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

Re: FET/IGBT drivers






Tesla List <tesla-at-pupman-dot-com> on 28.12.99 05:01:33

To:   tesla-at-pupman-dot-com
cc:    (bcc: Marco Denicolai/MARTIS)
Subject:  Re: FET/IGBT drivers



Original Poster: "Bill the arcstarter" <arcstarter-at-hotmail-dot-com>

>The bit about the 20 ohm shunt resistor presumes that the source of this
>bounce is (?) high impedence and thusly 'quenched' out by the 20 ohm
>resistance?  Did this ground bounce couple into the circut via a conducted
>or radiated path?

I have to explain a little bit: the resonant load SMPS topology should switch
ON/OFF the IGBTs always at 0 current but...

Think to my specs (e.g. with my current configuration, 2 drivers, up to 10 kV):
my specs are to reach 10 kV in say 1 ms. But my switcher pulse rate is 50 kHz:
this means that in 50 pulses' time I must reach 10 kV. Again, that means that
each pulse accounts for 200V (and this is trurely happening!).

Now, if I were using only integral pulses, I would have a voltage stability of
+/- 200V only. So the switcher chip can also "cut" the drive pulses when
reaching the preset voltage. But then the IGBTs are no more opening/closing
at 0
current, the transformer primary looks as an inductive load and I get
extracurrents typical of big inductive loads.

This is just what happens when I make a short (as the sparkgap): then we
are out
of the resonant load topology for a while.

With big inductive loads, there are 2-3 way the IGBT bridge can screw up:
typically it's always about:

1. trace inductance and therein produced voltage (some nH of inductance and 10A
in some ns can produce several tens of volts!). This produces the IC ground
bounce!

2. Rapid changes in the Vce (or Vds) of the IGBT that will capacitively produce
a Vge in the other IGBT switching also it on for a moment (this is called
"shoot-through current"). This will produce a current peak and will
probably get
you into problem n.1

A common cure for all the above is to slow down the IGBT turn on/off by
increasing the Rgate: but in that way you also increase termal losses so it's a
tradeoff.

In my case (see my schematics) when Q2 is turning OFF, pin U2/5 is still
grounded (by Q2). As the current through Q2 start decreasing, the inductance
between Q2/C and the load terminal LOAD_B produces a voltage across the trace.
That is a negative voltage in respect to GND that will drive U2/5 BELOW GND
level! The IR2112 will break at last when that voltage is over 25V.

I placed a 20 ohm resistor between LOAD_B and the node connecting U2/5, C1 and
D3: that way, even is LOAD_B goes below GND, C27 and C1 will keep U2's Vbs at a
value of 16V, and C27 will not overcharge (which was the origin of the
failure).

Hope this helped.


>Please hurry and get your circuit working again so you can tell us all your
>tricks!

I had more damaged components that I though (a varistor, a couple of diodes,
etc.): that caused, again, a damage in another driver module. Fixing more
vaporized traces and replacing more IGBTs...

Regards

______________________________________________________