Title: Overview of ATLAS Liquid Argon Frontend Crate Electronics John Parsons Nevis Labs, Columbia Universi
1Overview ofATLAS Liquid ArgonFrontend Crate
ElectronicsJohn ParsonsNevis Labs, Columbia
UniversityLAr Electronics ASSO Review, June 17
2002
ATLAS
2Overview of ATLAS Liquid Argon (LAr) Calorimeter
Readout
3Requirements of ATLAS LAr Frontend Crate
Electronics
- read out ? 170k channels of calorimeter
- dynamic range ? 16 bits
- measure signals at bunch crossing frequency of 40
MHz (ie. every 25 ns) - store signals during L1 trigger latency of up to
2.5 ?s (100 bunch crossings) - digitize and read out 5 samples/channel at a max.
L1 rate of 100 kHz - measure deposited energies with resolution lt
0.25 - measure times of energy depositions with
resolution ltlt 25 ns - high density (128 channels per board)
- low power (? 0.8 W/channel)
- high reliability over expected lifetime of gt 10
years - must tolerate expected radiation levels (10 yrs
LHC, no safety factors) of - TID 5 kRad
- NIEL 1.6E12 n/cm2 (1 MeV eq.)
- SEU 7.7E11 h/cm2 (gt 20 MeV)
4ATLAS LAr Frontend Crate Electronics Overview
- On-detector electronics
- Boards tested functionally on Mod 0
- ATLAS rad-tol boards being finalized
Controller 116 boards
Calibration 116 boards _at_ 128 ch
Front End Board (FEB) 1524 boards _at_ 128 ch
Tower builder (TBB) 120 boards _at_ 32 ch
5Module 0 Electronics Experience
- Full functionality Module 0 CALIB and FEB
boards were developed - Provided verification of the electronics design
concepts - Used in testbeam runs with Module 0 and
production calorimeter modules - Several CALIB boards and ? 6000 channels of FEBs
were produced - Have been operating reliably in testbeam for
past several years - Performance meets or exceed ATLAS specifications
- Prototype Controller board was evaluated in
dedicated test - Due to schedule, Mod0 boards developed with some
short cuts - Did not use final control signal (TTC and SPAC)
distribution - FEB used Cu output cables instead of optical
links - Did not pay strict attention to radiation
tolerance requirements - the main task remaining in the development of the
final ATLAS boards was to radiation harden the
designs, and in particular to replace several
FPGAs and other COTs with custom rad-tol ICs
6On-detector electronics control interfaces
- LAr on-detector electronics uses common control
signal distribution - TTC for 40 MHz CLK, L1Accept and other fast
signals - SPAC serial protocol for downloading/reading
back configuration parameters - Controller Board (one per 1/2 crate) is connected
optically to external TTC system and SPAC Master,
and fans signals out electrically to each of the
boards within its 1/2 crate - Redundancy has been implemented to increase
reliability, with no logic switching components
required on-detector - TTC and SPAC Master ? Slave (MS) signals sent on
two separate fibers each, with off-detector
selection of which will be active - On-detector electronics configured to drive both
of the SPAC Slave ? Master (SM) links - SPAC MS and SM redundancy preserved on SPAC bus
(though Controller is common and must therefore
be VERY reliable)
7On-detector electronics control interfaces
(contd)
- TTC
- TTC fanount within crate is point-to-point over
halogen-free USB2.0 mini-B cables connected
through the boards front panels - Each board has TTCRx chip to receive 80 MHz
encoded TTC signals - SPAC
- Manchester-encoded serial protocol combining CLK
and data with 20 MHz edges - Implementation uses separate (ie.
