MPA-LIGHT: Design and results of a 65 nm digital readout Macro Pixel ASIC prototype with on- chip...
Transcript of MPA-LIGHT: Design and results of a 65 nm digital readout Macro Pixel ASIC prototype with on- chip...
MPA-LIGHT:Design and results of a 65 nm digital readout Macro Pixel ASIC prototype with on-chip particle recognition for the Phase II CMS outer tracker upgrade
A. Caratelli, D. Ceresa, R. Francisco, J. Kaplon, K. Kloukinas and A. Marchioro.
PH-ESE, CERN
TWEPP ‘15, Wednesday the 30th September, 2015
2
CMS was originally designed to run atLuminosity = 1 x 1034 cm-2 s-1
HL-LHC will improve up toLuminosity = 7.5 x 1034 cm-2 s-1
Phase II upgrades prepares CMS for HL-LHCTechnical Proposal for the Phase-II Upgrade of the CMS Detector: http://cds.cern.ch/record/2020886
3
Main Outer Tracker requirements:
Up to pile up = 200
Contribution to Level 1 decision
L1 rate = 750 KHz
L1 Latency = 12.8 us
Reduce Material
Pixelated sensor
Larger bandwidth
Larger buffers
PT modules
New technologies
+ POWER
- POWER
4
PT module discriminates particlePass Fail
Stub
Particle are discriminated according to their transverse momentum
The module is composedby two sensor layers
5
Outer Tracker working principle:
Stub Finding
Outer Tracker Front-EndStubs only
Trigger-less readout: The electronics provides the stubs found at each event.
Triggered readout: The electronics stores the full frame at each event.
6
Stub Finding
TrackFind
Readout
Outer Tracker Front-End Tracker Back-EndStubs only
Outer Tracker working principle:
Dedicated talk: “Track Finding in CMS for the Level-1 Trigger at the HL-LHC”, M. Pesaresi
7
Stub Finding
TrackFind
CMS Level
1
CMS DAQ
Outer Tracker Front-End Tracker Back-End CMS
Level 1accept
Stubs only
L1 accept rate = 750 KHzLatency = 12.8 us
Full Frame
Outer Tracker working principle:
Readout
8
2 x PT modules:
140 mm70 mm
130 mm
130 mm
Z
Ф
2 Strip moduleStrips ~ 5 cm x 90 umr > 40 cmTID ~ 40 Mrad
Inner Tracker (RD53)Pixel Detector
Pixel + Strip modulePixels ~ 1.5mm x 100um Strip ~ 2.5cm x 100um 20 cm < r < 40 cmTID ~ 100 Mrad
9
Pixel + Strip module exploded view
ShortStripASIC x 8
Concentrator IC (CIC)Data aggregation
MacroPixelASIC x 16
Pixelated Layer
~ 32k pixel
LP-GBT and VTRx+Data transmission through the optical link
DC-DC convertersPS module is fully integrate entity
Strip Layer~ 2k strips
10
Pixel Strip module Front-End ASICs
MPA (1.2 cm x 2.5 cm)
FE
SRAMData
Formatting
I/OBinary Readou
t
Stub Finding Logic I/O
FE
Binary Readout
I/O
I/O
SHORT STRIP ASIC
FullFrame
Stubsonly
Strips
Pixels Level-1 accept
Power budget < 100 mW / cm2
(Total for both Front-End ASICs)
11
MPA-Light: the MPA prototype
Analog FE circuitry in 65 nm
Development of the sensor
Module assembly
MaPSA-Light
MPA-Light
12
MPA-Light ASIC floorplan16 * 100 μm
6.3
38
mm
2 m
m3
x 1
44
6 μ
m
z
X
Analog Pixel FE
Digital Pixel Logic
Analog Bias
Digital Periphery Logic
Bump Bonding Pads
I/O Driver
Row 1
Row 2
BB connectionto PCB/hybrid/detectorperiphery
WB connectionto PCB/Hybrid(staggered)
BB connection to detector pixels
BB connectionto detector GND
1.7 mm
13
MPA-Light functional description
FE
Hit Counter
Binary ReadoutSample Clock
(40 MHz)
3 modes of operation available
Stub finding
Digital-to-analog
Converters (DACs)
Calibration pulsesThresholdTrimming (per pixel)
Input
Test results are obtained without sensor i.e the input are from calibration capacitances
Memory
Components for testing
14
Analog Front-End schematicdesigned by J. Kaplon
Preamplifier Shaper Discriminator
Current consumption < 30 uA / channel
15
Calibration DAC measurement
