To improve re-use of time/frequency slots in a cellular radio system, it is desirable for the average interference levels seen by all users to be made approximately equal. We provide constructions based on orthogonal latin squares that guarantee different sets of users will interfere in successive slots. We illustrate how this may be combined with convolutional coding to provide large performance improvement with low decoding delay in a slow hopped system. I. Introduction In mobile cellular

more »

... o, the dominant impairments are multipath fading and interference from other mobiles. In a conventional time division multiple access system (TDMA), mobiles are assigned slots which they keep from frame to frame, with the interfering mobiles assigned slots in the same way. As interference levels vary widely between slots due to such factors as shadowing and geographic proximity, some mobiles suffer from persistently poor signal to interference ratios (which we shall refer to as C/I). Systems are generally designed for 90% and 99% worst case conditions. Therefore, the result of this uneven interference distribution is overly conservative restrictions on frequency re-use between cells, and thus reduced capacity. If instead slot assignments were arranged such that different interferers were encountered in successive frames or slots and repetition or channel coding used to provide averaging, then the worst case error statistics would improve. In [1] random hopping over eight frequencies was used to average over different interferers and channel conditions. As a result, all frequencies could be re-used between cells. In [2], a combination of coding, explicit frequency re-use restrictions, and slow frequency hopping was employed. Both yielded capacities similar to that of the direct sequence spread spectrum code division multiple access (CDMA) system analyzed in [3] . Here we provide a specific construction of the hopping patterns which when combined with convolutional coding leads to improved performance, without restrictions on frequency re-use between 1. This work supported in part by AT&T Bell Laboratories and ARPA contract DAAB07-93-C-C501.