Transcript The CMN module

ORIGIN OF THE CMN IN TOB/TEC MODULES
•A PLAUSIBLE EXPLANATION BASED ON MEASUREMENTS
PERFORMED ON 1 (AND ONLY 1) MODULE.
•Guido Tonelli
•Laura Borrello
•Mariarosaria D’Alfonso
•Lino De Maria
•Suchandra Dutta
•Alberto Messineo
•Giusy Valvo
•B. Caltabiano
C. Cerri, A. Profeti, P. Mammini
Guido_Tonelli _Tracker_Week_April_21_ 2004
1
Outline
•Introduction.
•Preliminary diagnostic analysis on TEC Module #30200020020516
•Electrical measurements.
•Study of the dependence of CMN and leakage current from the
potential difference DV between strip implant and metal.
•Analysis of the time evolution of the CMN vs DV.
•Simulation of the effect at the device level.
•Consistency of the explanation with other observations.
•Possible actions.
•Conclusions.
Guido_Tonelli _Tracker_Week_April_21_ 2004
2
The Common Mode Noise
Module 5081(Sensors 23974205
23974314)
Channels: 20 at 300 V and 420 at
30200020005081
Bias Current (nA)
2500
2000
Current(DB)
1500
Current(probing)
1000
Current(Bonded)
500
0
0
100
200
300
400
500
Voltage
One or more strips exhibiting a noise incompatible with the leakage
current; so large that it affects the entire chip (sometimes in unrecoverable way).
Strong correlation with increase in sensor leakage current with
respect to sensor QTC data.
Guido_Tonelli _Tracker_Week_April_21_ 2004
3
Attempt to use sensors grading (W7B)
1.5 mA
1.5 mA
AA
A
Kink in IV
Grade AA : Itot < 1.5 µA (and no kink)
Grade A: Itot < 1.5 µA (and kink)
Grade B: Itot > 1.5 µA
B (different scale)
1.5 mA
Contractual limit: Itot @450 V < 10mA
Guido_Tonelli _Tracker_Week_April_21_ 2004
4
CMN modules and sensors grading.
Sensor
2001-2
2002-3
Total
2003
Grade
#
CMN
%
#
CMN
%
#
CMN
%
#
CMN
%
GRADE A+
32
1
3.1%
4
1
25.0%
12
0
0.0%
48
2
4.2%
GRADE A
42
2
4.8%
11
1
9.1%
16
1
6.3%
69
4
5.8%
GRADE B
22
3
13.6%
10
2
20.0%
1
0
0.0%
33
5
15%
Total
96
6
6.2%
25
4
16%
29
1
3.4%
150
11
7.3%
Over 150 modules 11(7%) exhibit CMN.
Grade B sensors develop CMN in about 15% of the
cases. Grade AA and A develop CMN in about 5% of the
modules.
No additional selection criteria found so far to further
reduce this fraction.
Guido_Tonelli _Tracker_Week_April_21_ 2004
5
The Common Mode Noise
The general quality of the first 300 TOB/TEC modules is good but
the production yield is around 90-95% (to be compared with 95-97%
of the TIB modules).
In addition a part of the community attributes this effect to an
intrinsic weakness of the STM sensors and believes to have
collected enough evidence that this effect will propagate with time to
an important fraction of the tracker.
Hence the actions to revise the STM contract (7000 sensors
ordered to HPK); to slow down in using STM sensors to build final
modules; to stop the STM production to initiate a new qualification
procedure (new pre-series of 1000 sensors).
Guido_Tonelli _Tracker_Week_April_21_ 2004
6
Proposed explanations
• intrinsic micro-discharge due to STM implant technology
• effect of the leaky strips
• in-appropriate handling
• mechanical deformations (vacuum effect)
• sensitivity to humidity
• abnormal time structure of the leakage current
• degradation of the leakage current with time?
Most/all explanations were concentrated on some intrinsic weakness of
the STM sensors BUT NO CONCLUSIVE EVIDENCE HAS BEEN
FOUND SO FAR.
Guido_Tonelli _Tracker_Week_April_21_ 2004
7
The CMN module: TEC 30200020020516
Built and tested in UCSB. Re-tested in Vienna.
