Looking for DRSSTC help/advice

[GKnapp]

Hello all,

This is my first post, spurred by not being able to figure out what on earth is going on with my first DRSSTC attempt! The physical dimensions are meant to be roughly a clone of Loneoceans DRSSTC1, so:
~225 khz secondary frequency with topload,
~180 khz primary (ish--I know this is too detuned, but I'm having trouble even with a dummy load instead of the secondary)
UD2.7 driver (through hole version from profdc9)
Full bridge with FGA60N65SMD igbts, in profdc9's full bridge board
simple interrupter based on loneocean's sstc guide, does frequency from 0-270hz and pw from 0-250 or something.

Here's the situation: there's some really odd behavior that I can't figure out from the bridge. This is all true whether or not I am using a dummy load. At low bus voltages (0-20 volts), all seems to be well--I get okay-ish waveforms between gate and emitter. They're a little ragged but seem okay. When the bus voltage gets higher though, it starts to get weird spikes, the pulsewidths actually get larger (which I really don't understand) and then eventually it totally turns into goop. The relevant voltages in the pictures are on the order of something like: 20v, 30v, 50ish volts.





Other odd behavior: this is true whether or not I'm giving it feedback with a signal generator, or with a ct. Also, the UD seems to find feedback from somewhere that I don't understand: even with no ct or signal generator attached to it, it'll start oscillating once I get above a certain bus voltage. Spooky.

I've moved the interrupter far away and it has a fiber optic, so I'm convinced that it isn't interference from there. And this happens no matter if there's a secondary or not, which makes me think it's not interference in general, but I don't know.

The whole setup will light up a little neon bulb, barely, at the highest voltages. I've disconnected all non-essential elements from the bridge, and as far as I can tell none of the rest are blown. I would love suggestions of how to go about diagnosing the problem. I've only built a simple SSTC before, and so this is quite a bit harder for me to wrap my head around, as I don't actually have any electronics background other than what I've picked up from this hobby over the years!

[AstRii]

Did you shield your UD2.7?
Tesla coils create a very high EMI, it can cause a glitchy behaviour.
How far away is your primary coil from your driver?

[GKnapp]

It's quite close because I didn't want long wires going everywhere, and no I don't have it shielded because I wanted to test things were working before I started mounting them/sticking them in boxes. Maybe I'll try throwing some aluminum foil around stuff and see if that changes anything!

I was under the impression that interference wasn't really something that occurred with a dummy load instead of a secondary, but do you think just the primary switching might be messing stuff up?

[AstRii]

I would say it's definitely an option. The primary coil creates a lot of interference, after all the voltage/turn on primary is usually even higher than on secondary.
You can even see the EMI from primary with a camera, like on this video from Kaizer at 0:28

I would try to shield the driver and also to scope the feedback CT signal and try to locate where the glitches starts happening.

[GKnapp]

Throwing a bunch of aluminum foil around the electronics appeared to very slightly help, but not really by much. I'm still getting weird double humps on the scope at higher voltages:


As for the CT, I can't really tell what I should be looking for--it just looks like a complete mess, and when I stop the scope there are sort of sine waves going on. I'm using a 289:1 current transformer, which I thought would work well for a coil that I intended to stay below 300A...do you think it could somehow be related to my ct construction? ON the other hand, I was still having these problems trying to use a signal generator.

My shielding is relatively shoddy, but I still wouldn't think it should cause this much of an issue: I mean honestly this thing doesn't look like it has an ounce of shielding: https://www.loneoceans.com/labs/drsstc1/ and it's what I'm basing my coil geometry on. I don't tend to see people entirely encasing their driver boards in metal boxes, and most seem to get away with it just fine :/

[davekni]

The need for shielding depends on many factors, especially ECB layout and cabling.  Twisted pairs are best, or at least bundled together.  Bundle and/or twist power wiring (VBus supply and H-Bridge output) with itself, separately from signal wiring bundles, never together.  Wiring with open loop area both sends and receives magnetic interference.  Of course, sending is of issue for power wiring, especially within the H-Bridge and bridge-to-bulk caps.  Receiving is an issue for signal wiring.  Some of us do put electronics in shielded boxes, especially for designs with control close to the coil.  My first guess is that your issues are in wiring rather than in the ECBs themselves.

Pictures of your setup from various angles would lead to more useful and specific suggestions.

At least with repeatable issues, you can scope back through the driver to see where noise is introduced.  But, it's probably better to clean up wiring first rather than hunting for possibly multiple noise issues.

Good luck!

[GKnapp]

Thank you for the encouragement! That's something I hadn't considered--in my haste to see if it would work before I put in in a nice frame and did all the finicky wiring work, maybe I ensured it wouldn't work by not setting it up nicely enough with respect to its wiring and geometry.

