A Case Study in Ultrasonic Cleaning Damage

Authors:

Mr. Mallette received the BSEE and MSEE degrees from the University of Central Florida in 1975 and 1977, and the MBA degree from Pepperdine University in 1985.  He has worked for NASA. Martin Marietta and as a consultant.  He was an Instructor of Engineering at UCF before joining Hughes Space and Communications Company in 1978 where he is now a project manager: managing high-technology subcontractors.  He has written over 30 articles on devices, dendrites, optical detectors, circuits, atomic frequency standards, ground systems, satellite systems, finance, genealogy, antiques and one book on quilts.

Dr. Chen received the BS in Engineering (1986) and the MS (1989) and PhD (1993) in Materials Science and Engineering from the University of California at Los Angeles.  He has been with Hughes Aircraft Company for over 10 years, and is currently a team leader in Failure Analysis.  He has written and contributed to articles on internal nitriding, fracture toughness testing of ceramics, reliability prediction, and geology.  He is also a member of Sigma XL.

 Mr. Johnson received a BSEE degree from Texas A&M University and MSEE from the University of Southern California.  He has over 30 years experience in design and development of electronic equipment for space applications.

Abstract -- A single lot of 32 SAW (Surface Acoustic Wave) filters, exposed to an undocumented, post-seal, ultrasonic cleaning process, resulted in latent damage that was not discovered until late in the build cycle of a space qualified frequency synthesizer.  The integrity of space-qualified hardware requires the mandatory identification of root cause for all failures.  The 35th frequency synthesizer in the hardware build in a multi-satellite program was subjected to acceptance level vibration testing and was observed to have a spectrum breakup.  The synthesizer performed normally before and after the vibration.  The failure was isolated to the facture of a 1 ml diameter gold wire in a SAW filter.  The wire, although fractured, was in contact with the ball-bond and maintained good electrical contact.  SEM (scanning Electron Microscope) examination showed evidence of significant high cycle fatigue on the fractured wire and varying amounts of fatigue (none to major) on other wires in the same filter.  An investigation team prepared an Ishikawa (fishbone) diagram and eight plausible hypotheses were identified for possible root cause.  This unique investigation was compounded because the filter was built eight years earlier -- the investigation required significant deductive reasoning to reconstruct the facts.  A total of 31 filters (14 lot acceptance test samples and 17 flight filters) from eight lot-date-codes (LDC) were examined in the SEM; testing was used to simulate several of the failure scenarios, the build history was reconstructed and inter/intra lot evaluations were performed.  The analysis of the information led to a clear probable root cause and eliminated the reachback concern to other lots on other flights.  It was concluded that a single lot of 32 filters was exposed to an undocumented, post-seal, ultrasonic cleaning.

Background 

There was a breakup of the output frequency spectrum during the z-axis random vibration of synthesizer 35.  This failure occurred after the unit had passed y-axis vibration (only).  The unit was returned to the lab and troubleshooting was performed to try to observe the vibration failure.  The inability to recreate the failure resulted in the investigation relying on visual inspections and repeated low level vibrations.  Low-level vibration localized the failure to the DDS chain in the synthesizer.  Testing in this area of the tray resulted in the failure being isolated to a SAW module in the DDS chain.  A new SAW filter module was installed in the tray and the unit passed all subsequent testing including vibration. 

SAW Investigation 

The suspect SAW module that was removed from the tray was tested further under three axis random vibration and a sine sweep.  This positively isolated the failure to the SAW filter (within the SAW module) 

An Ishikawa (fishbone) diagram was prepared, and eight plausible hypotheses were identified for possible root cause.  After eliminating several bones on the fishbone by testing, the filter was delidded and internally examined (figure 1).  The two large rectangular devices with circular features are inductors and the smaller rectangular devices are capacitors; these are used to tune and match the input and output of the filter.  Approximately 23 gold wires are used to configure and tune the filter.  The number of wires varies slightly depending on the number of tuning wires on the capacitors.  The suspect filter was examined in the SEM.  This revealed a break in the output wire (figure 2); the break was located on the inductor end of the wire and showed evidence of classical fatigue.  The input wire also showed evidence of severe fatigue damage but it had not fractured completely.  Three other wires were observed with varying levels of fatigue damage, ranging from minimal to considerable.  This filter was from LDC 8852.

Build History

A review of the lot travelers showed no evidence of ultrasonic cleaning.  There was no pattern to the rejected parts to suggest some other problem. There was nothing in the travelers that showed any significant difference between LDC 8852 and the other lots.  The significant data that was found in the travelers is summarized in the bubble chart (figure 3) showing the relationship between LDC 8852 and LDC 8850.  Since the eight parts that were split off from the group that became LDC 8850 to become part of LDC 8852 also exhibited the problem. It is certain the problem was not caused by the bonding operation.  It is hypothesized that an undocumented ultrasonic cleaning procedure (applied to the entire lot at some point after final electrical test at the vendor [1988] and prior to module build [1996]) caused the problem.  There is nothing to have prevented the same procedure from being used more than once.  However, the reason for believing parts from other lots are acceptable is that no evidence of the high cycle fatigue has been found in parts from any other lot date code, and with the affected lot (LDC 8852), a very high percentage of the parts showed fatigue damage.

