Cap Location

From: 	Greg Leyh[SMTP:lod-at-pacbell-dot-net]
Sent: 	Sunday, January 11, 1998 6:20 AM
To: 	Tesla List
Subject: 	Re: Cap Location

> >It is preferable whenever possible to put the _spark gap_
> >across the transformer, so that when the primary circuit
> >fires, the only voltage that gets kicked back to the xfmr
> >is the IR drop of the gap, rather than the full voltage
> >swing present on the primary capacitor.  This results in
> >at least a 10 to 1 reduction in the RF voltage thrown back
> >to the xfmr.
> >
> >-GL
>   Tesla in the CSNotes said he thought the gap across the transformer was
> better than the cap across the transformer as you and Mr. Cox suggest. He
> based this on the fact that the current from the cap did not have to go thru
> the gap to get to the TC primary winding. 

???  Did he say where the current _did_ go, if not thru the gap?

>   The TC primary circuit contains an inductance. When a switch (gap) opens
> on an inductive circuit with current, there can be a large voltage transient
> created. This voltage is   V = L di/dt   and can be high enough to break
> down the secondary winding insulation in the power transformer. I believe
> this is what breaks down neon transformer insulation.

The voltage due to (L di/dt) would be a major problem indeed,
_if_ the spark gap were a perfect switch, and capable of 
suddenly interrupting 1000's of amperes.  In fact, in PSPICE
one has to disregard this spike, if a perfect switch is used
rather than a more accurate model.
However, in my experience with actual devices, rotary gaps are 
quite thyristor-like in their nature, requiring the gap current 
to decay to zero, or very nearly so, before the gap could even 
begin to open, and once again support a voltage.  There is a 
wealth of hard data on the recovery times for spark gaps in
'Gaseous Conductors', by Cobine.
In all of the _actual_ primary current waveforms that I have
measured, there has always been an _whole integer_ number of
beats before quenching.  Often times there were more beats 
there than I would care to have, but it was always a whole 
number, indicating that quenching always occurs at near zero
primary current.  Thus the (L di/dt) voltage is quite small,
and probably gets lost in the distributed reactance of the 
charging circuit, as I have not seen a noticable large spike
in the Vgap waveform at the time of quenching.