Right now there are plenty of alligator clip type connections and such, though I do believe I have all the signal wiring twisted in pairs. Could all my problems be from shoddy connections and wiring layouts?

In any case, tomorrow I'll take some pictures of my testing setup so you all can at least see the boards and such, and I'll try to clean up the wiring as much as possible.

[AstRii]

DRSSTC is a very fragile and very sensitive system
All your problems could easily be made by bad wiring. Don't forget that even a single cm of wire has inductance which at these frequencies is noticeable.

Alligator clips are the worst Although it may seem as a solid and stable connection, they have very high resistance, a year ago i used them to connect one end of primary coil to MMC and the system didn't start to oscillate at all.
I eventually had to give up the idea of using alligator clips to tune the primary and used solid solder connection which solved the issue

[GKnapp]

EDIT: turned the UD2.7 around 90 degrees, and that seems to have entirely solved the issue, at least up to 60 volts. Now it's time to stick a secondary in there and see what happens!


Here's my testing setup as it currently stands, actually much improved from yesterday if you can believe it. The three scope pictures show what's happening--the yellow is the gate-emitter trace, and the purple is the feedback ct. At low voltages all seems well:

but then at higher voltages I start getting feedback spikes right in the beginning:


and finally it turns into a complete mess at around 50-60 volts:



Based on this and what you all have told me, my guess is that the inputs of my feedback CT are too close to the bridge outputs, as they come in the same side as one of the bridge outputs based on the way I've stacked the boards. So when voltage gets too high, the CT goes crazy at least in the beginning and starts messing up the pulses from the UD2.7.

Unfortunately, the full bridge board from profcd9 has bridge outputs and bus inputs alternating on sides, instead of both on one side or the other, but if I stacked the UD2.7 with a different rotation, I could have the feedback come in on a bus supply side, not a bridge output side. I'll try this next and see if it helps!

If that doesn't fix the problem, I'd love to hear any other ideas.

More explanation on my setup since I know it looks a little confusing:
The alligator clips come from a dc supply for the bus. the red and black wires from a dedicated 24v supply for the driver. My cap (~.1 uF in theory, but more like .09 based on measurement) is directly soldered to one of the bridge outputs. The CTs have 17 turns of speaker wire around each toroid, and so are supposedly 289:1 ratios. The GDT is a bunch of twisted 28 awg magnet wire. I used to have it as CAT5 cable, but I originally thought that the GDT must be the problem, so I remade it. As far as I can tell, it works just fine, and having it around the brass support doesn't seem to affect it, though I know this probably isn't good practice. The interrupter isn't in the picture, but its fiber optic cable can be seen coming into the receiver in the back. Right now it's all in a milk crate for convenience. Wrapping most of the outside of the milk crate in grounded aluminum foil doesn't appear to make a difference in terms of interference.

So again, my new theory is that the CT wires are just too close to the power and bridge cables. Does this seem like a reasonable conclusion?

Thank you all again for all your help!

[AstRii]

It most definitely sounds as a reasonable conclusion, but don't forget that in a DRSSTC system, anything can be a cause of such odd behavior

I like the design; how close are all the components together, this will make for a low inductance and compact Tesla coil.

Also (probably not the cause of the issue), but i would get rid of all the excess leads of the THT parts.



As they are basically antennas waiting to pick up interference from the IGBTs switching high currents just a few centimeters below.

Good luck, i hope it will work soon

[GKnapp]

Thank you! Moving the CT wires farther away did indeed seem to fix the interference issue for the moment. I stuck my secondary in there, but still not really any output. It'll light a small neon bulb, barely, and seems to do this no matter what combination of primary direction and feedback direction I choose. At least the switching stays even though. It's still confusing though, because in theory it's a full bridge switching +- 60 volts. I would expect more from it... Based on Java TC, my primary and secondary shouldn't be all that far off, and I would certainly expect more than a barely flickering neon bulb at that much voltage.

Any advice on obvious things that I'm not trying? My secondary is grounded, and I've tried all combinations of primary lead and CT directions... Switching the phase on the ud2.7 just leads to nastiness.

[AstRii]

Even Java TC is not perfect, the primary resonant frequency depends on so many more factors than any calculator can take
Don't forget your primary is far from the bridge, that long cables can decrease the resonant frequency as if you wound another turn(s) of primary

I think you're probably just out of tune.
I always tune the primary to have exactly the same resonant frequency as the secondary and then just add primary turns to lower the resonant frequency until i get the best output.

[davekni]

Your wiring looks much cleaner than I was expecting for the issues showing up!  You have nice tight twisting on control signals, and only the one large loop on H-Bridge output that would be tricky to avoid given ECB connection points.

