WIFI signal generation for deterministic WIFI performance testing

[taydin]

I am testing and evaluation various WIFI radios to be used in an embedded project. Currently I'm using a 2.4GHz / 5GHz combo access point, Wi-spy to view the spectrum and embedded linux. I'm doing continuous iperf tests and later look at the bandwidth graphs.

Now I'm at the stage where I want to completely eliminate any variance that occurs as a result of "in the air interference" and want to test how CPU utilization and the SDIO link between the WIFI radio and linux performs. So, I want to feed the WIFI radio a deterministic, fixed, clean WIFI signal and want to monitor CPU utilization, interrupt rate etc on the linux side. What options do I have for this? For example, what would happen if I connect the output of the WIFI access point directly to the WIFI radio over a 50 dB attenuator. Something like this. AP output to SMA cable, 50 dB fixed attenuator, SMA to u.FL cable and then into the WIFI radio antenna input.

If the above would work, it would also be possible to use an SDR like hackrf for more sophisticated test patterns. There are also vector signal generators that also seem to allow this, but that's not an option for me, cause I would like to keep both of my kidneys if possible 

[taydin]

I guess the difficulty, or the part that I can't wrap my head around is, how to have both transmit and receive happen on the same cable. The AP's output should be attenuated coming to the WIFI radio, and the WIFI radio output should be attentuated going towards the AP.

[Rerouter]

remove your antennas and use Coax and appropriate attenuation, if you want to measure with another device, use a 3db splitter and make sure both your measuring devices are properly terminated, most wifi modules already will be, but your measurement gear may not.

As for test patterns, that comes down to why you want well defined patterns, most bugs come up on unpredicatable signals of different densities, so say polling your device once a second may behave very differently to you just throwing it garbage data as fast as it can handle.

instead of vector signal generators, instead you could approach this with arbitrary waveform generators and just load up a snippet of a wifi exchange you want to repeat. that way your only down half a kidney

[taydin]

But what would I put in between the AP and the radio? A 50 dB fixed attenuator would probably work only one way (not sure about that, never used one and don't know what's inside one). It would have to be a symmetric attenuator.

Regarding generated WIFI signal, the important thing is that it completely saturates the link so that I get the worst case interrupt rate and SDIO traffic. So I would want to experiment with various patterns to achieve that.

[taydin]

Ok, I've been continuing to research this. There are T type balanced attenuators described in the wikipedia article. Just need to find such an attentuator that accepts SMA on both ends.

[mmagin]

Resistive attenuators work the same in both directions.  They're commonly available with SMA connectors for operation at WiFi frequencies (some are even good to 11 GHz and greater).  Here in the US I've bought unused surplus ones on eBay at reasonable prices, though if I didn't have the means to test them myself (a good signal generator + a microwave power meter) I would have purchased ones from Minicircuits probably.