The type of coupler I am thinking of is indeed where the coupling lines
for
forward and reflected are in the same physical region. The point I was
making refers particularly to the HP 778D; where it is specified in:
http://cp.literature.agilent.com/lit.../5952-8133.pdf
"Data can be read from the two meters of the vector voltmeter and
transferred directly to a Smith Chart". Indicating that there is no
significant phase error over the nominal bandwidth of the coupler.

The old vector voltmeter has a phase offset ability to compensate.


Just read Wes' comments on the line stretcher on the above pdf. I must
admit I also wondered about that, but assume it is intended as a cal
adjustment for a short/open standard.

See my previous post on this. It IS used to get both samples "At the
measurement plane."

73, Steve, K,9.D;C'I

Ok, Steve, makes sense. Anyway, have re-visited Matthaei et al, and also
"Foundations for Microstrip Circuit Design", by T. C. Edwards (Edwards does
appear to contain an error referring to "j-omega", which should simply be
"omega"). Both texts agree on the expression for coupling on a single
section, quarter wave, TEM mode, coupled transmission line. Also realized
that the equation does contain a frequency parameter.

Running an analysis in MathCAD, for a 20 dB coupler with design center at
150 MHz, produces some interesting results:

Coupling at 100 MHz, and 200 MHz = -22.5 dB;

Maximum phase error -- at band edges = +/- 2.5 degrees.

Increasing the frequency range of analysis from 50 to 250 MHz shows a
dramatic drop in coupling amplitude, at these frequency limits, to 32 dB.
Phase error, however, does not seem to be effected very much; peaking at +/-
2.8 degrees, and following what appears to be a sinusoidal curve.

From experience I know that coupler directivity degrades significantly
beyond the design bandwidth. Edwards does state: " Values of directivity,
on microstrip, beyond 12 to 14 dB are difficult to achieve". Co-planar
structures are much better, and can easily be analyzed with Genesys' 2.5D EM
simulations. Without access to HFSS, or similar FEM programs, I doubt
directivity could be calculated for coaxial structures.

The main problem, with operation of a coupler beyond its design bandwidth,
appears to be its loss of coupling. This, combined with degradation of
directivity, would certainly account for very large phase errors.

Frank