You are not logged in.
Hi,
we would like to experiment with MIMO links that use various modes such as SISO,
2x1 STBC and 2x2 SDM (Spatial Multiplexing). OFDM ref. design v13.0 seems to be
enough for communication with all three modes. However, for better understanding we need
to export certain data from the boards to a user-space application running on the host PC connected
to one or more of the WARPs. This needs to be a wide range of different data such as MAC backoff
counters, packet delivery rate, number of corrupt packets as well as the fine grained PHY parameters
(I/Q samples, constellations, RSSI etc.) included in the WarpLab ref. design.
Can we practically use WarpLab with OFDM ref. design v13 just to gather these data and display them?
(with some modifications to the WarpLab .m files and OFDM design) If yes, could someone point us
to the necessary modifications so we can move on implementing them? I guess we can follow the same
approach by exporting these data to some registers and reading from Matlab.
Is using TeraTerm another alternative to display gathered statistics? We tested a very lame approach of
xil_printf'ing everything with a timer but this will not allow us to pretty-print the PHY parameters?
Any help on this issue is appreciated.
Cheers,
Mustafa
Offline
I'm not sure a single design can provide everything you described. The OFDM reference design does support the 3 PHY modes, and is capable of tracking performance statistics per-flow (good/bad packet counts, average SNR, MAC retransmit counts, etc). These statistics can be transferred to a host PC via Ethernet. But the design does not buffer raw I/Q samples; the over-the-air waveforms are created/processed in real-time. WARPLab does exchange raw I/Q samples with a host PC, but the transfer via Ethernet is much slower than real-time. I would suggest considering which of your experiments require raw I/Q samples in MATLAB, and which require aggregate performance statistics, then dividing the experiments between the WARPLab and real-time (OFDM ref design) design flows.
Offline
Thank you. That makes sense.
I have question just to clarify:
By saying that these statistics can be transferred via Ethernet, do you mean we can
encap the statistics at the board and send it to a small app (on the host PC) listening on some pre-defined port?
Can we encap the statistics with MAC, IP and UDP headers? If not, can we do something like a tcpdump on the host PC to filter out
the statistics packet from other sniffed Ethernet packets?
Thanks,
Mustafa
Offline
The base OFDM reference design doesn't use an IP stack for its Ethernet interface, so socket-based links would be tricky. Using raw Ethernet frames works great, however. We've used this technique for some experiments in the past. warpmac.c includes a function (warpmac_sendRawEthernetPacket(void* pkt, int length)) that will transmit an arbitrary byte array via Ethernet. Then using tcpdump or a custom app built with libpcap you can capture and process these packets. We use custom values in the ethertype filed of the Ethernet frame (see Ethernet II framing for details) to filter received packets on the PC.
Offline
Thanks so much.
Offline