Even 1nm resolution (10Å) would be useful. Of course, 1Å is nice as it could split the Sodium D lines and be used for Doppler measurements.
An imaging spectrometer is admittedly difficult. Here's how one might be manufactured, using a monochrome (ASI MM) camera:
- Camera acquires a slice of the image; simple drift (no tracking) results in a scan.
- Each row (or, maybe, column) acquires a spectrum for a corresponding pixel of the slice.
- Rather than a single prism or diffraction grating, there exist diffractive optical fibers that can be laid across (and adhered to) the sensor's face.
- The first several columns (conversely: rows) are used for only monochrome imaging.
-- As a star's image drift's across the first columns, an estimate of the angle and speed of the drift can be done by acquiring several images.
-- As the drift progresses, all columnar spectra are acquired even if the resulting image ends up being non-rectangular (trapezoidal).
- Even an ASI120MM, spreading spectra across the vertical axis (960 pixels/rows), could resolve a good 900 bands.
-- 960 - 900 = 60 rows for drift processing
-- Assuming just a visual band of 700nm through 400nm results in ⅓nm granularity: 900 band pixels ÷ (700nm - 400nm) = 3 band pixels per nm
--- 400nm was chosen to avoid "synonyms" between 700nm and 350nm and because the resulting range is a convenient value (900 rows and 300nm). Of course, the range could easily be 725nm through 375nm as could the drift count be something quite a bit smaller than 60 rows.