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

Re: DRSSTC - I think I heard Cross Conduction...

Original poster: Terry Fritz <teslalist@xxxxxxxxxxxxxxxxxxxxxxx>

Hi Bob,

At 07:10 AM 2/19/2005, you wrote: 
.........
>
> So before pulling the H-bridge to add the caps, I tried the caps across
the
> output first to see how well they stop the noise at that point

I think that idea must have been caused by a stray cosmic ray passing
through your brain and triggering daft idea neurons LOL (not offense
intended).

I have powered the thing up a hundred or so times now. Many times I have backed off thinking it was going to blow. But as time went on, I just forgot about it and never gave it a second thought. I had almost forgotten that it might ;-)) Blowing the IGBTs was sort of an expected test anyway, so now I got that taken care off. No other damage was done. Gate drives, other H-bridge parts, wiring, etc. all came through without a problem! In less than a week, I will have three new H-bridges ;-)) I already had most of the parts but never bothered to get extra boards...

....
> http://drsstc.com/~terrell/pictures/TraceBlown.JPG
Now I know what your layout looks like.
If you lay out an other board put the power input connection next to each
other and the same for power output.
Definitely not on opposite edges. Keep the traces next to each other  i.e.
minimize the loop area.
Try to think the cennections as a pair only just seperated sufficently for
insulation. Same for the decoupling/bypass cap loop.

The board is pretty dense and there is not a lot of room. Ideally, it would bolt down to a motherboard that has nice power traces well laid out... But for now, it has to be sort of modular. I was hoping the 10uF bass caps right there would kill the noise on the ~6 inch wires back to the main 4700uF filter bank. But now I will add the ceramics (or micas) in there too since the ~25MHz stuff is sneaking in.


With the power loops you appear to have, a scope probe and its ground
connection will form a nice pick up loop to them.
Try grounding the scope probe tip with its ground connection to form a loop
and see what you can pick up.

I usually use the differential probe (Tek P5205) When hooked to low impedance stuff (like directly across the buss) it is pretty good at rejecting noise. I have tried to sniff around with a shorted probe too but the probe is too big to get real specific. Best to just tap and see with the P5205. That is how the buss noise was found.


Nice tight twists on the signal wires, thats good given the big loops on the
power lines.

I thought you where using bricks. Do you have the spec on these wimpy
looking devices?

Bricks would have blown the side of the cabinet off and might cost more than $7.00 ;-)) The little IRG4PF50WDs blow very painlessly and are cheap. They can easily take the power in this case:

http://hot-streamer.com/temp/irg4pf50wd.pdf

I don't miss cleaning up all that gooey silicone stuff ;-)) The case of the IR FETs is actually designed to blow up "nicely". The tops just come off and allow the flash out while doing little real damage. I certainly don't need the huge heat dissipation of a brick. It would be cool to get the new super IR IRGP50B60PD type WARP-II IGBTs but they are not available yet.


Robert 

And Teslina wrote:

Come on Terry, blowing IGBTs is the nice thing. After
some time, the streamers become just boring.

teslina

They sure are louder than streamers too!!! I wish I would have had the video camera going!!

It looks like the traces may have seen like 50,000 amps probably for less then a uS. Pretty cool patterns and creases blasted into them. The trace that actually went is just a green film deposited on the buss caps now ;-)) The high current damage is actually very interesting!

I am beefing up the traces, but I don't think anything could stop it once it lets go. But the breakers and fuses did not trip so the IGBTs blew themselves out of the circuit very quickly. Good since that really helps limit damage. I am using the little IGBT as fuses in that case and they almost worked. Maybe I need smaller IGBTs ;-))

Cheers,

        Terry