wiki:OFDM/MIMO/Docs/Benchmarks

MIMO OFDM | Documentation? | Performance Benchmarks

This page presents some physical layer benchmarks for the WARP MIMO OFDM Transceiver. For higher-layer benchmarks (like throughput), please refer to one of the OFDM Reference Design applications.

Methodology

The tests below were conducted using the OFDM PHY included with OFDM Reference Design v14. The test setup consisted of two WARP kits (SISO Kit v1) whose radio boards were connected via coax cables and 80dB series attenuation. A laptop was used to generate traffic, consisting of 1454 byte Ethernet frames consisting of random data which changed with each packet. The nodes were configured to transmit these packets in one direction, with both nodes tracking statistics (Tx packet count, Rx good/bad packet count and BER).

Results

Packet Error Rate: PER is measured by transmitting a known number of packets, then counting how many are received successfully. A packet error occurs either when the receiver fails to detect a packet or receives it with at least one bit error in the header or payload.

The experimental independent variable is gain of the variable gain amplifier in the MAX2829 transceiver at the transmit node. The Tx gain is controlled by a 6-bit digital value, which maps control values [0,63] to Tx gains of approximately [-31, 0]dB. A control value of 63 corresponds to maximum transmit power, which is the default for the OFDM Reference Design. We sweep over control values of [15,63] in our tests. The plot uses an x-axis of 'Channel Gain', which is the sum of the Tx gain and the 80dB external attenuation. This axis is analogous to (and used as a proxy for) SNR.

PER for QPSK modulation for full-rate symbols:

Bit Error Rate: BER is measured by transmitting packets with payloads known to both nodes, then counting how many payload bits are received in error. The PHY records bit errors only when a packet is received with a valid header. Packets which are not detected and packets received with header errors are not included in the BER calculation. This biases the results, but is the fairest measurement we can take without modifying the PHY to ignore header values (length and modulation rate) for received packets.

The experimental independent variable is the same as for the PER test described above.

BER for QPSK modulation for full-rate symbols:

Last modified 14 years ago Last modified on Dec 4, 2009, 11:48:42 AM