40MHz Spark Gap Behavior
From: terryf-at-verinet-dot-com [SMTP:terryf-at-verinet-dot-com]
Sent: Sunday, April 12, 1998 10:52 PM
Subject: 40MHz Spark Gap Behavior
I have shielded my voltage probe against the radiation from the current
spikes and it is now working properly. I can now see the actual voltage
across the gap during the firing cycle which has given me a much better
understanding of what I have been observing. The voltage probe and the
current transmitter are now in die-cast boxes that are sealed with copper
tape (that stuff really does work). The voltage probe now only shows little
glitches at the current spikes (these probably are real voltage glitches).
The current spikes still remain very difficult to quantify. Their energy is
enormous and not only do they have a heavy 50MHz content but I now believe
they have much higher frequency components that also contain enormous
energy. This is playing havoc with the frequency response of the current
sensor. At very high frequencies, the skin and other effects raise the
impedance of the current shunt and make the spikes appear larger than they
are. How much larger? I really don't know. Here is a summary of what I
now believe is happening:
Just before the gap fires (~50ns), The voltage across the gap fluctuates
and seems to ring at about 50MHz for a few cycles. Then the current shoots
high to an enormous level. This spike, in my low power testing (2000 volts,
17nF, 120uH), appears to be around 4000 amps. It seems to have an ~50MHz
ring but it probably has frequency components well into the GHz. This spike
returns to lower levels after about 25ns and then rings at ~50MHz for a
variable length of time (~~300ns). At this point the voltage across the gap
remains low and current is passing through the gap much as if it were
shorted (I was wrong before in believing that there was not current flow
during this time). The current is around the level expected given the
reactances of the inductor and capacitor. The current continues to flow
until it reaches the first zero current crossing (this coincides with the
voltage of the primary cap being at it's peak). At this zero crossing the
gap is actually opening and the voltage across the gap rises in a small
spike. Then the current shows enormous ringing and again the high frequency
energies seen in the first spike reappear but at considerably lower levels.
This second burst lasts for about 100ns. This process continues for the
rest of the firing cycle with a current burst occurring at every zero
current crossing until the gap opens. Often the last current burst is very
Here is my best guess as to the nature of the bursts. The first burst
contains much energy due to the charges around and in the gap just before it
fires. When the air ionizes and brakes down, the charges are allowed to
transfer to a state of semi-equilibrium in an extremely short time. This is
accounting for the enormous currents and high frequencies seen. I suspect
the ~50MHz signal I see is related to some transition time. Probably the
air ionization time. During this time the impedance is very unstable and
the system oscillates for some time before the current is established and
stability is achieved. Everything is fine until the current drops to zero
or, more importantly, must change direction. It is easy for current to do a
180 degree turn in wire, but I suspect it is much harder to do in air. The
charges and ionized gases must completely reverse themselves. In this
process I believe that the initial oscillations and instabilities are once
again established and the burst is again created.
I have examined the energy and lengths of the burst and have concluded the
I believe the gap is dissipating the vast majority of it's power during
these bursts. The voltage and current clearly are in phase and energy is
being lost. Heat, light, and RF radiation must be huge during these bursts.
To take a guess at the numbers for the initial spike, if the current is
4000A at 2000V for 25ns the energy lost is very roughly 0.2 joule!! This is
a pretty high rough calculation for a system that only has 0.034 joule to
begin with. The energy being lost in the gap at other times seems very
insignificant! This would explain why primary systems seem to loose so much
energy on the first oscillation. Obviously, my number of 0.2 joules is not
correct but it does give an indication that the energy lost in these bursts
could be very high.
I also believe my initial thought, that these high frequency bursts could
account for the sensitivity of coils to good RF design, is true. It is well
known that a sloppy primary system with thin wire and long connections
looses much more energy than a thick copper primary system with short heavy
leads. This does not make sense given the few hundred kHz frequencies they
run at. The primaries should have milliohms or resistance instead of the
ohms of resistance they often test at. I have found that when poor wiring
is used, the current bursts change dramatically. They loose their sharp
fast appearance and become long and painful looking. The energy being lost
appears to be much much higher. You can see the voltage across the gap
start to rise and the burst can last a few microseconds which is a
substantial part of the cycle. The current and voltages in the non burst
areas seem to still be efficient. I do not know if the ~50 MHz signal is
making its way around the inductor but the effect of poor wiring on the
current burst is very obviously wasting much power.
I'll try to write up a nice paper on this as time permits. It is hard to
write when the experiments are calling!
There are still many questions and details that have not been answered.
Here is my list of unknowns:
Why do the bursts oscillate near 50MHz?
How can I get the shunt to work well at high frequencies? (I think I need
to add some capacitance to kill the 200MHz and above signals. They do make
fiber-optics up to 5GHz. :-))
Does the ~50MHz travel through the whole primary system?
What are the factors involved in the first giant current spike?
How will this affect the secondary system?
Can the spark gap be made more efficient in light of all this?
Does a rotary gap work better? (I did a little testing and it looked about
the same but more careful testing is needed.)
How does placing the gap in a vacuum or under pressure affect this?
Thanks again to all who have sent ideas and comments!