Tom,

Thanks for the link. This unit only goes to 60MHz so needs a lot of mod
to get it up to 900MHz where I am working.

I have found this one that does go up there
http://www.geocities.com/robert_laco...us_designs.htm

Some of the parts used in his design are getting difficult to obtain so
I am thinking of replacing the Cypress device and prescaler for a
LMX1600 PLL and driving the generator ramp via a PIC micro controller
instead. The VCOs, Splitters and Amp are all still available.

Before I start on this project I will make a manually adjustable version
(The directional couplers will not be redundant)

The reason the detector head is a separate item is that this will be
inside a small shielded area inside the main box that also has the micro
controller, LCD display, keypad and serial port.

Then I can simply swap the 900MHz coupler for the 433MHz coupler to
change bands.

I have completed the artwork for the Bi_Directional coupler and will
make the PCB today. Using RFsim99 I was able to play with combination of
track widths and gape between mainline and coupled lines to get the 20dB
coupling and 50 Ohms.

K7ITM wrote:
You should probably have a look at http://users.adelphia.net/~n2pk/.
Paul has done a nice job on a vector network analyzer project, with
some good additions from others, and you can probably pick up quite a
few nice ideas from that design, maybe even use a lot of it.

The single parallel strip really does pick up both forward and
reflected, and if you can terminate it properly at both ends
(especially the forward end) so that the load there does not reflect
back down that line segment, you can sense both forward and reflected
on the single parallel strip. Making the load really be 50 ohms isn't
easy when it also has to be a detector, though, so two strips may make
more sense for you. It should work fine to have them just be
symmetrical, one on either side of the main line.

If you really care about phase detection, be aware of the phase shifts
through line segments. Such couplers are sometimes called quadrature
hybrids. The sampled forward output will be in phase with the applied
input, but the through-line output will be delayed 90 degrees (a
quarter wave in the transmission line, obviously). The sampled
reflected output will be in-phase with the reflected that arrives at
the through-line load port--by symmetry. But that means you'd measure
a 90 degree phase difference between the forward and reflected, for a
case where they were in phase at the through-line output port. Hope I
have that all correct...if not someone will probably jump in and
correct it. It should be easy to calibrate out. Since the phase
shifts are frequency-dependent, if you want to cover much of a
percentage frequency range, be prepared to handle frequency-dependent
calibration. The coupler goes out of quadrature, and you probably
won't have exactly the same length lines from each coupler to the
detector.

Since the parts are really pretty cheap and replicating PC boards is
straightforward, you might think about the advantages of NOT having to
disconnect things, and the expense of putting good RF connectors on the
coupler, and maybe just build the detector into the coupler, and make a
separate one for other uses. A couple good SMAs and the hassles of
connecting and disconnecting a few times should pay for just buying
another $20 part. Half a tank of gas these days, huh?

Cheers,
Tom