Setting the num_of_obs_per_detector according to the length of the noise correlation

[anand-avinash]

In the simulations that @martamonelli has been doing, as well as for some future simulations, we are assuming that the 1/f noise has a finite correlation length. To enforce this, we need to manually create timeline chunks of appropriate size, and supply the number of chunks as a parameter to num_of_obs_per_detector or n_blocks_time in sim.create_observations(). This approach ensures that the timeline associated with a given observation instance contains exactly one correlation length of time-ordered data (TOD).

Should we consider implementing a feature in LBS that automates this segmentation of timelines? For example, we could add a parameter noise_corr_len_s (s for seconds) to .create_observations(). Using this parameter, the entire timeline can be divided into a certain number of chunks, and the parameters num_of_obs_per_detector (and/or n_blocks_time) can be set accordingly. We might also need this for our e2e pipeline.