FSBB-M and FSBB-A: Two Large Scale CMOS Pixel Sensors Building Blocks Developed for the Upgrade of the Inner Tracking System of the
ALICE Experiment Frédéric Morel (on behalf of PICSEL team of IPHC Strasbourg)
Outline
Starting point: STAR-PXL MISTRAL and ASTRAL
Description Test results
Conclusions
Starting point: Ultimate chip in STAR
TWEPP 2014 IPHC [email protected] 2
Physics data taking from March to June 2014
MIMOSA28(ULTIMATE)
Pictures Credit: Berkley Lab and IPHC
Towards Higher Read-Out Speed and Radiation Tolerance Requirements for inner and outer barrel (compared to STAR-PXL):
Main improvements required while remaining inside the virtuous circle of spatial resolution, speed, material budget, radiation tolerance move to 0.18 µm process
To enhance the radiation tolerance High resistivity epitaxial layer Smaller feature size process
To accelerate the readout speed More parallelised read-out Optimised number of pixels per column New pixel array architectures Smaller feature size process
Tight schedule (production in 2016) and large surface (10 m²) 2 readout architectures R&D in parallel
Synchronous readout: mature and robust architecture (MIMOSA28 like) Asynchronous readout: new and challenging architecture
TWEPP 2014 IPHC [email protected] 3
tr.o. sp TID Fluence T op Power Active Area
STAR-PXL <~ 200 µs < 4 µm 150 kRad 3x1012 neq/cm² 30 °C 160 mW/cm² 0.15 m²
ITS-in <~30 µs <~ 5 µm 700 kRad 1013 neq/cm² 30 °C < 300 mW/cm² 0.17 m²
ITS-out <~30 µs <~ 10 µm 15 kRad 4x1011 neq/cm² 30 °C < 100 mW/cm² ~ 10 m²
Sensors R&D for the upgrade of the ITS: Our Strategy
TWEPP 2014 IPHC [email protected] 4
R&D of up- & down-stream of sensors performed in parallel at IPHC in order to match the ITS timescalefor 2 final sensors (~3x1.3 cm²)
- Mature architecture: MISTRAL = MIMOSA Sensor for the inner TRacker of ALICE Relatively moderate readout speed (200 ns/ 2rows)
~ 200 mW/cm² , σsp ~ 5 µm for inner layers ~ 100 mW/cm² , σsp ~ 10 µm for outer layers
- Improved architecture: ASTRAL = AROM Sensor for the inner TRacker of ALICE Higher speed (100 ns/ 2rows) + Lower power
~ 85 mW/cm², σsp ~ 5 µm for inner layers
Modular design + reused parts optimising R&D time FSBB-M (Full Scale Building Block) : 1/3 of MISTRAL final sensor FSBB-A (Full Scale Building Block) : 1/3 of ASTRAL final sensor
Alternative sensor with asynchronous readout: ALPIDE Ref: Ping Yang @ PIXEL 2014
IPHC [email protected] 8TWEPP 2013
~1.2
cm
~1 cm
SUZE
~1.2 cm² array
Discriminators
SUZE
~1.2 cm² array
Discriminators
SUZE
~1.2 cm² array
Discriminators
Serial read out RD block PLLLVDS
~1.3
cm
~1 cm
Serial read out RD block PLLLVDSSUZE
~1.3 cm² array
SUZE
~1.3 cm² array
SUZE
~1.3 cm² array
MISTRAL_in ASTRAL_inFSBB_M FSBB_A
IPHC [email protected] 8TWEPP 2013
~1.2
cm
~1 cm
SUZE
~1.2 cm² array
