CALICE MAPS Meeting, RAL, 08/02/07 ================================== Present: Jamie Ballin, Jamie Crooks, Paul Dauncey, Anne-Marie Magnan, Marcel Stanitzki, Konstantin Stefanov, Renato Turchetta Phone: Yoshi Mikami, Owen Miller, Vladimir Rajovic, Giulio Villani (for part), Nigel Watson, John Wilson Minutes: Paul Paul introduced Jamie Ballin, a new PG student at Imperial, who will be working on MAPS for his thesis. Minutes of previous meeting: No corrections. For matter arising: The deep p-well contract has now been processed so the foundry have released the process design details. They have also specified the splits they need to check the deep p-well. They require one with no deep p-well, one with their standard process and one with a stronger p doping. They can also do only 5.5mu or 12mu epitaxial layer thicknesses. Given that the deep p-well is ~3mu, then the 5.5mu thickness is probably of little use but the four splits we will ask for will be the three deep p-well variations with 12mu and the standard deep p-well with 5.5mu. JamieC has not sent the design to the foundry for review yet. This will be done next week. They need ~two days to study it and then they will provide feedback in a phone conference. This will only cover the pixel design, not the logic. Renato has not yet confirmed the two external reviewers (Mark Pryddyrch and one of either Nicola Guerrini or Andy Clark) for the FDR on Wed 28 Feb. He should do this soon to be sure we do not need to move the date. Giulio was not present to report on the exact calculation of the total charge available for collection as determined from the simulation. The figure of 1300e- was thought to be approximate but an exact value would be useful. Sensor design: Renato went through the status of the layout using the design tools (so no slides are available). Contrary to what had been discussed previously, the design now has all the digital circuitry on the outer edges of the pixel; indeed some components overlap into the neighbouring pixels. This will require deep p-well right at the edge, where the charge collected is low. This change was to maximise the distance between the analogue and digital components and to keep the deep p-well clear of the collection diodes. The latter was recommended by the foundry. During the discussion, several points were decided: o The deep p-well must be symmetric within the pixel, so the coverage around each edge has to be the worst case of all four sides. o The n-well components requiring deep p-well should be kept as far from the corners as possible to minimise the loss of charge there. This means the layout along the edges should be squeezed towards the centre of the pixel. o Giulio reported in the last meeting that a 0.5mu wide deep p-well surround around an n-well still allows significant charge to leak into the n-well. It was thought that a width of ~2mu (i.e. around the same dimension as the deep p-well thickness) would be better. This should be implemented where space allows. o The four different pixel types will be the four combinations of shaper/sampler and n-well/metal capacitors. The latter were thought to be only available in the uppermost two metal layers but Jamie and Renato think they can make them from lower metal layers anyway. This is an area where feedback from the foundry during their review will be very useful. o The choice of the capacitors rather than varying the deep p-well was made as the former could prevent the pixel from operating at all while the latter would cause a degradation, not a failure. Giulio was called for this discussion and the conclusion was that the three process splits with differing deep p-well should give some information on the deep p-well effects. With four (or possibly six) splits and four pixel types, there are 16 (or 24) different pixels to test. It was decided that no more variations should be made. o There is not enough room for 7.2mu diodes or for eight diodes. Hence the only options available are the diode sizes of 3.6mu or smaller and the diode positioning. The latter can only be moved by a small amount. o The deep p-well will be needed in the centre and along the edges. The decision on whether to connect these regions into one large cross-shaped area, leave them as five regions, or connect the outer four regions with a "guard ring" into each corner was left for later. Input from Giulio's simulations on this would be useful. Jamie will send the GDS file to Giulio when the layout is complete. Sensor simulations: Marcel reported for Giulio on recent sensor simulations; see slides on the usual web page. Renato raised the issue of the deep p-well ground contact, shown as a small area at the top of the pixel. He was concerned that this might distort the potential seen within the epitaxial layer. There will be many such ground contacts across the whole pixel in the real sensor and so Renato was concerned