Ringing makes me scared / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

Ringing makes me scared

1 2 next

Move Thread

LAN_403

Josh Johnson

Thu Sept 17 2009, 03:30AM

Registered Member #793 Joined: Sun May 20 2007, 06:50PM
Location: Minneapolis, MN
Posts: 35

Ok, so I put together a DRSSTC with a full-bridge PCB that may have a bit too much parasitic inductance. I measured Vce on one of the IGBTs at low voltage, ~35VDC, with no secondary coil in place. Check out the attached waveforms. One is the full ringup/down and the other zoomed on an individual turn-off.
So far I have been running off rectified main voltage, 160VDC, with no problems. I now have all the circuitry in place to voltage double the input so I'll be switching 320VDC. Now here is where it gets scary. The spike at turn-off is roughly two times the DC bus voltage so at 320VDC that's 640V which smokes my poor ol' IGBTs. Now I have 400V TVSs very near the IGBTs but can I rely solely on them for protection?
The engineer in me wants to attack the problem with a new PCB layout, snubbers, etc. but the cheap, lazy guy in me wants to do nothing and have everything be fine.

Any opinions welcome.

Thanks,
Josh

...

Thu Sept 17 2009, 04:15AM

Registered Member #56 Joined: Thu Feb 09 2006, 05:02AM
Location: Southern Califorina, USA
Posts: 2445

It seems like with no secondary in place, lots of ringing would be expected, where do you expect all of that energy to go

Drop in a secondary (or lump of metal, anything to absorb some of the circulating power in the tank circuit other than the resistance of the fets/primary) and see if that helps

Arcstarter

Thu Sept 17 2009, 05:07AM

Registered Member #1225 Joined: Sat Jan 12 2008, 01:24AM
Location: Beaumont, Texas, USA
Posts: 2253

Attached is what my primary current looks like with no load. The second one is with a load.

The one with no load has 7 cycles of ringup, and 3.5 cycles of ringdown. With two MOTs as a primary load, 7 ringup and 2 ringdown cycles. That is at 60us on, and 180Khz.

My point is, if you do not have a load in place, you can expect all of that extra energy to go into the ringdown. This will be dissipated in the circuit in different places. With a load, like the last pic, there are less ringdown cycles, because the load absorbs most of that energy, dissipating it as (in my case) heat.

You should have a nice juicy low inductance low ESR film cap across the supply rails, as close as possible to the FETs. Do you have one? What are the specs?

A new layout also works, but if you do not have time at the moment, or just want to tinker with this one some more, just solder a cap on real quick.

If you haven't, you can measure primary current with a CT, 1 turn primary connected in series with the primary, and how ever many secondary turns there are, needs that divided by ten ohms of resistance in parallel with the output for burden. For example, a 1:100 CT needs 10 ohms of burden. Then you can scope that. Each volt measured on the scope is 10 amps.

Thanks to TheBoozer for all the help, who provided motivation and taught me most of what i know.
EDIT:

Just like i was telling myself *not* to do, i forgot to attach them. Here are the links...

No load.

Loaded.
Both are 100 amps primary current.

Josh Johnson

Fri Sept 18 2009, 05:20AM

Registered Member #793 Joined: Sun May 20 2007, 06:50PM
Location: Minneapolis, MN
Posts: 35

I tried adding a load (Al heatsink, steel box, secondary) none of these have any effect on the ringing.
The plots are taken with this 8.2uF film cap,

, on the bus.
The ringing decreases when this cap is removed and only the large electrolytics are connected.

Just to be clear, the area I am interested in is circled in red. The second picture of my first post is a zoomed in picture of that area. That is the ringing I am referring to, not the ~85kHz ringup /down cycle; obviously the ringdown time is effect by the addition of a load. The ringing in question is ~25MHz and what I believe to be caused by the parasitic inductance in the PCB coupled with the IGBT capacitance.

As you can see the peak of this ringing is more than 2x the DC bus voltage. If you scale that up and run with a 320VDC bus, 2x is 640V and my IGBTs are only rated for 600V.
I am wondering if I can rely on TVSs to protect my IGBTs in this case.

Steve Conner

Fri Sept 18 2009, 08:58AM

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

Have you tried the same experiment, but with the scope probe tip clipped onto the same place as the ground clip? This should obviously produce no signal at all on the scope. If you still see the big spikes, then you have a problem with common-mode noise, and they aren't real. In this case, wrapping the probe lead several times through a large ferrite ring will help.

If the spikes are real, they're most probably caused by too much stray inductance of your DC bus. The concave tops of your waveform also suggest you have a serious problem with stray inductance, and you need to tighten up your layout and/or install some more decoupling capacitors (but watch out for self-resonance)

Then again, this effect may not be real either, it could just be mutual inductance between other parts of the circuit and the loop composed of the probe tip and its ground clip. You often see people wrap the ground lead around the tip to minimize this loop area.

You can of course leave everything as it is, and keep an eye on the TVS to see how hot they get. If they don't get so hot that they burn out, no problem!