Discriminators
SUZE
~1.2 cm² array
Discriminators
SUZE
~1.2 cm² array
Discriminators
Serial read out RD block PLLLVDS
~1.3
cm
~1 cm
Serial read out RD block PLLLVDSSUZE
~1.3 cm² array
SUZE
~1.3 cm² array
SUZE
~1.3 cm² array
MISTRAL_in ASTRAL_inFSBB_M FSBB_A
Upstream Downstream
²
Amp cDS ADC SUZE Data trans
Steering, Slow control, Bias DAC- Column-level discriminators MIMOSA: based on architecture of MIMOSA28- In-pixel discrimination AROM (Accelerated Read-Out Mimosa)è 2 rows read out at once in a Rolling shutter mode
Pixel 1
Pixel 2
Pixel N
Power density, Integration time and Spatial resolution MISTRAL for outer layers:
Sensitive area ( 3 x 1.3 ) cm² 1 SUZE per final sensor N Col = No. of Columns (modulo 32) NFSBB = No. of FSBBs per final sensor N Col /NFSBB = No. of rows Pitch Row = 13000 µm / (N Col /NFSBB ) Pitch col = 30000 µm / N Col
Pixel pitches vs Spatial resolution (sp ) sp = 7 [Pitch col x Pitch row )/(22 x 66) ]k µm
with k = ½ or 1 Empirical formula for a specific tech-
nology and architecture based on the spatial resolution of tested chip with a pixel area of 22 x 66 µm²
No. of Rows vs Readout time (TRO) TRO = 200 x (N Col / NFSBB ) / 2 ns
No. of columns vs Power density (PD) PD = P / area P = (0.117 x N Col x 2) x 1.8 + 46.5 + 0.5 x N Col / 32 mW Popt = (0.09 x N Col x 2) x 1.8 + 46.5 + 0.5 x N Col / 32 mW
TWEPP 2014 IPHC [email protected] 5
Pixel Array
Common Digital Part
Pixel Array Pixel Array
Discri Discri Discri
~3 cm
1.3
cm
2.5 mm
Total of NC columns for NFSBB pixel arrays
Upstream of MISTRAL Sensor
TWEPP 2014 IPHC [email protected] 6
Pixel optimisation: Sensing node: Nwell-PEPI diode
Optimisation = f (diode size, shape, No. of diodes/pixel, pixel pitch, EPI) No accurate model exists need submission iterations
In-pixel amplification and cDS: Limited dynamic range (supply 1.8 V) compared to the previous process (3.3 V) Noise optimisation especially against random telegraph signal (RTS) noise
Sensing diode: avoid STI around N-well diode RO circuit: avoid using minimum dimensions for key MOS & avoid STI interface Trade off between diode size, input MOS size w.r.t. S/N before and after irradiation
Beam test of MIMOSA-22THRa Detection efficiency ≥ 99.8% while Fake hit rate ≤ O(10−5) 22×33 μm² binary pixel resolution: ~5 µm as expected from former studies Final ionisation radiation tolerance assessment under way
Designed by Y. Degerli (Irfu/AIDA)Layout of 2 discri. (1 columns)
~300
µm
Pixel level Column level
MIMOSA-34Sensing node opt.
MIMOSA-32FEEAmp opt.
MIMOSA-32NRTS opt.
MIMOSA-22THRa1&2Chain opt. => 1 D/col
MIMOSA-22THRbChain opt. => 2 D/col
sf
sf
Bias Bias
sf
sf
Bias Bias
Latch
latch
Vclp
Vclp
Vref1
Vref2
ф1
ф1
ф1
ф1
ф1
ф2
Out
Outbф1
ф2cs
Power
Vclp_pix
sf
Slct_Row
2 x FSBB_M0Verify full chain
and full functionalities1.
