CALICE MAPS Meeting, RAL, 18/10/07 ================================== Morning session =============== Present: Jamie Ballin, Paul Dauncey, Anne-Marie Magnan, Yoshi Mikami, Marcel Stanitzki, Nigel Watson, John Wilson Minutes: Paul GEANT4/Mokka: Yoshi showed some slides following up on his studies from the last meeting of the number of MIPs going through each pixel; see usual web page. He finds a very slight dependence of the number of MC contributions on the epitaxial thickness, which is nowhere near as big as would be expected if this number was related to the number of steps taken to cross the epitaxial layer. He is using versions Geant4.8.0.p1 and Mokka-06-00 and will probably have run with the default step size of 50mu. It is clear the number of contributions is not understood; it might be that looking through one event with both GEANT4 and Mokka printout turned on may clarify what the number means. Nigel reshowed his slides from 08/02/07, showing the number of MC contributions he finds looks consistent with Yoshi's for the detailed shower mode. With detailed mode turned off, the numbers look sensible; here the number means the number of separate particles contributing which enter the ECAL, so a number different from one indicates some interaction before the ECAL. Anne-Marie showed some slides on her findings. She has discovered the activeRangeCut was not used but it is simple to fix. It appears that the step size would have been different for 50mu and 5mu cell runs (since the step size was being set to the cell size itself), which may explain the difference in the energy spectrum seen when comparing these two. Unfortunately, the new versions of GEANT4 and Mokka she is using, Geant4.9.0.p01 and Mokka-06-04-p03, have major changes and this resulted in a bug in the cell index for several of her plots. [Note added after the meeting: new versions of the plots with the bug corrected have been added to the usual web page.] It appears that with the new version, the number of MC contributions does seem to be consistent with the number of steps taken in the epitaxial layer, but Anne-Marie should vary the step size and layer thickness to verify this is the case. PFA: Marcel showed the second part of his talk, slide 11 onwards; see usual web page. SLIC can easily switch to "tracker-like" hit recording, where all hits are recorded at their exact positions rather than summing over the cell and giving the centre position. This would result in very large output files but would allow detailed studies. Such a switch would in principle be straightforward in Mokka also but it was not known if it had been implemented. SLIC seems faster but this may be because the geometry description has been simplified. The underlying GEANT4 code is the same, of course, and this takes most of the CPU time. The number of MC contributions found in SLIC looks more like what would be expected in terms of the number of distinct MIPS going though a pixel. However, it would also be consistent with being the number of steps if the step size has been set to be bigger than the epitaxial layer thickness. Hence, it would be sensible to check how the number distribution varies as a function of the step size. For the same reason as for Mokka, it would be most useful to implement two simultanously sensitive detectors in SLIC for the MAPS and bulk silicon, so direct comparisons of the standard and MAPS ECALs can be done from the same output files. It was not thought that the hexagon geometry of the SiD ECAL would be a problem if the bulk could be used separately from the pixels, scaling the energy by 300/(300-15) to account for the missing silicon of the epitaxial layer. JamieB showed some slides on PFA studies, see slides 1-10 on the usual web page. He is attempting the "no-harm" study but sees a degraded energy value with MAPS, at least for a subset of the events. This may be due to not weighting the back layers correctly for the thicker tungsten. This would be seen by a layer-by-layer comparison of the distributions of the standard and MAPS energy deposits in each layer in the collection fed to PandoraPFA. It should also be less important for low energy (e.g. 5GeV) showers. Afternoon session ================= Present: Jamie Ballin, Jamie Crooks, Paul Dauncey, Anne-Marie Magnan, Yoshi Mikami, Marcel Stanitzki, Renato Turchetta, Nigel Watson, John Wilson Minutes: Paul Minutes and matters arising: Giulio's TWEPP talk is now on the conferences web page. The only missing item is the proceedings contribution for EPS from Nigel. Correction to the minutes, Owen reported at the last meeting that he had compiled George's semi-online monitoring code. Sensor design: JamieC has looked at data corruption. He sees every row in the sampler regions (regions 2 and 3) reports six of the seven groups out for timestamp 0, even though none of the channels is above threshold. This is 6*168*2=2016 words in total every bunch train, i.e. ~8kBytes. The reason is not understood. He also sees a low rate of other bit corruption which is strongly dependent on the power supply setting for the SRAMs. This is currently supplied externally but will be derived from the unused DAC channel in future. It needs to be set to 2.6V or above to minimise corruption, although there is still some number of channel bits set wrongly even then. First results on the analogue test structure (see later) suggest the gain is low by between a factor of ~2. JamieC also sees the power drawn is lower than