CALICE MAPS Meeting, RAL, 29/02/08 ================================== Present: Jamie Ballin, Jamie Crooks, Paul Dauncey, Anne-Marie Magnan, Renato Turchetta, Giulio Villani Phone: Tim Martin, Owen Miller, Nigel Watson Minutes: Paul Minutes of previous meeting: No comments. Political status: Paul reported that the results of the ranking of projects would probably be released at a meeting on Monday. However, it would then be three weeks until the actual budget allocations are known. Nigel reported that the LC-ABD collaboration have ILC/GDE components in their plan which might indicate that STFC are prepared to be flexible, if it is approved. Sensor studies: JamieC showed some slides on noise and comparator studies; see usual web page. He scanned over both positive and negative thresholds; the comparator is not retriggerable so the rate is seen to drop as the threshold becomes more negative. From the test structure, he measures a threshold of 400 "threshold units" (TU) as corresponding to an injected input signal of 210mV, which is the signal which gives a response equal to the Fe55 peak. This peak is for 5.9keV, and using the relation that 1e- = 3.6eV deposited, then this is 1600e-. Hence, for the sampler test pixel, then roughly 1TU ~ 15eV ~ 4e- for signal charge equivalent. Jamie believes that the comparator circuit saturates around 700TU, so it is not worth operating it above this level. This would correspond to around 10keV so this should be sufficient. He also thinks effects near the top end of the plots shown (1024TU) are not significant as they are right at the edge of the response range. Renato showed some estimates of the noise made from the rate response during threshold runs; see slides and spreadsheet on usual web page. He measures a noise of 10-15mV for shapers and 15-20mV for samplers. This is converted from TU using a factor of 1.8V/4096TU = 0.44mV/TU so these noise values are equivalent to 20-30TU for shapers and 30-40TU for samplers. Assuming the bulk samplers have the same response as the test pixel, this implies a noise for samplers of 450-600eV or 120-160e- equivalent signal charge. The expected noise was around 30e- or 100eV so there is a discrepancy of a factor of around five. JamieC also showed results on laser tests; see usual web page. With the laser intensity probably equivalent to an injected input signal of 300mV, he sees (with some work) the response of the samplers cuts off for at least one pixel around 600TU, which is consistent with the above estimates. He also sees the shapers cut off around 300TU for the same signal size, implying the expected factor of two difference in gain. This would mean the shapers scale is roughly 1TU ~ 30eV ~ 8e-. To make things more efficient, the laser system needs to be automated. All the hooks are available for the DAQ to drive the laser stage and fire it but they have never been tested. This should not be a lot of work but would need some time with Paul (and probably Matt). Once working, there are a lot of useful measurements which could be done. Nigel will look into the possibility of Owen coming down to RAL to help with this. Sensor simulation: Giulio showed an addendum to his previous work on the bulk pixel simulations, looking at efficiencies for higher noise; see usual web page. He also showed some plots of the simulation for the test structure pixels; see document on usual web page. Due to the absence of as many neighbouring pixels, the charge collected by the test pixels is O(10%) higher than in the bulk. The deep p-well simulation shows significant peaks at the diodes compared to the pixel centre, with around a factor of two difference between them. This was not observed in the measurements (shown at IEEE) where a smooth surface over the whole pixel was seen. This would imply the laser spot size was comparable with the pixel size, although it was nominally small, with a 4x4 or 5x5mu2 shutter. However, a large spot size would seem to be contradicted by the non-deep p-well measurements where the diodes were visible, which implies the laser spot was indeed small. This needs to be understood. Paul showed results from a simple diffusion model for the charge spread where the parameters had been adjusted to agree with Giulio's results to some extent; see usual web page. When the charge is absorbed slowly, then diffusion dominates and the two simulations agree. However, when there is a large absorbing area, specifically the non-deep p-well case, then the two disagree by an order of magnitude, with pure diffusion giving a much higher signal. Giulio commented that the main effects missing from diffusion would be recombination and the dependence of the diffusion constant on the charge density. For the latter, high densities tend to spread quicker due to charge repulsion so the diffusion constant increases with higher density. Beam test results: Nigel showed slides from Owen (who had had to leave) on measuring the MIP efficiency from the beam test; see usual web page. He searched for hits within a specified distance (the "margin") of the expected position on a sensor and sees an efficiency of between 8-15%. There is no correction for non-ideal alignment yet but the efficiency is stable for a wide range of margin cuts so this does not seem to be a significant issue. It was not clear whether the plots for layers 0 and 2 (and 1 and 3) were identical due to the analysis method or for some other reason. JamieB also commented that the runs being used were taken before the PMTs were moved to a separate USB_DAQ and hence would suffer from extra noise hits in at least the one layer which shares the PMT readout. Tim presented slides on his method for measuring the efficiency; see usual web page. He looks for pairs of clusters in the same region of the sensors in two layers and gets the efficiency from the ratio of single to paired clusters. This gives a value for the efficiency of around 4% although he has removed hits from around 50% of the sensor area to remove noisy channels. JamieB showed slides of his work on tracking, measuring the efficiency by comparing three-hit and four-hit tracks. He sees dead regions close to the memory columns but otherwise a reasonably uniform efficiency. His average efficiency for shapers and samplers at nominal threshold is 10-15%. Anne-Marie showed some slides on simulating the efficiency vs threshold, see slides at the end of JamieB's file. From the tails of the threshold scan, she estimates a shaper noise of 25TU, compatible with Renato's estimates above. Using the simulated efficiency to estimate the TU, then she gets 1TU ~ 30eV for shapers and hence a noise of 750eV, again very similar to Renato. The efficiency values are all converging but are still different by factors of around two so these need to be understood further. Test hardware: JamieC reported that four more sensors had been mounted but only one worked with no errors. This was given to Paul to take to Imperial for the usual modifications. The others failed with separate problems so they are likely to be individual faults, such as single bad wire bonds. Matt has ordered 30 more PCBs and base plates and these are expected to be completed by mid-Apr. More stand-off posts and screws should also be ordered to match. Next meeting: At RAL, on Mon 31 Mar at 13.00.