wiki:OFDMReferenceDesign/Applications/Bridge/Benchmarks/v14

OFDM Reference Design v14 - noMac Benchmarks

Performance Tests

The tests below use the same payload, MAC and PHY parameters described in the Timing section of the Bridge application documentation.

Peak Throughput

Methodology: the tests below are conducted using two WARP Kits (SISO Kit v1), with radios connected via coax cables with 60dB series attenuation. The FPGA boards are each connected via Ethernet to a MacBook laptop. The setup is illustrated below.

No image "wired_test_setup.png" attached to OFDMReferenceDesign/Applications/Bridge/Benchmarks/Files

This test measures the peak throughput between two PCs connected via the OFDM Reference Design's bridge application. The PCs use the OFDM/noMac link as a wired-wireless bridge, emulating a point-to-point Ethernet link.

We use iperf for throughput testing. One MacBook runs an iperf server, the other runs the iperf client, using one of the following calls:

#For UDP tests (transmit 20Mb throughput with full-length packets for 30 seconds):
iperf -u -c 10.0.0.1 -b 20M -l 1470 -t 30

#For TCP tests (transmit full-length packets for 30 seconds; TCP settles to its achievable throughput):
iperf -c 10.0.0.1 -l 1448 -t 30

Results: Throughput in Mbps, as reported by iperf:

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

Bit/Packet Error Rate

Methodology: the tests below are conducted using two WARP Kits (SISO Kit v1), with radios connected via coax cables with 60dB series attenuation. The FPGA boards are connected to a common Ethernet switch, along with a MacBook laptop running the refdes server application and client script. The setup is illustrated below.

No image "ber_test_setup.png" attached to OFDMReferenceDesign/Applications/Bridge/Benchmarks/Files

The PER is calculated by comparing the number of packets transmitted by node 0 and number packets successfully received by node 1. A packet is successful if it is received with zero errors in its header and payload. Packets received with header or payload errors, or packets which are not detected are counted as errors.

The BER is calculated in real-time by the PHY. This test uses a feature of the PHY which allows the receiver to compare an incoming packet with a packet already stored in a local buffer. In this test, the refdes server sends a packet via Ethernet to both nodes. Node 0 transmits the packet via the OFDM PHY. Node 1 copies the packet to a local buffer, which the receiver uses for its BER comparison. The receiver only updates its BER counters for packets which are received with an error-free header (packets with header errors cannot be received, as the modulation and length fields can't be trusted).

The independent variable in both tests is transmit power at node 0, swept over a range of 60dB. The power is controlled by setting the gain applied by the variable gain amplifier in the MAX2829 Tx path. The X-axes below are path loss, ranging from -60dB (maximum transmit power, 60dB series attenuation) to -90dB (minimum transmit power, 60dB series attenuation).

Last modified 14 years ago Last modified on Dec 1, 2009, 4:02:23 PM