Noise added by 2-bit compression using LOFAR test set A. Left image: results of calibration, 3C 196 subtraction and imaging without compression. Centre image: same, but before processing the visibilities were compressed using the 2-bit quantization scheme (16× compression) with the 2.5σ-truncated Gaussian distribution. Right image: difference between left and centre images. While the added compression noise dominates the noise in the image, the compression has not affected the sources systematically and the added noise is unstructured.