Traditionally, optical remote sensing payload design satisfies highly defined specifications arrived at by consensus of the
scientific constituency. Designs are constrained by required performance such as resolution, Modulation Transfer
Function (MTF), and Signal-to-Noise-Ratio (SNR). Payload designers satisfy the specification by performing hardware
and cost trades. This process may lack continuous feedback between the performance of the scientific algorithms and the
payload design, potentially missing optimal design points.
The traditional method has produced separate and specific designs for imagery (over-sampling ratio Q > 0.8) vs.
radiometry (Q < 0.8). Radiometers are scientifically precise, with highly accurate scene collection over a tightly defined
pixel size exclusive of other scene points, often across several spectral channels. Imagers reveal sharper features, but
have considerable "bleeding" of scene radiance into adjacent pixels, causing errors in application of multispectral
Recently, we created end-to-end models that optimize end scientific data products by considering the payload design and
data processing algorithms together, rather than simply satisfying a payload specification. In this process, we uncovered
optimal payload design points and insights.
We explore end-to-end modeling results that show an optimal single converged payload design, and data processing
algorithms that produce simultaneous radiometer and imager products. We show how payload design choices for
Instantaneous Field of View (IFOV) and Ground Sampling Distance (GSD) maximize SNR for multiple data products,
resulting in an optimized design that increases flexibility of space assets. This approach is beneficial as we move towards
distributed and fused image systems.