expected, implying the pixels may not be operating at the expected working point. This needs to be tracked down as this would influence a lot of the testing. Sensor testing: Five completed sensor PCBs were returned to Imperial last week. Matt tested them and no errors were found. It was decided to fit the originally specified inverter to maintain uniformity across all the PCBs. The new inverters are available and will be put onto the next round of boards. The PCBs were taken to RAL last Monday. Sensors have been glued (three with deep p-well, two without, all with 12mu epitaxial thickness) and bonded. Two required rework; the sensors were removed and new sensors mounted and they await bonding. The other three have been tested at a preliminary level by JamieC and seem OK so far. Only one of the five PCBs has been modified for the required fixes to the sensor. JamieC should try to do the other four before he leaves for IEEE but otherwise, it might be possible to get them changed at Imperial. There are five USB_DAQ V2 boards tested at IC. These are in addition to the two V1 boards at RAL. Another two boards are being assembled and should be available within a few days. There are also five adapter boards at Imperial, again in addition to the two at RAL. A sixth is ready for testing at Imperial. Matt has 17 aluminium mounting plates and will get 5 to RAL asap. There is no change to the cable status since the last meeting. In summary there are now at least five full systems available. These should go to the RAL sensor group, RAL laser setup, the Imperial source system, Birmingham to start on the cosmics setup, and Matt for debugging the firmware. Birmingham and RAL will be able to do firmware updates themselves when Matt releases new firmware versions so there should be no problems keeping the systems in synch. Similarly, if Paul can log onto all four PCs, then it should be straightforward to keep the software updated. Paul will need access to the Birmingham PC and Nigel will set this up. All future systems should go to Birmingham until they get a full set of four. This will require more sensor PCBs and the next batch of seven should be assembled on the right timescale, i.e. for a beam test in early Dec. The actual beam test may be later if it takes longer to bring up the multi-board system but it has to be before the end of the year as the DESY beam will be off from Jan 2008 until next summer. We should assume that we will need a group of 3-4 people at DESY during this period; Anne-Marie, JamieB, Yoshi, Marcel and Paul thought they would be available for at least parts of Dec. The quotation for the tungsten plates arrived during the meeting and will be around 400 pounds with a lead time of 5 days. This order should be placed straight away. There has been no significant progress on the DAQ control of the laser; the DAQ socket can read the settings but not write them. This would make a laser scan possible but painful, as runs would have to be stopped and started by hand in between each laser step. Nigel reported that Owen has compiled the semi-online monitoring code from George and now needs a test file to look at. There are many runs available and so Paul will put one in a web-accessible area straight away. Longer term, JamieB and Nigel will sort out distributing files on the web, which means Marcel and Paul will need to get Grid certificates. Paul put up a list of the known firmware developments needed for the USB_DAQ boards. There were no other additions to the list during the meeting. The highest priority items were thought to be the sensor control and readout (the first three in the list), the time tagging (fifth, essential for source, laser and cosmics) and the laser control (ninth). The "good" bunch operation (sixth) will make cosmics more efficient and the multiple USB_DAQs (tenth) will been essential for cosmics and the beam test but this can be left for a few weeks. Marcel showed some results from the analogue test structure; see pages 1-10 of his slides on the usual web page. He sees a good response with charge sharing across the pixel boundaries. The shuttersize here is the spot size in mu and Jerry is the name for sensor #1. The measurements shown were measured from the scope which was not sensitive to low levels and so the apparent pedestal seen is not necessarily real. Paul showed some slides on measuring monostable lengths and DAQ rates. The pixels appear to be quite non-uniform. There seems to be ~4000 words (hits) read out even for high thresholds; 2016 of these are the zero-hit words seen by JamieC but the rest are not understood. JamieB showed some results of the noise dependence on threshold, see slides 11 onwards of his talk. This shows very unexpected behaviour which is not yet understood. This was discussed but not thought to be due to the inability to load the sensor configuration data as this was reset at the start of every run. Conferences: Marcel and JamieC gave a runthrough of their IEEE talks; any further comments should be sent to them by email. If there are any further results, specifically from a sensor test structure without deep p-well or from the source tests, they could be included. Marcel will give effectively the same talk at the ALCPG meeting, with the results from his and JamieC's talks combined. He is also giving a non-MAPS general software talk at the same meeting. Next meeting: This will be on Wed 7 Nov at the usual times. We will meet in the sensor group meeting room, R76.