edfas.org 19 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 28 NO. 1 of 240 Hz while the laser position was over the ROI and 20 kHz everywhere else. All the steps described above occur in real time autonomously with no user interference required. The identified ROI covers ~17% of the total scan area and SAILS reduced the time to capture a single frame by 81%. EXPERIMENTAL ANALYSIS: ARRAY-BASED IMAGE RECONSTRUCTION Figure 7 presents the experimental results of actual SDL performed on a DUT. The left side of the figure displays the six-frame average with laser power set at 100% and 33%, while the right side illustrates the generated ROI masks with a 5% frame area, using two different approaches. When the laser power is at or above 40%, both approaches are capable of generating ROI masks with 100% site coverage. The primary difference between the two approaches becomes apparent when the laser power is reduced to 33%; the array-based ROI mask displays a site coverage of 28%, while continuous averaging yields 0% (as indicated in the table in Fig. 7). Although this experiment demonstrates that the arraybased approach generates better outcomes for weak site detection, further case studies are necessary to confirm these findings. A variety of factors, such as AOI drift, optical focus, and image resolution, may significantly affect the image SNR and overall quality. Nevertheless, the experimental results in this study indicate that the array-based approach provides higher site coverage compared to the traditional approach. A revised automated SDL workflow consists of three stages. In the first stage, a normal full area ROI is conducted for the initial six frames to acquire frames necessary for site detection. Subsequently, the frames are processed to generate two ROI masks, one with a 5% area and another with a 35% frame area. The 5% area ROI mask is applied first to enable rapid frame acquisition for strong sites. Once the strong site(s) have been captured, the process Fig. 6 Rough scan (top), smoothed scan (middle), and site detection algorithm generated mask (bottom). Red squares highlight areas in rough scan with anomalistic intensity, white polygons in mask were chosen as ROIs. Fig. 7 Six-frame averages at 100% and 33% laser power levels (left) and their corresponding 5% frame area ROI masks (right), with site coverage comparisons under different site intensity (bottom table). Fig. 8 Capture frame vs. SDL acquisition time for different ROI masks: Utilizing a combination of 5% and 35% ROI masks within the data acquisition workflow maximizes frame count for detecting both strong and weak site(s).
RkJQdWJsaXNoZXIy MTYyMzk3NQ==