In section 6.2 of PSTN-051, we present a comparison of the impact of taking visit pairs in the same filter versus different filters. I feel like I’ve seen some confusion about this, though, so here’s a summary:
In the v1.5 simulations, we generally take visits in 1x30s exposures (i.e. single snaps), and visits are acquired in pairs – two visits to the same pointing in each night, separated by about 22 minutes. Then we have two simulations that vary the filters used for those pairs —
In baseline_v1.5_10yrs.db: we take those pairs of visits in u+g, u+r, g+r, r+i, i+z, z+y, and y+y. (MAF output pages: science, SSO, metadata, glance)
In baseline_samefilt_v1.5_10yrs.db: we take each pair in the same filter (u+u, g+g, etc.).
(MAF output pages: science, SSO, metadata).
These are grouped into the ‘intranight’ set of simulations.
In baseline_samefilt we have fewer filter changes, and thus can achieve more visits. We fit about 4% more visits into the entire survey, which is also translated into 4% more visits per pointing over the WFD area.
More visits is beneficial for science which wants more visits and does not need colors within a night. Solar system discoveries can benefit from this in particular, as visits in the same filter on the same night have a better chance of detecting the same object. Overall completeness for fainter solar system populations increases by about 3%, depending on population details. Likewise, static science metrics tracking coadded depth, such as the number of galaxies or coadded depth in i band, increase by about 2%.
However, science which needs photometry in multiple bands over a short amount of time, such as SN detection, suffers greatly when filters are not acquired in mixed pairs. For example, the number of TDEs with well-measured colors drops by 60% and the number of detected fast microlensing events drops by ~15%.
This is discussed in PSTN-051, section 6.2. An illustration of the response of metrics from a variety of science cases is available here (Figure 59 from PSTN-051).
To create your own evaluation of the impact of mixed-filter vs same-filter pairs, please use the simulations mentioned at the top of this post and compare metric results between these two simulations.