In the IDL code, map convolution defaults to adding Poisson noise unless the keyword no_count_stats is set. However, this makes the assumption that the units of the map are detector counts. Note that even when the units of the source map are photons, the units of the convolved map multiplied by effective area are ideal counts not detector counts (because deadtime and other effects are not included), so it's not necessarily appropriate to apply Poisson noise in even that case.
I suggest that the default instead should be a straightforward convolution, and then the addition of Poisson noise can be an affirmative choice by the user (e.g., by setting a keyword like add_poisson_noise).
In the IDL code, map convolution defaults to adding Poisson noise unless the keyword
no_count_statsis set. However, this makes the assumption that the units of the map are detector counts. Note that even when the units of the source map are photons, the units of the convolved map multiplied by effective area are ideal counts not detector counts (because deadtime and other effects are not included), so it's not necessarily appropriate to apply Poisson noise in even that case.I suggest that the default instead should be a straightforward convolution, and then the addition of Poisson noise can be an affirmative choice by the user (e.g., by setting a keyword like
add_poisson_noise).