Notes on MAPS-DAQ meeting, RAL, 03/05/07
========================================
Present: Jamie Crooks, Paul Dauncey, Matt Noy, Vladimir Rajovic, Marcel
 Stanitzki


Schedule: Working from a completion date of mid-July, then the following
 schedule was drawn up for the sensor PCB:
  o Schematics review - Thu 17 May (two weeks)
  o Layout review - Tue 26 June (five weeks)
  o PCB fabrication start - 3/4 way through June (one week)
  o Assembled PCB ready - 1/2 way through July (three weeks)
 Clearly with some extra cost, we could speed up the last step if
 required but this decision can be made at the time the PCB is
 submitted.

 We need to make around 15 PCBs but will populate only a few initially.
 The cost was estimated to be around 100 pounds/PCB.

 We will need around 10 of the USB_DAQ boards also, together with some
 PCs to run the DAQ systems. The PCs do not need to be particularly high
 specification and most groups will already have one available; the RAL
 Microelectronics group may be the exception as the PCs will need to be
 running Linux.


Power: The sensor PCB board will only need 3.3V and will regulate this
 down to 1.8V locally. For all uses, the power will come in on a
 (suggested lemo) connector and so could be supplied from the USB_DAQ
 board or directly from an external power supply. This should help to
 avoid ground loops.


Signal connector: There are 61 digital inputs to the sensor and 51
 outputs. Very few of the input signals could be hardwired locally so
 effectively all these signals need to be sent from the USB_DAQ. There
 are also 4 serial control inputs which do not go to the sensor (to set
 the DACs, etc) making a total of 65 input signals. Of the outputs, 5
 are debug signals and are not required to be sent off the sensor PCB,
 leaving 46 required outputs to the connector. This is a total of 111
 differential LVDS signals, or 222 pins total on the connector. 

 Edge connectors would be convenient but finding a connector (or two)
 which can handle this large number of pins will not be trivial. Jamie
 showed a Samtec connector which might be feasible, although it was
 expensive at ~1k per cable. Also, the particular cable used is limited
 to 1m lengths, while it was thought we would need cables up to 2m long.

 The USB_DAQ can effectively route any signal to or from any pin so the
 cable signal layout should be defined by Vladimir to be the most
 convenient for the sensor PCB layout.

 The sensor PCB should keep the sensor in a safe state if the USB_DAQ
 cable is missing. The input signal state is defined by the LVDS
 converter and the chosen component sets the outputs high if the inputs
 are disconnected. This would be incorrect for at least the enables and
 so inverters will be needed for around six signals. This will mean for
 these cases that the signals sent from the USB_DAQ will need to have
 opposite polarity to the ones on the sensor.


Cosmics: The cosmics (and beam test) stack of four sensors should have
 the sensor spaced as close together as possible. In reality, 1cm
 spacing would be good although up to 2cm would be acceptable. This
 could be obtained by having a rectangular PCB with the four being used
 rotated by 90 degrees. The large components (connectors and power
 regulators) could be mounted only in non-overlapping regions on the
 PCBs.

 Vladimir suggested multiplexing the signals from each of the four PCBs
 into one set of the 111 signals. There would need to be two further
 signals to set the board address, but otherwise this was thought to be
 workable. The motivation would be to save on the number of USB_DAQ
 boards needed for the cosmics/beam test system; multiplexing would then
 require just a single USB_DAQ board. However, the firmware and software
 to run such a system might be significantly different from that for a
 single sensor, as for the other test systems. The overhead of debugging
 another system as well as the added complexity to the sensor PCB (which
 is needed on a short timescale) were thought to outweigh the cost of
 using four USB_DAQs. Hence, a simple, non-multiplexing PCB was chosen.


Sensor mounting: The PCB assembly at the manufacturers will only be for
 the standard components, not the sensor. The gluing and wire-bonding of
 the sensor will be done at RAL, although Imperial could take some of
 the wire-bonding on if the RAL technicians are not available.

 The same PCB will be used for tests with bump-bonded sensors. This will
 require the sensor to have an identical orientation, but be mounted on
 the opposite side of the PCB. This in turn requires the wire-bonding
 pad signals to be passed by vias through the PCB and connected to
 bump-bonding pads on the other side. The vias will need some minimum
 spacing larger than the pad separation and so a fan-out and fan-in will
 be needed.

 The sensor will be glued to the PCB with conductive epoxy. The
 substrate connection to ground will then need to be made using a jumper
 so that it can be tried either grounded or floating. An area 1mm wide
 around the sensor edge should be sufficient for gluing and this will
 need a metal contact for the glue. This means a square hole of 8x8mm2
 will be needed in the PCB. The test structure pixels will need a notch
 cut in addition to the regular square hole so that light can be shone
 onto the test pixel area. This will bring the hole very close to the
 bonding pads (particularly for the bump-bonding option) and the extra
 part of this hole needs to be defined with care. 

 The sensor does not have to be mounted centrally on the PCB. If a cross
 shape is to be made of the four PCBs for the cosmics and beam test,
 then clearly, the mounting holes have to be symmetric around the sensor
 position.

 The sensor PCB will often be used with light illuminating the substrate
 side. It will also need its components probed and so these two
 activities should be able to be done without moving the PCB. This will
 require the sensor to be mounted on the "underside" of the PCB, i.e.
 the opposite side to the components (assuming it is single-sided).


Miscellaneous: There should be no LEDs on the PCB sensor, as they would
 need to be disabled during normal use anyway.
 
 There should be plenty of probe points for scope work and in particular
 a good hook for a ground connection is needed. It would be useful to
 have a logical analyser connector on the PCB. This would need to handle
 around 30 signals to be maximally useful. Marcel will consider which
 signals would be most useful on such a connector.


Notes from Marcel added after the meeting:

a) The analog outputs (mainly for the test structures), will they be
buffered? Giulio was also concerned about noise problems here. Giulio
will follow up on this as well

b) For a potential testbeam, the area around the sensor should be free
of any active components, the more the better, since this board might
end up in a test beam. A rough number of 4 cm clearance around the
sensor was estimated (easier if the sensor is located at one edge)

c) For the upcoming measurements, it would be great to have the power
lines with a jumper, so you can measure currents as well.

d) To protect the chip and the bonds, a chip carrier package was
suggested