Issue link: https://iconnect007.uberflip.com/i/241303
article influence of via stub length and antipad size on the insertion loss profile continues marginally above 10 GHz. To quantify the effect of the two parameters' antipad size and stub length, the numeric output from the ANOVA evaluation is used. The data show, that the stub length accounts for 98% of the variation in the resonance frequency, where the antipad size has an effect of less than 2% (Figure 18). Another ANOVA was performed to investigate the influence of panel number, PCB number, antipad size and stub length on the absolute insertion loss value. This can be done for every frequency in the captured data (10 MHz to 40 GHz). Here only the data for 5 GHz and 10 GHz are presented as an example. For both frequencies, there is hardly any variation over the PCB number/location of the coupon on the panel. Some variation can be seen between the five manufactured panels. Again, the antipad size has a small influence, with the larger clearances causing less insertion loss. The main contributor is the stub length, causing an increase in the single ended insertion loss from around 4.5–5.5 dB at 5 GHz. The ANOVA main effect plots for 5 GHz and 10 GHz are shown in Figure 19 and Figure 20. Using the numeric output of the ANOVA at 5 GHz (frequency chosen as one example), shows the panel to be a minor influence causing only 1.8% of the variation. The antipad size is a second order influence with an effect of 11.3% and the stub length is again the major influence, being the cause of 83.2% of the variation, as shown in Figure 21. Differential Insertion Loss Testing The setup for the differential insertion loss testing can be found in Figure 22. A 4-port vector network analyzer was calibrated at the connector interface to the device-under-test with an electronic calibration module. The use of the eCal module lead to a significantly faster, easier and virtually error proof calibration process, especially for the 4-port calibration. On the test board, the interface to the VNA was provided with flange mounted compression type connectors. Figure 23 shows a screenshot of two different stub lengths superimposed. The upper por- Figure 20: Main effect plot for insertion loss at 10 GHz. January 2014 • The PCB Design Magazine 55