CALICE MAPS Meeting, RAL, 06/10/08 ================================== Present: Jamie Crooks, Paul Dauncey, Owen Miller, Marcel Stanitzki, Renato Turchetta, Mike Tyndel, Nigel Watson Minutes: Paul Minutes of last meeting: No comments. Sensor 1.1 status: Jamie showed some slides on the status of the V1.1 sensors; see usual web page. So far, four sensors, all from the same wafer, have been bonded of which three show power-ground shorts. The fourth seems to work so the other three are likely to be bonding, not sensor, problems. The bonder has a positioning resolution of around 5mu so there needs to be some tolerance of bond positions at this level. The nominal positions will be adjusted inwards by ~5mu to try to fix the errors. In total, we received: 60 sensors of 12mu epitaxial thickness, with deep p-well 60 sensors of 5mu epitaxial thickness, with deep p-well 60 sensors of 12mu epitaxial thickness, without deep p-well 40 sensors of 5mu epitaxial thickness, without deep p-well as well as a similar number of Python test chips. The SensorLoad program consistently reports errors which Jamie thinks is due to the faster LVDS converter components moving the clock timings and so this is something Matt will need to fix. It would be dangerous for Paul to take the only working sensor to Imperial so Jamie should try to send Matt enough information that he can tweak the timings in the firmware remotely. The change to the bias resistor network for I12IOUTBIAS means the DAC setting for the monostable lengths may need to be readjusted. The hi-res wafers were sent to the foundry last week. The 18mu ones had a problem with flatness but the foundry waived this for the shuttle run. The fabrication will not start until the end of Oct and so the 6-8 weeks needed takes us to close to the end of the year. This is then potentially very close to the time of the beam test and it is important to know if the hi-res have to be tested in beam. In principle the laser measurements would be sufficient. However the non-linear behaviour seen implies we don't really understand the response so comparing different epitaxial thicknesses would be difficult. (Jamie saw only a 40% increase in laser signal between 5mu and 12mu epitaxial layers previously but this was before the importance of focus was fully appreciated.) A beta source would be better as it genuinely deposits energy throughout the epitaxial layer but what information can be quantitatively obtained isn't clear. Sensor 1.0 tests: Jamie also showed some slides from the RAL students, Barnaby and Michael, who finished work last week. They have put much of their documentation on the SPIDER Wiki website which is not accessible from outside RAL. They attempted a 2x2mu2 laser scan but this was disrupted when the Windows laser control PC thought there was a network outage. Recovery brought the laser back with different settings. There seems to be some uncontrolled parameter(s) of the laser which are not set consistently. However, Giulio is not available to ask and the Labview code is too complex to rewrite directly in Linux. The scan takes around 3 days, as the scope takes ~1 min to average over 100 values of the analogue signal and there are around 4000 points. The idea is to compare the test pixels with and without deep p-well, so the only bonded non-deep p-well sensor (currently at Imperial) will need to be returned soon. Marcel showed some results on Fe55 calibrations; see slides on the usual web page. He sees a pedestal subtracted average of 119TU for quadrant 0 and 157TU for quadrant 1, with gain spreads of 10% and 20% respectively. The Fe55 peak width gives values very similar to the RMS noise values measured in pedestal runs and a correlation plot of the two values might be very revealing. For quadrant 1, with 157TU = 1610e-, then the calibration indicates 1TU = 10.3e- and hence the noise of 6TU is 62e-. This is close to three times the expected 23e- noise from the circuit simulation. In terms of gain, the threshold DAC is thought to give 0.4mV for each TU although Jamie will check this explicitly. This goes directly to the comparator input which implies the 10.3e- = 400uV and so the gain is 40uV/e-. This compares with the expected preshaper gain of 94uV/e- and so it is again off by more than a factor of two. The gain is mainly determined by the large 4Mohm resistor (which can be checked on the new sensor as it has been implemented as a test structure) and the ratios of capacitors, which should not be out by a factor of more than two when the spread is only 10-20%. Note, the expected noise in mV at the comparator input is 23e- x 94uV/e- = 2.2mV while the measured noise in mV is 6TU x 400uV/TU = 2.4mV. However, given that the gain does not agree, this can only be a coincidence. Paul reported that he had checked the noise on the non-deep p-well sensor and seen results compatible with the deep p-well sensors, so the noise did not appear to be influenced by the deep p-well. Nigel reported that he had now tried all the suggestions for the cosmics stack at Birmingham and still sees the random loss of a trigger signal after a few hours of running. He is currently running a system with one master USB_DAQ and one slave USB_DAQ and no PMT connections at all. Matt has been at CERN recently and so has been unavailable but he suggested the best approach would be for him to set up a parallel system at Imperial and see if it works there, before then trying the same at Birmingham. Paul will use then this system for beta source tests and it would be useful to borrow a small PMT paddle from Birmingham for this. By the time Matt has debugged a system at Imperial, it seems likely that the best approach would be for Birmingham to work directly with sensor 1.1. Conferences: Nigel will give a CALICE Si-W ECAL talk at the CLIC meeting at CERN, 14-17 Oct. The talk is probably on the 15th (although the schedule is not yet fixed) and MAPS will be a small part of this. Jamie will help Rebecca put together a poster for the IEEE meeting in Dresden, 19-25 Oct. There is an issue with copyright as we cannot reuse figures which appear in the IEEE proceedings. We should use older ones for this (and the poster) and keep the best for the paper later. This can include replotting the same data and remaking the circuit diagrams with four diodes and some colour. Marcel had given a talk at IPRD08 in Siena the previous week and has similar issues with copyright for the proceedings of this. He should coordinate with Jamie and Rebecca to be sure to use different sets of figures. Anne-Marie is down to give the LCWS08 MAPS talk in Chicago, 11-16 Nov, although the talk itself is not yet confirmed. We should not bother with ACFA as it is close to the beam test and there are more important things to do at the time. We have 40k total travel money for the whole of CALICE this year so people should charge travel to other budgets wherever possible. Marcel (and Jan) will charge the SiD meeting to Brian Foster's detector R&D fund, for example. Sensor 1.0 paper: There was a long discussion on the aims for the paper. It is unlikely to be the high priority once the new sensor is shown to work so it should be wrapped up within the next four weeks and submitted. In particular, unless it can be done quickly, it is not worth taking more laser scan data on the bulk pixels, but we should replot the existing data (as done by Jamie for his recent Naxos talk). The main feeling was that it should be split into two papers, with the motivation and physics performance going to NIM and the design and results to IEEE. [Note added after the meeting: the impact factors of these journals are NIM 1.114 and IEEE TNS 1.431. JINST has not been running for long enough to have a factor yet; however, the publication of all the LHC final technical reports will boost its impact.] The two difficult areas for the paper as it stands now are: - The justification for not discussing the pre-samplers further needs to be made more concrete. Some further results from these pixels should be shown, giving more reasons for them not being usable, and ideally some reason for their behaviour should be put forward. - The difference in noise and gain for the pre-shaper (discussed above) needs to be handled carefully. It might be that the resistor value will explain the gain difference, which would then be a definite explanation. The noise discrepancy might indicate the simulation noise models are wrong but this would be difficult to prove given the data available. Clearly, careful wording here wil be needed and this must be done when all results are known so as to be sure of what is known and what is guesswork. [Note added after the meeting: Following a later discussion, Barnaby and Michael should be included in the acknowledgements, rather than as authors, of the IEEE paper.] Next meeting: This will be on Mon 10 Nov, again at the earlier time of 10am.