Hi Takumi-san, I will try to provide brief explanations to your questions below: INL Q1-1) Correct, the INL error is an absolute error so it may be +/- 3ppm, which matches well with Figure 19. Q1-2) Yes...offset and gain calibration are typically performed first and then a straight line is drawn between the endpoints of the transfer function. The maximum deviation of the transfer function from this straight line is the INL error. Q1-3) This shape is partly due to the end-point INL calculation method but also because most CMOS switches are implemented using a parallel combination of NMOS and PMOS transistors. As the input voltage is swept, the signal will pass through one or both of these transistors and you see a symmetrical "S" shape for the INL error that depends on the input voltage. Noise Q2) Noise is measured on the bench. Either a fixed number of samples are collected or data is collected over a short fixed-time (10s or less). From the data, the RMS (standard deviation) or peak-to-peak (max - min) codes are computed and used in Equation 1 to calculate ENOB or NFB. I hope my explanations are clear. Please let me know if you have any additional questions! Best regards, Chris
↧