To check the ENVIS design one needs to put a wide angle, uniform light source at the focus of the lens and then observe the diameter of the beam coming out the front of the lens. This would show how much of the lens was being used to illuminate a particular part of the photocathode. I do not have a source that is wide-angle and uniform enough to do this, my fastest interferometer diverger is f/1.7, and even that is not uniformly illuminated.
The best place to have a narrowband filter is where the angle of light passing through the filter is the closest to normal incidence, or going straight through the filter. (In certain eyepieces, the ideal place might just be inside the eyepiece if the light is nearly parallel there, but this is not going to happen.)
This brings up the possibility of some interesting relay lens arrangements, which create parallel light, but to me that is overly complicated.
Putting a flat filter over the photocathode is the worst possible place because the angle is quite steep there with the very fast incoming light cone.
For larger lenses or telescopes, the best place would be over the aperture because the field of view is smaller and thus the departure from normal incidence of incoming light is pretty small. Of course this is impractical for larger instruments, but it could work for some camera lenses, and ~2x-6x objectives.
For a NV unit with ENVIS objective, the lens system gives the NV unit a field of view about 40 degrees wide, meaning that the light from the edge of the field is coming through the filter at an angle of about 20 degrees from normal incidence. This is a lot. I do not expect any narrowband filter to work exactly as it was designed with those angular departures.
Yes, band shift is what I am talking about. As the angle increases, the band shifts, and you effectively move away from the H-alpha wavelength, and this should cause the image to get significantly fainter. So, yes, I am attritubuting the vignetting that I see to band shift of the well off-axis parts of the field of view.
Yes, there is less band shift if the filter is moved in front of an eyepiece compared to putting it over the ENVIS lens, and that is what I said above. For an F/3 cone coming out of a telescope and into an eyepiece, the light covers a 16 degree cone, and that is a maximum departure from normal incidence of about 8 degrees, much less than 20 degrees for the light coming into the ENVIS.
I believe I have a way to confirm this phenomena the next time it's clear and I have time.
I will also have other updates in the near future.