edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 26 NO. 1 44 material. She responded that the milling tool as that of ASAP from Ultra Tec was used for the sample presented but emphasized that several tools need to be optimized for the material type of the packaging and chips within the device. Jim Colvin from FA Instruments opened up the discussion to various sample preparation topics including artificial intelligence, small to tall devices, and how to handle sample preparation and backside thinning. Overall, the audience drove many of the discussions and was able to share valuable expertise on sample preparation, which the attendees can employ with their application needs. Three topics prompted several insightful discussions in the OFI User Group this year at ISTFA 2023. Dan Bodoh started the User Group introduction with the section of FA Roadmap specific to die-level fault isolation. The topics discussed covered several of the roadmap focus areas including general leading edge technologies and 2D/2.5D/3D package FI. William Lo, Nvidia, presented the first topic related to an x-ray technique potentially termed XADA, or x-ray assisted device alteration, which is similar to LADA (laser assisted device alteration). Last year, an investigation on ionizing x-ray’s effect on device timing was presented. The setup was reviewed first by Lo followed by comparisons of the timing shifts of XADA compared to LADA. Shifts induced by XADA were not as expected so the floor was opened to discuss potential causes. Instead, a question related to the recovery of the alterations was asked. Lo confirmed the XADA device shifts are permanent over, at least, a work session. Annealing was brought up by the audience as a potential way to reverse the alteration, but Lo was not aware of any experiments that tried this approach, yet. Using different voltages during the experiment was also asked and discussed. Power supply noise was the major detractor when trying other voltage regimes. However, additional experimentation could be done to potentially address this. Furthermore, using newly available x-ray tools could prove beneficial. The conversation progressed further into the primary motivation behind XADA: addressing the loss of OFI techniques due to backside power delivery. Many were reassured that laser and photon-based techniques are still not viable as-is. Spot size and beam steering of the x-ray was also a question. Lo estimated the spot size in his experiment was between 10-20 µm, which was small enough for his devices that ISTFA 2023 OPTICAL FAULT ISOLATION (OFI), TEST, AND DIAGNOSTICS USER GROUP Chair/Co-Chairs: Dan Bockelman, Dan Bodoh, Neel Leslie, and Kevin Distelhurst dan.bockelman@intel.com, dan.bodoh@nxp.com, neel.leslie@thermofisher.com, kevin.distelhurst@globalfoundries.com “...LASER AND PHOTON-BASED TECHNIQUES ARE STILL NOT VIABLE AS-IS.” were spread out by 50 µm. A scanning pattern was used since there was no reliable way to position the x-ray spot over a device of interest. Fleming Bruckmaier with Quantum Diamonds Sensor Technology shifted the topic with his presentation on quantum diamond microscopy (QDM). QDM images magnetic fields are unperturbed by most semiconductor chip and packaging materials. Bruckmaier reviewed the sensor setup which is a diamond with nitrogen vacancy which, with the proper setup, emits an optical image captured by a camera. A use case with results and a comparison of this sensor with a SQUID sensor and thermal imaging followed. The first question after Bruckmaier’s presentation was related to sensor resolution. The spacing of the NV centers in the diamond does not limit the resolution. Instead, the thickness of the diamond and the spacing of magnetic field center to the sensor are the primary variables. A quick survey of the audience showed that less than five users of current magnetic field imaging techniques were present. Extending the use of this sensor to other use cases were discussed. A full wafer sensor was one possibility but growing full wafer, high-purity diamonds is the current difficulty. AFM tips with the sensor was suggested instead. The viability of the technique was discussed further. Analyzing signals at the transistor level in a flip chip package was suggested. Stand-off of the sensor from the transistors is the major limiting factor. However, QDM
RkJQdWJsaXNoZXIy MTYyMzk3NQ==