I'm doing some throughput testing with RB922 5Ghz radio cards, using 80-mhz wide channels.

Looking for lab-level performance, I am working with full enclosures, and using coax and attenuators between the AP and CL.

The combined RSSI is in the low to mid -40's. The on-board radios got me 530-550mbps in UDP mode, using NV2, and using devices on either side of this radio pair, not the radios themselves for the bandwidth test.

After the initial round of testing, I decided to disabled the on-board radio, and try out the 5.8GHz mini-pcix Mikrotik card (R11e-5HnD) in one of the units. I logged onto the CL, disabled the on-board radio, powered down, opened it up and added the pciex radio card. When connecting the 'N' connector antenna pigtails to the R11, I just took a stab-in-the-dark and connected them to whichever antenna connector on the R11.

My assumption was that if it's the wrong way, I should be able to tell, and can just swap the coaxes on the CL box. I logged onto the CL, copied the WLAN cfg from the disabled on-board radio to the R11 card. The CL then linked up to the AP (which had not been changed yet, and still using the on-board radio.)

I repeated the throughput tests. After coming up with the same 530-550mbps in UDP mode, which was similar to what I got with both using on-board radios, I thought I'd swap the coaxes on the CL to see what the loss was when the chains were crossed from AP to CL.

I was kind of bewildered that with the swapped chains, I was still getting the same signal level and throughput numbers!! (I had never really thought about it, nor tested it before, but) was under the impression that chain0 TX had to/is be received by chain0, and chain1 TX needed to be received by chain1, for the best results.

In effect, what I did was mis-connect a dual-polarized antenna, but on coax.

Has anyone ever done this testing?

Was I incorrect to assume chain0-to-chain0 and chain1-to-chain1?

Thanks in advance.

No, chains are not "hard-coded", it's fine for them to be crossed. Further more, in MIMO 2x2 (or NxM for that matter) most often every Rx antenna will receive signal from all Tx antennas (with different Rx levels) and receiver will then use (pretty complex) computation to make most out of received signal. It helps a lot to receiver if each Rx antenna receives signal from one of Tx antennas with much higher signal level - which happens in case of polarization diversity or as in your case with "coax-separated ether" ... or if received signals are shifted in time, which happens if paths have different lengths. This happens because antennae (both Tx and Rx) are at least a few wavelengths apart and if by some chance signal between different Tx/Rx pairs bounces from different obstacles than the signal between other antenna pairs (this is called "multipath" in radio communications). If all antennas (both Tx and Rx) are really close together, chance of experiencing multipath is very slim, only polarization difersity can work in this case. Same if there are no obstacles for signal to bounce off (i.e. line of sight with no obstacles inside Fresnel zone).