Performance Assessment of Mobile OFDM-based Systems： Variability within Given Wideband Channel Categories

An ongoing challenge in communications is to provide ever faster, broadband multi-media access to mobile terminals. In the mobile multipath environment, performance is affected by both frequency-selectivity and Doppler spread. Realistic channel simulators are required to compare competing systems on a fair basis. The RACE code-division testbed （CODIT） provides a realistic model based on extensive wideband measurements, ideal for OFDM （Qrthogonal Frequency Division Multiplexing） testing, which defines a range of indoor and outdoor environment categories. However, there remains considerable variability within any given category. To enable fair and reliable comparison, the variability within a given channel type must be quantified, since this determines the length of simulation time and the number of independent instances of a channel type to be averaged before stable and repeatable results can be inferred. This paper explains how the channel model was implemented, providing the important environment parameters. Variability is then quantified in terms of statistics of BER （Bit Error Ratio） from simulation runs of differing lengths for many different instances of nominally the same environment. An accurate estimate （-5%） of mean BER for a given CODIT environment typically requires averages over simulations of around 90-150 instances of the channel, with runs of at least 700 OFDM symbols.