August_EDFA_Digital

edfas.org 33 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 22 NO. 3 caused by the variation in brightness value, that is, large noise. In the ultrasonic image obtained by using pulse compression (Fig. 6b), the boundaries of the 100-μm defects were clearer than in the image obtained by using pulse waves, and defects smaller than 100 μm could be observed. For quantitative evaluation, the defect-detec- tion sensitivity was evaluated by comparing the standard deviation (σ) of the brightness value in the bonded area and SNR in each defect area of the ultrasonic images. Here, the signal values (S) of the SNR are defined as the brightness value in each defect part, and noise values (N) are defined as 3σ. Figure 6c shows A to B cross-sectional profiles of the bonded region. As a result of comparing the cross-section- al profiles, the image obtained by using pulse waves was noisy, and the standard deviation 3σ from the extracted cross-sectional profile was 0.03 for the pulse waves and 0.01 for the pulse compression. As can be seen from these results, pulsecompressionreducedthenoisevalueby66%. Figure 7 shows the relationship between defect size and SNR calculatedwith the ultrasonic images shown in Fig. 6. As shown in Fig. 7, the value of SNR was high as a result of applying the pulse compression. For example, at the 100-μm defects, the SNR was 12.2 for the pulse compres- sion and 2.3 for the pulsewaves. Here, defining the thresh- old at which defect detection is possible at SNR = 2, it was found that the 100-μm defects could be detected with the pulse wave method and 20-μm defects with the pulse compression method. This result proves that the spatial resolution was improved in the SAT images where the pulse compression method was applied. CONCLUSION The defect-detection sensitivity of ultrasonic inspec- tion was improved by developing a CMUT probe through hardware development and a signal processing technique using a pulse compression method through software development. To improve the lateral resolution of the CMUTprobe for the through-transmissionmethod, broad- band and highly sensitive CMUT cells were developed. An ultra-narrow gap with a height of 30 nm was imple- mented to enhance the receiver sensitivity in the CMUT Fig. 6 Ultrasonic images and cross-sectional profiles of bonded region obtained using pulse waves and pulse compression method. Fig. 7 Comparison of signal-to-noise ratio calculated from SAT images. (a) (b) (c)