CLAS12 RICH
Frontend
Electronics
Outlook
 Requirements
 Description of MAROC ASIC
 MAROC Tests and Results
 Front-End Electronics for RICH
M.Turisini (JLab)
P. Musico (INFN-Genova)
E. Cisbani* (INFN-Roma/ISS)
 Conclusions
CLAS12 RICH Review, 2013 September 5-6
* Speaker
Requirements (for PMT readout)
Single photoelectron sensitivity
~50fC
Number of channels per sector
25600
Anodes gain spread compensation
1:4
Event Rate
20 kHz
Dead Time
few%
CLAS12 trigger latency
8 μs
Time resolution
~1
ns
To disentangle direct and reflected photons (can be done off line)
Compactness/ Power consumption
Max area 1 m2, max thickness 10 cm, power consumption compatible with
air flow heat dissipation
Radiation hardness
Must operate in the same irradiation level of the MAPMTs
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
2
ASIC choice
•On-the-shelf components (no brand new development)
•Fulfill the requirements
•Existing expertise in the collaboration





VMM1/FermiLab
CLARO/INFN
APV25/CMS
DREAM/JLAB
MAROC3/LAL
not consolidated , interesting specs
early stage, few channels
not enough latency
CLAS12 Micromegas
ATLAS Luminometer
MAROC3 initially used for detector prototyping
then adopted as baseline solution
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
3
MAROC (Multi Anode Read Out Chip)
Single Channel (x64)
Originally designed for ATLAS
• Preamplifier, configurable (gain 8 bit)
★Fast line: 25 ns shaper + discriminator  Binary datum
★Slow line: 100 ns shaper + mem. cell + ADC  Analog charge
★Prompt Sum line: send to one of 8 embedded sum output channels
•240-pin
•16 mm2
PROs: 64 chs, designed for
MAPMT, existing expertise,
low power
CONs: Analog limited
latency, single channel time
resolution few ns
Outputs: 1 Analog (MUX), 1
digitalized charge, 64 Binary, 8
prompt analog sums
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
MAROC2
4
In House MAROC based DAQ
Original system developed for Medical Imaging with radionuclides
Many optical photons applications
Designed for H9500/H8500 PMTs
4096 channels, USB2.0 readout
Binary outputs for self-trigger
Front End Cards
Backplane
MAPMT H8500
Control Board
Not optimized
for
Single Photon
and/or
external trigger
Scintimammography Detector Head
(Italian National Institute of Health)
Modular design (FrontEnd + Bus + ControlBoard + USB int.)
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
5
Analog Response (Charge info)
Bipolar analog signal offers two optimal
Datasheet
sampling delays corresponding to the two local
maxima:
1. positive, largest dynamics, occurs early
2. negative, additional 100 ns to trigger latency
1
2
Waveforms well reproduced
Turisini – Musico - Cisbani
Dynamic Range Optimized (by ext. bias)
 permit to reduce PMT operating HV
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
6
MAROC Analog Noise
Common noise suppressed
offline significantly reduce
the pedestal RMS
Before
suppression
• COMMON NOISE seems to be the largest
component
• Reduced or negligible noise with short or
no cabling
After
suppression
Amplitude [ADC unit] 
! Noise conditions are site dependent !
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
7
MAROC from Analog to Binary
Mirazita Talk on
Prototype Test
MAROC analog output worked pretty well in RICH prototype test,
but cannot be used in CLAS12 due:
to limited latency (200 ns vs 8 ms required) and relatively slow readout
Single Photo Electron detection does not strictly need analog
readout (e.g. HERMES RICH electronics)

MAROC binary information (64 parallel outputs) can be a
valid alternative
C. Cuevas Talk
on FPGA boards
‣ Binary data latency depends on external logic! Feasible
‣ Stability/sensitivity of threshold to single photoelectron?
‣ Noise in MAROC fast (shaper) binary line ? Measured
‣ Readout Speed ? (not an issue) Estimated
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
Tested
Next
slides
8
Single Photo Electron Level
External trigger / Light Source
Clean SPE identification
with binary data (red)
Analog Timing: First local maximum
GAIN = 1
The smooth cut observed in the analog
distributions can originate from:
noise (both analog and binary combined)
binary line provides better information on
single photoelectron and «sees» good event
GAIN = 4
below analog pedestal
Analog saturated
Analog-binary amplitude mismatch due to
not-optimal relative synchronization between
analog and binary outputs (next slide)
Binary output behaves
as well as analog (or even better)
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
Turisini – Musico - Cisbani
9
Single Photo Electron Level
External trigger / Light Source
Common noise not subtracted
Red: signal in digital line
Binary Line Threshold (100 unit step)
Analog-Binary relative synch (25 ns step)
Charge [ADC unit]
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
10
MAROC “Binary Noise”
Binary pedestal – Open inputs: No PMT connected
Single Channel
“Normal” condition
MAPMT 64 Channels
Single MAPMT/64 chs
Black - #hit/#trigger
Red - Derivative
Mean
Narrow Transition Good!
Channel 
RMS
Single Channel
Large Common Noise
Noisy conditions (for comparison)
 Digital RMS < 1 thr unit (<3mV)
Analog Pedestal RMS 3-5 ADC unit (2-4 mV)
Digital/Analog comparable  see next for SNR
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
11
Dark Noise and SPE
Self-trigger
Dark Counts
(SPE)
SPE region extends over ~200 thr unit
(compare with Ped_RMS <1)
Red: signal in binary line
Charge [ADC unit]
“sharp” Gaussian cut,
no pedestal raising

