Research And Implementation Of High-Speed High Precision ADC Test Algorithm With Parameter Spectrum Estimation

In recent years, analog-to-digital converter has a rapid development toward high-speed and high precision, which is immensely promoting the research and innovation of ADC test algorithm. In these years, to realize the fast and accurate test of the high-speed and high precision ADC has been the current problem to be solved. The main reason is the IEEE standard requires long test time to finish the test of high-speed and high precision ADC, which limits applications in the field of engineering. Therefore, to research a new method of the high-speed and high precision ADC has now been one of the hot research topics in the mixed signal test society.In this thesis, the traditional parameter spectrum estimation is studied and improved, including frequency estimation optimized by iteration and dynamic input-output characteristic curve modeled by Chebyshev polynomials. For a given high-speed and high precision ADC, dynamic input-output characteristic cure of the ADC is expressed as a relation of both output code and analog input with a sum of first and second kind Chebyshev polynomials. Then least-squares fitting approach (LS) is selected to calculate the coefficients of the first and second kind Chebyshev using the known input signal and ADC output data stimulated by sinusoidal signal. Finally, the dynamic input-output characteristic cure of ADC will be acquired quickly and dynamic INL can be estimated. The reduced measurement accuracy is the price for fast test. For the purpose of improving of accuracy of estimation, the optimal estimation error of the proposed method is researched in details. A 14-bit and a 16-bit ADC are chosen to verify the proposed method by simulation, and a 14-bit AD9648 provided by ADI is selected as DUT to experiment. The experimental results show that the dynamic INL estimation error of 6 MHz and 10 MHz frequency input are 0.3849 LSB and 0.3885 LSB, which improves by 51.6% and 58.8% comparing with dynamic estimation method of frequency-domain. In the experiment, the samples of the proposed method reduce by 2% of improved traditional histogram, which shortens the test time significantly.Finally, these research results are of great practical value for the validation and testing of analog-digital mixed-signal devices.