Tracking Efficiency

The tracking efficiency has as been difficult to try and determine because we do not know how many triggered events should have tracks. It is clear from examining the drift chambers alone that not every event should yield a track since, for most of the time, cannot even see a clear track from the event displays. There's a few ways to do estimates for this, though none are exact. They have all given us numbers of about 75% or higher for our higher current runs if I remember correctly, so I believe that this is closer to the true value.

Using Elastics

One way is to try and determine elastic events using all detectors except the drift chambers. This can be done using spacial cuts based upon the shower position and neutron arm hits, time of flight cuts, total shower energy deposition, etc. Once elastic events are (crudely) selected you can examine how many hits are in an elastic peak and the ratio is your tracking efficiency. I calculated numbers around 85~90% when the experiment was running, though I haven't redone it. It should be noted that this can be done only for the H2 runs which were generally at very low currents.

Events Skipped by Tracking

You can also estimate a lower bound by examining events that are skipped in tracking (B.dc.trackskipped>0). This is based on the assumption that the tracking code will always find a track given an infinite amount of processing time (we limit this time for practical reasons) and that events we wish to do tracking on will always lie within the shower cut regions (which should be true if we choose them carefully). Given this, the only events that the tracking causes an inefficiency are ones that we have intentionally skipped. Since we believe that not all events skipped could yield tracks, this gives us a lower bound.

In this case:

Eff = 1 - (% skipped)/(%with tracks + %skipped)

Events with Large Number of Planes Firing

I used a somewhat rough estimate during the experiment based on the number of planes firing. You'll probably remember making these plots for each run while we were running. What we used as our indicator was how frequently we found a track when a specific number of planes fired within our shower region.

For example, let's say we found 400 tracks when all 15 planes fired within the shower region and we had 500 events where all 15 planes fired within the shower region. This would give us a tracking efficiency of 80%. The flaw in this method comes from the fact that we have noise introduced in the system. It is possible that this noise could cause 15 planes to fire when no track was present artificially diluting our tracking efficiency. This effective becomes more likely as your reduce the number of planes firing and we can see in the data that this "efficiency" becomes smaller as you reduce the number of planes firing.

If you know the background rates and the plane efficiency I believe you should be able to calculate the true efficiency using this method, though I haven't attempted to do it. I believe this method will once again give you a lower bound on the tracking efficiency if you believe that the tracks we want to measure are falling within the shower region. I remember that we would get typical numbers of around 75% for production runs at our highest currents for the 15 plane case, which from all other measures is probably reasonable.