Horrible waveforms advice--just bad test setup?

[GKnapp]

Hello all, trying to finally get a coil up and running. Finally put the primary and secondary on to check my feedback and primary phasing.

The coil works--starts oscillating at about 10 volts on the bus, and I can get about an inch spark to an object with 30 volts.

These are the G-E waveforms I'm getting. Before I put it all together with a primary and secondary for testing the feedback, I was getting not perfect, but acceptable square waves. I'm using an aliexpress ud1.3b so it has no phase lead, but seems to work quite well.

With it all put together, I'm getting these horrible things. My test setup isn't truly horrible, but it does have overly long primary wires so I can move them around. The gdt is pretty normal--15 turns or so of cat5 cables twisted together, and about as short as they can reasonably be to the transistors. Could these waveforms just be that my test setup has much too much leakage inductance, bad grounding, etc. or are they bad enough that it implies something else is wrong? Again, before I connected the primary and secondary, the waveforms were mostly okay.

[alan sailer]

I can see one possible issue with the scope leads. Since there are such large magnetic fields
generated by the primary current you can easily induce voltages on any conducting  loop. The
loop formed  by the ground lead of the scope probe can be a problem. It would be a good idea
to wrap the ground lead around the body of the scope probe to minimize this loop.

Steve Ward wrote a great paper on the pitfalls of measuring DRSSTC wave-forms.

https://hobbydocbox.com/Radio/67533485-Testing-and-verification-waveforms-of-a-small-drsstc-part-1-steven-ward-6-24-2009.html

Cheers.

[davekni]

Scope ground loop is a possibility as Alan mentioned.  However, I'd expect that to pick up mostly sine-wave field from primary current.  The distortion doesn't have that look.  I'd be more suspicious of a GDT phasing issue - one of the four outputs may have leads reversed.  It is possible that you got some output without running long enough to fry IGBTs.

Looks like the large white cap is your primary resonant capacitor.  What is the array of smaller rectangular capacitors on vectorboard, light gray in color?

[GKnapp]

After some more testing, I think I must have grounding issues. Even with only one scope between g-e, the bridge was drawing too much power and not working (a different issue from last time..). Not sure why this would be--I'm using a dc bench supply for testing, so everything should be isolated... I know not to use two scopes at a time, but I was able to measure across both igbts, and one starts up on the pulse and the other down for the pulse, so I'd assume I'm good on gdt phasing.

On the principle that if I don't measure it it isn't a problem, I removed all scope probes and it seems to work fine. Draws very little power at 30v, and making about an inch spark to a screwdriver. Based on prior experience, I'd think this means I have all the feedback and other phasing right, as it shouldn't work nearly so well on only 30v if it weren't correct...

Dave, the little array of capacitors is between ground and the negative bus and ~.1uf--on Mads' site, he explains that this can help protect some electronics in cases where there is an arc-over or something by allowing an easy path for the high frequency ac.

[davekni]

I know not to use two scopes at a time, but I was able to measure across both igbts, and one starts up on the pulse and the other down for the pulse, so I'd assume I'm good on gdt phasing.

Are you using two probes at a time?  If either probe is on a high-side IGBT Vge (grounded to a high-side emitter), that will cause problems.  One issue is if the other probe is grounded to a different IGBT emitter.  The other is that the 0.1uF from Vbus- to ground is about 8 ohms reactance at 200kHz, so is close to grounding Vbus-.  Yes, that Vbus- to ground capacitor is useful, though typically not necessary to use an array of capacitors.  You could disconnect it for low-voltage scoping.

[GKnapp]

I should have said probes not scopes--i know not to use two at the same time, but it was having troubles even with one, which I found odd. I pulled off the secondary, and it had no issues, so I'm a little mystified. But it was all seeming to work and relatively in tune, so I tested it at about 60vac and it made pleasant sparks without drawing too much power so I think I'm good.

Would it be better to reduce the capacitance for that capacitor?

Also, I'm using a ud1.3b and I believe it has a 10ohm OCD resistor. I read that I shouldn't turn the pot voltage higher than 3.75 ish volts for that model, so I think I can't have an effective ocd with my current setup (aiming for max 300a, 289:1 current and feedback transformers). Do I need to parallel or replace that resistor with something lower--and is there a best  way for me to test the ocd? I only have a signal generator that goes to +-2.5 volts.

[davekni]

Would it be better to reduce the capacitance for that capacitor?

0.1uF is fine, except if scoping a high-side Vge.  Even with a single probe, grounding a high-side emitter (an H-Bridge output) will cause problems.  A differential probe is best for scoping high-side Vge with Vbus powered even at low Vbus voltage.

Also, I'm using a ud1.3b and I believe it has a 10ohm OCD resistor. I read that I shouldn't turn the pot voltage higher than 3.75 ish volts for that model, so I think I can't have an effective ocd with my current setup (aiming for max 300a, 289:1 current and feedback transformers). Do I need to parallel or replace that resistor with something lower--and is there a best  way for me to test the ocd? I only have a signal generator that goes to +-2.5 volts.

Yes, you will need to reduce the 10ohm resistor to 3.75 ohms or a bit lower.  For testing, I'd run the coil with OCD set low.  If you want to check with the POT at 3.75V, test initially before reducing the 10ohm resistor value.  Then OCD will be ~120A.  Or, to test the final driver setup, loop the H-Bridge output lead through the CT 2 or 3 times (if the wire fits).  Then OCD will be 1/2 or 1/3rd of the final value.