The long electron drift times in the MDT's (0 to 500ns) together with the high background counting rates in the muon spectrometer could cause a significant drop in efficiency due to signal overlay. For that reason the use of a double track separation circuit on the MDT front end was studied.
For 2x pole/zero tail cancellation and a preamp with 15ns shaping time, the signal width caused by a single ionization electron is about 100ns which, assuming a uniform irradiation of the tube, results in a flat deadtime spectrum of 100 to 600ns.
In order to be able to see the leading edge of the overlayed signal one has to raise the threshold and overcancel the signal tail. Since raising the threshold causes a decrease in resolution and overcancellation results in baseline shift these options can only be added in a separate second channel. The pattern recognition program would start with the TDC hits from channel one and would look at TDC values from channel 2 only in case of a recovered inefficient tube.
Figure 7 shows separation efficiencies for a 3x pole/zero tail
cancellation circuit and different thresholds.
The signals induced on the wire were simulated with the
GARFIELD[7] program and
then sent through a simulated ASD chain with a stand-alone
program. Signals for randomly distributed distances were generated and
overlayed. To define the efficiency one looks for the leading edge
of the overlayed track within 
of the expected position,
where 
is the position resolution for single tracks.
The number of hits in the plot refers to the average number
of hits for a double-track event for the given threshold.
The efficiency goes up by raising the threshold until one starts to exceed
the signal. Clearly for a high efficiency one has to pay with a
degraded resolution of the overlayed event.
The curves on the bottom of the plot refer to fake hits which is
the fraction of hits within of the expected
position caused by a spike of the first signal without
overlaying the second one. The resolution numbers in the figure are the
position resolution obtained for the second track.
The simulation shows that by adding a second channel with higher threshold and strong tail cancellation one can reduce the deadtime from a flat 100ns to 600ns distribution to a 80% efficient deadtime of 100ns.