On the Physical Layer of Ultra-Wideband Systems

Licentiate thesis, 2005

This licentiate thesis consists of two published conference papers, one technical report, and an introduction. The focus is on communication at bit rates of 100 Mbps and higher for single-user systems at short range up to 10 meters.Three physical layers are investigated that use impulse radio, orthogonal frequency division multiplexing (OFDM), and carrier-based direct-sequence spread-spectrum, respectively. All layers use coherent detection. In all papers, the IEEE 802.15.3a channel model is used for the evaluation of the systems and their performance. In paper A, a comparison between direct-sequence impulse radio (DS-IR) and time-hopping impulse radio (TH-IR) at 100 Mbps is presented. The performance of suboptimal, fractionally spaced (FS), coherent Rake receivers with a pulse-matched filter is numerically evaluated with two channel estimation algorithms. TH-IR and DS-IR with antipodal modulation is shown to perform basically the same. The FS-rake receiver performs much better than chip- and symbol-spaced rake receivers. In paper B, an amplitude autocovariance function of the Fourier transform of channel impulse response is defined. A coherence bandwidth is obtained for every channel impulse response with this autocovariance function. The standard deviation of the coherence bandwidth is shown to be one third of the mean coherence bandwidth. Numerical simulations show that the performance of a multiband-OFDM system on the IEEE 802.15.3a channel model CM4 is close to the truncated union bound of the performance of convolutional codes on uncorrelated Rayleigh fading channels. Paper C provides a set of physical layer specifications for a carrier-based single-band direct-sequence spread-spectrum system so that the physical layer requirements from IEEE 802.15.3a are fulfilled. Performance results are also presented for a dual-band system that uses the same spectrum and sampling rate as the single-band systems. A channel impulse response gain is defined. It is shown that this gain, which is a random variable, can be approximated by the multiplication of two other random variables that have a log-normal and a gamma distribution, respectively.