Optimization And Implementation Of Adcs’ Static Test With Parametric Spectral Estimation

In recent years, the development of ADC is constantly moving in the direction of high-speed and high-precision, which makes the cost of testing ADC with standard test algorithms become increasingly high. And the need to collect two types of static and dynamic test data combined with the corresponding static and dynamic ADC test algorithm to measure all the parameters adds to the cost of testing ADC. Currently, the existing single test data collection to achieve all the parameters of ADC are two types, one is based on static parameters for dynamic testing, but such methods are based on the histogram test to achieve, So it needs a lot of points to be collected and only reduce little cost; the other is based on the dynamic parameters for the static test, and such algorithm is based on fit to achieve, So the desired sampling point is rare and can reduce a lot of costs, but the disadvantage of such method is low accuracy of the test result. As a consequence it will be great practical value to realize the test of all the parameters of ADC by a single data collection.Based on this, the parametric spectral estimation which based on the dynamic parameters for the static test is studied and is optimized combination with segments. Firstly, the model of the transfer curve of ADC is modeled by Fourier series and the first chebyshev equation in the optimized parametric spectral estimation. Then the transfer curve is dived several parts and we calculate the Fourier series at each segmentation. At follow, the transfer curve at each segmentation is connected and minus ideal transfer curve to achieve INL. Finally, a platform of ADC auto test which is based on 14 bit AD9648 is built and we verify the performance of the proposed method with this platform.It turns out that the INL estimation error of the parametric spectral estimation is 0.624LSB, if we regard 1.189LSB which estimated by histogram as the true INL of this platform. Compared with the parametric spectral estimation, the INL estimation error of the proposed method is 0.2953LSB.The accuracy of the proposed method improves 0.3303LSB. In addition, the proposed method does almost two millions fewer multiplications than the original parametric spectral estimation. And compare with the computation time of the original parametric spectral estimation which based on MATLAB is 2.605s, the computation time of the proposed method is only 0.2832s. The computational efficiency of the proposed method increases 90 percent.