Directivity of a Directional Coupler

[logbook]

I guess this is a similar question/problem to my last question on waveguide spacers.

(These are factual statements not academic thought experiments) I have a 10dB directional coupler with a claimed directivity of 30dB. I have four nominally identical terminations each claiming to have a VSWR of better than 1.06 (p=0.029). The terminations and coupler are from the same manufacturer. I tried to verify the directional coupler at a spot frequency using a power source and yet another directional coupler to give me constant incident power.

So I did the experiment as it says in my text books, just swap the coupler around and take 10*log10 of the power ratio. Easy. The trouble is the results don’t seem to tell me much at all. All the power ratios were above 28dB, which is consistent with vector summation of the error terms (it could be around 24dB worse case). The trouble is that it seems that I cannot set a limit on the directivity of the coupler at all. For example a 15dB directivity coupler and 15dB return loss terminations could give the same result. Certainly if the coupler directivity is poor the termination would have to be similar in order to get a good overall result, but that doesn’t give me a limit.

It seems that I need something else to separate the variables. Perhaps a sliding termination?

How do engineers normally handle this situation? (Other than getting someone else to calibrate the loads/couplers of course)

 

[Higgler]

Expensive vector network analyzers are needed to measure devices primarily since their test ports' effective load return loss becomes nearly perfect after calibration (-50 to -70 dB rtn loss, if you keep your cables steady, -40 dB usually if you bend your cables a bit). That's the best way (or actually the only good way) to measure your device due to the calibration method making the loads on the device nearly perfect.

Real life coupler performance for both coupling and directivity depends on your load quality at your connectors. Hence, you may not be able to get the results they state in real life without ultra low rtn loss loads.

kch

 

[logbook]

I suppose one has to get a VNA, hook input to output and tell it that S21 is 1 and S12 is 1. And additionally that S11 is 0 and S22 is 0. Then a piece of open waveguide should measure as a relatively low loss. Then various offset shorts should give defined reflections; a full 12 term calibration in fact.

I just found a VNA that will do the job. I hate to think how much it is going to cost but that is someone else’s problem!

 

[Higgler]

Have an experienced ebay shopper (familiar with scams) look on ebay for used VNA equipment. It's common for scammers to try and sell VNA's on ebay, and ebay's "security" staff does nothing to prevent most thefts.

kch

 

[logbook]

There's no way I can buy a second hand one. It's going to have to be new, state of the art, and seriously expensive. I'm thinking maybe £100k. Even the cheap ones are £40k and I need the frequency extension heads.

I think I must be missing out on some fundamental metrology considerations. I know standards labs use air lines as calculable impedance standards. This must get tricky for waveguides as the impedance changes with frequency ratio from cutoff. I suppose a standard straight guide is an impedance standard, like an air line, but the termination can't be checked at DC and its inductance calculated as you could with a coaxial standard.