Title: The deliverable: THE ALICE PHOS FRONT END ELECTRONICS moving data from PHOS detector to ALICE DAQ sy
1The deliverable THE ALICE PHOS FRONT END
ELECTRONICS- moving data from PHOS detector to
ALICE DAQ system -
PHOS FEE project, Beijing and Wuhan, Nov. 2002
http//www.fys.uio.no/elg/alice
B. Skaali, t.b.skaali_at_fys.uio.no
2 The PHOS detector in ALICE Status for PHOS
Physics requirements for PHOS FrontEnd
Electronics (FEE) General organization of the
PHOS FEE Timing synchronization with LHC and
ALICE, ALICE Trigger Data flow from Very
FrontEnd to Data acquisition FEE for the RHIC
PHENIX electromagnetic calorimeter Quick look
at US proposal for the PHOS (and EMCAL) FEE
Design cycle Integrating the FEE and Detector
Control Project road map
PHOS FEE project, Beijing and Wuhan, Nov. 2002
http//www.fys.uio.no/elg/alice
B. Skaali, t.b.skaali_at_fys.uio.no
3The ALICE detector
PHOS
4The ALICE detector longitudial view
5The PHOS detector
Courtesy Dmitry Budnikov, RFNC-VNIIEF, Sarov
(Sep. 2002)
6PHOS status March 2002
7Physics requirements for FEE - summary
8The PHOS FrontEnd Electronics (FEE)
- RD and construction of PHOS electronics has so
far been focused at preamp/shaper for the
photodiode (PIN, and now APD). A design of the
readout, trigger and TOF systems does not exist - As such, the PHOS FEE project is late, and is on
the watch list of the ALICE management - Time constraint a baseline system (Energy) must
be available when integration and calibration of
the first PHOS module starts in 2004 - Development resources of the European PHOS
partners are not adequate for design and
delivery-in-time - A US proposal for building a separate EMCAL for
ALICE, as well as delivery of the complete PHOS
FEE chain, has been put on hold, earliest release
of money is 2004 - Note also that the US groups have indicated
strong interest in using a non-US PHOS FEE also
for EMCAL! - ALICE management has therefore urged the PHOS
collaboration to adopt electronics and solutions
developed by other ALICE detectors
9General functional requirements for FEE
- Geometry PHOS module mechanics, space/volume,
heat removal capability - PHOS specific requirements
- energy resolution
- also individual/sector gain control of
photodetector via APD HV settings or variable
gain amplifier (?) - dynamic range
- L0 trigger generation ( L1 ROI ?)
- TOF capabilities
- ALICE requirements
- LHC timing and synchronization, interface to
ALICE trigger and the TTC system - Data flow and link interface to DAQ hardware-
and protocol-wise - Interface to ALICE Detector Control System
- voltage setting/monitoring
- cooling, temperature
- channel gain control (?)
10Module geometry for PHOS FEE
- 64 x 56 (3584) channels
- warm volume for FEE underneath the crystal
volume - (horizontal) area 130 x 150 cm²
- total height 32 cm, with 20 cm height (TBD) for
vertically mounted (daughter) cards, each card
serving a strip unit with 8 crystals via a
connector and cabling to the cooled (-25?C)
crystal volume - heat removal capacity 2 kW
- Total (single side) area for FEE (see following
mechanical design proposal) - 448 daughter cards
- plus horizontal area 19500 cm²
- in theory an enormous area available for the
FEE! - however, there are a number of constraints!
