More on TPAC1.2 trimming

1
More on TPAC1.2 trimming

• Paul Dauncey

2
Trimmed sensors 29, 45, 46, 48

• Same method as before
• Fix trims to some values
• Unmask 42 pixels per readout region
• 4 regions 168 pixels total in each run
• Do threshold scan to determine mean and width
• Do 168 runs to cover complete sensor
• Adjust trim to narrow mean distribution
• Picked target for pixel mean 100TU for all
  sensors
• Adjust trims using binary division of remaining
  range for each pixel
• Needs six set of runs to fix all six bits
• Following completed trim, run with all pixels
  unmasked
• Threshold scan similar to real running

3
Sensor 29 trim values
Trim
4
Sensor 29 shift relative to trim0
TU
External resistors adjusted compared to results
shown in last meeting
Trim
5
Similar for all four sensors
TU
29
TU
45
Trim
Trim
TU
TU
46
48
Trim
Trim
6
Sensor 29 effect of trimming on mean
After trimming
Trim0
TU
7
Sensor 29 mean (cont)
TU
8
Similar for all four sensors
29
45
One bad configuration column 168 pixels
TU
TU
46
48
TU
TU
9
Sensor 29 effect of trimming on RMS
Trim0
After trimming
TU
10
Similar for all four sensors
29
45
TU
TU
46
48
TU
TU
11
Sensor 29 dependence of RMS on trim
TU
Trim
12
Similar for all four sensors
TU
29
TU
45
Trim
Trim
TU
TU
46
48
Trim
Trim
13
Position dependence of RMS
29
Pixel y
Pixel y
45
Pixel x
Pixel x
46
Pixel y
Pixel y
48
Pixel x
Pixel x
14
Sum of all region 0 hits from trim runs
Rate
Normalised to hit rate/BX
TU
Paul Dauncey
14
15
Rate from runs with all pixels unmasked
Rate
19BX/BT
100BX/BT
1000BX/BT
8000BX/BT
TU
Paul Dauncey
15
16
Sum of all hits again
Rate
TU
Paul Dauncey
16
17
Exclude pixels with RMS gt 15TU
Rate
2 of pixels removed
TU
Paul Dauncey
17
18
Unmasked runs for 8000BX/BT mean
Paul Dauncey
18
19
Unmasked runs for 8000BX/BT scatter
Paul Dauncey
19
20
Four regions not exactly the same
Paul Dauncey
20
21
Not exactly reproducible
13/6/09
25/6/09
Paul Dauncey
21
22
The four sensors vary significantly
29
45
Run stopped artifically
46
48
Lost trim settings?
Paul Dauncey
22
23
Conclusions

• For trimming, the sensors look very uniform
• Same mean, noise distributions
• Same response to trim
• Very few dead pixels
• Will hit memory limit if running for 8000BX/BT
• Only for a few of the most noisy pixels
• Can reduce rate significantly by masking most
  noisy 2 of pixels
• When running with all pixels unmasked, see
  instabilities
• Not very reproducible for a given sensor
• Different sensor to sensor