Influence of channel transfer function on LMS algorithm based nonrecursive deghosting in analog TV

Ghosts are time delayed, attenuated and distorted versions of the intended television signal. This paper treats the influence of the channel transfer function on performance of an adaptive inverse modeling method, namely the least mean squares (LMS) algorithm. The zeros in the channel transfer function correspond to poles in the inverse model. With these poles inside or outside the unit circle on the z-plane, accurate enough modeling can be achieved with left-handed and right-handed sequences, using finite impulse response (FIR) filters, large enough and modeling delay. For non-repeated poles, the level of residual ghosts is greater, but line-by-line processing can make them smaller and their level decreases when the number of taps is increased. Repeated poles cause the largest magnitudes of residual ghosts, and they cannot be suppressed more than a level even with an increased number of taps.