From: Paul
Date: Mon, 29 Apr 2002 13:41:10 +0100
Subject: Re: [TSSP] Secondary voltage stress factor
Bert wrote:
> The great mystery still seems to be the root cause of "racing
> sparks" and the seeming link to overcoupling.
> Modern magnet wire insulation has a relatively high dielectric
> strength...
> With breakdown voltages like these, turn-turn voltage stress may
> not matter very much ...
Yes, I see what you mean.
> Unlike power arcs, "racing sparks" seem to be of comparatively low
> energy, and seldom seem to do permanent damage.
> Excessive transient voltage stress may stimulate an initial
> breakdown of the air immediately above the wire/insulation surface
I see now. This is more significant than the naive turn-turn
insulation breakdown resistance, so we need to be considering V per
unit length, rather than V/turn. The table values would be the same
in both cases.
> Its known that creeping leaders can be initiated at significantly
> lower voltages than in-air leaders.
Well that certainly seems to describe the phenomena of racing arcs,
and raises the nasty question of what those surface breakdown
gradients actually are.
So the routine might be:
* estimate topvolts using say, Vtop = Vpri * sqrt( Cpri/Cee)
* divide by the coil length to get average volts/unit length
* multiply by the table's 'stress factor' to get the maximum
V/unit length.
* compare with an estimate of the surface breakdown gradient for
the particular secondary coating, etc. I suppose this would
involve another 'factor', being the ratio of the breakdown
field strength above the surface to the 'free space' value
of 26kv/cm.
Looks like computing the stress factor table is the easy bit. Still,
the table would be useful for comparisons of different geometries.
I've just done some rough sums for Chris Swinson's flat secondary and
arrived at 30kV/inch, or 12kV/cm, which is well short of the 26kV/cm
that I've come to think of as the gradient at which leader formation
begins. But in view of your comments,
> (a factor of 4-5X!)
seems we might expect trouble well below 12kV/cm ?
> Perhaps as an offshoot of the proposed study, maybe we could then
> look at the worst-case transient voltage gradients occurring across
> the secondary. This might be done by looking at the transient
> analysis during ringup or by stimulation from an out of tune
> frequency source.
The difficulty there would be finding the worst case conditions. For
a given tuning setup we can calculate the time domain response in
about 30 mins, and extract from the that the highest gradient. But
then we would need to repeat the process for various degrees of
off-tuning and so on. I've just tried this with the model of Chris's
coil and I can get 2.0kV, 2.4kV, 1.9kV, per turn, all with different
Cpri. It would be very time consuming to evaluate all these and
unless the peak gradient happened to be a smooth function of the
tuning, we would never really know if we'd hit the worst case. The
other problem is that, as soon as we include primary options, the
number of systems that need to be processed multiplies dramatically,
and we would need more than the two parameters A and B to dimension
the table.
> This type of analysis would really make Terry's Sun Blade100 earn
> its keep.
Yes, I'm afraid Terry's Sun CPU is not likely to get much rest for
the foreseeable future. It's at around 50% capacity right now with
just the qvar stuff. I've got a few more things lined up too...
It's actually quicker for me to run this on Terry's computer than to
set it up on the system at work. For relatively small jobs like this,
the extra effort in setting up is not worthwhile, so it could prove
quite handy. I'll try to get the stress table stuff up and running
this week.
Oh, and while we're at it, the program could output a corresponding
table of the ratios Les/Ldc for each geometry.
--
Paul Nicholson,
--
Maintainer Paul Nicholson, paul@abelian.demon.co.uk.