From: Bert Hickman
Date: Fri, 03 May 2002 07:41:59 -0500
Subject: Re: [TSSP] Secondary voltage stress factor
Paul, Boris, all, I think Boris is exactly right. Since previous history (dielectric "memory") governs the distribution and intensity of stranded surface charges, trying to achieve a hard degree of predictability is probably fruitless. There are a number of studies that imply surface fields of 5-10kV/cm are sufficient to cause further leader growth once a leader has been initiated. But I believe Boris is also correct - the point of initiation will still require a local E-field in the 26-30 kV range to trigger initial ionization. E-fields of this magnitude might occur at localized defects in the winding insulation, imbedded particulate matter, or defects in the surface coating. These points might become sites of initiation when combined with with higher order resonance modes and nearby stranded surface charges of opposite polarity. I have noticed that when racing sparks begin they don't appear to be completely random, but seem to be somewhat repeatable from randomly scattered sites on the resonator. Maybe that's another clue... Adequately predicting the effects of stranded surface charges on dielectrics on partial discharges and dielectric breakdown is a very difficult (but very practical) design problem facing designers of HV components and equipment today, and there's lots of current material in the literature. Perhaps all we can reasonably do is predict that resonator geometries with higher voltage gradients should have a greater probability of seeing these problems. Best regards, -- Bert -- -- Bert Hickman Stoneridge Engineering Coins Shrunk Electromagnetically! http://www.teslamania.com Paul wrote: > > Bert wrote: > > > I would expect that the initiating conditions for surface > > discharges could exist at E-fields of 10 kV/cm, perhaps less... > > That's low enough to be within reach of some practical coils then. > > For a coil like Thor, which could reach over 400kV if the bang > were 20kV, the secondary is long enough to give an average gradient > of around 2.5kV/cm, and even when multiplied by a stress factor > this would still fall well short of a 10kV/cm threshold. > > Terry's small coil could reach 250kV with a 10kV bang, an average of > around 3.8kV/cm, and the stress factor for this toroided coil is 1.5, > so we might expect to find an occasional peak of 5.7kV/cm. > > Chris Swinson's flat coil 350kV over 55cm radius gives an average of > 6.4kV/cm, and with a stress factor of around 2.2 could reach 14kV/cm > occassionaly. > > (the above mentioned stress factors include the effect of primary > induction, which is not quite what we're constructing in the table, > eg the 'table' value for the above coil is 2.05 rather than 2.21). > > Forget the base current measurements. I'll settle for anecdotal > reports of primary bang voltage at which given coils begin racing > arcs. > > > http://www.sandia.gov/media/images/jpg/Z02.jpg > > And the energy we see dissipated in the picture is just the leakage! > A few slight mods and it would make a wicked can crusher :). > > > http://www.geocities.com/CapeCanaveral/Lab/8063/lichtenb.htm > > Yeah, they're the ones. Nice site. > > Boris wrote: > > > I'm not very optimistic in determining its start by > > simulation nor possibility for discovering all the > > critical factors that enable the phenomenon. > > Nor am I, for the reasons you give. I appreciate the reminder that > the dV/dx on the conductor is not the same thing as the grad(V) on > the dielectric surface. And the scale factor problem that you > describe is a serious obstacle. Treating the thin coating by the > method of equivalent surface charges is I think not viable. It is > tricky enough trying to account for the thicker dielectric of the > coil form. Numerous unknown factors must be involved in determining > the longitudinal breakdown voltage at the coating surface, and I see > no hope of predicting it from any set of first principles. This is > a case where we must collect data from actual coils and see if it > is consistent enough to establish a predictive recipe. What we're > after is an estimate of that surface breakdown voltage, in terms of > the coil dV/dx that coincides with it. Given such a recipe, we can > predict, for a given coil, whether racing arcs are likely. > > -- > Paul Nicholson, > --
Maintainer Paul Nicholson, paul@abelian.demon.co.uk.