Training and tracking of adaptive DFE algorithms under IS-136

The performance of wireless systems is severely corrupted under randomly time-varying channels. Efficient algorithms suitable to implement for fixed point DSPs are required to compensate for this degradation. The IS-136 standards define channel conditions under which wireless TDMA systems are required to work. Adaptive equalizers are unavoidable in order to satisfy these requirements. We consider training and tracking behavior for adaptive DFE structures under IS-136. It turns out that the essential difference between the performance of RLS and LMS algorithm is due to the training phase rather than the tracking behavior. For most of the conditions in IS-136, an LMS algorithm can track almost as well as an RLS algorithm when applied to a decision feedback equalizer, assumed that an RLS algorithm was run during the training phase. Simulations defined by the IS-136 standards corroborate our assumptions and validate theoretical results.