Characterization of arbitrary waveform generator by low resolution and oversampling signal acquisition

The paper deals with two relevant aspects of output section characterization of an Arbitrary Waveform Generator (AWG). The first concerns the difficulty of acquiring the output signal from Analog to Digital Converter with both higher linearity and resolution than the AWG under test. The proposed solution makes the problem simpler: oversampling with lower resolution. This result is achieved by means of an acquisition procedure to reconstruct the AWG output signal based on the zero crossing time sequence. Since the reconstructed signal is not uniformly sampled, proper algorithms in the time and frequency domain are proposed to evaluate also the dynamic parameters. The second is linked to the first and concerns the difficulty of evaluating the transfer characteristic. Indeed, the correspondence between the input code and the output sample of the AWG is avoided due to the acquisition procedure proposed, and the AWG operating mode. The proposed procedure allows the input code to be correlated with the corresponding sample of the output signal. It operates in iterative modality to achieve the valid statistical occurrences of the codes. Indeed, owing to the oversampling and the memory size of the acquisition hardware system, the number of samples could be lower than the number of expected output levels of the AWG. Numerical tests have allowed validation of the proposed iterative procedure in the case where the transfer characteristic is affected by nonlinearity, missing code and quantization noise. The experimental setup of the proposed procedure has allowed assessment of: (i) the effectiveness of the iterative procedure in evaluating the transfer characteristic; (ii) the lower resolution of the acquisition hardware system with respect to the direct acquisition. On the basis of the error caused by the numerical procedure and the noise introduced by each device of the experimental set up, the lower bound of the final uncertainty is deduced.