CALICE MAPS Meeting, RAL, 07/08/07 ================================== Present: Jamie Crooks, Paul Dauncey, Anne-Marie Magnan, Yoshi Mikami, Owen Miller, Matt Noy, Vladimir Rajovic, Marcel Stanitzki, Mike Tyndel, Giulio Villani, Nigel Watson Minutes: Paul Minutes and matters arising: Renato was again not present to report on when the foundry can be named. Jamie will follow up on this. Vladimir was able to get an extra three weeks in the UK and this period started last week. Nigel gave the MAPS talk at EPS on Thu 19 Jul. This went well with a lot of interest from the audience. Marcel will put the talk on the conference web pages. Sensor design: Jamie sent Paul the sensor user manual and this is now on the web pages. Jamie has looked through the foundry wafer quality information as mentioned in the previous meeting. There is a lot of very detailed information given there but the basic result is that all the wafers sent to us pass their quality tests. Two of the sensor PCBs have been received at RAL and are in the queue for sensor gluing and wire bonding. The two sensors to be mounted both have the 12mu epitaxial layer and deep p-well. They are expected to be returned in around one week. Giulio has not yet tried to observe the thickness of the epitaxial layer on a spare sensor. Sensor simulation: Giulio showed some slides on the status of the sensor simulation; see latter half of his talk on usual web pages. He is simulating in 5mu steps, giving an 11x11 array of 121 points total over the central pixel. He has completed ~40 points so far and with a rate of ~2.5points/day, he estimates it will take around 6 weeks to complete. The plot of the 36 values in one quarter "S1" of the pixel looks reasonably symmetric down the diagonal. Quantitatively, the signal size for the corner of ~500e- is higher than the physical maximum of ~1300/4 = 325e- where the charge is divided between the four pixels at the corner. This is because the collection diodes of the other three pixels are not being modelled (for reasons of speed). Hence, this simulation will mainly be useful for comparison with the laser measurements near the centre of the pixel, not at the edges. It is therefore not so useful for the physics simulation. The edges will need to be redone in more detail later. Sensor testing: The first half of Giulio's talk gave the status of the laser preparations; see slides. He has calibrated the laser using the Vanilla sensor and finds it can generate from 0.5-2.0 MIPS-equivalent with an error of around 15%. Three sensor PCBs were fabricated and populated commercially over the last few weeks. One has five missing coax connectors as these have not yet been delivered; they will be added at Imperial. The dip switches to set the board IDs were not included in the assembly but have since been added by Imperial. The zero ohm resistor to define the DAC PU pin state needs to be added later to each board. Some of the exact resistor values were not available and alternatives were used instead; 20.7 kOhm was substituted by 20.0 kOhm, a difference of ~3%. This means that there will be minor differences in the reference values if the later boards are made with the originally specified values. Matt has done quite extensive tests of these sensor PCBs (without the sensors) and his results were posted previously to the MAPS mailing list on 02/08/07. Following component assembly, the sensor PCBs have a bow of several mm, with the component side being bigger and hence being on the outside of the bow. Therefore, there is a risk of breaking the component joints if they are forced flat by bolting to the base plate. Matt will contact the assembly company to see if they have some idea of the risk of this. The bow also complicates the Birmingham mechanical holder for the beam and cosmics tests. They may need to only insert the first ~10cm of the PCB into the holder and provide separate cable strain relief. In principle, it seems feasible to remove a bad sensor and replace it with a new one on the same PCB. This requires both regluing and redoing the wire bonds. For the former, the important issue is to allow the substrate contact to the PCB. Heating the glue to make it soft and hence prevent ripping the PCB metal layer is a technique which has been used before. For wire bonding, then there needs to be enough of the bonding pad available for a second bond. The pads are 200mu, which is bigger than the Atlas hybrid pads where rebonding has been achieved. It would clearly be useful for the bonds to be made off-centre on the PCB pads so as to allow more space for a rebond. Jamie will follow up on this. Even if sensors can be replaced on the PCBs once (or even twice), then the three sensor PCBs which were made will limit us relatively soon. We should allow around one month to make more so as to not have to pay too much in the next production round. This should therefore go ahead as soon as the PCBs have been fully verified. Matt and Jamie thought this would be possible within a week (or possibly two); it should certainly be done by the end of Aug. The original number to be made was 15, i.e. the existing 3 plus another 12. However, it was thought that 12 more may not be enough, so around 20 PCBs should be made. Components for the 12 have been purchased and so these could be populated easily. The others could be held and populated later if needed. The need for more should become clearer when we get some estimate of the yield over the next few weeks. The baseplates will be made at Imperial. Vladimir sent the hole specifications to Paul and Matt and they will contact the workshop to make the plates. They will be very simple, with no countersunk holes. These will be passed to Jamie as the baseplates will be added to the PCBs immediately after the sensor mounting and bonding to protect the bonds. Some cables to connect the sensor PCB to the USB_DAQ adapter have been made. Due to various miscommunications with the company, they sent flat ribbon (i.e. non-twisted pair) cables. The minimum quantity for twisted pair cable is 500m. (We