I also realize now that the type of
cable being used is probably the worst possible choice. Just swiching
to a foam dielectic cable may solve the problem.

Jimmie

The physical dimensions of the coax changes with temperature. The
ratios of inner and outer diameter stays pretty much the same, so the
Z doesn't change, but the epsilon of the dielectric changes (it's less
dense at higher temperatures). since propagation speed goes as 1/sqrt
(epsilon), and the density goes as 1/t^3, the prop speed goes as 1/
(sqrt(1/t^3)) = t^(3/2)...

On top of that the coax gets longer as a function of T, so the time
delay/phase goes as T/(t^1.5) or 1/sqrt(T).

A CTE of 30ppm/deg C is about 0.05%/deg C for time delay. If your
coax is 1 wavelength long, a change of 5C is about a degree. (0.25% =
1 part in 400, 360 degrees/wavelength, etc.)

You can see that working at microwave frequencies, where coax might be
100 wavelengths long, trying to match to a degree is a challenge.