CALICE MAPS Meeting, RAL, 19/11/07 ================================== Present: Jamie Ballin, Jamie Crooks, Paul Dauncey, Anne-Marie Magnan, Yoshi Mikami, Owen Miller, Matt Noy, Marcel Stanitzki, Konstantin Stefanov, Renato Turchetta, Mike Tyndel, Giulio Villani, John Wilson By phone: Nigel Watson Minutes: Paul Minutes of last meeting: No comments Sensor basic tests: JamieC reported he had checked that a positive threshold was required and that the pixel trim threshold had the opposite sense, i.e. a higher trim value corresponded to a lower threshold and hence more hits. He thinks the trim LSB corresponds to 1.8mV, see spreadsheet on usual web page. This is to be compared with 0.4mV for the overall threshold LSB, giving a relative factor of 4.5 (although see below). JamieC also showed a spreadsheet plot of the bias currents against DAC settings, see the usual web page. This indicates they are very non-linear compared to the DAC values. He then showed a table of the estimated DAC settings and bias resistors needed to get to a reasonable operating point. These are quite far off the nominal DAC settings of 2048 which the sensor PCB was designed to supply. This is because the PCB assuming the bias current could be set by a resistor chain without the sensor being sensitive to the resulting voltage. However, the sensor only works with these inputs within a particular range of voltage, so the resistor changes are to reduce the voltage drop across the resistors and so bring the voltages into range. The most recent resistor/DAC changes, shown in blue, are new and have not been implemented by Matt on any of his boards. JamieC did these to bring them into the nominal range although he does not see any difference in behaviour with these changes. Laser tests: With the new operating point, JamieC and Giulio have remeasured the test structure pixel gain using the laser and see little change to the gain. It is still approximately 2-3 times lower than expected, despite the pixels now drawing four times the current. This is probably due to the gain being set by capacitors (and hence the fixed layout geometry) rather than the bias voltages; the noise should be more sensitive to the operating point. The noise from the test structure is estimated to have an amplitude of 20mV, while the MIP-equivalent signal is thought to be at 50mV. This gives a S/N of only 2.5, which would of course be disasterous if true for the bulk pixels. It was thought unlikely the external amplifier would be causing much extra noise. However, the laser signal is seen to saturate at around 250mV and Konstantin thinks this is due to the external amplifier being near the top of its range; he can fix this easily and this should be done before any further measurements. The laser signal could be different from the estimate, although Giulio is confident of his calibration. Hence, this should be cross-checked against the Fe-55 gamma source (5.9keV, giving ~1600e- within a very small volume ~1mu). With the laser calibrated, then a series of gain measurements using the laser and varying each DAC setting in turn should be done to see how sensitive the gain is to each. In addition, a signal injected onto the VRST when held open would provide a relative measure of the gain variation. Giulio is cross-checking the laser calibration with an unpackaged sensor but this appears to give a similar answer. He is also doing a sensor simulation of the three test pixels to compare with the data shown at IEEE. It would be good to compare with both the deep p-well and non-deep p-well results, but it will take until the end of the year to do even the first set. The limit is the CPU time of the farm used and he cannot improve the speed without making the grid coarser. Marcel reported that the laser-DAQ socket communication had been tested in the morning and was effectively working. Some work-around for the oddities of LabView had to be found and some readjustment of the data packet may be needed to give a sufficient coordinate system range, but the basic code was now functional. The DAQ readout of the sensor itself did not yet work but the sensor #1 being used has a readout problem with region 0. This will be swapped with sensor #4 (brought by Matt today) after the next round of laser tests as the current sensor is positioned accurately for illuminating the test pixels. Source tests: JamieB showed a plot of number of hits from region 0 of sensor #2 vs threshold setting, see usual web page. The red points are with the source and the blue points are without. Most of the features of this plot are not understood but the anomolous behaviour seems to be reduced by waiting between each bunch train. The high numbers of hits disappear completely if a 1 sec wait is used. This eliminates most causes. This test should be repeated on other sensors to check if they look similar; data from sensors #4 and #6 were taken over the weekend but have not yet been analysed. Paul showed a plot of noise rate for region 0 on sensor #4 vs the overall and trim threshold settings. The line of equal noise rate has a slope of around 1/8, indicating that the trim LSB is about eight times larger than the overall threshold LSB. This is about a factor of two larger than JamieC's expectation (above). Other test preparations: Matt reported that there are now ten working V2 USB_DAQ boards; one