hint of better SPE
separation than analog
(?)
Rate not corrected for CPU Busy
Turisini – Musico - Cisbani
(likely the fast shaper works in a sort of saturated regime)
Dark rate consistents with MAPMT
specs
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
12
Gain Equalization
Self-trigger capability can be exploited for effective calibration
(gain equalization, threshold selection, MAPMT gain compensation)
Threshold is sligtly
under estimated
Threshold is overestimated
(or gain must be increased)
Dark spectra for two different channels
The small dark count rate (10Hz) is quickly acquired in self trigger mode,
while in external trigger would require either a long acquisition time or an
external light source.
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
13
MAROC Front End - Heat dissipation
• MAROC power consumption
as small as 3.5 mW/ch
• Current implementation at
the level of 0.5 W/card 
total power for 400 cards is
~200 W
• First thermal test with 8 FE +
1 Controller in sealed
container of 20 liters volume
– no cooling.
• Two probe points: contact
with Front End (FE) and
close to the Control Board
(CB) – «hottest positions»
• Small temperature increase
~ 1°C/h in running
conditions (after initial rump
up) ~ 40 °C
Small fresh air flow in electronics housing expected to be enough to keep stable conditions
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
14
Cherenkov Light (BTF Test Aug/2013)
Online display from MAPMT analog spectra with binary response
Small arc of the Cherenkov ring is clearly visible
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
15
New MAROC based ASIC board
The existing electronics not suitable for CLAS12 RICH operation:
• not optimized for binary readout,
• limited speed through USB-2 interface,
• layout requires cabling to connect MAPMT
• no time information
Need to implement a new front-end board
(reuse large part of existing design)
•
•
•
•
Directly connected to MAPMTs (no
Low Voltage
cables): one MAPMT  one MAROC
64x parallel binary outputs as main
information
From
8x analog sum8 sent to fast amplifiers
MAPMT
and discriminator (NINO from
CERN/ALICE) for sub-ns timestamp
73 outputs lines
resolution
23 configuration lines (20 shared with other MAROCs)
1x analog MUX (or the internally
converted digital charge) for calibration
Few options (not exclusive) open for
(event by event or dedicated runs)
calibration/stability checks
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
16
DAQ electronics
INFN responsibility
JLab responsibility
C. Cuevas Talk
on FPGA boards
Modular design with 3 logic/physical layers
(similar approach of the existing electronics  straightforward porting)
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
17
Readout Speed and Dead Time Estimation
• When a binary channel has a hit the corresponding time is stored in 13
bits (1 ns resolution and > 8000 ns time range); 6 bits for addressing
• Each MAROC sum line has 14 bits information (0.5 ns resolution, same
time range); 3 bits for address
• Charge info for one channel, calibration/stability check (12bits ADC)
• Maximum hit occupancy 10% (exagerated! one order of magnitude larger
than GEANT4 Simulation)
• MAROC multiplexing factor 3 (up to 3 MAROCs share the same optical
link)
• Fast 2.5 Gbps serial link (assume 2 Gbps sustained)
Expected Readout Time (from MAROC to SSP):
[( 19 bits x 64 channels + 17 bits x 8 sums ) x 0.1 occupancy + 12 bits ]
x 3 MAROC/board / 2 Gbps
= ~220 ns
or <0.5% dead time at 20 kHz
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
18
Requirements fulfillment
Single photoelectron sensitivity
~50fC
Number of channels per sector
25600
Anodes gain spread compensation
1:4
Event Rate
20 kHz
Dead Time
few%
CLAS12 trigger latency
8 μs
Time resolution
~1
ns
Specs, slides 9, 10,
12 ...
64ch/ASIC
adequate
Specs, slides 4, 13
Slide 16, 17, 18,
Cuevas talk
Slide 16, 18,
Cuevas talk
Specs, slide 16
To disentangle direct and reflected photons (can be done off line)
Compactness/ Power consumption
Max area 1 m2, max thickness 10 cm, power compatible with air flow heat
dissipation
Radiation hardness
Must operate in the same irradiation level of the MAPMTs
Turisini – Musico - Cisbani
Specs, slide 14
MAROC build in CMOS 0.35 mm for
ATLAS radiation conditions.
No surprise expected
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
19
Alternative / DREAM ASIC
Dead-timeless Readout Electronics Asic for Micromegas
Single Channel (x64) - Design for Micromegas @ CLAS12
•Preamplifier, adj gain on 4 ranges (60fC,120fC,240fC,1pC)
•Shaper, adj peaking time 16 values from 50 ns to 1 μs
•Analog memory 512 cells, sampling rate 1-50MHz
•Discriminator, trigger pipeline 16 μs , sum of 64
•140-pin
•0.4mm package,
•17mm x17 mm footprint
PROs: analog pipeline,
designed for JLAB12
CONs: dynamic range
(?), time resolution
Output: Analog MUX and
Digital Sum
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
20
PMT DREAM interface
Dead-timeless Readout Electronics Asic for Micromegas
R&D from Micromegas group
Attenuation board for H8500
with various divider ratio for testing
TEST SCHEDULED 2013 October at INFN-FRASCATI
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
21
Conclusions
 MAROC chip (binary data) represents a valid choice for RICH

