Hi! Actually I am thinking about rf-oscillators. Could it be possible to double the frequency of an oscillator if I forward the signal through a wilkinson divider to split the signal into two paths and then feed those two paths into a ratrace-coupler? The mixing products should be f1 + f2 and f1 - f2. Thanks!

A ratrace coupler is not a nonlinear element so it can by itself not do mixing, or can it? Basically you can do frequency doubling with any non-linear element. Feeding your signal through a diode is enough already to get a component with twice the frequency (as well as other harmonics). All you need is a filter to select the harmonic you want.

Hm, wiki (http://en.wikipedia.org/wiki/Rat-race_coupler) sais: "Rat-race couplers are used to sum two in-phase combined signals with essentially no loss". Is that wrong? But I will also keep the hint with the diode and the filter network in mind. Thank you :)

Aah... the "sum" is not the sum of the frequency, just of the voltage, right?

OK, I got my mistake. In some radar applications ratrace couplers are used as mixers but they always have diodes at their ports. Now I understand why. The diodes do the mixing! :)

Yes, exactly. Whenever a device generates frequencies at its output which are not present at its input, there must be a non-linear device (such as a diode) involved, and a microstrip arrangement is to my knowledge never non-linear.

And yes, the sum is of the frequencies, not the voltages; you need to multiply the voltages to get the mixing effect. You can most easily see that by denoting the signal as sin(w*t) and then looking at sin(w*t) + sin(w*t) = 2+sin(w*t) (no new frequency components) and sin(w*t) * sin(w*t) = 1/2 (1-cos(2*w*t)), which contains half the amplitude in the difference frequency (the "1", which is DC because your two frequencies are equal) and half in the sum frequency 2*w, which is exactly what an ideal frequency mixer would give you.