Ultra-wideband Technology

Theory

• Channel measurement and modelling
• ISI effect is dominant, compared to fading effect.   Since the pulse is very short, the ISI or echoing effect (superposition of replicas) is very strong.
• Saleh-Valenzuela Model (S-V)
• For UWB channels, since the transmitted pulse is so short that every multipath is resolvable (i.e., different replicas/echoes do not collide in time), there is no need to use equalizer to remove ISI.  Perfect equalization only gives less than 1 dB gain over the non-equalization case.   (See Arjunan Rajeswaran's ICC 2003 paper)

For an ultra-wideband (UWB) system, since the signal bandwidth and the sampling rate are high, there are only a small number of paths that contribute to a resolvable tap/bin.   So central limit theorem does not hold and we do not see uniformly distributed phase.   It is observed in actual measurements that the phase only takes values of 0 or \pi.

Measurement results show that the amplitudes can be modeled as lognormal-distributed r.v.'s.  The reason why the amplitude does not follow a Rayleigh distribution, is that central limit theorem does not hold due to a small number of paths associated with a resolvable bin.   Two Poisson models are employed in the modeling of the arrival times of resolvable paths.  The first Poisson model is for the first path of each path-cluster and the second Poisson model is for the paths (or rays) within each cluster.  The power delay profile is modeled by a double exponential decay model; i.e., the average powers of the first paths of path-clusters follow an exponential-decaying law, and the average powers of the paths within each cluster also follow an exponential-decaying law.

• Since lognormal distribution is statistically better than Rayleigh distribution for the same SNR, i.e., lognormal has a larger tail than Rayleigh, we need fewer taps in the RAKE receiver to achieve the same BER performance.

• Modulation and detection
• Interference and capacity
• Regulatory issues
• Positioning services