6 cm
1.8 cm
Test Results of FSBB-M Transfer function measurement:
208 discriminators connected to 80,000 pixels SUZE activated normally pure noise measurement performed in obscurity
Perturbations due to cross-talk with modest impact on TN & FPN Submission of a corrected version next month
TWEPP 2014 IPHC [email protected] 7
TN & FPN for 17 different chips
Threshold Fake rate
4 mV 1.4 10-5
5 mV 4.3 10-6
6 mV 1.3 10-6
7 mV 4.9 10-7
8 mV 2.8 10-7
FPN in transfer function mode FPN in detection mode
TN ~ 0.87 mV FPN ~ 0.55 mV
TWEPP 2014 IPHC [email protected] 8
Upstream of ASTRAL sensor Thanks to the quadruple-well technology, discriminator integrated inside each pixel
Analogue buffer driving the long distance column line is no longer needed Static current consumption reduced from ~120 µA up to ~14 µA per pixel
Readout time per row can be halved down to 100 ns (still with 2 rows at once) due to small local parasitics Sensing node & in-pixel pre-amplification as in MISTRAL sensors In-pixel discrimination
Topology selected among 3 topologies implemented in the 1st prototype AROM-0 Several optimisations on the 2 most promising topologies in AROM-1
AROM-0
AROM-1
FSBB_A01/3 of final sensor
1.6
cm
0.9 cm
vRef1
vRef2
read
read
read buffLatch
latch
Pow
er
Powerclamp
vRef
3
Bias
Bias
Sel_
D
A1 A2cs
Power
calib
read
readca
lib
calibtest
Test Results of the Upstream Part of ASTRAL Sensor Preliminary lab test results @ 30 °C and @ 100 MHz (instead of 160 MHz)
Current acquisition board limitation AROM-1b/c
Discri alone: TN ~ 0.75 mV, FPN ~ 0.63mV Discri+pixel: TN ~ 1.1 mV, FPN ~ 0.66mV
AROM-1e tests are ongoing Discri alone: TN ~ 0.29 mV, FPN ~ 0.19mV Discri+pixel: TN ~ 0.94 mV, FPN ~ 0.23mV
TWEPP 2014 IPHC [email protected] 9
AROM-1e optimisation is validatedSome residual coupling effects are investigated
Summary
TWEPP 2014 IPHC [email protected] 10
MISTRAL: architecture validated FSBB-M:
Spatial resolution (22x33 µm² pixel) ~5 µm Integration time ~40 µs Power consumption ~200 mW/cm² Detection efficiency <~ 100 % for fake hit rate ≤ O(10−5) (beam test of FSBB-M in Oct./Nov.) Yield: 25 chips operational among 25 chips tested
MISTRAL for outer layers to be submitted in Q2/2015: Spatial resolution (36x62.5 µm² pixel) ~10 µm expected Integration time ~20 µs Power consumption ~100 mW/cm²
ASTRAL: architecture validation on going ASTRAL pixel front-end amplification (same as in MISTRAL) validation completed Downstream of ASTRAL (shares the same logic with MISTRAL) validation completed In-pixel discrimination validation completed ASTRAL exhibits 2 x faster readout and lower power consumption than MISTRAL
Integration time: <~20 µs Power consumption ~85 mW/cm² for inner layers
Conclusions
TWEPP 2014 IPHC [email protected] 11
ALICE-ITS sets a new challenge in pixelated devices Large detector areas may be equipped with granular and thin pixels (at affordable cost)
3 complementary sensor architectures developed 1st full scale prototypes fabricated in Spring ’14 1st test results
MISTRAL : rolling shutter, directly derived from the STAR-PXL sensor mature & robust ASTRAL : rolling shutter approach pushed towards its limits with in-pixel discrimination ALPIDE : asynchronous (like hybrid pixels), most challenging but highest potential
All 3 seem operational but their performance assessment is still under way Further optimisation of design parameters on-going Design finalisation in 2015 production in 2016
High potential 0.18 μm CIS process for HEP experiment Feature size, deep P-well, 6 ML, ”thick” and high-resistivity epitaxy, large reticule
New step in particle detection performances with CPS w.r.t STAR-PXL pioneering device Still room for improvement towards real potential of CPS: smaller feature size, more ML,
etc. New horizons to develop high potential CPS to reach sub-microsecond integration time, ultra
low power (few tens of mW/cm²), and high radiation tolerance.
TWEPP 2014 IPHC [email protected] 12
Thank You !