Status on
CMOS sensors
Auguste Besson
on behalf of
IRES/LEPSI: M. Deveaux, A. Gay, G. Gaycken, Y. Gornushkin, D. Grandjean,
S. Heini, A. Himmi, Ch. Hu, H. Souffi-Kebbati, I. Valin, M. Winter,
G. Claus, C. Colledani, G. Deptuch, W. Dulinski
(M6/M8 DAPNIA: Y. Degerli, N. Fourches, P. Lutz)
• Develop. of large CMOS sensors (3-T/pixel)
• Caracterization of the technology without epitaxy
• R&D on fast sensors.
• 2004 schedule and summary
November 2003 ECFA-Montpellier 1
History
CHIP YEAR PROCESS EPITAXIAL PITCH METAL PECULIAR
m m
M1 1999 AMS 0.6 m 14 20 3M thick epitaxy
M2 2000 MIETEC 0.35 m 4.2 20 5M thin epitaxy
M3 2001 IBM 0.25 m 2 8 3M deep sub- m
M4 2001 AMS 0.35 m 0! 20 3M low dop. Substrate
SUC 2 2003 AMS 0.35 m none 20 3M low dop. Substrate
M5 & M5B 2001/2003 AMS 0.6 m 14 17 3M real scale 1M pixels
M6 2002 MIETEC 0.35 m 4.2 28 5M col. // r.o. and integrated spars.
M7 2003 AMS 0.35 m none 20 4M col. // r.o. and integ. spars. (photoFET)
• MIMOSA 1,2,4,5 tested at CERN-SPS with 120 GeV/c -
• M6 tests in progress
• M7 available soon
• SUCESSOR 2 (SUCIMA PROJECT): beam test in 2003
40 m pitch, no epitaxial layer.
2003: M4, M5, M6 tests, M7 fabricated
November 2003 A. Besson, ECFA-Montpellier 2
3-T/pixel large CMOS sensors (M5)
• AMS 0.6 m (M1 like)
reticle size 19.4 x 17.4 cm2
512 x 512 pixels
(/ each of 4 matrices)
17x17 m pitch
• 4 sub-matrices per sensors, read-out in parallel
• 6 wafers (6’’) built in 2001
• 3 wafers thinned down to 120 m (2 in 2003)
• 2002 results:
Yield 20-30 %
det ≳ 99%; sp ~1.7 m; ~0.2%
November 2003 A. Besson, ECFA-Montpellier 3
3-T large sensors: 2003 (2)
• Beam test at SPS (2003)
3 sensors
120 GeV/c -
• Performance uniformity tests
between sub-matrices, sensors
diode size comparisons
Small diode
Big diode
(3x3 m2)
(5x5 m2)
November 2003 A. Besson, ECFA-Montpellier 4
3-T large sensors: results (3)
SMALL DIODES BIG DIODES
min max min max
Eff (%) 98.9 99 98.9 99.1
N seed (e-) 18.9 19.5 24 25.1
S/N seed 22.6 24.5 22.6 23.5
Q seed 279 +- 5 286 +- 6 353 +- 8 372 +- 4
Q 9pixels 970 +- 11 1032 +- 15 1213 +- 23 1245 +- 11
Q 25 pixels 1178 +- 14 1235 +- 19 1386 +- 27 1445 +- 14
sp (microns) 2.3 2.5 2.4 2.5
submatrices have similar properties
~1 dead column / 512 (i.e. ~0.2% det inefficiency)
single point resolution ≲ 2.5 m (still improvable)
• Effect of particle incidence
chip turned w.r.t. beam direction
charge cos as expected
November 2003 A. Besson, ECFA-Montpellier 5
3-T large sensors: M5-B (4)
• Mimosa 5-B
23 wafers produced in oct. 2003
• Slightly improved fabrication process (metalisation)
should reduce dead columns rate.
should improve rate of good chips (yield)
setting up thinning to 15 m (Nov 03)
(with a Si wafer on the electronics side for handling)
Application to bio-medical imaging (20-30 keV e-)
November 2003 A. Besson, ECFA-Montpellier 6
3-T large sensors: application (5)
• STAR: extension of the Vertex Detector (2006)
charm physics small radius, granular and thin detector
2 pixel layers ≳ 1000 cm2 R(layer 1) ≳ 2 cm ; R(layer 2) ≲ 4 cm ;
• Requirements
• pt~ 3 m
• What to improve ?