uni-directional) MS and SM lines - SPAC signals are fanned out via SPAC bus,
implemented as a PCB along the side of the crate
with through comb connectors to boards - Each board has SPAC Slave to receive/transmit
SPAC data using either 8-bit parallel interface
or one of two I2C interfaces provided by SPAC
Slave - Final SPAC Slave version includes separate
address and data partial checksums for better
fault and SEU tolerance - Some concerns
- Cabling of TTC cables on crate, while still
allowing extraction of single boards - Robustness/fault tolerance of SPAC bus (will be
tested in system crate test) - Robustness/fault tolerance of Controller
8Controller Board Overview (LPNHE-Paris)
9Connectors on Controller
Up to 22 TTC connectors (by pairs on the 2 faces)
Optical Part for TTC and SPAC
TTC (USB mini-B) connector pinout
1 GND 2 (D-)TTC- 3 (D)TTC 4 not
connected 5 GND
shield connected to GND
10Controller Board Status
- Test board produced to evaluate fanout issues
- Design of full prototype board underway delivery
to BNL by September - Radiation tests of SPAC Slave V3 and optical
receiver by end June
11Controller Board Responsibility Chart
12Calibration Board Overview (Annecy, Mainz,
Orsay)
- Generate 0.1 precision calibration pulses
- Rise time lt 1 ns
- Current pulse amplitude from 200 nA up to 10 mA
- Delay programmable from 0 to 24 ns in 1 ns steps
- Number of current pulsers per CALIB board is 128
13Overview of Main CALIB Components
DMILL
AMS
COTS
Enable
Spac
VDAC
6 CALogic
128 opamp 10 µV offset
1 DAC 16 bits
1 SPAC
128 Output signals
IDAC
TTC
CMD
16 driver
128 HF switch
1 TTCRx
2 delay
4 pos. Vreg and 1 (non-essential?) neg. Vreg
14CALIB Control Interfaces
- SPAC slave used only as I2C master, with one I2C
link used for TTCRx and Delay chips, and the
other I2C link used for all the other chips
15Digital CALIB Testboard
- Developed to test integration of digital ASICs on
CALIB board - Tests with TTCRx, SPAC Slave, 2 Delay chips, 6
CALogic all OK
168 Channel CALIB Prototype
- Include digital control plus analog chain for 8
channels - 3 boards received in April 02
- Design of full-sized 128 channel board is
underway delivery to BNL by Oct.
17CALIB Responsibility Chart
18Frontend Board Overview (Nevis)
- functionality includes
- receive input signals from calorimeter
- amplify and shape them
- store signals in analog form while awaiting L1
trigger - digitize signals for triggered events
- transmit output data bit-serially over optical
link off detector - provide analog sums to L1 trigger sum tree
19Overview of main FEB components
DMILL
AMS
DSM
COTS
- 10 different custom rad-tol ASICs, relatively few
COTs
128 input signals
32 SCA
16 ADC
8 GainSel
1 MUX
32 Shaper
32 0T
1 fiber to ROD
Analog sums to TBB
1 GLink
1 Config.
2 SCAC
2 LSB
TTC, SPAC signals
14 pos. Vregs 6 neg. Vregs
1 TTCRx
7 CLKFO
2 DCU
1 SPAC
20FEB Analog Interfaces
- Input signals
- 128 signals per FEB, using the two outer rows of
the two 96-pin VME-style DIN connectors - Middle row of 32 pins of each signal connector
are shared Ground pins (ie. one GND pin per 2
signal pins) - Signal connectors equipped with custom shields
and additional grounding contacts, with mating
springs mounted on crate baseplane - Mod 0 demonstrated this is necessary to achieve
coherent noise specification - Output analog trigger sums
- First level of summing provided by shaper chips,
configured via switches to enable/disable
individual channels - Next level of summing provided by plug-in Layer
Sum Boards (LSB), the output stages of which
drive the sums through the middle FEB connector
to the baseplane, where they are routed to the
TBB (or TDB for HEC) - LSB sums depend on calorimeter region, with up to
32 separate O/P sums - Remaining 64 pins are GND pins, with GNDs
separating adjacent signals
21FEB Control Interfaces
Left Gain selector 0
..
