This paper analyzes and compares two different approaches to base station sectorization: the traditional wide-beam trisector cell (WBTC); and the newer (but not well-characterized) narrow-beam trisector cell (NBTC). Unlike the WBTC architecture using three 100/spl deg/-120/spl deg/ antennas in a base station, the NBTC architecture (also called "clover-leaf") uses three 60/spl deg/-70/spl deg/ antennas at each base. We obtain new results on the quantitative performance of both architectures, taking into account the impact of (1) actual (rather than ideal) antenna patterns, (2) channel utilization factor, (3) site diversity, (4) non-ideal cell site locations, and (5) the influences of reuse factor and shadow fading. We show that the performance of NBTC systems exceeds that of WBTC systems, both in signal-to-interference ratio (SIR) statistics and signal coverage. The improvement in coverage translates to a reduced power requirement of about 2 dB for the same cell size, or equivalently, to a 25% larger cell coverage area for the same power. The improvement in SIR performance corresponds to a gain of 2-3 dB in the 90/sup th/ SIR percentile. This gain can be translated to significantly improved grade-of-service in cellular environments, with no attendant cost.