The Study Of Dynamic Accuracy Of High-speed Analog-to-Digital Converters

The high-speed analog-to-digital converters (ADCs) was widely used in all kind of fields. The request of dynamic accuracy was always a core problem in the data acquisition and treatment. The main dynamic performance involved in this article is the effective number of bit (ENOB) of ADC. This article introduces two practical measurements of ENOB: the sine curve fit test and the discrete finite Fourier transform (FFT) test. We analyze the two tests by means of theory and practical test in order to investigate their similarities and differences. Firstly, after the two tests of ENOB were analyzed in theory, it was proved that the two tests are equal in using sine wave as signal. We introduced the FFT test of ENOB in detail, studying the accuracy in amplitude of FFT and comparing some method of improving the accuracy in amplitude of FFT. It was discovered that, provided one period with few data number, it will obtain higher accuracy in amplitude by means of Hanning-window. We analyzed FFT spectrum of ideal analog data with quantity error, theoretical data with nonlinear error and error from limit slew rate of amplifier. We also analyzed FFT spectrum of data with combined errors. It was discovered that the quantity error make influence on all harmonics, error from limit slew rate of amplifier only make influence on all odd harmonics except first harmonic while the nonlinear error only make influence on all even harmonics. The FFT spectrum of data with combined errors is basically the combination of FFT spectrums with different errors. We introduced the sine curve fit test, simplifying the complexity by means of intuitive fitting. Moreover, we studied also a more complicated fit including first harmonic,second harmonic and third harmonic. After analyzing many experimental data, it is showed that the results of ENOB by way of simple sine curve fit are low and all less than the ones by way of FFT test. While the results of ENOB by way of synthesis sine curve fit are high and all more than the ones by way of FFT test because all second harmonic and third harmonic were eliminated. Because the values of ENOB by FFT lie between the upper limit and lower limit of ENOB by fitting test, the experimental data agree with the theoretical conclusion basically: the two tests are equal in using sine wave as signal. The improvement of the dynamic accuracy of one self-making data acquisition system with replaceable operational amplifier is introduced. Experiments showed that the gain-bandwidth of the subtracting operational amplifier is the key of hardware method. After analyzing third harmonic, it is showed that, in some range of definite frequency, the increase in the gain-bandwidth of the subtracting operational amplifier will effectively reduce the amplitude of third harmonic. Experiments showed that FFT filter smoothing is an effective software method for improvement of dynamic accuracy. However, larger number of the FFT filtering points makes reduction of the amplitude of the smoothed data of sine wave, which makes the improvement of dynamic accuracy by FFT filter smoothing limited.