On TPC cluster reconstruction

For a bias-free momentum measurement of TPC tracks, the correct determination of cluster positions is mandatory. We argue in particular that (i) the reconstruction of the entire longitudinal signal shape in view of longitudinal diffusion, electronic pulse shaping, and track inclination is important both for the polar angle reconstruction and for optimum r phi resolution; and that (ii) self-crosstalk of pad signals calls for special measures for the reconstruction of the z coordinate. The problem of 'shadow clusters' is resolved. Algorithms are presented for accepting clusters as 'good' clusters, and for the reconstruction of the r phi and z cluster coordinates, including provisions for 'bad' pads and pads next to sector boundaries, respectively.