Sensitivity considerations on denoising series of spectra by singular value decomposition

When acquiring series of spectra (T-1, T-2, CP buildup curves, etc.) on samples with poor SNR, we are usually faced with choosing between taking a few points with a large number of scans to maximize the SNR or more points with a smaller number of scans to maximize the information content. In this Letter, we show how low-rank decomposition can be used to denoise a series of spectra, reducing the trade-off between the number of scans and the number of experiments.

Automated summary

When dealing with samples that have poor signal-to-noise ratio, it is often necessary to choose between taking a few points with a large number of scans to maximize signal-to-noise ratio or more points with a smaller number of scans to maximize information content. In this study, we demonstrate how low-rank decomposition can be utilized to denoise a series of spectra, reducing the trade-off between the number of scans and the number of experiments.