TheBoozer

Fri Sept 18 2009, 09:04AM

Registered Member #1535 Joined: Wed Jun 11 2008, 11:37PM
Location: Northeastern Pennsylvania - USA
Posts: 117

I use TVS's and they don't clip these spikes for me. Maybe the the ringing is to fast for them?

Using a RC snubber reduced these spikes for me. They waste power. Mine was a 20ohm & .1ÂµF across each collector/emmiter. As far as component values, your mileage will vary. It was cheap and easy.

Seeing Finn Hammers's prediktor thread led me to believe that one can pretty much eliminate these spikes with proper gate timing. This is beyond the scope of my expertise for now...

Josh Johnson

Sat Sept 19 2009, 04:37AM

Registered Member #793 Joined: Sun May 20 2007, 06:50PM
Location: Minneapolis, MN
Posts: 35

When I first measured the waveforms the the first thing I thought (hoped) was "Ahh, it must be the ground lead on the probe", so I redid the measurement with one of the little spring-type ground wires:
Update202

The common-mode junk was no more than a couple of volts, so I think 90% of what I see is real oscillations.

I ended up shoe-horning in some by-pass film caps extremely close to the IGBTs. It looks ugly, it makes the waveforms ugly, but, happily, it cuts the peak voltage in the ringing down by 50%. I feel a lot better running at full 320V with this setup. I'll try it out tomorrow and see how it goes. It something does blow up it gives me a good excuse to design a better, lower inductance, bridge layout.

I use TVS's and they don't clip these spikes for me. Maybe the the ringing is to fast for them?

I have heard this from others as well. The TVSs I am using claim a response time of 1ns so I am thinking they should be fast enough.

Using a RC snubber reduced these spikes for me....Mine was a 20ohm & .1ÂµF

I am curious, what was the power rating of your resistor on these snubbers? I thought in an RC-type snubber the power dissipation in the resistor is directly related to the capacitance of the capacitor since the cap is charged/discharged every switching cycle. My calculations showed for my setup a 10nF cap the resistor would dissipate 5W and a .1uF, 50W!

Thanks to all for the replies!

TheBoozer

Sat Sept 19 2009, 11:21AM

Registered Member #1535 Joined: Wed Jun 11 2008, 11:37PM
Location: Northeastern Pennsylvania - USA
Posts: 117

The power resistors are 25watt. They are gold heatsink looking things. They can be directly mounted on the bridge heatsink.

My DRSSTC currently only has an IGBT average ontime of 1.5%. I think going beyond 10% would be suicide.

So the snubber runs pretty cool due to the low duty cycle. If this was a CW type coil, yeah, the resistors would probably superheat.

Steve Ward

Sun Sept 20 2009, 03:26PM

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

Hi Josh,

The ringing is caused by the parasitic L between the IGBT dies and the nearest bus capacitance (probably some sort of low uF film cap) and the junction capacitance of the IGBT/diode, which usually rings in the 10s of MHz range. Anyway, in some instances you are stuck with the internal L of the IGBT module and cant fix it, unless you go to RC or RCD snubbers on every IGBT, which is very annoying, and only works if its excessively lossy (because again, the internal IGBT inductance limits the snubber performance, so it must be very overdamped to work).

I say dont worry about the spikes too much, the energy content of them is small and ive done tests on some IGBTs that shows they do the same work as a TVS and just absorb energy that exceeds their Vce blocking voltage. All of my bridges (even ones with very low inductances) show this ringing. You may wish to consider a little more gate resistance for the IGBT turn ON.

BTW, your original post is mistaken, you said the ringing occurs at the turn OFF, its actually the *other* IGBT turning ON that produces the ringing due to diode recovery causing the current stop flowing in the diode. If your DRSSTC is like mine, it always switches late, so the IGBT never turns off with any current.

Josh Johnson

Sun Sept 20 2009, 09:15PM

Registered Member #793 Joined: Sun May 20 2007, 06:50PM
Location: Minneapolis, MN
Posts: 35

My biggest mistake in the design was the large physical distance between the IGBTs. This is a full bridge of TO-247s:
Picture of layout
I placed the big electrolytics and film cap in the middle, between the two IGBT pairs. I thought I could just make the traces wide enough to keep the parasitic inductance low.
The width of the traces help but I learned it really comes down to the size of the loop the current has to travel and in my case that was pretty long giving me these ringing issues.

I did some simulations and was amazed to find just how physically large of R and C you need to snub that first initial spike. Large R from the power dissipation and C from the relatively high capacitance and voltage rating. At first I thought I'd just slide in some RC snubbers to fix the problem but then when I actually calc'd up some values, no way they'd fit.

Yesterday, after adding some DC bus bypass caps right next to the IGBT leads I put about 10 minutes of run time on the coil at 320VDC on the bus with no problems. Hopefully everything holds up. I'm still on my first set of IGBTs and loving it.

Steve W.,
Thanks for your experimental evidence on these spikes. Good to hear things should be ok.
You are right that it is the IGBT switching on that causes the ringing. I was just referencing what the IGBT I had the probe on was doing at the time of the ringing.
Also, I did change the gate resistor from 5 to 12 Ohm which helped a bit, maybe lowered the initial peak by ~10%. I didn't want to go much farther as the turn-on time was getting pretty long.

1 2 next

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