There are a variety of testbeam results from 1996 available. As time permits, more of the work will be included here. At the moment, you can choose from the following:
Here are the first test beam results! We see beautiful peaks with very little noise. If you look at the Postscript file, you will see a typical pad in the detector as well as a sum of all pads. Notice that the sum is raw and does not correct for pad-to-pad variation in the pedestal. There are also no corrections for hit location, pulse sharing, common mode noise, etc. Despite these limitations, you see signal to noise for 1 MIP (~ 10 GeV pi-) of almost 30-1! If you look closely at the sum, you will also see some junk below the pedestal at the few % level. This results from a few bad channels that we saw at MIT as well and are investigating further. It appears to be an ADC problem, not a detector problem.
The three panels in the figure show minimum ionizing peaks from tracks going through the Si at three angles. The top panel has tracks perpendicular to the surface, while the next two panels show the distributions from tracks at angles of 30 and 50 degrees from the normal. The arrows show the locations of the peaks found by a fit. Within the uncertainty in the fit and the measurement of the track angle, the peaks occur at exactly the location predicted by the increasing track length through the Si with increasing angle.
Notice also the significant increase in the events between the pedestal and the peak. These are events in which the signal is shared between pads. The region in which this can happen is very narrow (~10-20 microns) in the transverse direction so pulse-sharing is a rare occurance for normally incident tracks. However, this pulse sharing region runs the full 300 micron depth of the wafer. Thus, when the Si is rotated by 50 degrees, the width of the pulse sharing region as seen by the tracks is hundreds of microns wide, while the projected width of the pad has been reduced to only about 600 microns (both statements apply to the horizontal dimension only).
This figure shows some of the results obtained with the tracking provided by the telescope. In each case, the X and Y position of tracks at the pad detector is plotted. Panel (a) shows all tracks in which one of the pads had a signal above 100. The outline of the detector is clearly visible. The gaps inside the boundary result from inefficiencies in the telescope.
Panel (b) shows all tracks for which no pad had a signal above 100. Most of these tracks point outside the detector. A count of tracks inside the lines yields an 'inefficiency' of 356/6237, or 5.7% for the pad detector. Some of these tracks may be incorrect and some of the events may be pulse sharing where two pads are hit but neither is above the cutoff of 100 so this is an upper limit on the inefficiency.
Panel (c) shows events with whole detector noise, in which all pads in the detector move up by the same amount. These seem to be randomly distributed. Panel (d) shows events with column common noise, in which only the three pads in a column move together. The overwhelming majority of these events are located below the active area. In this set of runs, almost all of the groups with large column common noise were in the lower half of the active area. In the region at the bottom of panel (d), there were detector pads which were not connected to anything but which had traces from pads in the active area passing above them.
Note the gap between the bottom of the detector outline and the noise tracks. In this region, there was one row of pads which were not connected to the preamp but which were grounded. Although the mechanism is not obvious, the plot clearly demonstrates that most of the column common noise occurs when a track hits a disconnected pad which has traces from active pads passing over it, i.e. it is a form of cross-talk. Note that checks for similar cross-talk between pads which were connected to the preamp showed only weak evidence for a tiny effect. These plots also explain the observation that events with large column common noise almost never had a valid hit on a pad anywhere in the detector while events with whole detector noise had valid hits at the normal rate.