CALICE MAPS Meeting, RAL, 19/06/07
==================================
Present: Jamie Ballin, Jamie Crooks, Paul Dauncey, Yoshi Mikami,
 Owen Miller, Marcel Stanitzki, Renato Turchetta, Mike Tyndel,
 Giulio Villani, Nigel Watson, John Wilson


Minutes: Paul


Sensor design: JamieC reported that the sensor will be shipped from the
 foundry on Wed 27 June and so should be at RAL in the first week of
 July. This will be before the readout PCBs are complete but there is
 little that can be done with the sensor in advance.

 JamieC is updating the sensor documentation and will put everything
 he produces on the "sensor1" web page. Currently, only some of the
 documents on this page are protected; this should be changed to
 protect all, with the exception of the pictures for talks. The
 documentation from the reviews also needs to be updated. 

 JamieB reported he had seen the foundry name in an unprotected pdf file
 somewhere on the web and this should be corrected if it is refound.
 Renato will contact the foundry to find out when we will be allowed to
 publically announce who fabricated the sensor.


Sensor simulation: Giulio showed results from the laser setup and the
 simulation; see his talk on the usual web page. Using an analogue test
 sensor in the laser, he has been able to calibrate the laser against
 a MIP. This might not be identical to our sensor due to backscattering
 from the metal layers but should be a good first approximation. He also
 gets a noise of 10-20% of a MIP which should be easily sufficient. The
 test plan is to understand the test structure pixel (which is a sampler
 type) using analogue readout and then compare to a standard sampler
 pixel in the sensor bulk with digital readout. This can then be
 compared with the shaper pixels with digital readout; there is no test
 structure of the shaper type. The test structure pixels have identical
 metal layers to the bulk pixels, so any backscattering effects should
 be the same. He wants to study charge diffusion and collection times;
 the laser pulse is nominally 4ns and can be focussed to 2mu. The test
 structures can be operated with only a small number of signals being
 functional. Specifically, they require the biases to be set and the
 reset signal to be operable. The biases are set by DACs on Vladimir's
 board using an SPI bus so this will need to be programmed. The reset
 is a single LVDS line. These should be able to be set using an NI
 board which PPD have and so they are likely to be able to look at the
 test structure without using the USB_DAQ, in case of delays. Jamie
 may share this system with PPD if necessary.

 Giulio has also run the simulation with the full detail of the layout
 from the GDS files. He turned on all checks for the first run which
 then took 47 hours to do one point. This should go down to around 24-36
 hours when operating in the usual mode. The main thing here is that it
 is proof that such large simulations can be done.


Sensor testing: Each test system will need a PC and power supply. For
 the latter, it would be useful to buy supplies which will be capable
 of being used for the second sensor round. The supply needed is clearly
 not well specified but could be approximated to six times the present
 sensor. However the LVDS drivers take around half the current in the 
 present system and there will be less of these in the next round.
 RAL Microelectronics and PPD don't need a power supply yet; Birmingham
 (and possibly Imperial) do, so Giulio and JamieC will distribute their
 suggestions on which to buy by email.

 For PCs, then RAL has a supply deal with Dell and hence it may be a
 cheap option to order all PCs through RAL. In principle, Marcel would
 be prepared to set them all up with Linux. However, it is not clear
 if this would be permissible when they are connected to the networks
 at Imperial and Birmingham. Hence, it may be necessary for the local
 system administrators to do the installation. The software should not
 depend on a particular compiler or Linux flavour. We should compare
 prices for a reasonable PC (~<1k), preferably with dual core, bought
 through RAL/PPD or the Universities. For transportability, Birmingham
 may want to consider buying a laptop rather than a desktop PC for use
 at the beam tests.

