Building my first DRSSTC

[drobotk]

I have some questions regarding grounding.

Do I connect the metal housing of the UD to the same ground point as the bottom of the secondary coil? Also, should I connect the UD's 0V rail to the metal housing or leave it floating?

When using a proper ground point for my coil and not mains ground, then should I still tie the mains ground to it at the power input? Or can I just ignore the ground wire from the power plug?

I have a computer in my workshop and I'm slightly worried I might damage it by having some crap going back into the mains.


Cheers

David

[flyingperson23]

What I do is tie mains ground to all metal housing and grounds, including 0v of the ud, and have the secondary base and strike ring on a separate rf ground not connected to anything else. Not sure if this is optimal but it works for me.

[alan sailer]

flyingperson23,

What is an RF ground in this case?  I have no agenda or criticism but am just curious.

I've always just picked a common AC (mains) ground point and bring everything to this point.
I avoid ground loops by running direct from the object to be grounded to the common ground
(ie no daisy chains).

To avoid crowding of crimp connectors the ground "point" is a hefty strip of copper buss bar
(2cmX3mmX10cm).

Cheers.

[flyingperson23]

RF ground would be a dedicated ground rod or counterpoise. Only using mains ground is definitely not recommended for bigger coils and tying everything together on one coil made my UD3 crash every ground strike. Haven't had a problem since I separated rf and mains ground.

[Mads Barnkob]

I tried to cover every possible aspect in the grounding and EMI part of the DRSSTC design guide: https://kaizerpowerelectronics.dk/tesla-coils/drsstc-design-guide/grounding-circuit-protection-and-emi/

[alan sailer]

flyingperson23,

Thanks.

That explains why I never had a problem with my single ground. Ground strikes worry me
so I set up the breakout so they don't happen.

Cheers.

[drobotk]

I have finally attempted first runs with a static load and at only 90VAC. I don't have a variac so not being able to ramp up slowly really added to the adrenaline! 

Here's what I've managed to capture.

First I tried to measure H-bridge voltage only - probe tip on 1 bridge output and probe ground on -Vbus



Then to measure primary current I attached the second probe (CH2) across the 51ohm feedback resistor in the UD2.7c. Currently CTs are 1:16:16, so here around 25A/div if I'm doing my math right.



So my thoughts now:
- Bridge output looks decent. Those spikes seem quite big, but from what I've heard, they don't get much bigger at higher Vbus so should be fine (?).
- Current measurement is very noisy. I didn't take a picture but when measuring only current by disconnecting the other probe completely I still get a lot of noise. So much noise in fact that I'm wondering whether those spikes are actually there or they are artifacts of bad probing.
- It looks like the driver is switching past the zero crossing instead of at or before it. I'm wondering if maybe there is some phase shift coming from my current transformers. I've made them with some random ferrite rings. I've had  3 same ones on hand and I figured it might work, but now I'm not so sure.
- Phase lead adjustment attempts yielded no noticable change in the scope captures.

As for probing, I didn't take pictures of my setup, but what I'm doing is wrapping the excess ground clip wires around the probe tips. Maybe there is something more I can do to improve my measurements.
Or perhaps my pocket oscilloscope thing (Hantek 2D72) is not the best for this.

I'd love to hear what you guys think.

Cheers,
David

[flyingperson23]

The spikes should be fine

Switching after zero crossings is the expected behavior without phase lead, as you have to wait for the switches to actually turn on. Then you use phase lead to switch earlier at the zero crossings.

The spiky current measurement looks like scope/probing error to me.

[davekni]

Current measurement is very noisy. I didn't take a picture but when measuring only current by disconnecting the other probe completely I still get a lot of noise. So much noise in fact that I'm wondering whether those spikes are actually there or they are artifacts of bad probing.

Presume "disconnecting completely" means that probe ground clip was disconnected too.  That was a great step to take.  Next step is to move other probe tip from active side of 51 ohm resistor to ground side where probe ground clip is connected.  See if spikes still show up when measuring ground.  If so, it is a scoping issue.  If not, it is more likely a real issue.  51 ohm resistor with excess internal inductance?  An error in H-bridge/MMC/primary circuitry creating capacitance across primary coil?  Or possible an issue with the scope itself, displaying very short nanosecond level spikes as wider.  Very short spikes could be due to CT inter-winding capacitance.

- It looks like the driver is switching past the zero crossing instead of at or before it. I'm wondering if maybe there is some phase shift coming from my current transformers. I've made them with some random ferrite rings. I've had  3 same ones on hand and I figured it might work, but now I'm not so sure.
- Phase lead adjustment attempts yielded no noticable change in the scope captures.

Spikes also suggest phase lag.  Perhaps phase-lead inductor on UD2.7 is too small a value to be useful?  Or input jumper is in wrong position, bypassing phase lead?

As for probing, I didn't take pictures of my setup, but what I'm doing is wrapping the excess ground clip wires around the probe tips.

