Keith Nagel wrote:
Hi Stephen,
it looks like it has something to
do with the signal itself, almost like the "pre-ringing" of a perfect
low-pass filter
Given that your cutoff frequency is 60MHz, the leading edge
of the signals you are seeing probably bear little relationship
to the actual state of the signal. I say this as the sparkgap
type circuit you are using can generate sub nanosecond risetimes.
And the measured risetimes just happen to correspond to your
scopes bandwidth limit.
Yes, exactly -- the scope goes up about as fast as it can, which is
surely a lot slower than the "real" edge.
Without more experiments it's impossible to say for sure what's going on
here but my first guess is self-inductance in the loop.
That's a good/reasonable guess. Are you interested in doing more experiments to
prove it out? As the current experiment seems to be showing
the opposite of the claim in the message header.
Yeah, I'll probably fiddle around some more with this, just because it
seems like the "pre-ring" is going the wrong way and I'd like to
understand why. I may dig up some 300 ohm antenna cable -- that's a lot
easier to deal with than bare wire and should be pretty similar, save
that it's (presumably) slower.
I should add that before I laid out the bare-wire loop I fooled around a
bit with insulated wires (zip cord, that sort of thing) and immediately
ran across the fact that you can't easily measure transit time of a
signal "through" a typical _coil_ of wire ... it tends to come out as
zero, as the induced voltage in the "other end" appears immediately.
So, the final rig was set up to keep the ends far enough apart so that
the induced signal would be small enough that I could see the "real"
(end-to-end) signal with reasonable clarity. In other words, I was
expecting to see an induced signal come out "early" before I looked at
the results; I was pleasantly surprised at how clean the trace turned
out to be. My backup plan if the noise was too bad and the induced
signal was too strong would have been to spread the loop out as linearly
as possible, with the scope probes at full stretch (would have
necessitated clearing some boxes away, that kind of thing). If that
wasn't good enough, next step would be to use "extension cords" made of
twisted pair, or, still better, coax, carefully measured to equal
lengths, to carry the signals from the ends of a long straight bare-wire
run into the middle, with the scope in the middle, and the signal fired
between the ends of the bare-wire segment; that would eliminate all
"short cuts". But none of that turned out to be necessary.
If the speed had been 'way low, OTOH, next step was going to be to try
to figure some way of suspending the wires from the ends, to get rid of
the foam and cement and make sure nothing but air was in the area.
(Back of the envelope calculations seemed to indicate running it right
next to the floor could have reduced the signal velocity by as much as
20%, but that "calculation" involved several levels of guessing.) That,
too, doesn't seem to have been necessary -- though getting a result this
close to "exact" first crack out of the box still seems highly
implausible; makes me wonder if I actually had two (or more)
compensating errors. If I were getting paid for this I'd repeat the
experiment a few times, using a fresh setup each time.
Better yet, I would consider using a high-frequency oscillator and
looking at the phase shift rather than using a single shot. But that
would necessitate figuring out the impedance of the line so it could be
properly terminated; otherwise the echoes would make hash of it. I
_think_ the impedance could be measured by putting a tiny resistor in
series with the (single-shot) voltage source and comparing the voltage
curves on either end of the resistor, but to be sure I'd have to start
with a line of known impedance (50 ohm coax, for instance) and see if it
worked.
The time it takes to do stuff like this "right" just goes up so fast
it's incredible.
BTW I didn't know what the "correct" delay was until after it was all
over and I was writing up the web page -- that's when I found out the
result was almost dead-on...
K.