Optimizing digital spectrometers for radioastronomical observations

TL;DR

This thesis develops a calibration procedure for FFT spectrometers that significantly improves mirror suppression by correcting frequency-dependent mismatches in time interleaved ADCs, enhancing dynamic range for radioastronomical observations.

Contribution

It introduces a three-step calibration method that optimizes mirror suppression across the entire frequency band, surpassing previous frequency-independent approaches.

Findings

Mirror suppression improved by up to 20 dB

Calibration reduces spectral artifacts caused by ADC mismatches

Enhanced dynamic range of FFT spectrometers for radioastronomy

Abstract

The goal of this thesis was to develop a procedure to optimize the mirror suppression, which reduces the dynamic range of the Fast Fourier Transform spectrometers (FFTS), and implement this procedure in the Field Programmable Gate Array (FPGA) of the FFTS. This is achieved by applying a calibration on the complex amplitude spectrum. Modern multi-beam heterodyne receivers are equipped with a large number of independent signal chains, which leads to a complex system that consumes space as well as power and is expensive. To reduce the complexity of the receiver array the intermediate frequency (IF) band is 4-8 GHz, which requires the FFTS to sample this band directly. The input bandwidth of 4 GHz of the current generation of FFTS cannot be sampled with a single analog-digital converter (ADC). To reach that bandwidth, multiple ADCs sample the same signal at different points in time. The method is called \"time interleaved sampling\". The characteristics of the ADCs differ from each other, which leads to mirror signals in the spectrum. These mirror signals can reduce the dynamic range of the spectrometers. Therefore the mismatches must be actively corrected. The current generation of FFTS use a frequency independent calibration, which optimizes the mirror suppression at one fixed frequency. This it is not optimal as some of the mismatches are frequency dependent and their impact increases with frequency. The procedure developed during this thesis work involves three steps. First the mismatches are measured over the frequency band. Then filter coefficients are calculated from these mismatches using a model of a time interleaved ADC. These filter coefficients are then applied to the complex amplitude spectra of the ADCs. The newly implemented calibration improves the mirror suppression up to 20 dB compared to the frequency independent calibration.