Calibration pulse[fC]
8.7 fC
0 fC
0 255DAC code
Nominal step = 0.035 fCMeasured step = 0.034 fC
Offset = 0.044 fC
On purpose for Radiation hardness
16
Pixel Threshold spread
100 250
Nominal Step ~ 1.5 mV
Threshold DAC
HitCounter
1000
1000
0
0
2 Trimming DAC steps = 7.5 mV
Best trimming withthe 5-bits trimming DAC
Pulse Injected ~ 3 fC
HitCounter
17
Gain and Noise from S-curves (1)
0.35 fC … 1.05 fC
60 200
Nominal Step ~ 1.5 mVThreshold DAC
1000
0
Optimization region
Calibration pulse amplitudes
HitCounter
2.45 fC
18
Gain and Noise from S-curves (2)
Equivalent Noise Charge = (sim. value < 200 e-)
SNR >> 20
0.35 fC 2.25 fCCalibration pulse
amplitude
Threshold[mV]
Gain =
30 mV
190 mV
HitCounter
1000
120 140
Threshold DAC
0
80
sigma(cerf) = ENC
mean(cerf) = Threshold
Complementary error function fitting
(sim. value: 85 )
19
MPA-Light functional description
FE
Hit Counter
Binary ReadoutSample Clock
(40 MHz)
3 modes of operation available
Stub finding
Digital-to-analog
Converters (DACs)
Calibration pulsesThresholdTrimming (per pixel)
Input
Test results are obtained without sensor i.e the input are from calibration capacitances
Memory
Components for testing
20
Binary readout allows FE studies:
3.4 fC
2.2 fC
1.1 fC
Peakingtime
~ 24 ns
0 ns 140 ns 0 mV
310 mV Peak Gain~ 91 mV/fC
ShaperOutput[mV]
Injectiontime [ns]
Shaper outputfor different charge
1 fC 9 fC
ΔWalkTime[ns]
Max delay < 10 ns
Injected
Pulse [fC]
FE walk time (Threshold = 0.5 fC)
0 ns
Simulations value:• Time walk: < 14 ns from 0.5 fC to 12
fC• Peaking time = 24 ns
21
0 0 00 0 01 1 0
0 1 0
1 0 0
1 1 0
Z
X
FromBinary readout
MPA-Light digital logic testing
1. No Processing mode:
Used to test Binary readout
and Pixel Clustering
22
0 0 00 0 00 0 0
0 1 0
1 0 0
0 0 0
Z
X
FromBinary readout
+Pixel Clustering
MPA-Light digital logic testing
1. No Processing mode:
Used to test Binary readout
and Pixel Clustering
23
0 0 00 0 00 0 0
0 1 0
1 0 0
0 0 0
Z
X
FromBinary readout
+Pixel Clustering
Centroid ListX = 3, Z = 1X = 2, Z = 3
MPA-Light digital logic testing
1. No Processing mode:
Used to test Binary readout
and Pixel Clustering
2. Data Processing mode:
Used to test Encodingand Stub Finding Logic
24
0 0 00 0 00 0 0
0 1 0
1 0 0
0 0 0
Z
X
FromBinary readout
+Pixel Clustering
Centroid ListX = 3, Z = 1X = 2, Z = 3
Stub ListX=3, Z=1, B=0X=2, Z=3, B=-1
or
0 1 00 0 0
MPA-Light digital logic testing
Tested with patternfrom FPGA
1. No Processing mode:
Used to test Binary readout
and Pixel Clustering
2. Data Processing mode:
Used to test Encodingand Stub Finding Logic
Further details on the Stub finding logic: http://iopscience.iop.org/article/10.1088/1748-0221/9/11/C11012
25
0 0 00 0 00 0 0
0 1 0
1 0 0
0 0 0
Z
X
FromBinary readout
+Pixel Clustering
0 1 01 0 0
Centroid ListX = 3, Z = 1X = 2, Z = 3
Stub ListX=3, Z=1, B=0X=2, Z=3, B=-1
or
0 1 00 0 0
MPA-Light digital logic testing
Tested with patternfrom FPGA
1. No Processing mode:
Used to test Binary readout
and Pixel Clustering
2. Data Processing mode:
Used to test Encodingand Stub Finding Logic
3. Strip Emulator mode:
Tested with oscilloscope
26
Total Ionizing Dose (TID) with X-ray
PS module TID < 100 Mrad
0
Pre radvalue
-15 %
Total Ionizing dose [log TID]
150 Mrad
Annealing- 11 %
Threshold DAC variation with TID
0
100 Mrad
∆ ThresholdStep [%]
27
Conclusion and future plans
MPA-Light fully functional
MaPSA-Light ready
MaPSA-Light Testing
MPA & SSA design
Final size MPA and SSA
28
Thanks to all theCMS TRACKER PHASE II ELECTRONICS TEAM
Davide Ceresa, A. Caratelli, R. Francisco, J. Kaplon, K. Kloukinas and A.
