Trip to Teledyne on 1/5/05

 

Richard Gobin and Charlie Young

 

 

All recent MCM’s have been rejected due to excessive cracked traces.  The primary purpose of this trip is to investigate possible causes.  Four people from SLAC participated:

 

Richard Gobin

Jerome Lepulu

Albert Nguyen

Charlie Young

 

There were four people from Teledyne:

 

Roger Briere

Brian Caplen

Peter Hansen

Jose Luis Saldana

 

 

Cracked Trace Problem

 

Hypotheses and Tests

 

Teledyne has hypothesized that cracked traces are due to improperly radiused corners on the bare PWB’s when SLAC deburred them by hand.  We tested this idea by bonding standard Parlex PA’s to undeburred PWB’s. 

 

It has also been hypothesized that the grain structure of the Pitch Adapters from Parlex, which is orthogonal to the length of the traces, leads to cracks when the PA is wrapped around the PWB.  We tested a second source of PA’s with no obvious grain structure.  These PA’s are made by Dyconex in Switzerland.  We decided to use undeburred PWB’s also. 

 

Pitch Adapters

 

Robert was in touch with Dyconex some time ago.  Three pieces arrived at SLAC earlier this week.  There has been no P.O., and no paperwork came with the PA’s.  So we cannot be sure of their manufacturing details.  We put serial numbers on them for identification.  Dyconex has put the same label on two of them, and a different label on the third.

 

Serial Number (by SLAC)

Label (from Dyconex)

Y0001

3093/C

Y0002

3093/B

Y0003

3093/C

We do not know the significance of these labels.

 

The PA’s have been measured by Tom Nakashima at SLAC.  He measured the distance between first and last tooling holes, and the locations of the first and last traces.  They are all within the tolerance of +/- 0.100 mm. 

 

Unlike Parlex PA’s, the Dyconex PA’s have no obvious Copper grain marks to the naked eye or under the microscope.  There are no bridged traces.  They are no dimples or other discernable defects.  We found some foreign matter on top that were easily removed with a Q-Tip.  

 

We took pictures of all 3 Dyconex PA’s and 2 Parlex PA’s under the microscope.  See Appendices.  The differences are obvious.  Note that the Parlex PA’s have a piece of Kapton tape, which rendered the traces reddish and made the grain marks less obvious.  We took pictures of both Kapton covered and bare regions. 

 

The history of Parlex PA’s came up.  We understand that the thickness of copper was changed from the nominal ½ oz to ¼ oz some time in late 2004.  We believe that Teledyne does not have the latter ones.

 

PWB

 

Four PWB’s were used in the tests.  They are all Rev-10, and have not been deburred by hand.  Richard and Jose Luis examined them under the microscope, and noticed some minor imperfections.

 

Application of Epoxy

 

We started with S/N Y0001, and immediately noticed that the layer of epoxy was abnormally thin.  Repeating the application did not help.  The epoxy was cleaned off while we investigated. 

 

We tried a scrap Parlex PA.  Epoxy thickness appeared normal to our untrained eyes.  We then applied epoxy to two Parlex PA’s, and put them on to PWB’s using fixtures #2 and #3.  They were inspected at this point and found to have large numbers of cracked traces, similar to what was observed after curing for the pieces glued in December. 

 

Peter Hansen examined the epoxy application machine at this time.  He noticed the squeegee was not parallel to the base plate.  We tried another scrap Parlex PA after Peter adjusted the machine.  Epoxy thickness is increased – it was much harder to discern the traces under the epoxy.  Therefore, the epoxy for the two Parlex PA’s in fixtures #2 and #3 was probably not thick enough.  However, we cannot quantify it.

 

One Dyconex PA was then mounted to gluing fixture #1.  There are many cracked traces, concentrated on the right side when looking at the component side.  Where there are problems, there is typically one crack per trace. 

 

A second Dyconex PA was mounted to fixture #4.  We inspected the traces as we rotated the fixture’s arm.  Cracks were first noticed when the arm was at approximately 45o.  The cracks are preferentially on the right side.  There is typically one crack per trace.  There was some lively but inconclusive discussion at this point between Brian and Peter whether the hinging of the gluing fixture was contributing to the problem. 

 

Here are the serial numbers and fixture information:

 

MCM S/N

PA S/N

Fixture

2122

Y0001

1

2124

Y0002

4

2142

2162

2

2066

2188

3

 

Curing and Subsequent Inspection

 

All four fixtures were put into the oven for the normal 90-minute cure cycle at 85oC.  We also cured the two scrap pieces, which were not attached to PWB’s.  Due to lack of time, the MCM’s were examined shortly after coming out of the oven without waiting for the normal cooling time.  They were examined prior to trimming. 

 

The two using Parlex PA’s looked similar to what we observed previously with deburred PWB’s.  Most if not all traces are broken.  There are typically multiple cracks per trace. 