PWO crystals, -25 C
warm volume, room temp
11Mechanics
12TTC - Timing Trigger Control
- ALICE Central Trigger Processor (CTP)
- PHOS Local Trigger (VME) Crate
- TTC signals to FEE
- LHC 40.08 MHz clock
- Overall synchronization
- ALICE L1 trigger and L1 trigger word
- ALICE L2 trigger and L2 message
- TOF reference
- Note! At the detector this clock will have a
jitter of some hundred ps, use PLL to clean - Interface TTC ?FEE via the TTCRX chip
13Data flow FEE gt ALICE DAQ
- FEE data sampling and event buffering
- latched and flow control from ALICE trigger
sequence L0 (1.2 ?s), L1 (5.5 ?s),
L2accept/L2reject - single event and multi-event buffer, detector
specific - ALICE L1 trigger data block and LHC bunch and
orbit included in event data structure - Data transport from detector RCU (Readout
Controller Unit) to ALICE DAQ RORC (ReadOut
Receiver Card) over (ALICE specific) Digital Data
Link (DDL) - DDL 100 Mbyte/sec bi-directional optical link
- PHOS one DDL per module (5 modules)
14The ALICE Digital Data Link
- DDL specs for H/W designers of detector FEE
- Source Interface Unit (SIU) must be mounted on
FEE card (Readout Controller Unit), data flow
from frontend buffer to DDL controlled by onboard
firmware - right top proto of Destination Interface Unit
(DIU) on the RORC side. Note dual fibres for
bi-directional traffic. - right bottom SIU proto
- Nov. 2002 new DDL (SIU and DIU) link cards in
CMC form factor now under test!
Courtesy CERN PhotoLab
15Luciano Musa
L0 1.2 ?sec (fixed) L1 5.5 ?sec
(fixed) L2accept lt 100 ?sec L2reject after
L1 For PHOS modules, L0 and L1 buffers are in
on-detector boards, L2 buffer probably on a
single RCU (Readout Controller Unit)
board. F2D-DL is the ALICE Digital Data Link
(DDL) RORC is part of the ALICE DAQ (Data
Acquisition system)
16PHOS Trigger
- PHOS shall supply a L0 trigger to the ALICE
Trigger system - Max delivery latency 800 nsec (the ALICE L0 is
issued after around 1200 nsec). - Traditional approach is to make fast (analog)
sums over a group of crystals and compare the sum
with a threshold. - Sliding summing window technique, illustrated to
the right. The yellow rectangle indicates a
cluster of crystals with output above threshold. - What about clusters that are distributed over
several groups, as illustrated by the green
rectangle?
17The PHENIX Electromagnetic Calorimeter
- PHENIX what can we learn?
- Physics performance Energy, Trigger,
Time-Of-Flight - Electronics system design
- Pb-scintillator calorimeter Pb-glass
calorimeter - read-out by Photomultipliers
- Pb-glass best granularity and energy resolution
- Pb-scintillator best linearity and timing
- Electronics
- Energy variable gain amplifier, gain can be set
remotely, two outputs low and high gain.
Analog stage ASIC. - gain control important for uniform amplitude
response for trigger summation - Fast trigger ASIC summing of 4 analog signals
plus results from neighbouring ASICs - channel TOF TAC started by signal above
threshold, stopped by beam clock, pipeline values
until trigger decision. Analog stage ASIC
The PHENIX Collaboration PHENIX calorimeter
18RHIC PHENIX calorimeter fast trigger
Within each ASIC the 4x4 sum signal is compared
to three separate thresholds, each remotely
programmable.
Courtesy the PHENIX Collaboration
19RHIC PHENIX Pb-scintillator - timing
PMT out risetime ? 5 ns Beam clock 9.4 MHz
Courtesy the PHENIX Collaboration
20The US proposal for PHOS FEE (and ALICE EMCAL)
- Based on the PHENIX system
- Includes a full chain from APDs to ALICE DAQ
- topology a three level system a PHOS module
56 x FrontEndModules
gt 7 x DCM layer1 gt 1 x DCM layer2 - includes some specially designed ASICs
21Designing and building the full PHOS FEE
- Activity so far (since 1997) preamplifier
(shaperADC)energy card for PHOS-256 prototype - Remaining Energy chain, Trigger, TOF and
interface to ALICE systems (DAQ, Trigger, DCS)
to be designed and built - ALICE/PHOS FEE
- US proposal 2001-02 (now on HOLD..)
