The dependence of small-scale fading on bandwidth is quantified experimentally in the 3.1-10.6 GHz band for indoor channels. The fade depth converges to 4 dB at 1 GHz bandwidth, with little reduction for further increase in bandwidth. A simple yet accurate empirical fade depth model is developed, enabling convenient evaluation of the link budget for a channel with given bandwidth.Introduction: Multipath propagation in wireless channels causes random fluctuations in the signal levels, leading to small-scale fading and system outage [1]. Recent research has analysed the impact of channel bandwidth on fading using information theoretic and statistical approaches. For example, an increase in the stochastic degrees of freedom with bandwidth in frequency-selective channels has been reported [2,3]. This Letter investigates the variation of channel energy with bandwidth up to 7.5 GHz, spanning the FCC-allocated ultra-wideband (UWB) spectral range [4], and proposes a simple model to characterise this variation. Fade depth is a measure of the variation in the channel energy about its local mean due to small-scale fading. This determines the system fade margin, a key parameter used in link level design. The variation of fade depth with up to 10 MHz bandwidth was analysed for code-division multiple-access (CDMA) channels in [5]. To our knowledge, there is no model describing the relationship of fade depth and bandwidth over a large bandwidth range, which would be of value to wideband and UWB system design.