Status on CMOS sensors - PowerPoint PPT Presentation

1 / 12
About This Presentation
Title:

Status on CMOS sensors

Description:

Hu, H. Souffi-Kebbati, I. Valin, M. Winter, G. Claus, C. Colledani, G. Deptuch, ... single point resolution 2.5 m (still improvable) Effect of particle incidence ... – PowerPoint PPT presentation

Number of Views:21
Avg rating:3.0/5.0
Slides: 13
Provided by: iresi
Category:

less

Transcript and Presenter's Notes

Title: Status on CMOS sensors


1
Status onCMOS 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
  • RD on fast sensors.
  • 2004 schedule and summary

2
History
  • 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

3
3-T/pixel large CMOS sensors (M5)
  • AMS 0.6 ?m (M1 like)
  • reticle size 19.4 x 17.4 mm2
  • 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 ?ltgaingt 0.2

4
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 (3x3 ?m2)
Big diode (5x5 ?m2)
5
3-T large sensors results (3)
  • 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 as expected

6
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-)

7
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
  • 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.)
  • Requirements
  • ?pt 3 ?m
  • 2.6 kRad/year
  • 2.1010 neq/cm2/year
  • read out time 10-20 ms
  • Power ? 100 mW/cm2
  • sensor thickness ? 50 ?m
  • Room temperature
  • What to improve ?
  • read out time ( 24 ms)
  • sensor thickness ( 120 ?m)
  • electronic noise (room T)
  • yield (not crucial)

8
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 !

9
No epitaxial layer (M4) (2)
  • Rad. tol. studies
  • 200 kRad (x-rays),
  • 1.4 1011 neq/cm2
  • S/N ? when T ?
  • If T ? 20C ? no obvious effects on efficiency
    and spatial resolution
  • Radiation effects are negligible at this level
    (200 kRad 1.4x1011 n/cm2)

10
No epi. SUCCESSOR 2
SB1 Charge (1,9,25 pixels)
  • 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
  • (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

Noise vs T
X resolution vs T
S/N vs T
?
11
RD 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, t25 ?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

12
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.
Write a Comment
User Comments (0)
About PowerShow.com