WARP Project Forums - Wireless Open-Access Research Platform

The current PHY is built for point-to-point links; this assumption defines a lot of its behavior. There will be a few complications with this approach.

-The PHY calculates checksums when transmitting a packet and appends them to the user-provided bytes. The receiver checks this checksum when the reception finishes. But if two nodes are generating independent payloads, the receiver will need to somehow separate the checksums from each transmitter in order to actually detect errors.

-The packet detection subsystem runs a correlation searching for the preamble's long-training-symbols. The peaks of this correlation determine the timing of the rest of the reception (defines FFT windows, etc.). If two nodes are transmitting, they either need to synchronize themselves very closely, or a different preamble/correlation approach would be required.

-The carrier frequency offset estimation/correction systems assume a single Tx-Rx frequency offset. If two independent nodes are transmitting, there are two different offsets to estimate and track. This is a tricky problem. In real OFDMA systems, I think the base station "trains" the CFO during the downlink so that during the multiple-access phase, every simultaneously transmitting node applies an inverse offset, effectively (hopefully) resulting in no CFO from the basestation's perspective.

You could explore a lot of these issues in simulation using the PHY Sysgen model. You would have to duplicate the transmit subsystem, then tweak each instance's subcarrier assignments, checksum blocks, etc.

There is some (admittedly meager) documentation at http://warp.rice.edu/trac/wiki/OFDM. I'm hoping to add to it in the coming weeks.