CALICE MAPS Meeting, RAL, 18/01/08 ================================== Present: Jamie Ballin, Jamie Crooks, Paul Dauncey, Anne-Marie Magnan, Yoshi Mikami, Matt Noy, Marcel Stanitzki, Renato Turchetta, Mike Tyndel, Giulio Villani, Nigel Watson Minutes: Paul Minutes of previous meeting: Paul raised the issue of the level of agreement between the Fe55 measurements and the response curve shown in an xls file by JamieC at the last meeting (see usual web page). The basic numbers are that the Fe55 X-ray gives 1600e- in silicon and the response seen on the analogue sampler test pixel is 207mV. The noise for zero signal is measured at 3.5mV, as shown by Marcel at the last meeting (again, see the usual web page). JamieC's response curve (brown triangles) shows the response at 1600e- should be 283mV, at which point it is almost saturated and so should not be very sensitive to the input energy and hence would be a bad point at which to attempt a response calibration. The predicted input signal which gives 207mV is around 700e-. In addition, assuming a linear response below 400e-, the plot indicates the noise measurement of 3.5mV would correspond to 10e-, while the predicted noise is around 25-30e-. It is clear these measurements do not agree with the predicted curve. It is therefore critical to map out the shape of the response curve asap. The laser would seem to be the ideal tool for this, as the intensity is known to be accurately linear with the control setting and the Fe55 1600e- point at 207mV can be used to set the overall scale of the laser response. Paul also raised the issue of the expected S/N. We have been showing plots for some time, made by Giulio, demonstrating a corner S/N of 15 for 1.8mu diode sizes. This includes in the OsC document submitted in July 2007. However, with a maximum MIP charge in the corner of ~260e- and a noise of 25e-, then this is a S/N of around ~10, not 15. Clearly, the plots used the wrong value for the expected noise and should not be shown in future. At the last meeting, Giulio had said the first sensor simulations for the test pixels (to compare with the laser measurements) would be ready by the end of the year. However, these were not run but will be needed for any publication so should be restarted soon. Note, the simulation will be needed both for deep p-well and non-deep p-well sensors. Project status: Paul showed some slides on the status of the funding and what we need to consider for a future programme; see the usual web page. The main issues are that we will be constrained although we should try to push for more funding if possible, and that we would have a much stronger case for future funding if we have publications from the current sensor. It was agreed that effort should be directed with a high priority towards papers from now on. Sensor design: JamieC showed some slides on remaining issues with the sensor and possible options for the next round; see usual web page. Jamie has considered whether a second test structure for the shapers could be added to the next design. This would be possible except the laser would not be able to be used; there is no access hole on the PCB and drilling one would not be possible without a risk of shorting the power and ground planes together. A single-seat submission for test structures would cost around 10k. Beam test: Nigel showed some plots from Owen from run 490061; see usual web page. This demonstrates he has the code now running at Birmingham. Marcel gave a detailed report on the beam test; see usual web page. One value not explicitly given there is the z spacing of the sensors, which was 1.5cm. The PMTs should have longer light guides in future as the PMT body is sensitive to MIPs and may have been in the beam halo; it is even possible that the stack position may have been adjusted incorrectly to centre the beam on the PMT bodies rather than the scintillators. Matt reported that the PMT readout works well at Imperial now with no changes from DESY, so the reason for no apparent correlation between the PMT and sensor data is not understood. If another round of PMTs is made, it would be useful if they could be made around 1cm longer at the far end from the cables, so as to enter the stack rails more securely. Whether this can be done by instructing the manufacturers to cut differently, or would require a change to the design file is not clear. Future possibilities for beam tests would be at CERN from around May onwards (where we would be limited to a few days), at DESY from Sep (where we would be limited only if there were other users) or at FNAL in May or Jul (where we would have to run in the periods the main CALICE beam test is operating and so would be parasitic, which would be hard). JamieB showed some slides on the efficiency for MIPs; see the usual web page. One issue discussed was whether this preliminary analysis using only timestamp correlations is valid; the geometry is not yet available for true tracking. In principle, any track hits must have coincident timestamps so this method should pick up every hit which would be used for tracking in a complete analysis. Unless the noise rate with beam is different than without beam (where the efficiency is found to be very low) then the efficiency should only get lower if full geometry is used (unless there is a bug, of course, although there is no clear evidence for this at present). Jamie should check for sensible position and timestamp distributions for the hits. Marcel reported on results for showers in the tungsten; again see the usual web page. The number of hits is below the expected level in the showers, which may have the same source as JamieB's efficiencies but could be due to other effects, such as the beam being off-centre on the stack. Again, the position of the showers should be plotted so as to eliminate effects unrelated to the sensors. Future work: We are clearly very short on PCBs to mount sensors, due to the shorts in the latest batch made. JamieC reported the wirebond head has been enlarged between the first and second batch and it was thought the bigger head was now breaking through the seal ring around the edge of the sensor. The bond head has now been replaced with a smaller one and some test bonds look a lot better. One real board had been bonded today and this would be tested after the meeting. Jamie has also arranged for the wirebond positions on the sensor to be moved inwards by 5mu so as to have a bigger stay-clear from the seal ring. The first trial of this gave a lot of broken bonds but this was thought to be unrelated. Jamie will continue to check the new PCBs and hopefully this means we will have more within a short time. The laser system is essential for measurements where a reproducible signal is needed at high rate. The highest priority is to measure the response curve to compare with JamieC's simulation, as discussed above. This will allow us to understand the gain and whether the observations of JamieB and Marcel from the beam test data are due to low gain or not. The next measurement needed, as discussed in several meetings previously, is to check the operating point for the bulk pixels. This would require scanning each bias setting in turn to maximise the S/N. Following this, a rescan of the test structure is required. The original measurement shown at IEEE is not of publishable quality as it does not cover the whole test pixel area or the same area for deep p-well and non-deep p-well. This is a very time-consuming measurement and it was decided this should be (semi-)automated so it can be done in a reasonable time, even if this takes some effort to set up. We will need a sensor simulation to compare (see discussion above) so these should be coordinated to ensure they cover the same area of the test pixels. In addition, besides the deep p-well and non-deep p-well measurements, a 5mu thick epitaxial layer sensor should be measured to look for differences (and to check that we really have used 12mu thick epitaxial layer sensors up until now). However, the laser system is currently occupied (by non-CALICE work) until the end of January; given the urgency of the above measurements, it would be good if some workaround, to allow us access before then, could be done. At Imperial, the Tl (beta) source measurements (as shown in various meetings previously, and in the last OsC document) seem to show no definite structure as a function of threshold. This means they cannot be used to set a gain scale. Measurements will continue with Fe55 and possibly the X-ray generator. Analysis of the beam test will also continue. The Birmingham group will reassemble the cosmic ray stack which they were taking back after the meeting. They will also analyse the cosmic data taken over the Xmas break. Conferences: Marcel is attending the SiD meeting in SLAC next week and it was decided he should show qualitative results only, given the lack of quantitative understanding of the bulk pixel response so far. Next meeting: This will be at 13.00 on Fri 8 Feb.