Shaper 0
Shaper 1
Shaper 31
Left Gain selector 1
TTCRx
23
Left Gain selector 2
5 MHz
I2C_0
Left Gain selector 3
4
Configuration controller
Right Gain selector 0
data70
subadd60
4
SPAC slave
n_read
n_write
Right Gain selector 1
Right Gain selector 2
I2C_1
8
8
4
Right Gain selector 3
4
Left SCA controller
Right SCA controller
Regulator 19
Regulator 0
DCU 1
DCU 0
22FEB Optical Links
- one GLink output link per FEB, with rate of 1.6
Gbps - link collaboration includes SMU, ISN,
- Stockholm, and Taiwan
- FEB integration complete (ROD integ. underway)
1.6 Gb/s
231/4 Digital FEB
- Test setup for FEB digital ASICs
- Complete operation of SCAC, GainSel, Config,
SPAC, TTCRx, MUX, GLink - Allowed quick testing/integration of new ASICs as
they became available - Integration was carried over to FEB design, which
worked properly first time
DMILL TTCRx
Optical Tx
Agilent GLink
DMILL SPAC
DMILL MUX
DMILL Config controller
DSM/DMILL Gain selector
DSM/DMILL SCA Controller
24FEB Prototype
- 128 channels/FEB
- large 10 layer PCB
- components on both
- sides to achieve density
- layout, with STm
- Vregs, done Oct. 01
- first FEB assembled
- and successfully
- tested Nov. 01
- pos. STm Vregs
- added Dec. 01
- 2nd FEB assembled/tested and shipped to Orsay,
Mar. 02 - Need neg. Vregs before launching 20 FEB
preproduction
25FEB Responsibility Chart
26Organizational Issues
- Meeting the specifications, particularly
radiation tolerance, has required development of
a LARGE number of custom ASICs and other devices - Final PCB designs, while the responsibility of
one (sometimes 2) labs, must integrate custom
devices developed by a number of different
institutions - Due to radiation concerns with COTs, custom
devices need to work together without resorting
to glue logic - For this integration to work as well as it has
required careful attention to interfaces during
ASIC design and development - Occasional workshops, including engineers, to
develop and agree on architecture - Many meetings and discussions
- Design reviews for each individual device
- Documentation and agreement on interfaces
- In limited (critical) cases, simulation of ASICs
together (eg. SPAC and Config) - As we move toward production, the complexity of
the distribution of responsibilities will present
us with a new set of challenges
27Moving Toward Production
- Final prototypes of each of the various boards in
the frontend crate are currently being developed
and/or tested individually - However, a system crate test using all the
various boards together is necessary to validate
the designs before launching production - Are boards functionally compatible with each
other? (eg. TTC and SPAC) - Does crate performance meet specifications? (eg.
coherent noise) - Are environmental issues under control? (eg.
power, cooling) - Test will be performed at BNL, with first boards
delivered for Sept. 02 - Some concerns
- Coupling of schedules increases schedule risk
- Continued problems with radtol negative Vreg are
causing a significant delay - Since we cannot wait too long to produce the
radtol ASICs, most of the components will be
submitted for production before completion of the
system crate test
28Component Production
- Lab which developed a given component (eg. ASIC
or plug-in) is responsible for the production and
delivery of that component - Separate PRRs to be held for all major custom
components - So far, have been held for preamps, HEC
preshapers, shapers, LSBs, SCAs - To achieve cost savings, many radtol ASICs will
be produced together on shared wafers - DMILL Analog Reticle opamp, DAC, BiMUX
- DMILL Digital Reticle SPAC, CALogic, Config,
MUX - DSM Digital Reticle SCA Controller,
GainSelector, CLKFO - PRRs for the ASICs on these shared wafers are
expected by end 2002 - Some concerns
- Sharing of ASIC wafers for production increases
schedule risk - Careful tracking required during ASIC packaging
and delivery
29Component Quality Assurance and Tracking
- QA plan developed and presented as part of PRR
to be archived in EDMS - For LSB example, see B. Clelands Trigger Sums
presentation - For plug-in assemblies (preamps, preshapers,
LSBs) this includes - Initial inspection
- Pre-test
- Burn-in at elevated temperature
- Re-test
- Custom ASICs will be individually tested, but not
burned in - Serial numbers are to be printed on the
components - Results of component testing and QA procedures
are to be entered in database and tracked via
serial number
30Example of Component Tracking
- BNL document for following preamp acceptance
procedures
31Example of Component Quality Results
(Final sample of 50000 channels 50W/1 mA preamps
from INFN-Milano)
Gain
Peaking time (5-100)
Equivalent Noise Current
Acceptance limits
32Component Qualification and Testing
- In most cases, lab which developed a given
component (eg. ASIC or plug-in) is responsible
for radiation qualification and - for functional testing of production parts
- prior to delivery for assembly onto PCBs
- For shapers and SCAs, where gt 50000 chips
- needed, ISN Grenoble developed robotic tester
- Shaper testing completed (last week)
- SCA testing to start soon
- Remaining ASICS must be tested manually
- Some concerns
- PCB complexity makes re-work difficult, requiring