The standoff through GDT core might be a problem.  Depends on whether the circuit board mounting holes are electrically connected or floating.  If mounting holes are connected together through ECB copper, then the standoffs form a shorted turn on your GDT.  If holes are floating, there's no issue.

Magnet wire for GDT (with it's thin insulation) makes for very low leakage inductance (good), but has two possible issues.  First, as you raise bus voltage, there is some risk of insulation failure.  Second, inter-winding capacitance is higher, injecting current from high-side Vce transitions into gate-driver circuitry.  I'd suggest going back to CAT5 or other such wire.  However, that may not have anything to do with your existing noise issues.

What cores are you using for GDT and CTs?  The CT feedback signal does not look much like a sine-wave even with lower voltage before oscillation starts.  I'd suggest focusing on that CT feedback signal at low voltage for debug.  It should look much more like a sine wave.  Actually, during the first half-cycle of H-Bridge output, the CT waveform does look like a sine-wave at ~300kHz.  The driver isn't picking up the initial transition and running at a lower frequency.  Looks like the initial issue may be phase-reversal of feedback CT wiring.  Try swapping feedback CT output leads.

Concerning the extra lead lengths below the ECB, cutting them off is a good idea, more to avoid accidental shorting.  At the frequencies and impedances involved here, the antenna effect will be negligible.  Magnetic coupling is by-far dominant (except for cases like GDT capacitance of extended adjacent wires).

[GKnapp]

Hi Davekni,

I was embarrassed by how it was yesterday so actually cleaned it up a bit! And moving the CT farther from the bridge outputs solved the noise issue for now. The CTs and GDT are n30 epcos material. The brass supports are floating. For the moment I think that the gdt is working fine, but I may well go back to CAT5 in the future... Now the waveforms look like this:


Swapping the CT leads doesn't appear to make any difference in the coil output. In fact, I can't detect really any differences in output from swapping the CT leads or the primary leads, or where I'm tapping the primary. It's very confusing to me, so I figure I must be making some sort of basic mistake. All configurations barely light up a neon bulb, some brighter than others, but there's not really a big difference anywhere between 4-10 turns which makes no sense to me.

Does it matter which end of the primary the CTs are on? Currently I have the end of the primary not connected to the resonant capacitor going through them.

[davekni]

Yes, there seems to be some basic issue that's preventing a clean primary tank circuit resonance.  For initial low-voltage testing, I'd suggest removing the tin-can load, leaving the primary coil unloaded.  That will provide the highest Q.  The can may add too much damping.

Next, have you scoped both H-Bridge outputs simultaneously to verify that they are 180 degrees out-of-phase?  If they are, and you still don't get a sine-wave from feedback CT output w/o tin can, then I'd wonder if your 0.1uF resonant capacitor is faulty.  Do you have specifications on that cap?  Most designs use an array of smaller caps (called MMC) to achieve reasonable total AC voltage and current capability.

BTW, my DRSSTC is permanently built into 2.5 six-gallon milk crates.  They are great building blocks for people like me who aren't mechanically skilled.  (Half-crates are available commercially.)

[GKnapp]

This is what I get scoping the bridge outputs--I scoped them before the capacitor. The first image is just the plain primary, the second image is with the secondary. This is all happening at about 60vdc on the bus. I don't really know enough to interpret what it should look like... the individual curves look like they're going opposite directions, but maybe I'm misinterpreting. My multimeter has a capacitance meter built it, and it still says the capacitor is 98 nf, which is what it should be... it's a cornell dubilier 942C 2kv capacitor. I could use four of them instead I suppose, but I've seen designs using just one that seem to work fine.

[johnf]

Are you still using that tin can down the center of your primary??
It is a shorted turn that is damping your circuit.
try taking it out and scoping again low volts of course

[GKnapp]

Hi JohnF,

In my latest post I wasn't using the tin can--picture one is with nothing in the primary, picture two is with the secondary.

[davekni]

Looks like several scope triggers plotted on top of each other (as some scopes have modes to do), so it's hard to tell.  However, I suspect it's showing an issue with GDT phasing.  Looks like both IGBTs of one half-bridge are being enabled together instead of alternately.  Run with zero bus voltage, trigger on one gate signal, and scope the other three.  One low-side IGBT should have the same gate-drive polarity as the opposite high-side IGBT.  The two IGBTs within each half-bridge must have opposite gate-drive polarities.

[GKnapp]

I think that might be it! The documentation of the board was confusing to me--the silkscreen has "+" and "-" markings on all gdt inputs, and I believe that I assumed this meant that I should have one phase be "+" and one "-" for all the gates, but now looking back on it, I'm sure I need to have it opposite on either the high or low side. I'll switch this up and then report back