CALICE MAPS Meeting, RAL, 17/09/07 ================================== Morning session =============== Present: Jamie Ballin, Paul Dauncey, Anne-Marie Magnan, Yoshi Mikami, Marcel Stanitzki, Nigel Watson Minutes: Paul Organisation: As this was the first time the MAPS meeting had been split into two parts, then there was a brief discussion on the appropriate place for some items. Beam test software should be discussed in the afternoon meeting, although the simulation for this (GEANT4/Fluka) should be discussed in the morning. Conferences will be discussed in the afternoon. Overview: Paul showed some slides on the general scope of the simulation work and what we should be aiming to do; see slides on usual web page. Concerning the number of MIPS per pixel, Yoshi reported that he thinks he understands the apparent bug which gives a peak at four truth particle contributions to SimCalorimeterHits. The default step size is such that a single MIP will take four steps to cross the epitaxial layer. Each is recorded as a separate contribution, even though they all come from the same physical particle. This would mean that there is no sensible way to determine the number of different particles crossing each pixel from these data. It may be possible to use the SiD simulation for such a study as the implementation is quite different. However, Marcel reported there are technical problems with the installation at RAL which prevent him from doing large scale studies there. Marcel and Anne-Marie had a productive discussion at Imperial following the last meeting and came up with a conceptual redesign for handling the SimCalorimeterHits (and subsequent digitisation). This will allow PandoraPFA to compare reconstructed and truth energies at a detailed level. This structure may allow us to better determine the number of particles passing through each pixel, on the assumption that there is never more than one in the 5x5mu2 sub-pixels. The software is still being implemented so this cannot be studied yet. A factor not mentioned in Paul's list for the hexagonal vs square optimisation would be the noise since a hexagonal structure would almost certainly require six diodes rather than four, with a corresponding increase in trace length and hence increase in input capacitance. The noise should be independent of epitaxial thickness and so is not an issue for the optimisation of the latter. It was not clear if there was a major technical issue with implementing the saving of the bulk silicon energy deposits as SimCalorimeterHits although it is clear it would not be easy to do. Yoshi and Anne-Marie will try to work through this in the next few weeks as long as it does not get in the way of producing other results in time for IEEE. If bulk hits are accessible, we would need to implement our own (simple) version of the ECAL digitisation but this would then open up a large number of useful comparisons which could be done on the same events. In particular, this would allow studies where only particular layers of the ECAL, such as the first few, are replaced by MAPS. Simulation sensitivity: Anne-Marie reshowed some of the slides from the previous meeting concering the sensitivity to the GEANT4 low-energy models, see the usual web page. Nigel had given a CALICE talk at the recent GEANT4 workshop in Hebden Bridge and had got some feedback from the experts on some of the issue of low energy cut-offs. There are no data below ~250eV so none of the models can be extrapolated to this region. PFA: Marcel showed some slides on recent work; see usual web page. The news on GEAR files being produced is good; although not complete, at least the ECAL is now included. The PFA resolution seems to be sensitive to charged particle loopers, as shown by the B field dependence, although this seems to be rather flat generally. Marcel says the different resolutions for different B fields are significant statistically, as the errors are around 0.2%, although they are probably not significantly different in terms of physics performance. We need to decide on what detector we will use for the comparison with MAPS. It was thought the LDC01 detector with the ECAL cells reduced to 5x5mm2 would be the most appropriate. This needs to be reached in several steps from Mark Thomson's benchmark model, LDC00, so we can understand which steps introduce any differences seen. Afternoon session ================= Present: Jamie Ballin, Jamie Crooks, Paul Dauncey, Anne-Marie Magnan, Yoshi Mikami, Matt Noy, Marcel Stanitzki, Konstantin Stefanov, Renato Turchetta, Nigel Watson Minutes: Paul Minutes and matters arising: Marcel needs to be sent the conference contributions from Giulio for his talk at TWEPP, JamieB for his poster at the RAL summer school and Anne-Marie for her LCWS proceedings. Nigel gave the MAPS talk at EPS and will be writing proceedings for this. He should send these to Marcel when they are ready. Sensor design: JamieC had sent a report to the email distribution list recently so gave an update since then. He has seen valid time codes, bank addresses and hit codes from two of the four columns (the pre-shapers) on the sensor he is testing. The others (the pre-samplers) give hits but the values were corrupted. It is thought to be unrelated to the pixel type but to be due to there being insufficient voltage to drive the write transistor which sets the memory. This in turn is due to having to operate this part of the circuit at 2.5V rather than 3.3V, which is because of an error in the design. This hypothesis still needs to be verified. One test which was done was to load all memory values with 1's rather than 0's as it was thought it would be more reliable to change a 1 to a 0 rather than vice versa. However, as this failed to cure the problem, it may be the error is due to something else. It appears to be related to a droop on the power supply as the corruption starts in the centre of a column if the voltage is set too low. The power supply fix to the sensor PCB previously suggested by Matt seems to cure the monostable problem although this still needs to be implemented on the other two existing PCBs. Matt also needed to change the bias resistors for one of the DACs. The new round of PCBs will be assembled with the new resistor values and the extra "fix" resistor will be added by hand later. A noise of ~10mV peak-to-peak was measured from the test structure although the pixel was being held in reset so it is not clear if this is a good measure or not. It would also be good to measure the noise rate from the bulk pixels. This should be possible for all pixels; the memory write pointer is not affected by the corruption error and so the number of hits should be valid, even if the information within each hit is not. The configuration data load and unload on the sensors has been tested and verified. Sensor testing: No signal has yet been seen from the laser as the pixels were inadvertently