OFDM Reference Design v14 - CSMA Benchmarks

Timing

The basic packet exchange which defines the maximum throughput for CSMA is DATA-ACK-DATA. The timing of this exchange is our implementation is illustrated below. Please note the figure is not drawn to scale.

This figure assumes the following parameters:

• SISO antenna configuration
• 2 OFDM symbols for channel training per packet
• 24 byte MAC header at QPSK (2 OFDM symbols)
• Full rate modulation of QPSK (12 bytes per OFDM symbol) or 16-QAM (24 bytes per OFDM symbol)
• 1484 byte payloads (1470 byte IP datagram + 14 byte Ethernet header)

This figure includes a random backoff period imposed after a node receives an ACK. This period is designed to prevent two fully-backlogged nodes from simultaneously attempting to transmit a new packet immediately after the previous DATA-ACK exchange completes. The backoff imposes a wait of an integral number of slot durations (a slot is 22µs in v14), with the slot count drawn randomly from [0,7].

Given this timing, the average minimum period for transmitting a new DATA packet is 1263µs (for QPSK full rate) or 775µs (for 16-QAM full rate), implying a peak data throughput of 9.31Mbps for QPSK and 15.2Mbps for 16-QAM.

Performance Tests

Peak Throughput

Methodology: two WARP Kits (SISO Kit v1), with radios connected via a cable and 60dB attenuation, FPGA boards each connected via Ethernet to a MacBook laptop. One MacBook runs an iperf server, the other runs the iperf client, using one of the following calls:

#For UDP tests:
iperf -c 10.0.0.1 -u -b 20M -l 1470 -t 30

#For TCP tests:
iperf -c 10.0.0.1 -l 1470 -t 30

Results:

Full Rate	TCP	UDP
QPSK	XX	XX
16-QAM	XX	XX

Packet Error Rate

Bit Error Rate