Calibration for the KEK Beamtest in February 95

Author: Riko Wichmann
This page was created July 13, 1995, last modified January 31, 1996

Introduction

The calibration was performed by fitting s-curves (integral of an error function) onto the efficiency as a function of threshold extracting their meadian value and the sigma. These fits were done for all channels and all chips for different calibration voltages which are related to a certain charge injected to the channels. The median value of these s-curves plotted versus the injected charge then gives us the response function of each channel which we fitted using a 3 parameter function.
These parameters m1, m2 and m3 give us the calibration of the detector system, meaning a relationship between the threshold voltage of the amplifier chip and the charge on the amplifier input. From the so obtained fitting parameters we can already calculate the small signal gain and and the noise of the detectors.

Characterization of the Runs

The calibration was performed for the three groups of threshold scans using the whole range of injected charges (0.75 fC - 5 fC). Here the runs will be labeled by the run number of the smallest charge:

run708: contains results of run number 708 to 716
57 ns delay optimized for n-side timing
detector board NOT yet in beam position
bias voltage + 50 V and - 50 V
run723: contains results of run number 717 - 725
same timing as in run708 (n-side timing)
detector boards moved to beam position
bias voltage + 50 V and - 50 V
run726: contains results of run number 726 - 734
35 ns delay optimized for p-side timing
same bias voltage as before

The hits per event were determined by oring two timeslices.
Here we present some results for board 0.

General Results

Threshold curves of channels which could not be fitted by a s-curve, show -1 for the response curve parameters m_i and for the noise. This means normally, that the channel was always on over the scanned threshold range.
Bonded channels can be separated from non-bonded by
1. larger value of m_3 for bonded channels
2. larger noise for bonded channels
For chip 0, strips were "ganged" in different ways. Strips ganged in the same way were separated by a cut on the channel number from other gang patterns and by a cut on the noise from non-bonded LBIC channels.
We obtained essentially the same results for run708 (not in beam position) and run723 ( boards in beam position).
There is no real difference in the fit parameters m_i and the noise using n-timing ( run723 ) and p-timing (run726).
For the p-side noise, we extracted the following values:
1. 2 strips ganged (chip 2) : 0.23 fC
2. all channels bonded (chip 3): 0.2 fC
3. non bonded channels (chip 2): 0.11 fC
For the n-side noise, we extracted the following values:
1. 4 or 3 strips ganged (chip 0): 0.24 fC
2. 2 strips ganged (chip 0) : 0.22 fC
3. no ganged strips, floating neighbors (chip 0) : 0.16 fC
4. all channels bonded (chip 1): 0.18 fC
5. non bonded channels : 0.11 fC
The noise of the n-side is less than the noise for the p-side. This is understandable in terms of type inversion of the detector bulk after irradiation.
We found a uniform value for the 3 fit parameters m_i for each chip.

Average Calibration Parameters

The following table gives the average of the response curve parameters over channels 10 to 54 (to exclude effects of the side) which were bonded to the chip. The separation of bonded and non-bonded channels was performed by a cut on the noise (bonding map).

run   chip#  |    m_1  |    m_2  |   m_3   | Remark
-----------------------------------------------------------------------------------------
723:   0     |  32.49  |  159.2  |  563.7  | averaged over channel 10 - 54
             |  40.55  |  157.4  |  548.6  | 4 strips ganged
             |  37.22  |  158.2  |  570.3  | 3 strips ganged
             |  37.15  |  160.9  |  559.4  | 2 strips ganged
             |  38.91  |  156.7  |  567.6  | alternating bonded/non-bonded LBIC channels
       1     |  47.22  |  161.0  |  560.3  | all channels bonded
       2     |  51.15  |  149.5  |  719.1  | 2 strips ganged
       3     |  44.68  |  149.2  |  697.9  | all channels bonded
-------------------------------------------------------------------------------------------
726:   0     |  37.38  |  160.4  |  524.7  | averaged over channel 10 - 54
             |  40.70  |  158.0  |  503.9  | 4 strips ganged
             |  36.38  |  161.9  |  538.4  | 3 strips ganged
             |  37.65  |  159.4  |  516.0  | 2 strips ganged
             |  37.12  |  162.5  |  535.4  | alternating bonded/non-bonded LBIC channels
       1     |  44.67  |  167.8  |  524.0  | all channels bonded
       2     |  50.01  |  152.0  |  718.5  | 2 strips ganged
       3     |  42.87  |  152.4  |  679.5  | all channels bonded

Some Plots displaying the Results

Here are some postscript plots illustrating the above stated results:

the fit parameter m_i as a function of channel for
run723: chip 0, chip 1, chip 2, chip 3
run726: chip 0, chip 1, chip 2, chip 3.
the noise for one channel as a function of injected charge and the noise at appr. 1 fC injected charge as a function of channel for
run723: chip 0, chip 1, chip 2, chip 3
run726: chip 0, chip 1, chip 2, chip 3.

Conclusion

Since the results for n-timing and p-timing are the same for the n-side and the p-side using ored time slices, calibration runs should be performed using only one timing. We suggest using the n-timing, since the p signal is wider.
To make sure, that the noise (especially on the n-side) is not influenced by bad statistics, calibration runs should be done with enough events. We suggest 500 events per threshold. A systematic investigation of the dependence of the n-side noise on the number of events is on it's way.
It looks like one can use a uniform calibration of each chip by averaging the parameters m_i over channels. To be sure about the accuracy of the uniform response function, one should compare this numerical response function with the one given by the data points.

Complete Calibration Data

The complete calibration data for all 15 chips can be downloaded as gzip'ed tar-files from my directory /pub/wichmann/kek_feb95 on the scipp ftp server (use anonymous ftp if not using the above link).

Break out of frames to bookmark page thereafter.