held in reset continuously during the last attempt. This will be tried again soon. Matt will order more sensor PCBs as soon as Exception reply to his request for a quote. He will fabricate five PCBs with a 5-day turnaround and twelve PCBs with a 10-day turnaround. The first five will be populated with a 5-day turnaround and seven more will be done with a 10-day turnaround. The remaining five PCBs will be left unassembled for now as parts are not yet in hand. This means the first five boards should be ready by early Oct. Matt should then test them before sensors are mounted so they should be ready for bonding by the week of Oct 8. Aluminium baseplates for all 17 PCBs will be made at Imperial on the same timescale. JamieC will try to get the sensor bonded to these five PCBs as soon as possible after they are returned. However, as the slots cannot be booked in advance, it is not possible to know when they will be done. Once started, they should only take two days (one for gluing and setting and the second for the actual bonding). He will get the non-deep p-well sensor (which has faults) replaced this week. Matt now has got a total of four sets of sensor PCB-to-DAQ adapter cables made and tested and will get a few more sets made as needed. He has also ordered ten micro-USB-to-USB cables of 2m length which should arrive within days. This is in addition to two 0.5m cables already in hand which should be sufficient for all the tests planned this year. The IDC-to-BNC cables for the laser communication still need to be made and Matt will also organise this. Two more DAQ adapter boards are being assembled at Imperial and are ~60% complete; they should be finished by tomorrow. Following this, more will be made at the rate of around 1/week. There have been problems with the assembly of the USB_DAQ PCBs at Imperial as the temperature in the reflow oven is not high enough. This means some sensitive components have to be added afterwards by hand. The third USB_DAQ board (actually the first version 2 type) is part assembled and should be complete by Wed. Following testing, a third system should be ready by the end of the week. Following this, further USB_DAQ boards should be able to be assembled at the same rate as the adapter boards, namely 1/week. This assumes no more problems are found with the current assembly methods; if they are, then we should consider commercial assembly. It was decided that the third DAQ system should go to the laser setup as they should be able to use a board full-time from now on. The fourth will go to Imperial for DAQ software development and setting up the source measurement system. All subsequent systems will then go to Birmingham for the cosmics stack. At the rate of 1 system/week, then Birmingham will have four systems only at the end of Oct. They will then need to debug the system and move it to DESY, at the latest by early Dec. This is not unfeasible (particularly as they can do much of the debugging with one or two systems) but is tight so the schedule should be watched carefully. Matt has had problems with email communication to Vladimir so there is no progress on the memory extension card layout. Marcel reported that he had made progress with the laser-socket communication and can now read the values out, although writing needs more work. The dimensions of the tungsten plates for the beam test have to be specified by the Birmingham group to ensure they are consistent with the mechanical support of the PCBs. If the sheets are mounted on aluminium plates with the same thickness as the PCBs, then these can be slid into the PCB structure in the same way as the PCBs, allowing a flexible stack to be built up. This will need at least 8 slots (to allow for four alternating PCB and tungsten layers) although 12 slots might be more flexible in case extra space is needed. The transverse dimensions of the tungsten are then constrained by the size of the PCBs in the stack (~10x10cm2) and a size of around 5x5cm2 was thought adequate. The thickness depends on the slot spacing which is determined by the clearance due to the wire bonds. This was thought to be around 2mm but possibly 4-5mm should be allowed. The sensor bonds will need to be protected by potting when they have been shown to be functional and the total height could then be higher than now. A tungsten sheet thickness of 3-4mm should be reasonable and this corresponds to around 1X0 (3.5mm). We will need up to 10X0 so would need 12 sheets if 3mm thick. Clearly, many of the sheets will need to be clamped together outside of the sensor mechanical holding structure. The sheet dimensions are then likely to be 5x5cm^2 and 3mm thick, but nothing should be ordered until Nigel has checked with the mechanical technicians building the stack. Note: the potting compound will be added by the same technician who does the bonding so the sensor PCBs will need to be returned when they have been verified as working. Paul raised the issue of (semi-)online monitoring. The CERN beam tests have a histogram system written by George Mavromanolakis which runs on DAQ binary files as they are being taken. It has a lot of pre-packaged histograms which are useful for quickly checking data quality. We could reuse this package although we would need to replace most (all?) of the histograms. The alternative would be to convert the binary data to ROOT and then run interactive ROOT for displaying histograms. However, it was not clear if this could allow displays during an ongoing run. Owen was volunteered to contact George and look into how his display works, to see how straightforward it would be to adapt. Conferences: It would be very good to get some results from the laser in time for IEEE, even if only analogue readout. The RAL people are aware of this and are planning for such tests as soon as they can. We should get one MAPS talk in the calorimeter session of ALCPG and also one talk in the simulation session. The latter is likely to include a discussion on user feedback on the software. Only Marcel and JamieB are available to attend this meeting and so the default will be that Marcel takes on the former and JamieB the latter. The following week is IEEE, where JamieC and Marcel will talk. Marcel expects the ALCPG MAPS talk to be a summary of the two IEEE talks. The ALCPG meeting starts on Oct 22, so practise talks for this and IEEE will need to be done towards the end of the week starting Oct 15. As Marcel will be in the US at this time, these will need to be done by phone. The next meeting known is the Asian ILC regional meeting at Sendai University on 3-6 Mar 2008. Yoshi was a student there and would be interested in giving such a talk. JamieB went through a draft of his talk to the CALICE-UK meeting in Cambridge on Thursday. Please send any further comments to him asap. Next meeting: This will be on Thu 4 Oct, starting at 10.30 for software and 13.00 for the rest of the meeting.