This phenomenon is quite reasonable. The fault filter works in a way like this: there's an internal high frequency clock F_filt for detecting the input voltage level, say if the input voltage stays one within continuous N periods of F_filt, the input is deemed to be one. For instance, suppose the input voltage is examined at each rising edge of F_filt, and N is set to 3, now see the attached pictures, the red waveform is the input signal on fault pin, and the blue one is F_filt. Both situation 1 and 2 can pass the filter, there's an error of +/- one period of F_filt.
In the case of 8036, F_filt is 96MHz/19 = 5.05MHz, so the period of F_filt is about 0.2us, which applies to your case.
This phenomenon is quite reasonable. The fault filter works in a way like this: there's an internal high frequency clock F_filt for detecting the input voltage level, say if the input voltage stays one within continuous N periods of F_filt, the input is deemed to be one. For instance, suppose the input voltage is examined at each rising edge of F_filt, and N is set to 3, now see the attached pictures, the red waveform is the input signal on fault pin, and the blue one is F_filt. Both situation 1 and 2 can pass the filter, there's an error of +/- one period of F_filt.
In the case of 8036, F_filt is 96MHz/19 = 5.05MHz, so the period of F_filt is about 0.2us, which applies to your case.
