CALICE MAPS Meeting, RAL, 06/09/06 ================================== Present: Jamie Crooks, Paul Dauncey, Bradley Hopkinson, Anne-Marie Magnan, Konstantin Stefanov, Renato Turchetta, Mike Tyndel, Nigel Watson Minutes: Paul Minutes of last meeting: No comments. Interim Design Review: This will be on Wed 4 Oct. We need to get at least two, preferably three, external reviewers. As this is a technical review of the schematics, then the reviewers need to be engineers rather than physicists, so Tim Jones (who was a reviewer for the PDR) would not be so appropriate for this review. However, Mark Prydderch, the other PDR reviewer, would be good for the IDR. Konstantin then volunteered as a second. Renato will contact Mark as well as another engineer from their group, possibly Steve Thomas. The review only needs half a day so we will start at 1pm and plan to finish by 5pm at the latest. We will not need the same level of formal documentation as for the PDR. The main documents which need to be produced are the schematics and the Spice-level simulation results. Jamie will produce these and will aim to have at least the schematics available by Fri 29 Sep so they can be put on the web before the weekend. Conferences: Giulio is presenting material on the CALICE MAPS work at the Siena Innovative Particle and Radiation Detectors conference; see the link under the 28/02/06 meeting. He is currently away and expected to return in mid-Sep. There will not be another meeting before the conference so he will need to circulate his talk to the rest of the group some days before his presentation but Mike thought he would not necessarily need to do a practice run-through beforehand. In general, anyone giving more than a trivial mention of our work outside the group should distribute their slides some days beforehand for comments. Renato will be giving general overviews on MAPS work at two conferences in the near future, Vertex06 in Perugia and Radiation Effects in Florence (see links on web page), but these will only have a very small amount of material on CALICE. Sensor design: The tender process had to be extended by two weeks over the summer due to holidays and so the deadline was 11am this morning. All three manufacturers (Foundries A, B and D) who were approached had submitted tenders but Renato had not yet looked at them to see if they were reasonable. Renato and Jamie met with Foundry B again and had a phone meeting with Foundry D. They raised the issue of deep-p implants under the n-well circuitry (to prevent signal charge collection). This should be technically possible but the manufacturers would need some development to validate this as part of the process. This would be unlikely to be done on the timescale needed by us. Konstantin stated that it should be possible for us to calculate the process parameters needed for the one extra step needed. If the manufacturers would then be prepared to set up the process as we specify, then we could produce wafers with this feature directly. This would clearly have some extra risk for the functioning of the circuitry above the p implant area. The date of the Foundry B shuttle run is now known to be 22 Jan 2007. Unlike last year, there is no second round in Feb so this means it is essential to make this date to remain on schedule if we go with them. Mike raised the issue that there is a CERN framework agreement with Foundry D which gives a fixed cost for wafer production when made to a specific process. CERN effectively puts together its own shuttle runs and divides the cost between the users, which means the actual price paid depends on the number of people involved. This would be either a 0.25 or 0.13mu process, not 0.18mu. We need to know the design process we will use by the time of the IDR which is only four weeks away so it is very late to consider this. Hence, Renato will contact the CERN leader (Alessandro Marchioro, Alessandro.Marchioro@cern.ch) asap to find out further details. Jamie showed some slides on his latest ideas for the shift register and memory logic. He thinks reasonable values for the number of pixels handled together would be between 36 and 60, given the physical constraints of bussing the signals. A value of 36 would give a 10% inefficiency and higher values would clearly improve this. A value of 56 might be convenient for several reasons and Jamie will see if this is feasible. This value needs to be divided into groups where the bits from the whole group are saved to memory if any are hit; the optimal division is not clear but Jamie should pick reasonable values which are convenient, e.g. 8x7 for 56 pixels total. Renato went though some ideas for reducing the reset noise through effective correlated double sampling (CDS) schemes. He thinks noise values of 20e ENC might be possible. However, these schemes would need a reset after each time there is a hit within the group. This would introduce deadtime for at least one beam crossing following the hit. This