Strange oscillations

[AstRii]

Hello,
I ran my DRSSTC at very low bus voltage of 30V to scope the primary current and gate drive waveforms and there are some weird oscillations occurring on the primary current.
Could it be just the interference from the secondary? And if not, what could cause them? Can it affect the IGBTs or the coil's performance somehow?
Thank you

[davekni]

I'm guessing (and for your sake, hoping) it's a ring and not oscillation.  Either way, here's one method for helping chase down the location:

Make a small coil of a few turns of wire ~20mm diameter.  Ideally connect this to the end of a coax cable, other end to a scope channel with 50 ohm termination (internal to the scope or external at the scope input).  If coax isn't an option, connect the coil to a scope probe.  Trigger on the other scope channel with your bridge output as in your posted pictures (or current sense).  Use the coil to probe for peak ring signal - probably around the bridge and snubber cap(s), but possibly around the gate driver.

As Mads has pointed out in other posts, such rings can change significantly between low and normal bus voltages because the IGBTs have much higher parasitic capacitance at low voltage.

BTW, I'm presuming the square-wave scope trace is a bridge output, even though it shows 25V rather than 30V.  (Or, is it a gate-drive waveform of some sort?)

[T3sl4co1l]

Probably hard switching, because load current is low and inverter capacitance is high?

Tim

[AstRii]

The square waveform is the drive waveform, one pole of GDT referenced to GND, it should be about 24V.
The primary current on the scope is about 50Apk. There shouldn't be a hard switching when the primary current and gate drive are in phase or am i wrong?
The coil has these parameters:

Resonant freq. ~218kHz
IGBTs: FGY75N60
MMC: 6000V 42nF
Driver: Steve Ward's UD 2.7
Snubber caps: 4 1500V 82nF in parallel
Bus capacitance: 11700uF 450V

[Hydron]

It's not real - the current isn't actually doing that.

What you are seeing is noise being picked up from the primary switching - there is a very high dv/dt so pretty easy to couple into your measurement CT.

I did a comparison of DIY and commercial (Pearson) current transformers here:

https://4hv.org/e107_plugins/forum/forum_viewtopic.php?177823

Unfortunately this was prior to HVF being around and 4hv is pretty broken so you can't get the large size of the images, but you can probably make out that I was seeing the exact same thing with a DIY CT, and even a 0.01V/A commercial CT picked up some noise that wasn't there when a 1V/A CT was used.

If I manage to find the original source images I'll repost the topic here, as it may be of interest (though I didn't find anything particularly surprising!)
Edit: found the images easier than expected, will post them shortly so people can avoid trying to squint at thumbnails.

[Mads Barnkob]

Probably hard switching, because load current is low and inverter capacitance is high?

If you do look at the inverter output voltage, you will see large switching spikes due to the high inverter output capacitance at low C-E voltages. Look at the IGBT datasheet for Coes curves to see why, as C-E voltage increases the output capacitance falls in the range of 5-10 times lower at 100V than at 30V.

[AstRii]

I have some more waveforms for you. I still use about 30V DC Bus voltage.
First screenshot is (yellow: gate - emitter and red is gate drive waveform)
Second screenshot is (yellow: colector - emitter and red is gate drive waveform)
And just for curiosity i scoped DC bus voltage on the 3rd screenshot (red waveform, yellow is again C-E)
The C-E waveform is looking very strange with about 55V peak spikes.
Spikes are at 166% of bus voltage, so at rectified mains voltage (325V) it should be about 512V spikes (if the spikes are linear), the IGBTs are rated 600V. Do you think it's worth a try to run the coil at full power?

[Mads Barnkob]

The C-E waveform is looking very strange with about 55V peak spikes.
Spikes are at 166% of bus voltage, so at rectified mains voltage (325V) it should be about 512V spikes (if the spikes are linear), the IGBTs are rated 600V. Do you think it's worth a try to run the coil at full power?

This is exactly what I just described above.

[AstRii]

I'm so confused right now. I thought the spikes will go out at higher voltages since the capacitance drops. Yet.. this is at 80V DC bus voltage, the yellow waveform is C-E on low side IGBT and red is C-E on high side IGBT.

[Mads Barnkob]

That looks very odd, not sure what to suggest right now...

Your buswave.jpg attachment shows a correlation between ripple on DC bus and the inverter output voltage?

[AstRii]

Yes it is the bus voltage and C-E scoped. I also scoped the inverter output and it looks pretty decent. There are 84V spikes on 80V DC bus voltage, the red sine wave is the primary current. So it seems that only the voltage on C-E of those IGBTs looks strange.

[milkman2046]

Hello
I think this situation is caused by the voltage lagging behind the current caused by the inverter bridge entering the capacitive load area, so this oscillation is generated. The solution is to adjust the phase shift compensation of the DRSSTC driving board to make the voltage enter the inductive area ahead of the current or just cross the zero off section.

[Hydron]

To follow up my earlier post, here is the repost of my old 4hv thread, showing the same "strange oscillations" with the my unshielded DIY CT:
https://highvoltageforum.net/index.php?topic=914.0

[T3sl4co1l]

Show a picture of your layout.

Probably hard switching, because load current is low and inverter capacitance is high?

If you do look at the inverter output voltage, you will see large switching spikes due to the high inverter output capacitance at low C-E voltages. Look at the IGBT datasheet for Coes curves to see why, as C-E voltage increases the output capacitance falls in the range of 5-10 times lower at 100V than at 30V.

Give or take layout, you don't even need to look at the inverter output to see the spikes or ringing -- nanosecond transients are easily carried throughout the circuit.  Careful construction and shielding are necessary.  Random clip leads are the devil's work.

