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

THOR PROBLEMS AND SOLUTION (was Re: FET/IGBT drivers)






Tesla List <tesla-at-pupman-dot-com> on 25.12.99 03:34:35

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



Original Poster: Kennan C Herrick <kcha1-at-juno-dot-com>

>To Marco Denicolai & all:

>Re Marco's IGBT-driver misfortune, I might have had the same problem:  I
>was using some IR 2110 high/low-side drivers in a prior manifestation of
>my FET t.c.  I found that they got zapped repeatedly, and so I finally
>went to a transistor H-bridge driver of my own design.  I think the cause
>was, that the 2110 and such-like incorporate a high-Z charge pump to
>drive the "top" transistor.  The electric field of my t.c. was getting
>into the IC and messing up that charge pump.  ...At least, that's my
>theory.

>Ken Herrick

You hit to point! While the IR2110/2/3 datasheet is full of nice diagrams, you
have to get some Application Notes and Design Tips from both IR and Motorola
(which manufactures a pin-to-pin compatible device) to get a good picture about
possible problems and cures. I found at least 5 Application Notes about these
problems.

The PCB traces small inductance (and I mean 1-2 cm!) is anyway big enough top
generate dV of 20-60 V on the chip ground terminal when the IGBTs switch
rapidly. This can kill your IR2110 (or 2112, a 200 mA 600V version I use) in at
least two ways:

1. Difference from digital to analog ground greater than +/- 5V -> latchup
2. Overcharging of the charge pump (over 20V) -> latchup

Yesterday I measured the voltage across a 0.1 uF capacitor charged to 7 kV
while
I shorted it through about 30 meters of wire (to simulate some inductance).
There was NO voltage rise, no overvoltage at all, and my IR2112 were still
burning died.

Then I measured the IR2112 ground bounce: from 10 to 30 V spikes! I also
checked
how the IR2112 died: they got I short between the charge pump pin and ground.
That indicated the chips where killed by overvoltage on the charge pump (>20V),
not by a burst on the 600V!

I tried a fast diode between ground and the H bridge central leg: no help, a
diode was too slow (I am talking about 10-100 ns spikes). Then I tried the
Motorola cure: a 20 ohms resistor between the chip pin and the H bridge central
leg. THAT WORKED! I made a lot of shorts up to 10 kV: long sparks, heat, UV and
pieces of melted metal all around. No problems, IGBTs and IR2112 were still
there and working fine.

Than I removed the long wire and I replaced it with a barely 1 m HV wire: tried
again shorting the capacitor. No problems at all: until now I have not been
able
to burn the IR2112 or the IGBTs.

But...

1. When closing back the frame box of one of the drivers (after I added the 20
ohm resistor), a screw went too near a 400 V phase line: result, BANG when I
switched it on! I got
- a 2 x 2 cm area of the PCB vaporized: all to rewire by hands
- all chips on the board to be replaced
- a 20V zener, the 5V regulator and the 16V regulator to replace too
- a 16 A fuse blown

After thanking Father Christmas also for this present, I repaired the
driver and
I was again verifying the switcher capacity to stand shorts. Then I decided to
measure with the oscilloscope both the HV output (one 4 kV probe) and the
measured and scaled HV output (0-3V, normal 250V probe). But... I connected
wrong the 250V probe: ground clip to signal and signal clip to ground. Result:
- reference voltage shorted to GND
- controller will see always OV as the HV voltage: therefore will never stop
switching
- the HV output will rise over 4 kV reaching 10-12 kV at least

My 4 kV HV probe proved not to stand 10 kV: sparked through breaking all its
isolation. For solidarity my measuring board's chip decided also to make
suicide... :(

Another series of thoughts about Father Christmas and I replaced the chips,
disconnected the HV probe and tested again the whole stuff: it works and it
stands repeated (even continuous) short circuit on the charged capacitor.

Probably tomorrow I'll go back to the university (where my old TC still is) to
test the switcher with a real TC primary and capacitor.

Regards and congratulations if you were so patient to read all the above.










_______________________________________________________________
YOU'RE PAYING TOO MUCH FOR THE INTERNET!
Juno now offers FREE Internet Access!
Try it today - there's no risk!  For your FREE software, visit:
http://dl.www.juno-dot-com/get/tagj.