- Design study June 2002 (Oslo)
- Design study Sep 2002, and Design of FEE for
PHOS, Nov. 2002 (I. Sibiriak) - The above studies are way short of a real design
specification!
Oct. 2003 ALICE milestone final design PHOS
FEE 2004 Start integration and calibration of
first PHOS Module
However, the PHOS FEE project offers challenging
tasks in developing state-of-the-art electronics
in a short time, with close contact to the
international high tech community at CERN!
22PHOS FEE baseline
- Physics requirements main issues previously
listed - ALICE (protocol) interfaces summary on previous
foils - Some interfaces not yet frozen, for instance
Central Trigger processor, Detector Control
System (DCS) - PHOS module geometry and FEE
- main mechanical parameters fixed, but internal
organization of FEE warm volume remains to be
defined - detector module is built of 448 strip unit for
8 crystals with APDs and preamps. Signals/power
connection to each strip unit via a common
connector with cabling into cold volume. - first FEE stage could be one vertically mounted
card per strip unit - a number of horizontally mounted (mother)boards
for processing sectors with say 14x16 (224)
crystals, giving 16 sectors per PHOS module - PHOS preamplifier (Bergen, Kurchatov)
- output dynamic range, signal characteristics
- any modification regarding current output 7V
dynamic range?
23PHOS FEE design baseline I
- ALICE Technical Co-ordinator in order to deliver
in time, re-use electronics / concepts already
developed by/for other ALICE detectors! - Electronic development resources are urgently
required, in particular in digital (programmable)
logic - However, there is also a mechanical engineering
task to be coordinated with the FEE layout - We have to design and build a FEE system, not a
number of individual electronics blocks - Beam tests of prototypes and construction of
test benches - TOF performance
- communication with timing/trigger system via the
TTC system - First milestone March 2003 a preliminary design
(or design options) and time planning has to be
presented by PHOS Project leader V. Manko to the
LHC Committee. - Economy
24PHOS FEE design baseline II
- Off-the-shelf components only, no ASICs
- No time for ASIC development, neither is it
necessary - For maximum flexibility and low cost, use
programmable logic wherever possible - FPGAs offers a wide range of circuit design
facilities - Have Megabits internal memory capacity, replaces
separate memory chips - Processing offers an inherent parallellism
- IP (Intellectual Property) processor and
interfacing cores can be bought (for a price ) - Technology trend increase in speed, capacity,
matched with advanced development tools - Minimize interboard cabling a source for
crosstalk, signal degradation, etc.
25Key system issues
- Event / trigger data to ALICE
- Crystal Energy, TOF value
- Crystal / crystal block gain setting / bias
voltage (TBD) - Crystal sector trigger generation
- PHOS module temperature readings, etc (how
many?) - PHOS L0, possibly L1 ROI (Region Of Interest)
- Needed a redesigned shaper with outputs for
Energy, Trigger and TOF channels - Energy current protodesign with shaping ? 2
microsec for best signal/noise and for ADC
amplitude digitization - sampling digitization shaping time ??
- Very large dynamic range gt multiple gain outputs
(PHENIX have low and high) - Programmable gain for output equalization as an
alternative to individual bias regulation ?? - Trigger and TOF channels fast, the same signal
may be used for both channels - Segmentation for trigger calculation
- dividing the 3584 crystal channels into a number
of individual trigger logic matrices
26PHOS preamp output
Intrinsic rise time with APD capacity ? 10
ns With beam signals the risetime is longer,
around 50 ns. This increase reflects the
scintillation decay times (several components) of
the PbWO4 crystals plus internal multiple
scattering. On the right are shown preamp outputs
after LED pulses into the crystal. These
measurement also show that the rise time is
amplitude dependent. (B. Pommersche, Bergen)
27PHOS Trigger Zhongbao Yin, Univ. of Bergen
28Trigger generation an example
Trigger decision in 300 ns Sector with 16x14
crystals, first stage analog sum 2x2 ref. Oslo
Design Study
29Trigger logic with sampling
Measurements show a rise time of around 40-50 ns
for (fast) signals from current PHOS electronics
(preamp). However, this does not cause a problem
for trigger logic, as time sliding summing window
can be included in the FPGA trigger firmware..