Simulated Exposure to Cleaning

This test was designed to verify that ultrasonic cleaning could cause damage to this part and in the same manner as seen in the failed filter (from LDC 8852).  Two parts were inspected by SEM, sealed, immersed in an ultrasonic cleaner for 10 minutes, delidded and reexamined in the SEM.  Six wires were found to have identical damage as seen in the LDC 8852 filters.  The ultrasonic cleaning experiment positively proved that an ultrasonic cleaning step would have caused the fatigue damage described above.

Structural Analysis 

A structural analysis of the gold wire was performed and it identified that the gold wire (size, lengths, geometry and hardness as in the LDC 8852 filters) was susceptible to energy in the ultrasonic region.  However, it would have to be of sufficient magnitude (which is lower for the higher modes) and of sufficient duration.

A finite element model was generated based on the expected configurations of the wire.  This gold wire model consists of 76 bar elements and 78 nodes.  Fatigue analysis was then performed to determine the maximum stress according to the wire exposed to a 10-minute ultrasonic environment, similar to the simulated exposure test above.

The structural analysis has shown that it is possible for this failure to occur during the cleaning process.  Based upon the wire bond resonant frequency graphs, it shows that the critical resonant frequency range is around 2,000 to 7,000 Hz for a gold wire with a 1 mil diameter and 120 mil length.  The model analysis results indicate that only one mode (2570 Hz) is within the critical range.  It is assumed that 2570 Hz was the resonant mode and the wire was excited for 10 minutes during the ultrasonic cleaning process.  Therefore, the fatigue cycles was calculated to be 1.54 x 10 cycles.  This yields a maximum strain of 0.0015.  The equivalent tensile stress and the equivalent tensile deformation were calculated t be 18,000 psi and 0.19 mil, respectively.

Physical Analysis 

A total of 31 filters from LDC 8852 were inspected.  Of these, 24 showed damage clearly due to fatigue, one part showed damage most likely attributable to fatigue, and two showed very minor damage that may be caused by fatigue but was difficult to distinguish from wire bonder collet damage.  The remaining four filters did not show evidence of fatigue damage.  Table I summarizes the number of wires showing fatigue and the worst-case level of damage found within each filter.  The damage ranged from none (grade 0) through minimal, moderate, considerable and severe (grades 1, 2, 3 and 4, respectively) to completely fractured (grade 5).  An example of a grade 4 (severe damage, but not fractured) is given in figure 4. 

Impact to Other Lot Date Codes 

There were seven other lot date codes purchased from the vendor.  All available filters (up to three flight parts) were inspected in the SEM for evidence of fatigue.  No evidence of fatigue was seen in any sample from any of these lot date codes.  There were 27 filters evaluated -- five flight filters, eight virgin samples, and 14 lot acceptance test samples that had been destructively tested for wire bond pull strength.  However, the date from the lot acceptance test samples (with bond wires that had been destructively pull tested) may not be definitive, due to the difficulty in ascertaining fatigue damage in pulled wires.  It was found that the fatigue damage could not always be clearly identified, due to distortion from wire elongation.  With prior knowledge of the existence and severity of fatigue damage on a particular wire, the fatigue damage could be identified after wire pull testing, but without prior knowledge, the identification of fatigue damage would become less certain.

Disposition 

The hardware corrective action was to remove the SAW filters from LDC 8852.  The entire lot was purged from all hardware and stores and downgraded to non-flight status.  The filters removed from the flight synthesizers were replaced with SAW filters from the different lot date code  All synthesizers have been retested and there are no problems with the new SAW filters.  The SAW filters from other lot date codes have been exonerated by test end inspection.

The corrective action for the misuse of ultrasonic cleaners is to educate personnel and place warning signs on all ultrasonic cleaners at HSC. Samples of the warning signs have also been sent to the vendor.  The key is education.  Many engineers and technicians are not aware of the potential damage that ultrasonic energy could impart to electronic components.  The educational process musts be "on going" as new personnel are brought into the work force.  This has been entered into the HSC lessons learned database.

Conclusions 

The gold wire failure mechanism appears to be high cycle fatigue, probably caused by unauthorized, undocumented ultrasonic cleaning to an entire lot of completed SAW filters.  The filters were removed from hardware and flight stores.

Acknowledgment

The assistance of Jerry Meldrum in the SEM examination of many SAW filters is greatly appreciated by all the authors.

References 

1.            Harmon, G. G., Wire Bonding in Microelectronics, International Society for Hybrid Microelectronics Reston, Virginia, 1991.