AVR-Chat

> 1.  How are you measuring your shunt voltage and waveform?  A scope I presume.

Yes

> 2.  Your shunt will be somewhat inductive at these frequencies and currents.

Yes

> 3.  Your frequency components will appear because of the pulsed
> waveform, so you'll need to be aware of that.

Right

> 4.  Have you tried to cancel fields by running the cable into and out of
> your shunt in opposite directions?

I haven't looked at this yet, it's a bit mechanically difficult at this point.

> 5.  Could the difference be caused by an induced field in your probes?
> 1KA at ~1mSec is a huge field to contend with.  200uSec rise time makes
> things even more interesting.  Physical routing of your wires might
> reduce any problems of this nature as well.

Could be.

> 6.  Have you thought about "skin effect?"  That makes life even more
> interesting at high currents and itsy bitsy rise times.  If you
> calculate this for 60 HZ you'll be amazed at the difference it will make
> in the resistance of a wire with frequency.  If you can do an FFT
> analysis of your wave?  Then take the highest significant harmonic
> component and calculate the skin effect you might be in for a little
> surprise.

The part that would be interesting is the 100uS risetime, Anything
else isn't all that relevant for me.



> 7.  16 AWG copper wire has a DC resistance of 4.02 milli-ohms per foot.
> Can you measure or estimate the length of your wire?

About 6 meters.  With 400V pushing, the DCR shouldn't be interesting.

> 8.  Can you take a comparable length of copper wire, stretched out
> straight, and get the same or similar readings?

That wouldn't have enough inductance to keep the current "down" to the
kA region.


> 9.  You're right about the timing of your pulse not being great enough
> to cause serious heating.  How about your duty cycle?

WAY below .1%  :)    Minutes between pulses at this point.

> 10.  The TCR of copper is +0.39% per degree C at 20C.  Not significant,
> but still a factor.
>
> 11.  Don't forget about your magnetic pulse and it's affect on item 5
> above.  We built a small transformer impulse test system once that went
> on a pilot line in the lab.  They were supposed to build a steel
> shielded room, but it wound up being copper as any normal RF screen room
> would be.  Impulse testing is basically generating about a million volts
> then jumping an arc to the transformer to simulate a lightening strike.
> We struck that baby off the first time ... and blew up every instrument
> in the lab with the magnetic pulse.  Did I mention that we only did that
> one time?

Whoops.. :)

> I'm going to be gone for a couple of days but this ought to give you a
> couple of days worth of experimentation.  Let's see what results you get
> from the above and then look deeper if it's still a problem.


This setup was just a proof of concept.  My only sore point is that I
don't really know what currents I was getting up to.
But I know it wasn't 2.5-3X the theoretical limit.

The shunt I'm using is from All Electronics,
http://www.allelectronics.com/make-a-store/item/SNT-25/25-AMP-SHUNT-50MV-25A/-/1.html



I did notice this morning, that my ground leads to the scope were
several inches apart on the rig. I've changed that to single point,
and I'll re-test.
That may have hosed the measurements a bit.