LHC clock is 40 MHz. The beam crossing interval
in p-p is 25 ns, in Pb-Pb 125 ns. All output data
Energy, Trigger, TOF must be referred to a
specific LHC beam crossing .
30Time Of Flight timelines
Note this is an ideal case. What is the time
jitter due to different shower characteristics,
internal multiple light scattering, etc? Beam
test and simulation (difficult) data are needed!
31anti-neutrons
32System design options
- Convential design
- revised shaper with slow (2 ?s shaping) E channel
with peak digitization, plus fast outputs for
trigger / timing - trigger fast digitization and FPGA firmware
- timing TAC, stop by LHC clock, buffer until L0
decision (Sibiriak) - alternative fast FPGA firmware
Nov. 2002 Sibiriak note
- TPC type design
- revised shaper with shorter (200 ns?) E channel
shaping - Energy ALTRO 40 MHz sampling, 2 (or 3) ADC
channels for 2 (or 3) E-gain shaper outputs into
ALTROs 10-bit ADC - TOF timing calculate directly from E signal (?)
- trigger use ALTRO 40 MHz sampling and FEC
onboard FPGA - use a modified TPC FrontEnd Controller with more
channels for larger trigger sector, current
version has 128 chs
33The convential system Oslo Design Study
summary
- Low number (16) of identical motherboards for
14x16 crystals - Motherboard including Energy ADCs mainly contains
a number of off-the-shelf ADC and two medium
sized FPGAs (of 2002 speed-grade) - Buffer memory in FPGA
- Moderate bandwidth requirements
- One of the motherboard with firmware for
multi-event buffering and communication with
Readout Controller Unit. The RCU is probably best
mounted outside the module mechanics for easy
access to ALICE data link - Minimized cabling
- Power budget well below PHOS design specs
34TPC RCU as PHOS FEE back-end
DDL SIU/DIU
First version of a generic card for TPC and HLT,
designed by Heidelberg. It can be configured as a
Readout Controller Unit (RCU) and HLT/DAQ Read
Out Receiver Card (RORC). DDL SIU/DIU is mounted
as mezzanine. For HLT the FPGA (lightly coloured
chip) will also contain the track finder
co-processor. The rest of the card is mainly
memory. To be used as TPC RCU it is complemented
with a mezzanine card (designed by Oslo),
attached through four connectors in the middle of
the PCB. This mezzanine contains the GTL bus
drivers for read-out of the TPC FrontEnd Cards,
PROFIBUS chip for Detector Control, and TTCRX
chip for trigger and timing. The PCI bus
connector is only used for testing and debugging
35TPC RCU board design
The data processing functionalties in the FPGA
are not indicated here. For PHOS that would as a
minimum imply formatting the event to be shipped
to DAQ. The data format over the DLL must conform
to the common ALICE standard.
36TPC RCU (FPGA) Firmware
most of the components also relevant for PHOS
37PHOS FEE and Detector Control System
38PHOS FEE project road map
- Organization
- Main partners Kurchatov and Wuhan
- Challenge a fair sharing of responsibilities
and benefits - Others Bergen, Protvino , Oslo
- Steering committee B. Skaali, V. Manko, I.
Sibiriak, .. (Wuhan), A. Klovning, L. Musa. US
contact T. Awes. - The project requirements are sufficiently well
understood to get a first order estimate of FTE
(Full Time Engineering) years required for 2003
04/05 (definitely not less than 10!) - Road map
- March 03 design options with respective plans to
LHCC - 2003 building and testing of prototype systems
- new shaper
- ALTRO based test bench
- TOF performance (both for ALTRO and Sibiriak
proposal) - Trigger logic (lower priority in startup phase)
- end 2003 (milestone Oct!) full plan, including
mechanical structure and Detector Control System