Marchioro.PH-ESE-ME, CERN
TWEPP ‘15, Wednesday the 30th September, 2015
29
Backup slidesDavide Ceresa, PH-ESE-ME, CERN
TWEPP ‘15, Wednesday the 30th September, 2015
30
Reduced thanks to:Bend bits reductionTight Stub thresholdDifferent LP-GBT transmission mode
Good stub losses lower than 0.5% in all the
tracker at Pile Up 200
Stub transmission to tracker back-end is clearly critical for the first barrel layer
simulation and studies by S.Viret
Tracker Back-end data transmission
31
Tracker power budget is limited
Working temperature cooling: ~ -30C
Total Power Allocated per 16 MPAs and SSAs:
4WTotal Power Allocated per MPA + SSA: 250 mWRough Power Estimation:
Analog 70 mWL1 Memory 70 mWSSA 40 mW
Remaining 70 mW
I/OClock DistributionData TransportStub Finding LogicL1 Data LogicOutput Interface
32
Powering reference scheme
0.8 – 1V used to power digital logic which can work also at low voltage
slide and studies from G. Blanchot, F. Faccio and S.Michelis
33
Stub Finding logic efficiency results
0 20 40 60 80 100 120
0.4
0.5
0.6
0.7
0.8
0.9
1
Module # ( 1 - 125)
Eff
icie
ncy
(%)
MPA efficiency per moduleMPA efficiency per layerMPA efficiency per layer with Z communication
Module efficiency
Layer Efficiency
Layer efficiency with interconnection
Vertex Z-Z
34
Stub Finding logic efficiency results
XV
Module A
Module B
Vertex Z
Recovered
Without an interconnect technology (ex: TSV) between the two sides of the module, tracks crossing the middle will not be identified as stubs
35
Stub Finding logic efficiency results
VV
Module A
Vertex Z
VTilted layout solves the problem and decrease the number of modules, but complicates mechanics
36
Flipped MaPSA concept
• Only bump bonding in MPA design• No extra material between sensors• Better cooling of MPA
• Digital IO through sensor periphery• Temperature gradient across the sensor surface • Larger sensor can create problem in final production
STRIP SENSOR
MPA
SSA
MPA
SSA
EXTENDED PIXEL SENSOR
Cooling
BASELINE
STRIP SENSOR
PIXEL SENSOR
SSASSA
MPAMPA
Cooling
PROPOSAL
37
MPA-Light test system overview
PowerSupply
Ethernet Cable
MPA-Light
Test system controlled by Python routines Communicates
with IP bus
Interface BoardVoltage and Current
generator and monitorsdesigned for the CLICpix readout
by Szymon Kulis
FPGAHandle configuration,
readoutand control the interface board
Load boardExchangeable board for the
ASIC wire bonding and decoupling
developed by A. Caratelli
38
Analog Front-End simulation resultsdesigned by J. Kaplon
Other simulation results:Gain from S curves: 85 mV/fCTime walk : < 14 ns with threshold at 0.5 fC and signal from 0.75
to 12 fCNoise(Worst case): < 200 e- (SNR >> 20)
Spread90 mV pk-pk
Peaking time24 ns
Montecarlo simulations 100 runs for 2.5 fC
39
Binary readout allows FE studies
40 MHz Clock
Hit detector output
27 28
Discriminator Output
Shaper Output
Changing the calibration injection point and using the synchronous readout we can characterize the shaper of the
analog FE
Observing when the time stamp change we canextract the rising time of the shaper from zero
to the set threshold
Threshold
G. Blanchot, F. Faccio, S. Michelis - 2S and PS Powering Meeting
Service Board with Staged DCDC scheme
30/SEP/2014
Input power: low profile Molex Picoblade 2 poles connector (stands our current rating).
The filters will stay out of the shield area, must be on sides.
Wirebond based IO implemented so far.
The same low mass ECCA based coil is used for the 3 converters.
The new LV DCDCs still based on teh FEASTMP geometry. However we can consider a smaller package for this chip.
The 3 DCDC stages can be fitted in the available board space without excessive compromise.
Wir
ebonds
Input
Filters
16V caps
FEASTMP
LV caps
New DCDCs
Filters
Wir
ebonds
PS module
Forum 2014: Tilted Barrel for CMS Phase 2 Upgrade
11
129
1
2
34
5
8
1. Silicon strip sensor2. Silicon pixel sensor3. Macro-Pixel ASIC
(MPA)4. Al-CF sensor spacer5. CFRP base plate6. FE Hybrid7. Opto-Link Hybrid8. Power Hybrid9. Short-strip ASIC (SSA)10. Concentrator IC (CIC)11. Hybrid CF support12. Al-CF Hybrid spacer6
710
1111
11
12
Support and cooling via a large-area base plate
42
More realistic tilted geometry
Forum 2014: Tilted Barrel for CMS Phase 2 Upgrade
Z = 0
• Short cylindrical section in the center• Constant tilt angle in several successive rings
o Helps avoiding clasheso Also, reduces support structure variants.
43
Material budget full tracker
Forum 2014: Tilted Barrel for CMS Phase 2 Upgrade
S. Mersi at al., Performance of Tilted Inner Barrel, CMS Upgrade Workshop 1 April 2014