had a fault but this seems now to have been fixed. There are also ten adapter boards for the USB_DAQ boards; eight have been tested and the remaining two will be checked at Imperial. The last 12 sensor PCBs and their baseplates have been sent to RAL for sensor mounting. These are still missing some components but these can be mounted at Imperial after the sensors are added. There is a gap in the bonding schedule which we can fill with the sensors. The PCBs have not yet been tested so there is some (probably small) risk in this. The 12 will be made up with ten sensors with a 12mu epitaxial layer and deep p-well and two with a 5mu epitaxial layer and deep p-well. The existing eight consist of one non-functional board, five with a 12mu epitaxial layer and deep p-well and two with a 12mu epitaxial layer and no deep p-well, giving an eventual total of fifteen working "standard" sensors. Matt will look into fixing the faulty PCB; given the other boards being produced, this is not a high priority right now. It would be useful to keep track of the sensors numbers and locations; a spreadsheet or database should be created. This will also prevent accidental mis-setting of the same number on two sets of sensor PCB dip-switches. Owen reported the Birmingham system runs to take data but they had not yet got George Mavromanolakis's code to give histograms from the files produced. This may be due to this code requiring event records while the MAPS produce bunchTrain records; a global edit of the former for the latter may help. The mechanical cosmics/beam structure should be ready within a day or so. This will need to have a mount so that it can be stood on a horizontal table without any fixings and will have a hole in the back so it can be oriented either way round with the beam hitting the epitaxial layers either first or last in each layer. The four USB_DAQ boards will be held in an adapted NIM crate. Marcel reported that due to a delay in the purchasing department at RAL, he had still not received the tungsten plates, but they were expected very soon. JamieB showed some slides on his GUI for the DAQ; see slides on the usual web page. This will need a more up-to-date version of java than currently on the PCs; he will install this himself, although Nigel has already done so for Birmingham. Beam test preparations: It would be best to take a complete, working system to DESY. This consists of: - As many sensors and PCBs as are available - Five USB_DAQ and adapter boards (four plus a spare) - The NIM crate to hold the USB_DAQ boards - The 5V power supply, rated up to 12A - A bench power supply for the PMTs (20V) and external 2.7V bias - Spare power supplies - The mechanical frame, tungsten plates and holders - The PC, screen, keyboard and mouse It was thought that a commercial shipping company will give enough flexibility, even given the uncertainty in the schedule, so Mike and Marcel will investigate doing this through RAL. The decision on whether to go out to DESY or not will be made at or before the next meeting on Mon 3 Dec. If only the shapers are working at this time, it was still thought worth going as the mechanical holder does not require the sensors to be rotated; they can all be mounted with the same orientation, albeit spaced further apart. If we decide to go, then the shipping is likely to take the rest of that week, so we should assume the equipment will arrive on Fri 7 Dec. We could start assembly and running over the weekend. The people available for at least some parts of the following two weeks were JamieB, Anne-Marie, Marcel, Yoshi, Owen and Paul. Matt and JamieC would also be able to come out if needed. We will need to negotiate with the ECAL tests which are going on at the same time for beam. We need to contact Erika to arrange safety training and for borrowing any further equipment, including a table for the mechanical structure to sit on. Marcel volunteered to take on the organisation and Paul will connect him with Erika. We will also need office space although there are several UK offices for Zeus or H1 groups which have been used by CALICE-UK visitors in the past. A rough draft of the beam test program was made. The mechanical setups needed will be: - No tungsten, to see "MIP-like" straight tracks - Sensors with no interleaved tungsten but behind varying total thicknesses of tungsten, to attempt to distinguish soft photons from MIPs (which shoud leave tracks in all sensors) - Sensors with interleaved tungsten, as is normal for a sampling ECAL The DESY beam gives electrons only. It can go from 1-6GeV and so in principle, each of the above (except the first) would be done at six energies, 1 to 6GeV in steps of 1GeV. Also, the second two would be repeated with the sensor orientation to the beam reversed to see the effect of the epitaxial layer at the front or back. It will also be necessary to scan at least the threshold for all runs to map out the signal and noise shapes. Conferences: JamieC has a draft of his IEEE proceedings and this will be put on the usual web page; it needs to be submitted by the end of the week so comments should be sent to him. Marcel has not yet finished his contribution but this should also be released in time for comments to be returned by Friday. Next meeting: This will be on Mon 3 Dec, again starting at the slightly later time of 13.30.