readout
 Laboratory/Beam tests showed a clean SPE detection
Porting of current MAROC electronics in CLAS12 DAQ
framework expected to be reasonably simple
•
Time measurement will be the really new aspect
 DREAM option left open (till test on PMT coupling Oct/2013):
•
•
•

Provide multi-sample analog information
Synergy with current development for microMeGas
Expected easier maintainance (both HW and SW)
Next steps:
•
•
•
•
continue analysis of BTF data,
finalize MAROC test (internal ADC, timing),
DREAM test,
design the new front-end board for the RICH and start testing it.
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
22
End
CLAS12 RICH Review, 2013 September 5-6
Components and Costs
Components Cost Details
Card Type
3
Integrated Circuits
Cost (EUR) Quantity
Extended CostNotes
MAROC
100
3
300
NINO
20
3
60
Ext. ADC
5
1
5 To be confirmed
Delay Line
10
1
10
Voltage Regulator
5
15
75
DAC
10
3
30
Passives
Resistors
0,02
180
3,6
Capacitors
0,04
240
9,6
Inductors
0,04
15
0,6
Misc Hardware
1
30
30
Multilayer PCB
40
3
120
Connectors
SAMTEC TMM-118-03-G-D
2
12
24
SAMTEC
3
2
6
LV Power
2
1
2
Assembling and Test
35
3
105
Total
EUR
USD
780,8
1054,08
USD/EUR
24
Expertise in SiPM readout
NINO : ultra-fast, low power, amplifier and
discriminator for TOF (ALICE)
SiPM to NINO implementation
Use of NINO chip as a preamp in precise time
measurement with SiPM (TOF-PET application)
Could be extended to SiPM for RICH
More integrated solutions to be considered
22Na
SiPM width spectrum (O-scope)
SiPM matrix
HPTDC
22Na
SiPM width spectrum (HPTDC)
16 x NINO
(128 channels)
Turisini – Musico - Cisbani
P. Musico INFN/Ge
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
25
External Light Source Test Setup
BNC Model 6010
490 nm ± 40 nm
Light guide in front of a single PMT
LIGHT
TRIG
Externally driven or Self
Trigger DAQ
BLACK BOX
light guide
PMT
MAROC3
BUF SHAP ADC
FPGA
PA
-HV = 1100 Volt
PC
SHAP DISC
INFN ELECTRONICS
USB 2.0
high density cable
• Assume analog output as reference
• Measure “digital” noise with PMT on, no light (and other configurations)
• Compare/Correlate analog and binary information, with internal (self) and external
(need synch) triggers
•
Measure range (in threshold) of the ~single photon signal by threshold scan to
estimate SNR
26
In House MAROC based DAQ
Original system developed for Radionuclides Medical Imaging
Many optical photons application
Designed for H9500/H8500
4096 max channels, USB2 readout
Binary output used for self-trigger
MAPMT Hamamatsu H9500
Front End Card
Backplane
Not optimized for Single Photon and ext. trigger
Control Board
Modular design (FrontEnd + Bus + ControlBoard + USB int.)
Turisini – Musico - Cisbani
Front End Electronics - CLAS12 RICH Review – JLab - Sept/2013
27