| Arbitrary waveform generator is a widely used stimulus source in many fields such as communication and radar,as it can generate diverse test signals according to user demands.The core of arbitrary waveform generator is direct digital synthesis,which requires a digital-to-analog converter to convert digital waveforms to analog waveforms.However,the hold characteristics of digital-to-analog converter can cause amplitude-frequency rolloff of the output signal,and the non-flatness amplitude-frequency characteristics of the following analog channel will worsen the signal distortion.Therefore,this thesis studies the error correction method of the abovementioned problem,designs a pre-correction module of a 5GSPS sampling rate arbitrary waveform generator to correct output flatness within ±1.5d B for 1GHz band signal.The main contributions are as follows:1.Design of high-speed parallel filtering structure.An amplitude-frequency correction module based on multi-channel parallel filter is proposed to break the limitation that a 5GSPS high-speed data cannot process directly in arbitrary waveform generator.A 16-parallel design can reduce the filter rate to 312.5MHz.To reduce the resource consumption of the filter,an optimized design method based on iterated short convolution algorithm is researched,which reduces the number of sub-filters in the parallel structure.Compared with the traditional 16-parallel filter structure,this method can save 68.36% sub-filters.2.Optimization design of filter coefficients.Acquisition of the target band’s amplitude-frequency response is a premise for designing the correction filter.Spectrum analyzer can aquire multiple points’ amplitude-frequency response information,but its accuracy in low frequency range isn’t well.To solve this problem,a joint measurement of the spectrum analyzer and power meter is used to obtain the target band’s amplitudefrequency response.A linear programming based maximum-minimum design method is used to optimize the filter coefficient,take minimizing the frequency response error between the designed filter and the ideal filter as the optimization goal,a 248 order filter can be designed.Analyze the quantization error of the filter,quantize the coefficients to16 bits so that the frequency response error between the designed filter and the ideal filter is less than 0.01 d B.3.Design and implementation of a filter coefficient reloadable amplitude-frequency correction module.An amplitude-frequency correction module based on a transposed structure sub-filter is embedded in the digital-to-analog conversion interface module of the arbitrary waveform generator system.To meet the portability requirement of the system,a fast multiplier based on the Booth-Wallace algorithm is designed to implement multiplication in the filter.To solve the problem that analog channel’s different amplitude bands have different amplitude-frequency responses,the filter coefficient can be reloaded using block RAM.The implementation of a 16-parallel filter is completed? timing optimization techniques such as pipelining and parallelization are implemented to make the system timing more easily satisfied.Tests are conducted on multiple amplitude points of a 5GSPS sampling rate arbitrary waveform generator,and the results show that the pre-correction module can effectively compensate the amplitude non-flatness of the system.The module can correct the about 10 d B amplitude non-flatness of 750 m Vpp,300 m Vpp,and 50 m Vpp amplitude points within 1GHz to below 1.5d B. |