Received in Pisa on March 14-th.
Investigations started on March the 22-nd.
Guido_Tonelli _Tracker_Week_April_21_ 2004
8
Preliminary diagnostics: initial tests
I @ 450 Volt = 13,3 mA (1-2mA expected from QTC)
Noise of strip #96 @ 400V = 60 ADC counts ! Ch 96 noisy already
@ 30V; CMN switched on @80V ! Strip #360: normal leaky strip (no
CMN). The results reproduce perfectly the data obtained at UCSB.
Guido_Tonelli _Tracker_Week_April_21_ 2004
9
General optical inspection
Several small irrelevant mechanical damages spotted.
The only relevant defect is a scratch on the bias ring (and on one
strip-normal in terms of noise-). The scratch is mostly on the inner
part of the bias ring. Not conclusive.
No significant mechanical scratch on the back of the sensors.
Guido_Tonelli _Tracker_Week_April_21_ 2004
10
Optical Inspection of the critical strip
A suspicious defect on the poly resistor. Considered not relevant.
Guido_Tonelli _Tracker_Week_April_21_ 2004
11
Study of the mechanical deformations
The z positions of the two sensors have been measured under a
CMM with the module lying free on the granite table. The spread of
the measurement points is several hundreds mm. No evidence of
mechanical stress on the sensors similar to deformations due to
vacuum.
Guido_Tonelli _Tracker_Week_April_21_ 2004
12
Behavior with humidity and time
CMN
CMN
Measurements done using dry air in the test box (1% humidity APV ON)
and high environment humidity (in the probe station APV OFF and in the
test box APV ON). Some trend toward a reduction of the leakage current
with time at high humidity but still a factor 2/3 higher and CMN still
present in data. Basic independence of CMN from humidity and time.
Guido_Tonelli _Tracker_Week_April_21_ 2004
13
Study of different ground connections
A direct connection to ground of the silicon sensors was used.
Special bond to connect the sensors
directly to ground.
Removed the bonding
bias ring-pitch adapterground through the
hybrid
Different shielding schemes used (clamshell/CF plate/support plate
grounded or floating).
CMN still present in data. Basic independence of CMN from shielding
and grounding scheme.
Guido_Tonelli _Tracker_Week_April_21_ 2004
14
CMN due to the leakage current time structure?
The same module was studied in Vienna to look for anomaly in the
leakage current time structure.
CMN module (@450V; 13,3µA) Residual system noise
Zoomed View, red line depicts average system noise
CMN present already in data taken at 80V!! Not conclusive.
Guido_Tonelli _Tracker_Week_April_21_ 2004
15
Study of the interaction sensors-APV25
We then decided to concentrate the attention on the interaction
between sensors and read-out electronics at the module level
(talk with Lino at CERN at the end of the CMS week).
Investigation of the over-metal effect in CMS modules.
The issue is particularly important for large pitch thick detectors
(OB2).
It was found to be critical already during the R&D phase when most
of the work was done on HPK multi-geometry structures
(see Lino’s talk to review the issue).
Guido_Tonelli _Tracker_Week_April_21_ 2004
16
The idea.
+ 0.75V
1.5MW
2.2kW
R
The input of the APV is slightly positive +0.75V; for normal strip
leakage current Ileak<10nA the implant is practically at ground (a
few tens of mV for a total Ileak= 10mA ).
A metal over-hang at positive potential may induce breakdown.
Guido_Tonelli _Tracker_Week_April_21_ 2004
17
Cross-check
+ 0.75V
Measurement
points
1.5MW
2.2kW
R
We measured the potential difference of the two points during
module operation using special micro-bonding to external wires.
Guido_Tonelli _Tracker_Week_April_21_ 2004
18
Measurement of the bias-ring voltage
Guido_Tonelli _Tracker_Week_April_21_ 2004
19
Measurement of the input voltage of the APV
Several APV input channels measured during normal operations
with different detector leakage current.
Results consistent with expectation: DV=0.75V +- 40mV
Guido_Tonelli _Tracker_Week_April_21_ 2004
20
The measurement set-up
R
By using an external variable resistor to connect the sensors to ground
we can exploit the total sensors leakage current to increase the
potential of the bias ring. Resistors from 100kW to 7MW were used.
