[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: drsstc - catastrophic failure! And Thor



Original poster: "by way of Terry Fritz <teslalist-at-qwest-dot-net>" <Marco.Denicolai-at-tellabs-dot-com>

Hello all,

 > -----Original Message-----
 > From: tesla-at-pupman-dot-com [mailto:tesla-at-pupman-dot-com]
 > Sent: 28. huhtikuuta 2003 20:19
 > To: tesla-at-pupman-dot-com
 > Subject: RE: drsstc - catastrophic failure! And Thor
 >
 >
 > Original poster: "Stephen Conner by way of Terry Fritz
 > <teslalist-at-qwest-dot-net>" <steve-at-scopeboy-dot-com>
 >
<SNIP>
 > I would also change the current limiting circuit- I believe
 > the circuit
 > Marco originally used could cause second breakdown in the
 > IGBTs, because it
 > clamps the current instead of feeding back to the controller chip and
 > ending the pulse.
 >
 > I'd appreciate any comments from the original inventor :)
 >
 > Steve C.
 >

I tested that BJT connection almost 2 years ago. I think I even have
somewhere an o-scope hardcopy with the clipped gate pulse. Here are my
points:

- The thing was designed to protect from the joined failure of two IGBTs
on the same H-bridge leg. Limiting starts at 30 A, while (if I recall)
bridge working current is around 12 A peak
- you are right, feedback to the driver chip is the best thing, but
- the environment is so much noisy that you can easily get false alarms.
If you suddenly shut off the drive, the original spike will only get
amplified and possibly kill the IGBT or the driver chip. There is some
literature about this. I think my linear clamping is a better choice
- to avoid spurious alarms you can low-pass the feedback signal, but
then you loose in response time and the IGBT can fail before you shut
off or clip the drive signal.

Moral: this is IMHO a compromise. I haven't got strong evidence that
this solution actually saved any IGBT but some evidence I have got that
it doesn't kill them. The truth is that it is still a kind of mistery
what was the mechanism that resulted in the IGBT pair shorting Vcc to
GND and, thus, committing suicide. I believe it really was the IR2112
going into latchup and forcing both IGBTs into on state. Limiting the
spikes around it prevented the latchup.

At today, Thor rev. B has performed 6 shows (the last one yesterday
night) and no failure yet. This sets a record, even if it surely makes
many of you readers smile or even laught... :)

I have almost got ready the new Controller board that allows me to make
a predetermined number of bangs (from 1 to 4080). I have also almost
ready the proximity switch to detect the RSG electrode presentations
(and timing)

Best Regards