 There need to be a hardware and software interface between the laser
 and the DAQ. The hardware interface was previously specified as TTL.
 Matt Noy has asked if this should be converted to differential, 
 specifically LVDS, to avoid problems with pickup from the laser, which
 were previously flagged up as being a major issue. Matt needs to know
 whether to build a TTL converter onto his interface PCB and so this has
 to be decided by next Monday. Giulio and Marcel will check the signal
 quality this week and email Matt by Monday at the latest. The software
 interface is the exchange of run and configuration data from the DAQ
 to the LabView laser control; see Marcel's talk below.

 The source, cosmics and beam tests will need scintillators for timing.
 Matt Noy's interface board will include two input for PMT signals
 direct to discriminators built onboard, with the threshold set by
 DACs. The status of these inputs will be recorded on a 50 or 100MHz
 clock, giving the time of hits to 10 or 20ns. Both Birmingham and
 Imperial have HV units in hand.

 Nigel showed some slides of the mechanical design for the cosmics and
 beam test sensor stack; see the usual web page. They have a stay-clear
 for the wire bonds of 4mm although JamieC estimates 2mm would be
 sufficient. It would be good to check the stack in both orientations
 to the beam, i.e. epitaxial layers both upstream and downstream, which
 can be done by flipping the stack over. However, to keep it symmetric,
 this will require a hole to be cut out of the base support to match
 the one on the top; this will be done.

 The beam test will also need tungsten plates to allow the sensors to
 be at least up to shower maximum. This means up to 10X0 are needed,
 and since tungsten has X0 = 3.5mm, then this is a 3.5cm stack total.
 Three plates of 0.5cm, 1cm and 2cm thickness would be ideal as this
 would allow 0.5cm steps up to 3.5cm. Tungsten is roughly $75/kg and has
 a density of 20g/cm3. With three such plates of 10x10cm2 surface area,
 then this is a total of 10x10x3.5cm3 = 350cm3 or 7kg and hence could
 be ~$500 and so affordable. Paul will check if we can borrow tungsten
 sheets from the French groups.

 Paul reported on the DAQ software system, which has been developed
 using the present CALICE beam test system; see slides on the usual
 web page. Suggestions for new run types should be sent to him. Marcel
 commented that flat files were prefered to a database for configuration
 and this should be implemented. There will be an issue with keeping the
 code synchronised as it is also still in use for the main CALICE DAQ.
 This means we have to be very careful about committing changes to the
 central code storage; Paul has been the only person to do this so far
 and this may be the best way to proceed for now. There is a local area
 for changed code which allows development without overwriting the
 original files and this should be used. At the end of the meeting,
 Paul showed a quick demonstration of the DAQ system running with a
 simulation of the sensor on his laptop.

 Marcel reported that a basic socket interface from Linx to the LabView
 laser control had been established; see slides on usual web page. The
 PPD group have continued their studies with the Vanilla sensor to debug
 the laser and can see significant charge spread.


Physics Analysis: Marcel also reported that he has made progress with
 PandoraPFA but is still not able to get sensible results using the
 output of Anne-Marie's digitisation. He also suggested a different
 digitisation algorithm which could speed up the processing. However,
 this involves several approximations where it is hard to estimate the
 size of the neglected effects. Hence, it will be necessary to run the
 full digitisation to compare with any approximation and hence see how
 they compare before the approximation could be trusted.

 Marcel has also started looking at pi0s, as this was a study suggested
 during the talks with SiD at LCWS. He can reconstruct single pi0s and
 will generalise to Z events.

 Owen showed some transparencies of his beam interaction background
 studies; see usual web page. He intends to parameterise the energy
 and position of the hits so as to be able to generate background
 quickly in the full simulation, without going through the large CPU
 needed to generate them from the machine simulation each time. The
 background will be detector dependent and so will need to be done
 for LDC and SiD separately. The main thing to compare is an analysis
 with and without background, so as to get a measure of its influence.
 Varying the background for the difference machine models will then
 only need to be done if the effect is significant.