The CMN should disappears when restoring the correct potential
difference between metal strip and implant.
Guido_Tonelli _Tracker_Week_April_21_ 2004
21
CMN vs Vbias@ R= 0MW
Strip noisy @ 30V and CMN switched on @ 80V
Guido_Tonelli _Tracker_Week_April_21_ 2004
22
CMN vs Vbias@ R= 2MW
Strip noisy @ 150V and CMN switched on @ 350V
Guido_Tonelli _Tracker_Week_April_21_ 2004
23
CMN vs Vbias@ R= 5MW
Strip noisy @ 350V and NO CMN UP to 500V
Guido_Tonelli _Tracker_Week_April_21_ 2004
24
Summary table
R
LEAKY STRIP ON
CMN
ON
CMN FINAL
0 MW 30V
80V
2.3 ADC counts
1 MW 80 V
150V
2.0 ADC
2 MW 150V
350V
1.0 ADC
5 MW 350V
_____
______
Guido_Tonelli _Tracker_Week_April_21_ 2004
25
CM subtracted noise vs Vbias and R
Guido_Tonelli _Tracker_Week_April_21_ 2004
26
Normal behavior of the entire module
Strip 96 is now a normal
“leaky strip”.
The module is a grade A
module.
Guido_Tonelli _Tracker_Week_April_21_ 2004
27
Normal response to Led & Pulse Shape
Fiber spot
Guido_Tonelli _Tracker_Week_April_21_ 2004
28
IV vs R
The most impressive evidence is the strong dependence of the total
sensor current from the small potential applied between implant and
metal. A factor 7 reduction in the leakage current which is now
compatible with QTC data.
Guido_Tonelli _Tracker_Week_April_21_ 2004
29
Why a so small potential difference can
affect the sensors ?
•Because it affects the field distribution at the
edge of the implants.
•The metal over-hang at a positive potential
with respect to the implants favors the
breakdown.
Guido_Tonelli _Tracker_Week_April_21_ 2004
30
Metal over-hang
Metal lines 4-8mm wider than the corresponding implant strips (metal over-hang)
are adopted to increase the breakdown performance. We expect higher fields at
large pitch & small w/p. Over-metal moves the high field region from Si to SiO2
Vbrk(Si) = 30 V/mm
Vbrk(SiO2) = 600 V/mm
BUT THE METAL SHOULD BE AT THE SAME OR LOWER POTENTIAL WITH
RESPECT TO THE IMPLANT OTHERWISE IT COULD BE DANGEROUS
Guido_Tonelli _Tracker_Week_April_21_ 2004
31
Metal over-hang at positive V
The high density of field lines at the
I
implant edges yields to breakdown. 10mA
The strip leakage current increases of
orders of magnitude.
A current of a few mA flowing through 10nA
the poly resistor (1.5MW) increases the
strip potential to values higher than the
metal strip. We jump to the following case.
Guido_Tonelli _Tracker_Week_April_21_ 2004
V
32
Metal over-hang at negative V
The density of field lines decreases and I
the strip exits from breakdown.
10mA
10 nA flowing through the poly resistor
(1.5MW) are not able to maintain the
implant potential positive with respect 10nA
to to the metal.
We step back to the previous situation.
Guido_Tonelli _Tracker_Week_April_21_ 2004
V
33
Huge CMN: intermittent breakdown
The previously described mechanism can explain why this
“leaky strip” is so different.
A “normal” breakdown (10mA through a single strip) is not
sufficient to account for these effects (60 ADC rms noise and
CMN on the chip).
It is not a normal breakdown it is an intermittent phenomenon.
To study the time evolution of the process we analyzed the data
taken on the module in different conditions..
Guido_Tonelli _Tracker_Week_April_21_ 2004
34
Data taking in multiframe-mode
Huge variations (oscillations) in the noise behavior.
Silent periods followed by explosions.
