Pixel centroid characterization with laser speckle and application to the Nancy Grace Roman Space Telescope detector arrays

TL;DR

This paper presents a novel speckle-based method to characterize pixel centroid offsets in infrared detector arrays for space telescopes, achieving sub-pixel accuracy and aiding in precise astronomical measurements.

Contribution

It introduces a speckle pattern projection technique to reconstruct pixel centroid offsets and assesses its effectiveness on real detector data.

Findings

Residual centroid offsets are approximately 0.0107 pixels RMS.

Quadratic fitting reduces residual offsets to about 0.0097 pixels RMS.

Method effectively distinguishes signal from noise in pixel offset measurements.

Abstract

The Nancy Grace Roman Space Telescope will use its wide-field instrument to carry out a suite of sky surveys in the near infrared. Several of the science objectives of these surveys, such as the measurement of the growth of cosmic structure using weak gravitational lensing, require exquisite control of instrument-related distortions of the images of astronomical objects. Roman will fly 4kx4k Teledyne H4RG-10 infrared detector arrays. This paper investigates whether the pixel centroids are located on a regular grid by projecting laser speckle patterns through a double slit aperture onto a non-flight detector array. We develop a method to reconstruct the pixel centroid offsets from the stochastic speckle pattern. Due to the orientation of the test setup, only x-offsets are measured here. We test the method both on simulations, and by injecting artificial offsets into the real images. We use cross-correlations of the reconstructions from different speckle realizations to determine how much of the variance in the pixel offset maps is signal (fixed to the detector) and how much is noise. After performing this reconstruction on 64x64 pixel patches, and fitting out the best-fit linear mapping from pixel index to position, we find that there are residual centroid offsets in the x (column) direction from a regular grid of 0.0107 pixels RMS (excluding shifts of an entire row relative to another, which our speckle patterns cannot constrain). This decreases to 0.0097 pix RMS if we consider residuals from a quadratic rather than linear mapping. These RMS offsets include both the physical pixel offsets, as well as any apparent offsets due to cross-talk and remaining systematic errors in the reconstruction. We comment on the advantages and disadvantages of speckle scene measurements as a tool for characterizing the pixel-level behavior in astronomical detectors.