Sounds like good technique.  If above steps don't help, a picture might show some other issue with setup.

[drobotk]

Hi again

First of all, I'm really sorry for keeping this thread quiet, I just kinda suck at documenting things.

Here's what happened since the last updates.

I spent around 2 days trying my best to adjust phase lead, but no matter what I did and how I measured, I'd always see spikes on the bridge output. Feeling defeated (and losing patience) I decided to set it by eye as close to ZCS and call it good. I figured I'd blow these IGBTs anyway and I really wanted to finally make some sparks after so many years of waiting and building.

Following that I ran more static load tests, this time at full 325V on the bus. I cranked my interrupter to the highest BPS and longest on-time and it ran fine for a good few minutes. The heatsink barely got warm.

That gave me a lot of confidence so I decided it was time to put the coil in and get first light. Here's a recording from one of the first runs.

In the video you can hear my friend telling me that he saw a racing spark across the secondary winding, followed up by me asking if it was just a single quick spark, to which he said yes. I didn't think much of it and nobody has ever seen another racing spark since then. You can also hear me saying something along the lines of "I'm barely cranking it", talking about the on-time on the interrupter.

In the days that followed I tried to tune it a bit more. I eventually got to a point where the arcs would constantly hit the ceiling, so I lowered the coil from the bricks it was standing on and also moved it a bit forward (the roof is sloped, so it got further away from it).

This enabled me to tune it even further, reaching about 60-70 cm of sparks. This unfortunately lead to the inevitable IGBT failure I was dreading.

I'm still very pleased with what I managed to squeeze out of those small cheap IGBTs. I think with proper ZCS and better tuning, I could've kept using them. TO-247 IGBTs are obviously way too small for this size of coil though.

Specs at death (limited accuracy on some):
Secondary: 32.5cm of 0.2mm wire on 11cm diameter, ~1500 turns
Primary: ~7.75 turns of 2.5mm2 wire (lol)
MMC: 6s10p of 47nF 1000V for 78nF 6000V
Fres: ~120kHz, primary at ~100kHz
Inverter: Full bridge of DGTD65T60S2PT 650V 60A
Driver: UD2.7c
OCD: 200A only
Bus: ~1mF charged to 325VDC from a single phase 230VAC

Thanks to all of you who helped me with your insightful answers to all my questions, I'm sure I would've blown the bridge way sooner without your valuable advice

David

[drobotk]

I began to look for bricks I could buy, and I found a really tempting offer for two SKM400GB066D bricks - $84 for two pieces with snubbers included.

I'm wondering about the suitability of those for my coil. Can such big bricks handle 100kHz? Also, they are only 600V, which wouldn't allow me to ever try 3 phase power.

What do you guys think?

Cheers

David

[Mathieu thm]

I began to look for bricks I could buy, and I found a really tempting offer for two SKM400GB066D bricks - $84 for two pieces with snubbers included.

For the bricks, i think the SKM400 are oversized for this coil. for comparison i have a 160mm coil with skm300,  and I have a coil similar to yours with skm150 that works at 400A+ OCD at 125Khz.
For me there would be no benefit to go above skm200 for this coil.

[drobotk]

Hello again

I managed to get 3 pcs of 2MBI100NC-120 100A (200A) 1200V IGBT bricks for free from scrap. They are very old and slow, so I wasn't sure if they would work in my coil at 100kHz.

I decided to test for shoot-through by measuring across the supply terminals of the half-bridge brick.

Test setup (EDIT: I forgot about the antiparallel diodes in the IGBT symbols, whoops!):


At first, RG was 5.6 ohms. This is what I captured:


After seeing this I decided to try to increase dead-time by changing RG to 12 ohms. Gate signal got less steep, but didn't improve the situation.


What I find puzzling though is that there seems to be something happening even at the first gate drive edge:


Both gate signals, for reference:


Makes me wonder whether my setup and/or thinking is even correct. Is this a correct way to measure shoot-through? And if what I'm seeing is indeed cross conduction, how would I go about eliminating it? Increase gate resistor even further or ditch these bricks all together in favour of some faster ones?

Thanks

[davekni]

Makes me wonder whether my setup and/or thinking is even correct. Is this a correct way to measure shoot-through?

Reasonable idea.  But realize that you are testing at low voltage and no load current.  Times change with more realistic operating conditions.  With normal current and phase lead (turn-off before zero current), goal is to have IGBT turn off and Vce rise to Vbus before opposite IGBT turns on (and before current reverses polarity).

Unless you have two differential probes, would be better to have resistor at top in your schematic, with scope probes grounded to 0V node.  Even if power supply is isolated, will have capacitance to ground.  Would also be better with much smaller resistor value to get more realistic switching currents, such as 1 ohm instead of 10 ohms.

The current you are measuring is likely mostly due to charging and discharging IGBT/diode capacitance.