I started a Pull Request to explore what support for coolTLUSTY would look like.
One hiccup so far is that the spectra appear to adopt two similar but distinct wavelength coordinates. Here I show the pixel separations expressed as a pixel spectral resolution, as a function of the wavelength coordinate.
Distinct spectral coordinates is fine, since the SpectrumCollection object of Specutils supports ragged arrays like this out of the box (if I recall correctly). But it does introduce some special considerations, since we've assumed shared wavelength coordinates in some places. For example the dashboard only updates the flux coordinates, which saves some overhead and lowers latency.
One key strategy could be to add a grid.resample_wavelengths() method that resamples the spectral axis to a uniform coordinate though regular-old interpolation. This has the demerit of adding interpolation error, but overall should be fine in these noise-free model spectra. Also resampling won't work for $\lambda<0.5~\mu m$ coordinates don't overlap, but few practitioners have Y-dwarf data in these blue wavelengths so I think that's ignorable.
The new coolTLUSTY paper is out, yay! https://arxiv.org/abs/2303.16295
I started a Pull Request to explore what support for coolTLUSTY would look like.
One hiccup so far is that the spectra appear to adopt two similar but distinct wavelength coordinates. Here I show the pixel separations expressed as a pixel spectral resolution, as a function of the wavelength coordinate.
Distinct spectral coordinates is fine, since the SpectrumCollection object of Specutils supports ragged arrays like this out of the box (if I recall correctly). But it does introduce some special considerations, since we've assumed shared wavelength coordinates in some places. For example the dashboard only updates the flux coordinates, which saves some overhead and lowers latency.
One key strategy could be to add a
grid.resample_wavelengths()
method that resamples the spectral axis to a uniform coordinate though regular-old interpolation. This has the demerit of adding interpolation error, but overall should be fine in these noise-free model spectra. Also resampling won't work for $\lambda<0.5~\mu m$ coordinates don't overlap, but few practitioners have Y-dwarf data in these blue wavelengths so I think that's ignorable.