S21 on outputs of 4 way splitter.

[waveboy]

Dear Engineers,

I recently attended an interview for a Microwave job and was shown a graph of S21 (magnitude in dB) against frequency (2-6 GHz) for the splitter's 4 outputs. (it was not an isolated splitter)

I would be most grateful if any Engineer could help with the interviewer's question...

For the question i was asked why one of the 4 splitter output traces was sometimes above that of the other three.

The graph showed all of the splitter's 4 output traces. The traces were all contained between -7dB AND -6.5dB.
However, though three of the traces were virtually coincident across the frequency band, one trace was about 0.2 dB above the the other three over the 4 to 5 GHz interval. Over other parts of the band, it was also slightly above the other three traces.

If anybody has even the slightest idea as to what type of things could have caused this deviation in S21 for one of the outputs then I would be most grateful indeed.

Best Regards.

 

[biff44]

This is a generic question designed to see how your mind works and your general understanding of microwaves. Obviously, a theoretical power splitter would have equal power outputs at each frequency. So the answer has to do with something that is different from theory and practice.

If it were me I would ask some questions of my own.

Is this 4 way power divider a true 4 way, or is it composed of three 2 way power dividers? If it was a true 4 way, how did they connect the isolation resistor from ouput 1 to the farthest output 4? If these were not wilkenson type splitters (for instance a branch line hybrid) then the outputs are NOT theoretically equal in power, over the frequency band.

Is the layout perfectly symmetrical (does each output have the exact same path length from its output to the common input, or is one path a little shorter than the others)?

Have you done a VWSR test at each of the output ports to see if each output is well matched to 50 ohms, or did 3 of the ports have worse VSWR than the fourth better port?

You should ask if the same output detector was used to measure all 4 outputs, or if more than one dectector was used (and possibly had a different calibration error)? Were all other outputs properly terminated when measuring the individual output ports?

Ask to look inside of the unit to see if there were any assembly difference, like solder blobs near a connector, etc.

In other words, just as a lot of bs about the unit to show that you can identify what a problem might have been, and ask enough to either determine if that was the problem or not, or needs to be investigated more.

 

[waveboy]

Thanks for your reply biff44,

May I ask another qu?.....you questioned whether one output was a little shorter than the others. -By this, am i right to believe that you are inferring that if one path was a little shorter, then it may have given the higher S21 value? (By 'higher' i mean 'less negative')

...The interviewer did seem to be guiding me toward saying something about the "phase" of the outputs.

-Regarding phase, I am cognisant that a slightly shorter path would mean a slight advance in the phase of that output.....though the graph that i was shown depicted the magnitude of S21. (It did not have the phase information).

Regards.

 

[biff44]

By longer path, I mean that a signal that travels along a longer path of microstrip would have more loss (loss/length x length).

It is possible that the signals phase up a little better on one output port than the other. Did he show you what type of splitter it was internally?

 

[logbook]

Obviously the correct procedure is to ask the interviewer at the end of the interview what the correct answer was. I usually offer answers to my questions as a matter of course, unless the candidate is too hopeless and I then might not bother.

It is interesting that BIFF was talking about 50 ohm matching. I assumed it was a waveguide structure rather than a coax structure, but it makes no difference.

The question one could ask is why is the output between -6.5dB and -7.0dB. In other words why is it not -6.0dB? The answer is of course genuine loss or just reflection. 0.2dB variation is only about 4% of power. If the matching or machining is off slightly this sort of error might easily be found. Of course if this is just a simulation then with a totally symmetric structure you would expect more "prefect" results. In this case one has to question the simulator setup. It is easy to mesh the problem inadequately and get a simulation error as a result.

 

[waveboy]

Thanks for all your replies. I was not actually physically shown the splitter which produced the aforementioned S21 plot. I was just told that it was of the type that is made up of three 2-way dividers. The interviewer drew an extremely simple "stick" diagram of such a splitter.

 

[biff44]

Yes, an as this is 2006 I also assumed they did not ask him any questions on maser amplifiers or parametric varactors!