The EMC situation is, uh, further complicated, by the fact that we're talking about Tesla coils that are literally intentional sources of spark-discharge noise.  Good luck telling which is which, but it's all still there -- squiggles from everything, the GDT, inverter, secondary.

Tim

[davekni]

The  ~1MHz ring on your C-E waveforms is probably due to inductance from your snubber caps to the larger bulk cap.  If you scope Vbus at the IGBT terminals, you'll see similar amplitude ringing as across C-E.

Is this a half-bridge design?  If full-bridge, are two snubber caps placed on each IGBT pair, with separate wiring back to the bulk VBus cap?

1MHz is roughly 160ns per radian, so about 0.5 ohms impedance for 328nF, or 1 ohm for each IGBT pair (for each pair of 82nF caps).  A 50A step into 1 ohm impedance would make a 50V ring.  Depending on your wiring configuration, a full-bridge may have even worse Vbus ring at the end of a burst, when the MMC/Tesla Primary coil energy is dumped back into the VBus bulk cap.  (That's the issue I had with my initial DRSSTC VBus wiring.)

If you can share a picture of your IGBT/snubber/bulk-cap wiring, that would help.  Copper planes work much better than wires for this.

[AstRii]

I didn't want to use wires yet i didn't have access to any copper bus bars so i designed a simple PCB for the job.
There are 4 snubber caps 82nF 1500V in parallel, diodes and resistors for slow charge and fast discharge of the IGBT gate, and 400V TVS diode on every IGBT.
The IGBTs are placed from the other side on a heatsink.

[davekni]

Looks like you have differential scope probes, or a scope with individually-isolated input channels.  That's wonderful for debugging circuits like this!  Do you mind sharing the scope and probe models?

Circuit boards are great for this power level.  You have a nice layout, with gate drive coming from one side of the IGBTs, and the power connections on the other side.  That's the closest to Kelvin connections that can be done with TO247 packages.  If you ever redo the ECB layout, fill all the unused space with VBus- on one side, and VBus+ on the other side within the center power area.  That will lower inductance dramatically.  You could get some inductance reduction by manually soldering copper foil to the existing ECB.  Would probably require removing the snubber caps to add foil under them.  This rework could be done all on one side, with overlapping VBus- and VBus+ foil separated with mylar or polyimide tape.  (My DRSSTC uses ECBs with added copper foil for 40 TO247 IGBTs, 10 parallel per leg.)

Studying your scope traces further, it appears that there is another issue that is exciting the 1MHz snubber/trace_inductance resonance.  There's evidence of cross-conduction (shoot-through) within the bridge.  In particular, the high-side IGBTs appear to turn on before the low-side IGBTs turn off.  The resulting high-current spike across VBus pulls it down locally (at the snubbers), then it rings back during the remaining part of each half-cycle.

The FGY75N60 IGBT spec. shows a dramatic slow-down in turn-off time with increasing gate resistance.  Driving the gate to ~22V may also contribute to the slow turn-off.  The spec. shows typical performance to be fine at 12Vge, so nominal 15Vge should be sufficient.  That would allow more headroom for overshoot in the gate waveform, allowing you to reduce the resistor in series with the diode below 5 ohms.  Anything you can do to reduce leakage inductance of your gate-drive transformer will also help.  How is your gate-drive transformer constructed, including wiring from the UD2.7 board to the power ECB?

Also, your inverter2.jpg image shows quite a bit of asymmetry between the high and low IGBT Vce waveforms, with the low-side appearing to turn off much more slowly.  Is there a chance that one of the low-side gate-drive resistors or diodes is bad (open or high resistance)?  Is there anything else asymmetrical about the gate drive, such as in the transformer?  Are the gate-drive transformer leads well paired (twisted) from the transformer to each IGBT connection and to the UD2.7?

Thank you for the detailed scope plots.

[AstRii]

Actually i don't have any differential probes or isolated scope. I only use a boost converter to step up voltage from isolated lab bench power supply 12V to 80V.
I checked the diodes and the resistors and they seem to be fine. The gate charging resistor is 15R and gate discharging resistor is 5R6. My gate drive transformer is a 1:1 on a ferrite core with approximately 15 turns.
It's all connected with cat5 twisted pair cable.   

[davekni]

How did you scope the inverter output and high-side Vce?  Did you ground one side of the inverter output?  Stray capacitance across the "isolated" supply can distort waveforms when grounding active (non-DC) circuit nodes.  It might be helpful to see scope traces of both inverter outputs with the scope grounded to VBus-.  Hopefully that will make it clear if there really is cross-conduction or if it is an artifact of grounding active nodes.  A scope trace of VBus at the IGBTs would also be helpful - ground to a low-side emitter and probe to a high-side collector, both on the IGBT leads (lead stubs on the back side of the ECB if you can clip to them).

Is the gate-drive transformer wound with CAT-5 cable, then separate CAT-5 cable from the transformer to UD2.7 and from the transformer to power ECB?  Within the transformer, is one wire of each twisted-pair used for each gate, with the remaining four wires (one from each pair) connected in parallel for the primary?  That's the best configuration.  It also helps to use four twisted pairs in parallel from UD2.7 to the gate-drive transformer, as any inductance in that wiring is four times as significant as output-side inductance.

Hope you don't mind all the detailed questions.  I don't know any other way to help with debug remotely.

[AstRii]

I scoped the inverter output directly on the output of the inverter (OUT1 and OUT2 on PCB). And high-side VCE across IGBT leads C-E.
This is the inverter output referenced to Vbus- (yellow is Vbus- and OUT2, red is Vbus- and OUT1)