CALICE MAPS Meeting, RAL, 04/10/07 ================================== Morning session =============== Present: Paul Dauncey, Anne-Marie Magnan, Yoshi Mikami, Owen Miller, Nigel Watson, John Wilson Minutes: Paul Paul started by reshowing his slides from the previous meeting with a list of the topics we should be concentrating on. GEANT4: Yoshi showed some slides on GEANT4 results; see usual web page. He clearly shows the low energy shoulder in the pixel energy spectrum is due to charge sharing, as expected, but the peak at ~1keV when working with 5x5mu2 subpixels is still not understood. The number of contributions to the SimCalorimeterHit has a peak at around 4 to 5 which was thought to be the number of steps for a single particle to cross the epitaxial layer, implying a step size of around 5mu. It is important to understand this issue as it is at the heart of the non-linear effects and hence the pixel size choice. The discussion concluded that Yoshi should study whether the MCParticle information from the getNMCcontributions() identifies the contributing particles as the same or not. Also, he will look at how the number of contributions changes as the epitaxial thickness is changed by -5mu, +5mu and +10mu. It would also be good to vary the step size but this may not be easy due to the software layers of Mokka and GEANT4. Paul suggested the only way to be sure would be to print out a single event in full from GEANT4 so the literal step size can be seen. This should be done with muons to be quick and keep down the amount of printout. In addition, changes to LCIO might make these studies simpler. Nigel thought Frank Gaede had produced a version of SimCalorimeterHit which included the truth particle momentum, as for SimTrackerHits. This would then allow a reasonable check to see if the momentum of the contributing particles to the hits were very similar (and hence the same truth particle) or not. However, this version is not in a release yet. Afternoon session ================= Present: Jamie Crooks, Paul Dauncey, Anne-Marie Magnan, Yoshi Mikami, Owen Miller, Matt Noy, Konstantin Stefanov, Renato Turchetta, Mike Tyndel, Nigel Watson, John Wilson Minutes: Paul Minutes and matters arising: No comments. Giulio's TWEPP talk has not yet been put on the conferences website. Sensor design: Jamie showed some slides on his findings; see the usual web page. Using the laser on the test structures, he sees a response from both capacitor combinations. (Note, there are no shaper test pixels.) The gain is not yet measured; this would require an absolute calibration which Giulio is confident he has but we should cross check it with the source studies anyway. The next steps will be the vary the laser position and intensity. There is no quantitative noise measurement yet but this should be feasible. Because the bulk pixel readout is not ready yet, then finding the test structure with the laser is a slow process. A cross has been etched with the laser on the PCB, 500mu from the test structure. [Note added after the meeting; unfortunately, the marked board is no longer functional.] The question was raised of why the sensor errors were not found by the simulation before fabrication. This is partially because a simulation of the whole sensor is not feasible so only isolated parts could be modelled at any one time. The errors could then have only been picked up by things like checking the currents being drawn in the simulation; Jamie's error 3 would have had a very large current. In principle, this would be possible, but was not noticed this time round. There are potentially big changes for the second round; the pixel size could change with a corresponding change to the number of pixels feeding into each logical column, the pixel layout may change if the shape is changed (to hexagons?), there are many changes already known to be needed for reducing the I/O count, putting the I/O onto bump-bond pads, etc. Would there be a way to reduce the number of errors in the second round, e.g. by improving the simulation facility? Renato thought this would not be possible but the question of design checks does need further thought. There needs to be a second round sensor Interim Design Review towards the end of the year. This should also include a post-mortem on the first sensor design. Sensor testing: After a few manufacturing delays, the next round of 5 sensor PCBs will be finished by Wed 10 Oct. After testing, Matt will take these to RAL on Mon 15 Oct so they can have sensors mounted. The sensor mounting and bonding cannot be booked in advance but should take around one week. There are another 12 PCBs on order to be fabricated and 7 of these will be assembled in the short term. The next 5 baseplates should be ready by Wed next week also. Of the three existing PCBs, all three now have deep p-well sensors after a new sensor was mounted on a previously used board successfully. Of the three, one has a fault in the readout of one of the four regions, while another is drawing a high current in the LVDS converters. A firmware switch to limit the readout to particular regions will be needed to continue working with the former and so this one will be returned to Imperial so Matt can work with it there. There are now 6 sets of sensor cables and 5 sets of the sensor PCB power cables. Matt has also made one set of the laser cables but these have not been tested. He has also received the ten 2m USB cables. There are now 5 USB_DAQ adapter cards with 4 tested, and they continue to be made at a rate of 1.5/week. There are now 4 USB_DAQ V2 boards, 2 of which are tested and the others will be ready by Mon. The next four subsequent ones made will then go to Birmingham. There has been no progress on the memory extension card and there was no information on the status of the laser control. Potting the sensor wirebonds will only be done after the next five sensors have been mounted and fully tested. The sensors have many biases and it is not clear how we find the optimisation point. Minimising the noise alone is not a good measure; the best would be signal/noise; for this, probably the only system with a high enough rate would be the laser, but this will need to be carefully calibrated beforehand against a source. Beam tests: Nigel confirmed the 50x50x3mm3 tungsten plates would work with the Birmingham mechanics for the beam test. These should therefore be ordered asap. There is 5mm clearance from PCB-to-PCB (i.e. 6mm from PCB top surface to top surface) which allows for 2.5mm either side for wirebonds and the components. Hence, the 3mm tungsten plates will need to be sunk into the 1mm aluminimum holder plate and not simply mounted on the surface, so they protrude 2mm above. The mechanical structure will have 20 slots, allowing quite a bit of flexibility in the positioning of the tungsten and PCB layers. The whole structure including tungsten will be quite heavy and will need to be able to stand vertically on its edge in the beam test. This means at least the lower baseline edge needs to be far enough away from the PCB that the downward-going cables have sufficient clearance. It will also need an extra detachable baseplate to support it. Owen has downloaded and compiled George's semi-online monitoring code. He hs not tried to modify it for MAPS. Conferences: For ALCPG and IEEE, we should aim to present results from the analogue laser in terms of linearity, noise and a 2D scan to show charge spread. If any results on the bulk pixels on noise and a signal from the source can be obtained, this would also be useful. Marcel will be back next week to work on this. The practise talks for ALCPG and IEEE will be at 16.00 on Thu 18 Oct. If needed, people can phone in. AOB: Jamie showed the last slide from his talk, concerning the possibility of production of some test sensors funded by an internal grant. This currently would only fund one seat (5x5mm2) but an further 35k would buy three more seats, allowing a full sensor to be made. The CALICE budget does not have 35k spare so this is probably not possible. Any sensor other than literally for bench testing would have to be pin-compatible with the PCB we have. There is also a danger that this would divert Jamie from the second round of sensor design. To minimise the dead area around the edge of the sensor, then putting bump-bond I/O pads over the active region would be very useful. This has been done (it is called "pads-over-logic") but Jamie does not have experience of it. With e.g. 100 pads on top of a 20x20mm2 sensor, then each could be ~2x2mm big; this would allow normal solder paste BGA-like connections to a PCB. Mike raised the question of the technology for non-wirebonding. He knows of a process for using kapton flexi-cables called "tab bonding" which might be an alternative to bump bonds. Next meeting: This will be on Thu 18 Oct, starting at 10.30 for software, at 13.00 for the hardware and at 16.00 for the IEEE practise talks.