the simulation might be displaying unrealistic effects, such as the amount of charge collected in the central n-well. As there is no room for the 7.2mu diode size, then this should not be simulated with any urgency, as suggested for future work. Physics and background simulation: Owen Miller showed some results on simulating machine background rates; see slides on the usual web page. The basic result is that there are ~43M pixels hit per bunch train. This corresponds to around 10k hits per bunch crossing. Each of these will on average give ~4 pixels hit after charge diffusion. The angular distribution is very forward peaked, with the endcap seeing an order of magnitude more than the barrel. This should be compared with the noise rate; for 10^12 pixels and a 10^-6 noise rate per pixel, then there will be 10^6 noise hits per bunch crossing. Hence, it seems for the DAQ rate, the noise is likely to completely dominate. However, in terms of filling memory locations on the pixel, then the worst case area, i.e. the innermost endcap region, might have a comparable rate. Owen will check how dense the hits are in this region. There are new machine parameters from the Beijing meeting this week which indicate an increase in backgrounds (albeit nowhere near enough to change the above conclusions). However, Owen and Nigel have not yet been able to reproduce the expected increases. Yoshi reported on his work on clustering; see slides on the usual web page. He hopes to use the shower shape to distinguish between photon and electron showers. This would be suprising as normally these two are considered very similar so results at the next meeting on this are awaited. Yoshi's results are all done with the "ideal" simulation with no digitisation. For clustering, the precise pixel hit distributions could be important so any results cannot be trusted unless reproduced with the full simulation. Hence, he was urged to move to using this asap. Marcel showed various items; see slides on the usual web page. He and Giulio have tried to come up with some simpler parametrisation of the charge collection using a potential-equivalent generated by each diode. The issue is how to develop the transfer function from potential to charge collection; e.g. for the case shown, the lowest potential is in the centre but this is the position of highest charge collection. Marcel stated that he thinks a factor of two improvement would be possible by optimising the diode placement. Marcel also reproduced the effect seen by Yoshi in meeting on 12/07/06, where the reported number of hits in each pixel peaks near four. These are energy deposit hits, where a single particle can generate multiple hits while crossing a volume. The peak at around four hits might be an effect due to the step size, i.e. a 15mu epitaxial layer may need an average of four steps given the GEANT4 step size parameter setting of 5mu. This should be checked, most obviously by changing the step size parameter. Anne-Marie pointed out there are two GEANT4 parameters; the range cut and the active range cut; only the former seems to have any effect. The effect has been seen by Yoshi not to depend on the silicon thickness however. The overall conclusion is that we are not yet able to unambiguously determine the total number of incoming particles. Nigel showed some work he had also done on the same effect; see slides on usual web page. The number of hits looks a lot more like the expected distribution if the detailed shower mode is turned off; this may be because the hits belonging to each particle are summed into one hit. Anne-Marie reported on her studies of resolution; see slides on usual web page. She sees the best resolution is obtained for 20GeV with ~100mu pixels. There seems to be a discrepancy between the absolute energy scale used in GEANT and the well-known average excitation energy in silicon of 3.6eV per electron. The average energy deposited in the epitaxial layer is 3.3keV while the approximate total charge generated in Giulio's simulation is 1300e-, which implies 2.5eV per electron. This has no effect on the simulation results, as the energy deposited is purely used as a scaling factor to Giulio's charge spreading. Anne-Marie sees an unphysical peak when dividing the pixels into 5mu subpixels, which is not present for larger subdivisions. The peak position corresponds to approximately 1/3 of the MIP value. As this results from crossing a 15mu epitaxial layer, then the unphysical peak seems to come from particles crossing through the 5mu subpixels. She has checked this effect was not due to rounding error in using floats so it is currently not understood. Next meeting: The FDR is on Web 28 Feb, starting at 10am in R76. The next standard meeting will be one week later, on Wed 7 Mar. This will probably be held in a PPD meeting room as the number of people now attending is getting too large for the R76 room.