need 3 cables for each of the ten USB_DAQs and with each cable being 2m, then the total length needed is 60m.) Hence, Matt has used the flat ribbon cable and finds good quality signals, albeit with the relatively inactive interface being used so far. Giulio is concerned about how the cables can fit into the laser enclosure so Matt will send him one made-up cable to check. Two USB_DAQ adapter PCBs were manufactured commercially and then assembled at Imperial. Matt also reported on the testing of these in a MAPS mailing list posting on 26/07/07. After the first was populated, it was noticed that the orientation of the lemo connectors for the PMT inputs would mean the lemo cable would prevent the memory extension card from being inserted, so the connector orientation was modified for the second card; subsequent cards will be made the with the modified connectors. Also, one pin pair needs correction wires but otherwise the boards seem to work well. Components for a total of ten have been ordered and at least eight more PCBs will be ordered when tests are complete. The USB_DAQ board has a micro-USB connector so we will need micro-USB to standard USB cables. Matt will order 15 of these, all 2m long. There are ten USB_DAQ PCBs existing. Components are in hard to populate at least three of these and the rest will be ordered soon. The assembly of these will be done at Imperial. The USB_DAQ memory extension card is being designed by Vladimir. The schematics are done and he has moved to the PCB layout. The board should be simple as it is mainly tracking from the connector to the components. The memory component to be used is 2M deep x 18 bits wide, so effectively 4MBytes, and each board will hold two components. However, there is a new component with an extra memory address pin, so 8M deep and hence doubling the memory. The board will be made to be compatible in case this is needed in the future. The USB_DAQ FPGA should have sufficient memory for all the near-term tests and it is possible the memory extension will not be needed before the sensor V2, but it would be useful if it is available earlier. Marcel has cloned all the PCs and they are ready to be taken to the separate institutes. The Birmingham and Imperial ones will be put in Jamie's lab until they are collected so as to allow them to be used (and get them out of Marcel's office). The software is of course in development and Paul will have to have access to all the PCs to do updates in the future. Marcel reported that Konstantin has had no luck finding tungsten sheets. Marcel has obtained quotes for tungsten sheets from a company. We need to decide on pure or alloy tungsten, where the former is expensive. (Longer term, we would need to worry about the magnetic properties also, but this is not so relevant for the current tests.) It was decided that 90-95% tungsten is fine for the beam tests as the difference is small and easily modelled; the impurities are mainly nickel and molybdenum and will be known. We need four weeks lead time to order them. We should aim for a beam test in ~Nov so this is not yet critical. We will need around 10X0 = 35mm in total so e.g. ten sheets each 5mm thick would be good. The main issue is the size as they cannot realistically be machined after they are manufactured. The size needed is mainly driven by the mechanical structure for the beam test, which is being built at Birmingham. If we want tungsten sheets between the sensors, then the structure will need to hold them. Hence, the decision on the exact size and thickness of the sheets was left for the Birmingham group. Matt and Jamie will spend the next week doing "proof-of-life" tests on the sensor following the mounting and wire bonding. We should be much better informed about yield, etc, after this time. Physics studies: Nigel has been running Fluka and showed some results in a talk; see slides on the usual web page. Fluka is more normally run in an integrating mode over many events to get averages, although it is possible to get event-by-event results also. The main issue is whether we use Fluka and adapt the existing digitisation (charge spread, threshold, noise, etc) code to interface to Fluka output, or whether we convert the output to LCIO and then use the digitisation directly, or whether we use Mokka. The control of the low energy photon cutoff is more obvious in Fluka than Mokka so we will have more ability to study this if it is an issue. Also, setting up geometries in Fluka seems more straightforward than in Mokka, although this may be just due to the experience of the people involved. The simulation will need to be run through the DAQ program to make output in the same format as the real data and this could be written to input any reasonable format (LCIO, Fluka or flat files). The digitisation will need to be part of this procedure, so this is probably not a limit. Nigel and Anne-Marie should consider the implications of the various options and report back at the next meeting. Marcel reported on SiD benchmarking. The area of tau identification using the ECAL, where fine granularity should help distinguish pi from rho, is not yet covered in SiD. The two Imperial students, Jamie and Hakan (who is not working in MAPS) are gearing up to work on this. Marcel has put some instructions on running LCSIM on the SiD-UK web pages; see link from the meetings web page. It is similar in concept to Marlin in terms of processors to handle events, etc. Nigel reported that he had used the GLDC ILC installation software for CondDB, MySQL, etc, and it seemed to work quite straightforwardly. Conferences: Giulio has a 15 min MAPS talk at TWEPP, Prague, Sep 3-8. We should try to have at least a noise response from the sensor by then and maybe even a laser pulse if possible. Giulio will give a practise run-though of his talk at the next meeting. The US region ILC meeting (ALCPG) will be at FNAL in Oct, the week before the IEEE. We should aim to give a MAPS talk at this meeting if someone is available to attend. Next meeting: Fri 31 Aug at 1pm in a PPD meeting room.