would need some calculation to see what the effect of this would be in terms of inefficiency. Sensor simulation: Giulio had send some slides, which Konstantin went through (see usual web page). This developed into a long discussion on the required results from the sensor simulation. At the previous meeting two months before, it had been clear one of the critical things to determine was whether the S/N increased or decreased when increasing the collection diode size; more charge may be collected proportional to area and the capacitance (and hence noise) may only increase as the square-root of the area, so an improvement may be possible. Giulio is now running a simulation with doubled diode size but this has not yet finished so no results are yet available. From the discussion, the following list of studies was thought to be needed: 1 Biasing the collection diodes at 2.5V rather than 1.5V 2 Charge collection vs. diode area 3 Charge collection vs. diode distance from the central n-well 4 Charge collection for a 3x3 collection diode array (minus the central diode) 5 Moving the n-well to the centre of one of the pixel side edges 6 Moving the n-well to one of the corners of the pixel 7 Moving two n-wells for neighbouring pixels together at the centre of the pixels' shared side 8 Moving four n-wells from neighbouring pixels together at the pixels' shared corner Some of these are illustrated in two files on the usual web page; the numbers in the diagrams correspond to the above list, with 0 being the standard layout. Konstantin will try to do a preliminary study of some of these with the 2D simulation in order to reduce the number of studies for which Giulio's full 3D simulation is needed. Also, some of the above are not as symmetric as the standard layout and so will need a longer simulation run than usual. Note, items 5 and 6 require no extra simulation but only a different summation pattern for the individual diodes and so can be done from existing data. Any layout with the n-well not centred will incur a longer signal path and hence capacitance, which would add a non-trivial amount of noise. However, this is not the dominant noise source and so such geometries should not be ruled out. Some other options were raised, such as spreading the n-well out into a thin line to act as a pixel border. However, the direct influence of the n-well or collection diodes due to their voltage is only ~2mu into the epitaxial layer. This leaves the other 80% of the layer free for diffusion so such a border does not really act as much of a barrier to crosstalk. Physics simulation: Yoshi had sent some slides, which Nigel went through (see web page). These were a followup of items raised at the last meeting. He had extended the pixel size range and it appears that there is a significant effect from particles crossing the pixel boundries even for 50x50mu2 pixels. He sees a 3% systematic reduction in the total energy deposited for very small pixels but otherwise the energy is independent of the pixel size, as required physically. Bradley, a summer student working at Birmingham, showed slides of his work on clustering with MAPS (see web page). He compares the standard ECAL 1x1cm2 diode pad clustering with MAPS 50x50mu2 clustering and sees very similar results. The algorithm used is one developed for the standard ECAL and so could probably be tuned further for the MAPS case. Unfortunately, Bradley has now finished his time with the group and so this work would need to be picked up by someone else to take further. Anne-Marie showes somes slides on the status of the post-GEANT4 digitisation software (see web page). There are many oddities, including the energy scale of the MIP peak and multiple particles in the events in addition to the one requested, which need to be understood. The addition of noise, threshold and output format in LCIO are still to be implemented. The oddities would be best tackled with Yoshi and Nigel so Anne-Marie will meet with them to sort out these issues. Anne-Marie also pointed out that Giulio's simulation gives results for charge diffusion based on charge deposit positions at the corners of the 5x5mu2 subpixels, whereas the GEANT4 hits integrate the energy deposits over the whole subpixel. These should be made consistent, i.e. the values should be changed so the points used by Giulio are centred on the subpixels used in GEANT4. Some rough numbers are than the 200GeV electron events Anne-Marie generates have ~60k 5x5mu2 subpixel hits and take around 1min/event for the GEANT4 step and 5sec/event for the current digitisation step. CASE studentship: The Birmingham/RAL CASE PhD studentship has now been filled. Owen Miller will start working at Birmingham in October. The exact details of how he will share his time between there and RAL are still to be worked out. Next meeting: The IDR will be on Wed 4 Oct, in R76 starting at 1pm. The next regular meeting will be the following day, Thu 5 Oct, also at 1pm in R76, as usual.