Source, either triboelectric (snow rubbing against the antenna) or induced (snow not so much, but if it was warm weather, it could be melting off the wires, carrying charge; though, this would be dissipative rather than active; a feedback mechanism is needed to generate high voltages, i.e. Kelvin dropper).
It could also simply be charged by the natural ionospheric potential, though that's always present, and no one ever gets shocked by rain or snow normally, so this doesn't seem relevant. (Also something about induced electricity not working that way, but offhand, I'm not entirely sure that's precisely true?)
He describes an unbalanced dipole, which... isn't a good idea in the first place, but seems to be relevant. If the feedline is ungrounded for its entire length, then both elements can charge up and the voltage is carried indoors. The leakage path can be through the radio and isolation transformer as he suspects, though this is missing one thing: the series capacitor was not arcing over, and it doesn't look like the kind that should break down internally (last time I overvolted a Y-type ceramic disc like that, it sparked along the body, between the leads), so it should also be visible if this is the case.
I think the more likely case is: industrial grade ESD. Either the whole cable charges up, or the center relative to the shield. Either way, a breakdown path is found: from shield to ground through the jacket (I imagine the antenna tower (wait, does he have a tower? I don't recall) provides plenty of opportunity for this, or perhaps near the house entry, or elsewhere along the floor?), or from center to shield, sparking at either end, or at any connector (standard 50Ω BNC is good for a couple kV so this is quite likely given just a little accumulation).
When a spark occurs, an electromagnetic wave is launched, as the point of suddenly-zero volts radiates outward at light speed, give or take.
Between center and shield, of course the signal quality is quite good, and it reaches the end of the transmission line only a little rounded off. The rise time is 10 or 20 nanoseconds. There, it reflects off the high impedance, doubling the electrostatic voltage into a p-p swing. The peak current available is the voltage divided by the transmission line impedance, and the pulse duration is fairly long (about the length of the line, but it rings down as the wave bounces back and forth between the spark and the open ends, or between open ends if the spark dissipates afterward).
Between shield and ground, the impedance is higher: perhaps 100Ω close to metal structures, up to 200Ω for runs not close to much of anything conductive and grounded. The elements for example are about like this.
(Note that resonant dipoles are lower impedance. They are series resonant, so the feedpoint impedance at resonance is lower than the characteristic impedance. Indeed, the ratio is the Q factor, so a "400Ω" line in space resonates down to a 70Ω feed for a Q factor of 5.7, which sounds pretty reasonable. For transients, we can't account for reflections when they haven't happened yet -- until the waves bounce back, the line looks like about 200Ω, give or take.)
With a free conductor (the coax shield), a lot of the wavefront is lost to space -- radiating EMP to anything in proximity, hence all the screwed up equipment. This introduces HF losses that tend to round off the wavefront. There may also be dispersion due to the circuitous path taken by the cable -- when its path curves around, the wave can couple into other parts of itself, sooner than it would otherwise. Hence the apparent "give" in the speed of light -- I wasn't just using a turn of phrase there.
Dispersion, by the way, is group delay varying with frequency. The basic importance is that, because a pulse contains many frequency components, they end up spread out in space, after traveling through a dispersive medium -- and so the pulse ends up smeared out over a longer time with lower peak amplitude, and probably ringing and whatever.
In either case, when the wave reaches the radio end, the fact is the radio is only touching one conductor of the coax, not both. This again fully unbalances the wave, so that it doesn't actually matter which way the spark originated; and is why I didn't give greater detail above. Now that the picture is complete, I can also finish quantifying it.
If the center conductor emits the spark, it emits it with respect to the shield. The radio is not connected to the shield, therefore the wave escapes from the coax and wraps around the shield, propagating back up the line along the outside.
If the shield carries the spark, it carries it up to the radio, which provides some load impedance, which reflects some signal back into the coax and back up the shield.
In either case, the transient / instantaneous impedance seen by the radio is the sum of impedances: coax Zo (50Ω) and shield (100-200Ω).
We can, at last, determine the magnitude of the spark. If the voltage is at least 2kV, the peak current is at least 2kV / 250Ω = 8A. A very typical ESD current, but lasting much longer than traditional ESD because the cable is much longer. The ringdown waveform will be irregular due to the differing velocities and impedances of the two propagation modes, the uncontrolled environment of the shield (it will randomly be close to large metal objects acting as local grounds, making for a lumpy impedance along its length; which makes it also another source of dispersion and loss), and the antenna elements.
The more likely scenario, to me, seems like: the whole cable is being charged. That doesn't need a mechanism for asymmetric charge between the antenna elements, and allows high voltages despite the presence of BNC connectors. The far-end reflection flipping a large peak-to-peak swing helps out further. The coupling capacitor inside the radio can become charged to some kV, most likely due to leakage current through the isolation transformer (including if it sparked internally). The capacitor getting charged is important, because it allows the full Vp-p edge to be utilized by the spark.
Finally, while the tuning coil might be 50 or 300 ohms or whatever, that's only true at its tuned frequency (and with the tubes heated, for that matter). At very high frequencies (we're talking 10ns edges here -- substantial frequency content above 50MHz), the impedance will be rising (inductive), due to the all the stray wiring. It could be kΩ up there. And since we know the wave impedance from the transmission line is under 250Ω, that's not going to attenuate the wave all that much.
Consequently: the spark happens somewhere on the line, the wave propagates into the receiver, the capacitor couples it dutifully to the coil, and the nearest terminal sparks to the chassis! It isn't very important that the coil is 30Ω DC, nor that it has a tuning capacitor across it -- connected by 10cm or so of hookup wire, remember.
So, what to do about it? Well, most obvious, put a leakage path on the cable. Interestingly, one possible mode would be completely avoided if a proper dipole were used -- the required balun shorts the elements together at DC, preventing differential charging. A ground leakage path is still needed.
Interestingly, this situation might occur -- seemingly almost intentionally, once you know about this -- when a spark gap is used to "protect" an antenna. The spark acts to sharpen the rise time of, say, induced lightning surge, and now what would otherwise be a, say, 12kV surge with the risetime of a lightning bolt (10s of us?), is now say, 6kV of brutal cable-discharge ESD (<10ns risetime)! Not to say spark gaps are a bad idea, just that there can be considerations beyond simply putting it there and assuming done is done.
Personally, in EMC work, I've found that you can often cause spurious upsets, at much lower test voltages, by holding an ESD gun's tip
just away from the target surface. This causes the tip to spark multiple times, just as the generator is starting its pulse. I haven't measured the waveform, but I'm sure it's well outside the standard. It's interesting, and shows why procedure must be followed correctly!
Tim
Home
Help
Search
Login
Register
