CALICE MAPS Meeting, Coseners House, 06/02/09
=============================================
Present: Rebecca Coath, Jamie Crooks, Chris Damerell, Paul Dauncey,
 Marcel Stanitzki, Jan Strube, Renato Turchetta, Mike Tyndel, Nigel
 Watson, John Wilson, Gary Zhang

Minutes: Paul

Minutes of previous meeting: No comments.


Sensor tests: Jamie showed some slides on sensor measurements; see usual
 web page. The full analogue scan is automated but takes 80 hours to
 run. All his plots of mV vs TU have the mV value corrected for the
 0.9*0.9 gain reduction of the source followers and so represent the
 actual value at the comparator input. These measurements were done on
 sensor V1.1; in principle sensor V1.0 could be different as the number
 of trim bits was changed. Jamie will check the V1.0 comparator, both
 on the sampler test pixel of sensor V1.0 (as the comparator is
 independent of shaper or sampler) and on the old design test pixel on
 sensor V1.1.

 Marcel showed some slides on test and bulk pixel Fe55 measurements;
 see slides on usual web page. For the reset measurements, the 7.89mV
 value is the equivalent of the noise, not the error on the mean. This
 represents an upper limit on the pixel noise, given that the reset
 pulse itself may not be constant and so may give an extra contribution.
 In slide 8, the measurements are for sensor V1.1, for slide 9, the
 green histogram is V1.1 and the red is V1.0 and for slide 14, red is
 non-deep p-well while green is deep p-well. Finally, slide 15 shows
 sensor V1.0 data from Quad1, i.e. the design used for sensor V1.1.

 Paul showed a spreadsheet of the various TPAC calibration numbers. With
 the assumption that the lower Fe55 peak is the one found in the bulk
 pixel measurements, then the various calibrations are consistent to
 around 30%, while the noise values vary between 18e- and 36e-, where
 the expectation is 25e- to 30e-. This is a significant improvement on
 the previous discrepancy of a factor of 2-3 between different methods.
 Mike has assembled a similar table and he and Paul will crosscheck the
 values.


Future tests: It would be useful to get a relative rate comparison
 between the low energy peak and high energy peak events. Given the
 fits, then the integral of the fitted Gaussian should allow the two
 peaks to be relatively normalised. This would help identify the source
 of the low energy peak; the diodes occupy 4*1.8*1.8/50*50 ~ 0.005 of
 the total pixel surface area so the relative rates would give an
 equivalent surface area for the low energy peak, assuming a similar
 depth profile. Marcel will try to extract these numbers from his fits.
 Suggestions for the origin of the low energy peak are from the
 relatively uniform region in the centre of the pixel (so as to get a
 peak), from corners (as the peak position is around 1/4 of the total
 Fe55 signal), from MIPS (as it corresponds to the average MIP energy,
 allowing for charge diffusion) and from interactions in the n-wells
 (which may depend on the presence of deep p-well although this was not
 clear).

 Another approach to interpreting the peaks would be to see if hits can 
 be found in the neighbouring pixels for Fe55 events in either of the
 peaks. If the high energy peak is from the photon being absorbed in the
 diode, there should be no neighbour hits. If the low energy peak is due
 the photon interacting in the epitaxial layer around the centre of the
 pixel, and hence giving standard charge diffusion, it might be possible
 to see the neighbour hits. However, neighbour hits are likely to only
 be visible within a very narrow range of thresholds; the expected
 fraction of charge in the neighbours can be up to 20%, but is usually
 less than 10%. For Fe55, 10% would mean 160e- and using the latest
 calibration numbers, this would require a threshold around 40TU, which
 might mean there is a large number of noise hits. This measurement
 would be best done with sensor V1.0 to avoid the comparator feedback at
 such low thresholds. Birmingham will try to look at this.

 To confirm the interpretation of the Fe55 peaks, then it would be very
 good if the preamp gain could be reduced by a factor of around half.
 This should then put the high energy peak within the range of many of
 the comparators before they saturate; it is thought it would currently
 be around 500TU. Jamie will see if this can be done.

 It would also be good to find an alternative gamma source to Fe55 with
 a lower energy photon. Ideally a photon with an energy of around 2keV
 would be used; this would match well to the comparator range and would
 be close to the expected signal from a MIP passing through the centre
 of the pixel.


Sensor V1.2 fabrication: Mike has raised the 35k needed to make a V1.2
 of the sensor. This will have the M2 and CS mask changes to fix the
 comparator feedback and the row addresses, respectively. A fix to cure
 the configuration load errors (and hence remove the need for the 
 external bias circuits) is not trivial and will not be done. Jamie will
 aim to have the design complete by the end of Feb and submit it at
 that time. It will be treated as an engineering run and so does not
 have to be scheduled for a shuttle run. The sensors should be returned
 around the end of Apr. We need to make a minimum of 12 wafers, so there
 will be four splits:
   3 wafers 12mu epitaxial layer, deep p-well
   3 wafers 12mu epitaxial layer, non-deep p-well
   3 wafers 12mu epitaxial layer, high res, deep p-well
   3 wafers 18mu epitaxial layer, high res, deep p-well
 The design changes needed are relatively minimal and Jamie has
 implemented most of them for the bulk pixels already. There should be
 an FDR, but this will be part of the next meeting so it can be short if
 that is all that is required.


Next meeting: This will be on Fri 27 Feb starting at 10.00. The meeting
 will include the sensor V1.2 FDR.