it won't work if there's too little or too much. Winding the primary and secondary coils on opposite core limbs is one way to get the required leakage inductance. I think Steve Ward prefers to wind both coils together and use an extra series inductor to make up the required inductance.

Finn: are you sure your winding window is optimally filled? :P

Finn Hammer

Fri Aug 24 2007, 07:48PM

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

Steve Conner wrote ...

Finn: are you sure your winding window is optimally filled? :P

I guess I am, sort of.
At one point I thought about winding both coils edgewise with the flat wire, but I found out there was ample cross section this way.

Now about those gate waveforms.
I don``t get it. I am not passing a lot of current and not swiching very fast, still this ringing.
I don`t recall having these kind of problems with the DRSSTC`s, which use similar components, but where I whack the gate open as fast as possible,and the current reaches 400A.
Marco D. slows down his gates to 1100nS, but even this does not work here. I am starting to suspect the long leads on the pcb btwn. gate drivers.

If this is not the case, perhaps there is too much leakage inductance in the gate transformer? Perhaps I should reduce the amount of turns.

But that requires a calculation of the volt seconds of the transformer and I can`t do that. Yet!.
Anyone have an explanation of how to do that calculation?
After optimizing the amount of turns, I could reduce the leakage inductance to half by making the primary as shield, surrounding the 4 gate coils, according to James theDatastreams survey.

Cheers, Finn Hammer

Marko

Fri Aug 24 2007, 09:25PM

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

About leakage inductance: If you resonate it with a series capacitor, and drive it at 1/2 the resonant frequency, with 50% duty cycle, it becomes your friend.

Ok, I ''know'' that much about SLR (what comes from steve's site :p)

I simply wanted to get an idea of difference in leakage inductance with primary-under-secondary and separate style transformer.

What really limits current (with ghetto SG3525-style controllers) when you draw arcs or short the transformer?
There must be a lot of secondary leakage inductance doing this.

If you aren't using a (insanely complex) resonant controller like Marco larger inductor is also needed to stabilize the inductance to a point so it can be tuned by a fixed controller like SG3525. (at least from what I 'figured out').

Apparent inductance drops proportionally with load, so I imagine fixed controller to have trouble without larger inductor.

Finn: I see you copied the transformer much from Denicolai's design. I'm lost with his resonant controller, but how do you keep the thing in tune with only leakage inductance of primary?

Marko

1 2 3 4 next last

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