You are not logged in.
What resources (aside from additional AGC cores) do we need in order to extend the switching diversity to 4 rx antennas? Do we also have to modify the packet detection?
Offline
The packet detector already supports four radio inputs; you can see this in the source ofdm_pktDetector_mimo.mdl model in the repository. You would need to instantiate two AGC cores, and modify the PHY Rx inputs to accommodate four I/Q pairs (instead of the current two).
Offline
We tried increasing the TX/RX antennas to 4 by introducing a second parallel data pipe in the ofdm core and the end result was too large for the FPGA fabric. Is it possible to process data from the two additional antennas using the same (I)FFT block in the ofdm core? If yes, I guess we have to double the OPB clock to 80MHz. What other changes would be required, specially w.r.t. clocking of other pcores?
Offline
The FFT could be run faster (I actually just committed a version of the 2x2 PHY that runs the Rx FFT at full speed). The Tx and Rx FFT cores could also be merged, if you're going to stick to half-duplex. These are things we're hoping to optimize in the PHY at some point.
Are you trying to build a 4x4 MIMO system (i.e. all four antennas operating in the same channel)? If so, it'll be more complicated than 2 parallel 2x2 PHYs, especially in the equalizer where the detection will require 4x4 complex matrix operations.
Offline
My system is 4x4 but it performs RX and TX beamforming instead of MIMO. Therefore I don't have to deal with the 4x4 matrix computations.
Offline
I haven't built a beamforming system before. If you're beamforming across 4 antennas, you really only need a SISO equalizer, right? You'd need a 4-way channel estimator (requiring a 4-way FFT) to feedback beamforming coefficients, but once you start receiving actual payload, there's a single spatial stream requiring just a simple (basically SISO) equalizer. If this is all true, you could build it using pieces of our current PHY, but it will take some serious modifications to get it running.
Offline
You could say the equalizer is SISO, however I need to combine the 4 received data streams on 4 RX antennas into a single stream before passing it to the equalizer. So I always need a 4-way FFT. I currently have a working Simulink design using a second FFT block to process the two additional spatial streams (antennas 3 & 4) but the 18x18 multipliers end up being overmapped (105%). That's why I need to run FFT block at double its current rate and possibly reuse the channel estimators of the first antenna pair for the second pair. Do you think this would be feasible?
Offline
You can definitely run the FFT at the model's fastest rate; the latest PHY does this (for the Rx only) and works great over the air. You could also probably adapt the current MIMO channel estimator to work for 1x4 instead. It currently estimates four taps per subcarrier (A-A, A-B, B-A, B-B) when in MIMO mode. It seems like you should be able to rework it to estimate four SIMO taps (A-A, A-B, A-C, A-D) with the same resources.
Offline