excellent pre-testing and rejection of faulty
parts - Achieving the necessary very high quality of
automated PCB assembly process requires no damage
to pins - Most of the critical ASIC testing is being done
long before the start of PCB production, meaning
there is no chance for feedback to the testing
procedures
33PCB Design/Testing Issues
- Each PCB includes many custom components
developed by other labs - Responsibility charts have been created for all
PCBs, identifying individuals responsible for
delivery of each of the major components (see
earlier slides) - Steps to ensure successful design and integration
- Usually, dedicated jigs developed first to test
integration of custom ASICs - All final schematics made available via WWW
- Relevant contact person asked to verify schematic
related to their ASIC - Board design documentation and reviews (eg. later
this week) - System crate test before launching board
production - All final PCBs will be fully tested before
delivery to CERN - Test procedures still being developed
- Take FEB as example, since it is by far the PCB
of largest quantity - Nevis and LAL are working on developing and
defining the final FEB test procedure (see next
slide)
34FEB Testing Procedures
- FEB test steps as currently envisioned include
- Receipt and visual inspection of assembled PCBs
at Nevis - Boards will be fully assembled in industry, but
NOT including preamp/LSB plug-ins - Labelling of FEBs with serial numbers
- Initial test without preamps/LSB
- Full testing/debugging of digital part
- Partial testing/debugging of analog part (from
shaper outputs) - Limited burn-in (eg. overnight? room
temperature?) - Does not appear practical to perform a long
burn-in on the full set of FEBs - Re-test
- In parallel, full testing (ie. with preamps/LSBs)
at Nevis of sample (5-10) of FEBs to allow quick
feedback to Assembler - Shipment to LAL of FEBs which passed Nevis tests
- Installation of preamps and LSBs of appropriate
combinations - Full testing/debugging/calibration of analog part
- Shipment to CERN of FEBs which passed LAL tests
- Installation and testing of cooling plates
- Functional re-test
- Ready for installation/commissioning in the pit
35Radiation Tolerant Voltage Regulators
- CERN-driven development of rad-tol Vregs by STm
has long been identified as a critical item,
particularly for the FEB development - Each FEB requires 14 positive Vregs and 6
negative Vregs - As one backup possibility, we asked MDI in early
2001 to consider an SBIR to develop a hybrid
radtol Vreg they did so, but it was NOT approved
by DoE - We irradiated with protons samples of 9
different commercial Vregs, including several
advertised as radtol (but only wrt ionizing
radiation) - Only the Intersil HS9S-117RH positive Vreg
survived as a backup (? 75 unit cost) - We worked on designing our own radtol Vreg with
COTs, but with little in the way of encouraging
results - STm delivered first positive Vreg prototypes in
Dec. 01, which we have since then been using
successfully on the FEB ? ve Vreg issue would
appear to be resolved - In May 02, STm announced they were still having
stability problems with the neg. Vreg design, and
an additional design iteration was required - STm has not yet announced a new schedule for the
negative radtol Vreg - Past experience would imply a delay of order 6
months (assuming no further problems are
encountered!)
36Radiation Tolerant Voltage Regulators (contd)
- Intersil has recently announced a new neg.
Vregulator, radtol to 300 kRad - We acquired a few samples (85 each)
- On May 24, we irradiated 2 samples at HCL ?
results look encouraging - STm remains the most attractive solution
(assuming availability) - Cheaper (10 unit cost versus ? 50 - 70??, about
10k devices needed for FEBs) - Higher current (3 Amp versus 1 Amp)
- Better voltage range (we need 1.7 V, while
Intersil rated for lt 2.5 V) - Same control logic as STm positive Vreg
- Next steps
- Complete evaluation of Intersil devices
- Further statistics needed for irradiation
qualification ( HCL no longer available!) - Produce jig to interface to allow FEB electrical
measurements - Discuss procurement/financial aspects with
Intersil - Keep pestering STm for updated delivery schedule
for their development - In September, re-visit situation and decide how
to proceed - Dates of PRRs will be fixed in September once
this issue is resolved - (given the Vreg delay, FEB PRR will not be
before Summer 2003)
37Summary
- Building on the success of the Module 0
electronics development, we are finalizing the
designs of the final ATLAS LAr readout boards - The system crate test is the remaining major
milestone before moving to production - Prototypes of the large number of required custom
ASICs have been successfully developed and
integrated - Most devices will move into production before the
end of 2002 - Continued problems with STm negative radtol Vreg
are now dominating schedule - We are aggressively pursuing a backup, and
working to minimize delay - As we move toward production, we are working to
formalize the documentation, QA and tracking
procedures that will be required to sucessfully
manage the complex production process