 

The two using Dyconex PA’s looked significantly better but still clearly unacceptable.  The left side is largely OK.  In fact, there are extended regions where the surface looked great.  We used light reflection pattern to conclude that these areas have very smooth and uniform curvature.  Cracked traces are concentrated on the right, typically one crack per trace.  As mentioned in a previous report, there are two lines of bright reflected light.  They are parallel and a fixed distance apart in the good region.  In the area of cracks, these lines of light are definitely closer together, and the cracks are within the lines.  [This may have been the case with Parlex PA’s, but it is hard to say given the number of cracks per trace.]  This suggests that the radius of curvature is reduced in the region of cracks.

 

Comments and Suggested Actions

 

Recent tests on gluing PA’s to PWB’s have all shown problems with cracked traces.  Here are some observations within the limits of low statistics.

 

  • When do traces become cracked?

1)     Cracked traces exist before and after trimming. 

2)     Cracked traces exist before and after cure cycle.

3)     Traces are cracked when PA’s are mounted to gluing fixtures.  We have looked at just one PA during this step. 

4)     There are no cracked traces initially. 

  • PWB issues.

1)     Manual deburring is not the basic problem. 

2)     There are surface outline imperfections. 

  • PA issues.

1)     Dyconex PA’s have no noticeable grain structure.

2)     Dyconex PA’s are better than Parlex PA’s in terms of trace cracking, but it does not solve the problem.  

3)     We do not know the specifications for the Dyconex PA’s, e.g. copper thickness.

  • Geometry issues.

1)     There is a correlation between smaller radius of curvature and trace cracking.

2)     This could be due to PWB geometry, epoxy, fixture, etc.

 

We suggest having a meeting to decide on the course of action.  Here are some ideas that came up during our discussions.  Some things may already have been done but we are not aware of them.

 

  • Dyconex.

1)     Find out what the specifications for the 3 boards we received.  Is there any difference between 3093/B and 3093/C?

2)     Acquire more PA’s. 

  • Parlex.

1)     Continue working with them to control dimensions.

2)     Work on reoriented grain direction. 

  • PA in general.

1)     Find out minimum recommended radius of curvature for our structure.

2)     Possibility of plating with a preformed radius.

3)     Test trace cracking dry. 

  • PWB.

1)     Measure radius of curvature from Diamond Tools.

  • Gluing fixture.

1)     Understand the detailed geometry of existing fixture.  For example, does it tend to stretch the PA when the arm is rotated?  What tolerance on, say, the pivot point, could lead to stretching?  

2)     Investigate alternative fixturing.

  • Analysis of failed boards.

1)     Select good and bad regions with Dyconex PA’s. 

2)     Section good and bad regions. 

 

 

Other Items

 

Rework

 

Teledyne has finished rework on all 20 pieces.  Albert tested them, and they all passed.  They will be shipped to SLAC later this week after inspection by Teledyne and by Jerome.    

 

PA Pull Test

 

As reported before, the pull tests have been done.  Brian has not finished analysis of the data. 

 

Degassing of Die Attach Epoxy

 

Roger Briere inquired about the degassing specifications.  The instructions we got from NASA and gave to Teledyne say to degas at 25” mercury for at least 10 minutes.  Roger has been in touch with the manufacturer, who recommends at most 2 minutes at 27” to 29” of mercury.  They advise against extended degassing as it affects the viscosity of the epoxy.  We have an action item to get feedback from NASA on degassing requirement. 

 

Encapsulation Test

 

Teledyne will put down dead ASIC’s and apply encapsulant to scrap MCM’s made in December. 

 

Clean Room

 

Construction appears to be over; there is no plastic sheet partitioning the workspace into clean vs construction.  Work is going on everywhere.  However, water leaks in the roof are not fixed.  We observed three areas with approximately 3’x3’ pieces of plastic that catch leaking water and drain into trash cans. 


 Appendix I.  Parlex PA S/N 2162

 

The pictures are from around U3 and U21 respectively.  The upper darker region is due to an overlay of Kapton tape. 

 

 

 

 

 Appendix II.  Parlex PA S/N 2184

 

The pictures are from around U13 and U21 respectively.  The upper darker region is due to an overlay of Kapton tape. 

 

 

 

 Appendix III.  Dyconex PA S/N Y0001

 

The pictures are from around U5, U11 and U25 respectively.  There is no Kapton tape. 

 

 

 

 

 Appendix IV.  Dyconex PA S/N Y0002

 

The pictures are from around U3, U9 and U16 respectively.  There is no Kapton tape. 

 

 

 

 

 Appendix V.  Dyconex PA S/N Y0003

 

The pictures are from around U8, U15 and U20 respectively.  There is no Kapton tape.  The last picture showed an imperfection, which we later determined to be some foreign matter on the surface.  It was easily cleaned off.