This page contains some of the results of our analysis of the Sep. 99 testbeam data. Further information can be found on:
(Tranparencies presented at the 4th international symposium on development and application of semiconductor tracking detectors, Hiroshima, March 2000)
(Tranparencies of the Jan. 2000 SCT week, in PostScript)
(ATLAS ID internal note ATL-INDET-2000-004summarizing the test beam results)
(First, the basic ideas of the analysis are explained.
The table shows a summary of the most important numbers obtained. More detail can be found in the plots, available for download in PostScript format.

      Modules

Three modules were present in the testbeam:

    ABC1: single sided ABC+CAFE/P module, entirely bonded to 12 cm

    CGNT1: single sided ABCD-no trim module, entirely bonded to 12 cm

    CGT1: single sided ABCD-trimDAC module, entirely bonded to 12 cm

      Analysis

Four scans were taken: EDGE circuitry ON and OFF for 80 and 240 V detector bias.

In order to be able to study FE electronics properties, we wanted to avoid any interference from bad bonding or detector related problems. This means we had to discard any channels showing deviating efficiency or noise (any channel having a factor 4 more or less hits in the hitmap was rejected. ~80 for ABC1, 13 for CGNT1 and 0 for CGT1 !!) Also, runs showing oscillation or common mode were not considered for that particular module.

All data were taken in ANYHIT compression mode. For the efficiency we simulate LEVEL mode by selecting one timebin (depending on TDC value). The TDC range taken into account is a sliding window, maximising efficiency for every threshold. The noise occupancy is calculated by adding the hits in three timebins for the noise events taken in the EOB.

The calibration used is the one done by Gareth Moorhead just after the testbeam, further information on http://www.ph.unimelb.edu.au/epp/tbSep99. There are various indications that the calibrations of the CERN/Geneva modules might be inaccurate up to about 0.2 fC.

These figures show the response curves of the three modules (PostScript ABC1, CGNT1, CGT1 )as obtained in H8 and by the home institutes

The noise figure is determined by a Gaussian fit of the occupancy versus threshold (obtained from events taken between spills, with absolutely no beam). As the center of the Gaussian is left free, the results should not be affected by a possible error in the offset, but are of course very sensitive to a badly determined gain. The numbers quoted below are the average of the 4 scans taken. As we average three timebins, the noise number is not very sensitive to the edge circuitry (in fact no systematic differences are observed). 

The spatial resolution quoted is the sigma of a Gaussian fit of the residuals. Fitting with a trapezoidal function and calculating its RMS does not significantly change these results.
Numbers are evaluated at a threshold of 1 fC as given by the calibration.

The timing study is based on the combined efficiency versus TDC plot. As data was taken in ANYHIT mode, hit information is available from the triggered clock cycle and the previous and next time bins. We can use this property of both FE chips to extend the time axis to get efficiency as a function of the charge deposition moment with respect to the sampling moment for a full 75 ns range.
The plot obtained can be used to verify the correct functioning of the edge sensing circuit (one expects the module to be efficient in a 25 ns time interval, as the tail of the shaper pulse is effectively cut after one time bin). The different efficiency versus TDC curves for the two bias voltages reinforce the attribution of the effiiciency loss to ballistic deficit.
Finally, one can plot the rising and falling 50% TDC values and the threshold in mV in order to reconstruct the median pulse produced by the shaper in both voltage cases.

    Monte Carlo

The results obtained in the test beam can be described fairly well by a Monte Carlo.
 

     Results

The directories below contain the most important results.

Efficiency and noise plots
 
Spatial resolution and average cluster width:

Timing results:efficiency versus TDC plots, pulse shape reconstructions:
 
 

     Laboratory measurements

Laboratory measurements could provide a much simpler and more controlled way to determine the exact shape of the pulse that enters the discriminator (in the laboratory the exact charge and its deposition moment can be controlled).
As a check of consistency I added two plots of efficiency versus strobe delay for a fixed threshold of 1 fC and various input charges.

      Summary of results
 
These numbers are an indication and should not be considered final. Errors are still being estimated.