• 2.6 kRad/year
• read out time (~ 24 ms)
• 2.1010 neq/cm2/year
• sensor thickness (~ 120 m)
• read out time 10-20 ms
• electronic noise (room T)
• Power ≲ 100 mW/cm2
• yield (not crucial)
• sensor thickness ≳ 50 m
• Room temperature
M5 performances are close to the STAR requirements
started a collaboration with LBL (and BNL)
first MIMO⋆1 prototype in summer 2004 (new TSMC 0.25 m tech.)
November 2003 A. Besson, ECFA-Montpellier 7
No epitaxial layer prototypes (M4)
• Properties:
AMS 0.35 m witout epitaxial layer.
Low doped substrate
increases e
120 GeV/c - SPS beam tests
Eff ≳ 99.5 %
resolution sp ~2,5 m (new)
Fabrication processes with
epitaxial layer is not mandatory !
November 2003 A. Besson, ECFA 8
No epitaxial
layer (M4) (2)
• Rad. tol. studies:
200 kRad (x-rays),
1.4 1011 neq/cm2
• S/N ↘ when T ↗
• If T ≲ 20⁰C no obvious effects on efficiency and spatial resolution
• Radiation effects are negligible at this level (200 kRad ;1.4x1011 n/cm2)
November 2003 A. Besson, ECFA-Montpellier 9
SB1 Charge (1,9,25 pixels)
No epi. : SUCCESSOR 2
• SUCCESSOR 2: (M4 like)
bio-medical imaging, SUCIMA project.
(no epi. layer , AMS 0.35 m)
40x40 m2 pixels
beam tests (oct. 2003)
different sub-structures tested
Noise vs T
• (3T pixel, Self-Bias pixels with 2 different
diode sizes)
eff ≳ 99.9 %
sp ~5-6 m (~2 x M4 with 20 m pitch)
best performances for large diodes SB
X resolution vs T S/N vs T
?
November 2003 A. Besson, ECFA-Montpellier 10
R&D on fast sensors
• M1-M5 1M pixels read-out in 1-10 ms
• FLC 1st VD layer must be read-out in 25-50 s (beamstrahlung)
potentially tremendous data flow:
e.g. 15 bits/pixels, t~25 s 500 Gbits/s/106 pixels !
main goal: fast signal treatment AND data compression
integrated in the sensor
• Fast // read out of short columns
• Different prototypes with different signal treatment:
M6 (with DAPNIA): tests in 2003,
individual pixels and discri work fine, but large spread of
pixel caracteristics (pedestal, noise, gain ?)
M7: available soon, tests in 2003-04
M8 (with DAPNIA): submitted in Nov., tests in 2004
November 2003 A. Besson, ECFA-Montpellier 11
Summary
• Large sensors (M5) (1M pixels, AMS 0.6 m )
ready to be used for a real detector
2nd fabrication (23 wafers) with a better yield expected
thinning down to 15 m in progress
application to extension of STAR Vertex detector in 2006
• No epitaxial layer sensors (M4, SUC 2)
validated for m.i.p. detection (eff ≳ 99.5%, sp ~2,5 m)
fits industrial CMOS fabrication tendancy
• Fast response sensors (M6, M7, M8)
studies: fab. techno., charge collection system, signal treatment
architecture read out speed, efficiency, zero sup., power diss. etc.
• 2003/2004 schedule
M5-B tests yield, thinning
M⋆1 available in summer 2004, tests in autumn
fast sensors: 2 prototypes
M7 and M8 tests
charge collection studies ionising radiation tol.
November 2003 A. Besson, ECFA-Montpellier 12