> W6RMK wrote:
> Hmmm.. are we talking direct hits? The lightning literature is fairly
> full of the electrical properties of lightning strokes, and the current
> pulse typical stroke is substantially less than 1 ms long (classic model
> is 2 us rise (10%-90%), 50 us fall to half max magnitude). There are
> some super strokes with longer durations and/or large "continuing
> currents" but they constitute substantially less than 1% of the overall
> strikes at a given location.
> Yes, we are talking direct hits, and those numbers I gave were for the
> very large positive strikes which are possible, but rare in comparison
> to negative strikes.
I suppose (like everything else) it comes down to a cost benefit
analysis. Do you spend X dollars to protect against 95% or 2X to
protect against 99% or 10X to protect against 99.999%.
I also stated that you have to decide what the
> something that doesn't cover all the possibilities, so there is an
> element of risk involved.
> The current pulse you choose for the excitation is only good at the
> point at which it is applied. At other parts of the system the current
> waveforms will be different. It's somewhat like pinging a complex
> electronic circuit consisting mostly of many inductors which also have
> mutual coupling, and asking what is the current waveform in this one,
> but it is more complex than that. If you put a current into a SPICE
> model of an inductor you instantaneously get the same current out the
> other end of the inductor. If you do that to a physical inductor of
> significant size, like an antenna, you do not get the same instantaneous
> current out the other end. To decide what you have at any point in the
> system you have to either analyze the whole system or make some good
> decisions about reducing the complexity by eliminating the insignificant
> items. If you get any arcing, waveforms change significantly. It's not
> enough to just know the peak currents in the system. These have to be
> applied over a time period. (i^2)t would be good information.
True enough. However, I still think there is value in a back of the
envelope calculation to get general magnitudes. The rough and ready
lumped inductor model I did (literally on the back of an envelope)
wasn't far off from what 'TTT calculated with a full up model, within
the assumptions stated. Sure, the incident impulse might be different
in some cases (e.g. positive megastrokes with thousand amp continuing
currents), but the 30kA 2us thing is probably an ok model for the vast
majority of hits.
> There is one other major element of this system that requires a huge
> simplification, modeling the ground system and its impedance. That will
> be very non-linear and frequency dependent. There are such things as
> ground saturation occurring that I don't think are very well
> understood. Just choosing a resistive value for ground impedance is a
> gross simplification. Even if you can't model ground saturation, wire
> and rod lenghts should be modeled.
> I don't know of any single tool that can perform all the calculations
> needed for this system. If NEC didn't have its limitation of handling
> closely spaced wires, and intersection of widely different size wires,
> and could handle an arbitrary source instead of just a sine wave, and
> could perform transient analysis instead of only AC frequency analysis,
> then it could do the calculations. That's a lot of if's.
The folks doing NEC models do it by running the model at a series of
frequencies, then combining those appropriately (e.g. with a FFT) to
simulate the time domain response. Just like the Vector Network
Analyzer does when displaying group delay or time domain reflectometry.
Indeed it's not a single tool, but it's pretty easy to put all the
pieces together with some appropriate scripting. (realistically, the
most painful part of the using NEC scheme is modeling the actual
lightning stroke... you have to run multiple models, using little
segments to represent parts of the stroke, then, once you get the
transient response of the system for each piece of the stroke, you sum
them all together.. and that relies on what's probably the biggest
limitation.. you're assuming it's linear so that superposition holds)
> neglect all inductive coupling and non-linear ground effects, SPICE
> could do the calculations, but that would generate only a rough guess
> for an answer.
> Also don't forget that the typical strike usually consists of more than
> one stroke. That may be important depending on what you are calculating.
Fusing current.. As you said above, the action (i^2*t) is probably a
more relevant measure.
> You can assume a current in a tower, and assume a rise time, and assume
> a tower inductance, and then calculate a voltage, but the answer doesn't
> give you any useful information.
I'd differ with you there. It's useful, in the sense of providing a
bound or loose estimate. And that might be sufficient for a planning
process for a ham station. Or, if one is relying on empiricism (A1BCD
has this installation and it works for him so that's what I'm going to
install at W1XYZ) it allows you to better understand whether you should
rely on the one anecdote.
The industry, of course, (and hams too) can also draw on lots of
experience. That's where most of the "standard practice" comes from,
but, as the theoretical knowledge expands, the standard and regulatory
practices evolve (viz, the NEC not allowing just a single driven rod as
a ground any more).
TowerTalk mailing list