Sequential readouts have identical noise when `!SIM.random.seed` is set

[oczoske]

What happened?

This is similar to identical to #298 , #295, #97.

PoorMansHxRGReadoutNoise and BasicReadoutNoise call np.random.seed(random_seed) each time that apply_to is called on a detector, with random_seed the same every time. ShotNoise calls rng = np.random.default_rng(self.meta["random_seed"]) with the same effect. Moving np.random_seed and the rng initialisation into __init__ should solve that, at least if several readout() are done in sequence (what about parallelisation?). That the various noise effects use the same seed is not ideal either, but may be not so harmful since different noise distributions are used.

Code to reproduce

cmd = sim.UserCommands(use_instrument="METIS", set_modes=["img_lm"])
cmd["!SIM.random.seed"] = 42
metis = sim.OpticalTrain(cmd)
metis.observe()

read_01 = metis.readout()[0]
read_02 = metis.readout()[0]
np.all(read_01[1].data == read_02[1].data)
#  np.True_

Exception traceback

ScopeSim bug_report()

In [41]: sim.bug_report()
Python:
3.12.9 (main, Feb  5 2025, 01:31:18) [GCC 14.2.0]

Installed Python packages:
           anisocado: 0.5.0
         astar-utils: 0.6.0
             astropy: 7.2.2
      beautifulsoup4: 4.15.0
            docutils: 0.21.2
             httpxyz: 0.31.2
                lxml: 6.1.1
          matplotlib: 3.10.8
      more-itertools: 10.2.0
               numpy: 2.2.6
           packaging: 26.2
               pooch: 1.9.0
              pyyaml: 6.0.3
               scipy: 1.17.0
            scopesim: 0.12.0b0.dev0
       scopesim_data: could not be loaded.
  scopesim_templates: 0.8.2a2
         skycalc-ipy: 0.7.0
             synphot: 1.7.0
                tqdm: 4.70.0

IRDB installation

Cloned and linked IRDB repository

Use case

Pipeline development