Report from Module Bonding Working Group Meeting

1
Report fromModule BondingWorking Group Meeting

• Salvatore Costa
• Università di Catania and INFN Sezione di
  Catania

2
Outline

1. Global info on production bonding of the Modules
2. Global review of Module bonding quality
   indicators
3. Short summaries on specific topics discussed at
   the WG Meeting

3
Globalinfo on production bonding review of
Module bonding quality
4
Production Summary
after Hyb prod resumed with stiffener added to
kapton cable
5
Production Rate

• It continues to be globally true that Bonding is
  not a bottleneck in the production flow.
• In all centers modules are bonded within
  single-digit number of days from reception

6
Module Pull Tests Performed
Results in DB from 13 Centers out of 14 (up from
9 three Mtgs ago)
Modules pull tested after 01Dec2003 (since last
Mtg)
Center PA TA PA-Sen Sen-Sen
Bari 77 (67) 80 (69)  n/a
Catania 47 (45) 47 (45) n/a
Firenze 9 (9) 5 (5)  n/a
Padova 0 (0) 34 (33)  n/a
Pisa 7 (1) 34 (27)  n/a
Torino 114 (63) 46 (40)  n/a
Fermilab 110 (109) 29 (7) 29 (8)
Santa-Barbara 42 (36) 42 (35) 42 (35)
Aachen
Hamburg 20 (15) 20 (15) n/a
Karlsruhe 0 (0) 20 (0) n/a
Strasbourg 0 (0) 27 (27) 27 (27)
Vienna 7 (1) 16 (10) n/a
Zurich 0 (0) 39 (30)  n/a
7
Pull test results
5g
BA CT FI PD PI TO FL
FL SB SB HA KA ST ST VI
ZH .
ss ss
ss
8
Bonding problems

• We had no reports of pathological problems.
• Non-severe problems
• Catania changed to a new toot type for last
  module bonded with parameters that result into
  reliable bonding, but still not optimized, had
  much higher pull forces (13-15 g) but all
  lift-offs vs. heel breaks.
• Firenze complained that the filter capacitor gold
  pad was particularly dirty in a few recent
  modules, resulting into failing bonds.
• Karlsruhe got their first 38 CMS modules (TEC
  R3) reported standard bonding failures in
  about 2/3 of the modules, traced to
  vacuum/support problems
• feedback given for jig improvement
• Vienna warned that Hybrids in their modules
  (bonded at Fermilab) have power bonds in groups
  of 3 vs. 5 present in Hybrids bonded at CERN.
• Zurich reported that out of 108 TEC R4 modules
  they bonded, a few had floating sensors and
  needed special care when bonding.

9
Selected highlights from theBonding WG Meeting

• of 08 June 2004

10
Pull test data comparison

• Direct comparisons between centers must involve
  correction for loop angles
• Launched a campaign to document loop in PA test
  areas, typically involving these steps
• Document the original bond
• Straighten bond to (near) triangular shape
• Measure the angle at the bond foot
• Determine correction factor for pull strength
• 5/14 centers have already responded and provided
  their loop doc at the Meeting.
• Will gather missing info and then use it to scale
  data in PA TA plot.

11
Pull Force correction factor
Correction factor a 1 / (2 sin?)
q (º) corr. factor a
10 2.88
20 1.46
30 1
40 0.79
50 0.65
60 0.58
70 0.53
80 0.51
90 0.5
For symmetrical bond
12
Loop angle ? Correction factor

• Catania

Correction factor 1
13
Loop angle ? Correction factor

• Padova

? arctan(525/900) 30º
Correction factor 1
14
Module Repair Centers

• We were requested to discuss the criteria to
  decide when to send a Module to the new official
  Repair Centers, from a Bonders point of view.
• I will report in detail to TPO. General remarks
• We perceive these centers more as debug places
  where teams of dedicated people study in deeper
  detail Modules with non-trivial problems, so that
  normal production can continue at the usual rate.
• Strictly speaking, trivial mechanical Module
  bonding failures should not be matter for these
  Centers a) there is nothing to study b) bonding
  centers (operators) are either able to repair
  them right away or, if they cant, then most
  likely the debug centers cant either.
• To repair damage to APV-PA bonds (typically
  caused by handling accidents), send Module back
  to the Hyb bonding place.
• For all other cases, which means a variety of
  electrical misbehaviors that may or may not be
  related to bonding problems, report the problem
  and let the experts at the Debug Center decide
  if they think they want to study that Module
  report procedure

15
Module grading in Tracker(DB)

• ModTest WG has clear criteria to grade Modules
  according to number of bad channels and other
  electric properties grade A, B, C, F
• Gantry is in the process of implementing a
  similar grading scheme based on the achieved
  alignment of components on the Module
• Bonding de facto has already a similar scheme,
  although we never called it grading.
• From the Bonding operation, we issue a negative
  flag in TrackerDB only if the Module is
• Declared unbondable or however unusable at
  Pre-bonding inspection
• Declared unusable at Post-bonding inspection
• Currently we never issue a negative flag based on
  quantitative variables
• Number of unbonded strips
• Number of bonding failures
• Pull test results
• We do assign, however, different positive flags
  based on number of unbonded strips (grading) 0
  lt1 unbonded Sensor strips
• 
  1 1-3 unbonded Sensor strips
• 2
  gt 3 unbonded Sensor strips
• These thresholds were decided quite arbitrarily
  about 1.5 years ago.

16
Grading from Bonding (preliminary!)

• We agreed on
• Keep Flag -1 if module found unusable at pre- or
  post-bonding inspection
• Not use pull test values for grading because it
  would require pull test on all modules
• Adjust our current grading based on unbonded
  strips in order to match the ModTest criteria
• 0 lt1 unbonded Sensor strips (matches Grade
  A)
• 1 1-2 unbonded Sensor strips (matches Grade
  B)
• -1 gt 2 unbonded Sensor strips (matches Grade
  C) and perhaps -2, -3, for different reasons
• We debated but concluded we need more time before
  we can come up with a well-thought scheme
• Including levels of grading based on indicators
  of the mechanical quality of the bonds, such as
• The number of repaired failures
• This is delicate because to be meaningful it
  must include a breakdown of the reported repaired
  failures based on reasons for the initial failure
  and repair manner.
• The number of bias bonds that could actually be
  made

Any changes to the grading scheme can only be
implemented with a new DB I/F version. We will
have to make a script to automatically change in
DB the flags for existing data where appropriate
(should be just a few).