Discrete component hardware neural net for drift chamber track finding

The muon shift chambers at the D0 detector at Fermilab use both the transit times of pulses and the charges induced on shaped cathode pads to reconstruct a track position along a wire. A simple three-layer neural net (Padnet) was built to map the two input pad charges into a position within the shaped pad repetition length. Another net (Tracknet) was made to reconstruct tracks from three decks of chambers. Tracknet has six inputs, three from the Padnet outputs and three from the transit time measurements. Tracknet maps these six inputs into a single output which is proportional to the distance from the colliding beam interaction axis. Tests are underway with a small section of a standard muon chamber to measure the operation of the nets. Very preliminary results are shown