 Yoshi also showed his work on simulation; see usual web page. He has
 discovered the cause of the apparent material discrepancy between MAPS
 and the standard ECAL simulation, so this is no longer an issue. He now
 finds the standard and MAPS ECALs have effectively the same resolution,
 presumably limited by the physical shower fluctuations and the ECAL
 sampling geometry. He also finds the "missing energy" problem is not
 dependent on the rangeCut parameter

 Paul has circulated some questions related to the simulation before the
 meeting. One concerned modelling a potential soft photon background
 which might not be visible in the present GEANT4 simulation as the
 photon cutoff is too high; specifically, tracking photons down to 1keV
 would be needed. The other issue is which simulation to use for the
 sensor tests; in particular the beam test, although the source test
 may need simulation to relate the signal size seen to a MIP deposit.
 Nigel thought the best approach might be to use Fluka as it can be
 run at quite a basic level and so gives the user control over many of
 the parameters, including the cut-offs. EGS would be another
 possibility but there is no experience of this within the group. It was
 thought that Mokka would be too unwieldy for our purposes.

 The OsC has requested a study of the effects of doubling the pixel size
 on the clustering and non-linearity. Results are needed for the OsC
 meeting on Tue 3 July. As we do not have the diffusion values for
 larger pixels, these studies will need to be done using just
 SimCalorimeterHits. Anne-Marie is presumably best placed to do the
 cluster study and Paul will follow up on this with her next week. Yoshi
 volunteered to do the non-linearity study, although he will need some
 of Anne-Marie's code to do this. The OsC also asked about remaining
 uncertainties in cost. The beam test PCB, for which a total of 60k
 has been allocated, is probably the biggest cost uncertainty (apart
 from having to do an additional sensor fabrication, of course). The
 bump bonding of this PCB has a lot of unknowns and is probably the
 biggest part of the PCB cost uncertainty.

 JamieB showed some studies using PandoraPFA; see usual web page. He
 finds the results are quite dependent on the calibration but that he
 does see a good Higgs signal. There are lots of small effects related
 to barrel/endcap differences. The next step is to redo the whole study
 with MAPS.


Reviews: The next review is of Vladimir's PCB layout on Tue 26 Jun. For
 this, he should provide the updated schematics and the individual
 layer layouts, in electronic form if possible. It is worth having the
 review, even if the layout is not completely finished. Note, Vladimir
 is due to return to Belgrade at the end of July, but Birmingham are
 trying to get an extension for another two months.

 JamieC pointed out that Vladimir's PCB only has mounting holes at the
 sensor end, but that the cable end will also need to be supported when
 used in a single board setup. Vladimir will need to add at least two
 more holes at the opposite end of the PCB. We will then need a support
 board to protect the sensor on the underside. This could be made
 quickly at Imperial when the hole positions are finalised.

 Since scheduling this review, it became clear we would need an adapter
 board for Matt Noy's USB_DAQ board and Matt has arranged for this to be
 laid out while he is away. Hence, if possible, we should review the
 adapter board on the same day. Paul will check with Matt when he
 returns from holiday. Procurement for both boards will be done through
 Imperial.


Conferences: Marcel reported that Nigel will give the MAPS talk at EPS.
 Otherwise there is little news.


AOB: Paul raised the issue of the coordinate system. The row and column
 numbering is fixed by the readout memory bit settings. Following those
 would give a non-standard xy coordinate system when viewing the sensor
 from above. Hence, we will work with y in the opposite direction to the
 row numbering and x in the same direction as the column numbering. Paul
 will make a diagram giving the numbering and coordinate systems; this
 will be added to the meeting web page. It may be useful for Vladimir's
 PCB to have the coordinate system visibly marked on the PCB near the
 sensor area to avoid confusion.


Next meetings: The next meeting is the review of Vladimir's sensor PCB
 layout on Tue 26 Jun. This will be in R76, starting at 1pm. Depending
 on Matt Noy's availability, we may start earlier if the USB_DAQ adapter
 board is to be reviewed on the same day.

 The next regular meeting will be Mon 16 July at 1pm. Marcel will find a
 conference room in PPD.