K7ITM wrote:
(snip)

To help resolve that, I did a Spice
simulation; I modelled a transmission line with ten "L" sections
cascaded. Each was 1uH series, followed by 100pF shunt to ground. I
put a 100 ohm load on one end and fed the other end with a 2.5MHz sine
wave with 100 ohms source resistance. Sqrt(LC) is 10 nanoseconds per
section, so I expect 100 nanoseconds total delay, or 90 degrees at
2.5MHz. That's what I saw. Then I added unity coupling among all the
coils, and to keep the same net inductance, I decreased each inductor
to 100nH. The result was STILL very close to a 90 degree phase shift,
with a small loss in amplitude. In each case, the current in each
successive inductor shifts phase by about 1/10 the total. Although the
simulation is less than a perfect match to a completely distributed
system with perfect flux linkage (and just how you do that I'm not
quite sure anyway...), but it's close enough to convince me that
perfect flux linkage would not prevent behaviour like a transmission
line, given the requisite distributed capacitance.

(That was from a "transient" simulation, 10usec after startup so it
should be essentially steady-state; but I'll probably play with an AC
sweep of both cases as I find time.)

I look forward to your tests with turn-to-turn capacitance added to
the model as well. It will no longer match the 100 ohm source and
load over as wide a frequency range, but it will look closer to the
real thing above the self resonant frequency. You may be able to
measure the phase shift versus frequency up to resonance, and test
Cecil's idea that you can measure the self resonant frequency and use
that to predict the phase shift at much lower frequencies.