ABCD2T module with irradiated chips

Here you can find the results from a test on ABCD2T irradiated chips connected to 12 cm unirradiated silicon detector (ABCD2NT chips are not considered in this analysis)

Module description:

Hybrid: Oslo barrel type with alumina substrate
Chips: 3 ABCD2T chips and 3 ABCD2NT chips
Detectors: 2 CSEM barrel detectors with a total leakage current of 50 uA at 80 V
Comment: the hybrid has been irradiated with 24 GeV protons from PS at Cern with a total fluece of 3x10**14 p/cm**2. The analysis of the 3 ABCD2T irradiated chips can be found in the ATL-INDET-99-019.

Set up:

PC connected to a VME-MXI-2 vme module
SEQSI
DRAFT
Pcb from Jan Kaplon's design
Software from Carlos Lacasta

Results:

The module is stable over a wide range of preamplifier biases. For preamplifier biases > 230 uA the chip's performances start to deteriorate: the module is still stable but the gain and offset distributions get wider and do not really fit to a guassian curve. The detailed analysis is reported for the preamp bias=202.4 uA and shaper bias = 31.2 uA .The gain and the offset are calculated fitting 6 points from 1.5 fC to 3.5 fC with linear curve. The 3 chips have been trimmed at 2 fC. The default run mode is edge sensing circuitry ON and 01X data compression.

Distributions as function of the preamplifier bias for chip0 (the other two chips do not change within errors): Gain for chip0 , Offset for chip0 , ENC for chip0
There is no sign of instability. Here you can find the plot of the sum of the s_curves of the 3 chips [ps] . The s_curves do not reach 100% effinciency (they do for charges > 2 fC). This is due to a combination of the noise effect and the edge sensing mode effect. In fact things get better if we run in level sensing mode [ps] .
The spread of the 50% distributions is about 3 mV for 1.5 fC after trimming: 50% versus channel number , spread of the 50% values
The gain is around 45 mV/fC with a sigma of 1 mV: Gain versus channel number , spread of the gains
The spread of the offset is around 4 mV Offset versus channel number , spread of the offsets
The ENC calculated at 1.5 fC is around 1630 ENC. Few channels on chip2 have a higher noise because are close to the unbonded strips (the second half of the module corresponding to the ABCD2NT chips hasn't been bonded) ENC at 1.5 fC versus channel number , ENC at 1.5 fC

The offset values obtained with the previous method can be compared with the ones obtained from the following scans:

Threshold scan with 0fC injected charge: chip0 , chip1 , chip2 .
Noise scan: chip0 , chip1 , chip2 .
Comment: the tails of the distributions are due to hot and untrimmable channels. In fact if , for example, we take away from the distribution of chip0 for 0fC injected charge one hot channel and 2 untrimmable ones, we obtain the following distribution chip0 . There is no difference in the offset values obtained from two noise distributions. This would lead to the conclusion that the DAC doesn't inject a small amount of charge (indipendently of the loaded charge) that would affect the offset values obtained from the linear fit. However the difference of the offset values obtained from the noise/0fC distributions and the linear fit is of about 20 mV. If we trust the offset value obtained with the noise/0fC scan, we get at 1 fC (about 100 mV) a noise occupancy < 10**-4.

Strobe delay scans have been made for charges between 1 fC and 5 fC (the strobe delay value doesn't change for charges > 5 fC). The threshold value is 80 mV, that corrisponds to 1 fC as caluculated from the linear fit:

Plots of the strobe delay in ns as a function of the injected charge: chip0 , chip1 , chip2
Distribution of the time walk defined as the difference between the value at 1.25 fC and the value at 5 fC: chip0 , chip1 , chip2
Warning: in order to calculate the time walk as from specification we had to go down to 1.25 fC for the injected charge value. The curves at these low values of the injected charge do not follow the expeted shape for the reasons already mentioned (see s_curves at 1.5 fC). Here it is the result of the time scan at 1.25 fC summed over the 384 channels ps . The shape of the sum rappresents quite well the shape of the single channels. We plan to estimate the time walk running in level sense mode.

Authors: Brendan Dick and Daniela Macina