Guido_Tonelli _Tracker_Week_April_21_ 2004
35
Time evolution of the CMN strip
#96
#95 and #97
#50
#106
Neighboring strip (95/97) and far away strips (50/106) plotted for
comparison
Guido_Tonelli _Tracker_Week_April_21_ 2004
36
Comparison with a normal TIB leaky strip
TIB
TIB
Noise amplitude distribution.
TEC 0MW
Guido_Tonelli _Tracker_Week_April_21_ 2004
TIB
37
Area vs Dt and number of flips
TEC@0MW
TEC@0MW
Guido_Tonelli _Tracker_Week_April_21_ 2004
TIB
TIB
38
TEC plots @ R=1MW
Amplitude
Number of flips
Guido_Tonelli _Tracker_Week_April_21_ 2004
Area vs Dt
Un-distinguishable
from TIB in all
variables.
39
Simulation
To understand better the mechanism in our devices, we have asked
STM to perform a detailed simulation for OB2 devices (need of using
the appropriate doping profiles).
The exercise was done for two different oxide charge density (roughly
corresponding to Flat Band Voltage of 1V-recent productions- and 5Vold sensors-). Details in Lino’s presentation.
Back bias
voltage
Qoxide
AC pad potential
Case
1.0·1010
charge/cm2
Ground
A
B
C
D
+ 500 V
1.5·1011
charge/cm2
Guido_Tonelli _Tracker_Week_April_21_ 2004
+ 10 Volts
Ground
+ 10 Volts
40
Electric field (detailed view)
Qox=1.5·1011 charge/cm2
Case D
Case C
Aluminum
Oxide
Guido_Tonelli _Tracker_Week_April_21_ 2004
41
Electric field across the implant
Case D
Case C
Case B
Case A
The combined effect of oxide charge
density and positive voltage on the
metal may account for more than a
factor 2 increase of the peak. A few
percent increase may account for a
small probability effect.
Guido_Tonelli _Tracker_Week_April_21_ 2004
42
Consistency with other observations
•The effect escaped our QTC measurements because in the
standard set-up the metal is floating (see Alberto’s talk).
•The appearance of CMN is more likely in old (high flat band) and
lower quality (B-C grade) sensors (the sensor grading could be
already an indication that some strips are close to breakdown).
•In OB2 should be more likely than in OB1.
•Observations by F. Hartmann (effect reduced after irradiation).
•Observations by T. Affolder (effect disappeared if the AC metal is
connected to ground).
•Observations by M. Poettgens (increase of the sensors leakage
current only when connected to the electronics).
•Remember that we are dealing with a small probability effect:
•5% of the modules exhibiting CMN means 2.5% of the sensors, or
2-3 strips over 51.200 (less than 10-4).
Guido_Tonelli _Tracker_Week_April_21_ 2004
43
Actions (1)
I would require the technical endorsement of the sensor group to
recommend a list of actions to be approved by the TSC/TIB.
1) Review and repeat our measurements on more CMN modules;
increase the statistical significance of our tests; add further
tests.
2) Implement/optimize a new procedure of testing with the metal
fixed at some positive potential (+5V ?) to identify potentially
weak strips (Alberto’s talk) and operate a screening.
3) Discuss with STM the possibility to adopt the same procedure
for the 1000 sensors to be delivered as new pre-series before
using them to build modules.
Guido_Tonelli _Tracker_Week_April_21_ 2004
44
Actions (2)
4) Use ourselves the same screening procedure on the sensors already
delivered that could be probably used without any risk for CMS.
5) Perform a stress test on the several hundreds of good TOB-TEC
modules already built. The test could be done by biasing the APV
with respect to a virtual ground brought at a positive potential of a
few volts with respect to the sensor ground.
6) Cure the CMN modules by simply removing the connection to the
APV of the noisy strip and bonding it to ground.
7) Study feasibility (and drawbacks) to use this feature to increase in
general the breakdown performance (or the lifetime) of the tracker
by means of a NEGATIVE BIAS APPLIED TO THE OVERMETAL
(APV INPUTS). We have discovered a new way of operating AC
coupled silicon detectors for extended breakdown performance
Guido_Tonelli _Tracker_Week_April_21_ 2004
45
Conclusions
As usual the work never ends!
Guido_Tonelli _Tracker_Week_April_21_ 2004
46