Al hilo de lo que se comentaba sobre el uso de filtros de banda estrecha con conos de luz de gran ángulo, copio un mensaje de Bob Ayers en el grupo narrowbanding, en donde se trata bastante bien el tema:
An f/1.8 converging beam is hard work for a narrowband filter.
The game is that light going thru the filter at an angle "sees"a different filter thickness, so it sees a different passband..
With both straight and off-axis light in the f/1.8 beam, the effectivewidth is widened because some light sees one passband, some sees a slightly different passband.
The formula for how the passband changes depends on the materials that the filter is constructed from (its refractive index)so there is no all-purpose formula, you have to know the fiter properties to get the right result.
griot filter passband shift
and look at the discussion in "Chapter 13" of their tuturial on "Wavelength dependence on angle of incidence"
There is also a discussion at
where they write
"The bandpass of an interference filter shifts blueward as the incident angle deviates from the normal. This can be useful as it enables a blueward tuning of the filter to be done. Typically a 5 degree tilt shifts the central wavelength of a 656 nm filter by about 0.7 nm while a 10 degree tilt shifts it by 2.7 nm. As most
imaging is done with a converging beam the bandpass will broaden and shift blueward depending on the f/ratio ..."
Consider the above only an estimate -- but still note that a f/2 beam is converging at a max angle of arctan(1/4) = 14 degrees and you will see that you have a problem at f.1.8 ...