20KW DRSSTC / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

20KW DRSSTC

first previous 6 7 8 9 next last

Move Thread

LAN_403

hvguy

Fri Apr 06 2007, 12:38AM

Registered Member #289 Joined: Mon Mar 06 2006, 10:45AM
Location: Conroe, TX
Posts: 154

Thanks Steve. Your right, it would be nice to perform this test on some bricks. The problem for me is all my bricks are either completely dead or good. I killed three 40N60s in the test and donâ€™t really have any bricks I can nuke while testing

If youâ€™ve got some half dead ones go for it. I would expect the results to be similar to this test but who knowsâ€¦ Thatâ€™s why we need to test one. Also, I like the idea of recreating the 10MHz ring. BTW that happens to be about the same Frequency as my buss spikes.

Finn, thatâ€™s basically it. An IGBT, MOSFET, or diode will â€œavalancheâ€ if a certain reverse voltage is exceeded. At that point it begins to conduct, causing the waveform across the device to appear clipped. If this voltage is exceeded to quickly and/or with to much energy the junction will fail.

It is defiantly clear that these parts have some avalanche capability. Thing is a spike of significant magnitude will kill the device; it just takes a lot of power. I am really curious to see what happens when the spikes are very fast, like the 50ns or so rise times Steve and I are seeing.

hvguy

Wed Apr 11 2007, 04:51AM

Registered Member #289 Joined: Mon Mar 06 2006, 10:45AM
Location: Conroe, TX
Posts: 154

Well another night of near perfect weather and another failed run. After all the modifications I though Iâ€™d break 20' for sure, as it turns out I killed an IGBT (ouch) and never put more than 400v into the bridge. Iâ€™m not sure what caused the error; the logic and gate drive are still functioning perfectly and the TVS strings show no signs of heating or damage. I will replace the brick and set up for some more bench testing to try and find the problemâ€¦

I do have this theory though: if you want your IGBTs to live a long, meaningful life run them within the data sheet specs. That means current rating too. I have done a lot of testing with small IGBTs and bricks and found them to be most reliable at Â½ their voltage rating and equal to or less than their peak current rating. The only problem with this philosophy is that several people have coils that run outside of these ratings and have never failed, so there are still some unknowns. I will be calling Powerex tomorrow to see what they think the max pulsed current could be given the pulse width of these coils.

Obligatory cool blown IGBT pic

ShawnLG

Wed Apr 11 2007, 05:33AM

Registered Member #286 Joined: Mon Mar 06 2006, 04:52AM
Location:
Posts: 399

"Obligatory cool blown IGBT pic "
Thats not so cool.

...

Wed Apr 11 2007, 05:36AM

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

yikes, I wonder if you could find a way to remove that goo (man that is nasty stuff) and see if the individual dies still work... It looks like only a few of them exploded

Steve Ward

Wed Apr 11 2007, 06:50AM

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

Oooh! ooh! ooh!

Looks like the IGBT didnt completely let go like they did for many of my failures. I also experienced a very similar "low power" failure where only parts of the IGBT failed, and similar to your picture, some of the dies had only small patches that failed. This seems to be some "other" failure mode (where the ENTIRE IGBT is obliterated) than what is most commonly experienced, i think. Or perhaps it just has to do with probability and chance... My similar failure was also at reduced input voltage, and i was very confused at the time. I was running the CM300DU-24F modules with the RTC (real time current limit) and i blamed the failure on the IGBT falling out of saturation (since i didnt disable the RTC at this time). I think the current was high enough (over 600A) for me to even consider that conclusion, but the current was also low enough that i didnt think it would have caused a problem.

If only we worked for powerex, we could test these failure modes all day while making big sparks

hvguy

Wed Apr 11 2007, 06:57AM

Registered Member #289 Joined: Mon Mar 06 2006, 10:45AM
Location: Conroe, TX
Posts: 154

Thatâ€™s basically were Iâ€™m at. Itâ€™s technically possible the IGBT fell out of saturation or there was some other "this is what happens when out side of SOA" failure. These failures are so damn frustratingâ€¦ BTW the current limit was set at ~2.4KA. Glad Iâ€™m no the only one who gets the strange fauilures

Now that would be a hell of a job! If only I got paid to destroy silicon.

Steve Conner

Wed Apr 11 2007, 09:44AM

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

Richie Burnett and I argued a lot over this some years ago. We came to the same conclusion Aron did: if you want reliability, you need to stick within the datasheet ratings. To be sure that an IGBT was safe for higher ratings, you would need to recharacterise it using a similar procedure to what the manufacturer does, which means a lot of computer modelling and trashing hundreds of devices.

No hobbyist (that I know of) has the resources to do that. Or the knowledge: the things that Powerex know about IGBT failure modes that we don't would fill several phone books. And besides, the manufacturer probably set the rating based on what his computer models predicted would give acceptable reliability, as opposed to, say, 10% below the point of instant destruction.

So, if you run IGBTs over their ratings, you should probably think of it in the same way as building a race engine for a car, or cranking up the boost on a turbocharged engine over stock. In other words, the warranty no longer applies and you're on your own. The only difference is that IGBTs don't have anything like knock, leaking head gaskets or little flakes of metal in the oil, to give you advance warning of a failure.

One final thing: Look at the guts of that brick, ponder the longness and thinness of the internal busbars, and try to picture how even (or otherwise) the current distribution will be at high frequencies. Is it just coincidence that the dies nearest the power rail inputs (and hence to the filter capacitors) are all intact, but the ones at the other end (possibly a few hundred volts of L*di/dt worse off) are fried?

Do you think you could make something that shared current better with a bucket of TO-247s and a bunch of sheet copper? I do.

Marko

Wed Apr 11 2007, 12:06PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Oh, god.

2400A isn't 'all that much' considering what people do to little 40N60 and live on.

Also, hvguy, you said that failure happened at only 400V into bridge? I find it difficult to believe that it's just overcurrent.

Was your OCD actually tripping for 2400A?

It's just a 16y old teenager's opinion, but simple problems like iGBT blowing off before actually being over SOA usually have simple causes.

I would heavily inspect that thing for simpler errors before attributing the failure to overcurrent or bus voltage spikes again...

Do you think you could make something that shared current better with a bucket of TO-247s and a bunch of sheet copper? I do.

Hi steve

It is generally difficult to paralell a number of physically large dies and keep good current sharing just with them alone.

I think your idea is not bad at all; using a large bunch of small IGBT's (30N60's or etc) wich would still be cheaper than these bricks, and parallel them along with a split MMC bank, wich would really eliminate current sharing problems!

thedatastream

Wed Apr 11 2007, 02:34PM

Registered Member #505 Joined: Sun Nov 19 2006, 06:42PM
Location: Yorkshire!
Posts: 329

Is it possible that opening the top has introduced some contaminants that have caused or accelerated this faliure mode?

This app note

from Microsemi / APT implies that Non-Punch Through IGBTs can be paralleled because of the positive temperature coefficient.

Marko

Wed Apr 11 2007, 02:38PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

At high pulse currents like this most IGBT's will go PTC because of stray ohmic resistances.
Stll, splitting the MMC bank is much more *right* way to paralell IGBT's!

Is it possible that opening the top has introduced some contaminants that have caused or accelerated this faliure mode?

The dies look like they are sealed in some kind of resin, and if IGBT isn't hermetically sealed and is made to be unscrewed and opened I wouldn't worry about that too much.

first previous 6 7 8 9 next last

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