Narrowband and fast telescopes
For some time I used regular narrowband filters in combination with my fast newtonian telescope. Well aware so called highspeed filters are recommended for faster optics, I always felt okay with my results.
But then I tried to image a detail from the Tulip Nebula with the slower APO but had low expectations, as I had to spent 20 hours with the fast newtonian on the same object for a decent result. However, after checking the first narrowband subs and test stacks, it was quite obvious, that it should be possible to obtain a similar result from way less total integration time:
The Newtonian still has a big advantage when collecting luminance and broadband color, the APO's narrowband stacks exposed more details and smoother gradients despite the shorter integration time. This could not be explained by the longer focal length.
Could this be caused by the filters? Would highspeed filters fix this issue? But what exactly is a highspeed filter?
Compared to a regular filter of same quality, a highspeed filter only differ in a single characteristic, its central wavelength of the bandpass is shifted a tiny amount towards longer wavelengths (red). The transmission plots of the filters used in this test (Antlia 3nm) look like this:
The nominal wavelength of Hydrogen Alpha emission line is 656.3nm, my sample has a central bandpass wavelength at 656.45nm (center of 50% transmission). In contrast the highspeed filters central wavelength is located at 657.70nm, shifted for about 1.4nm. I've used the transmission plots packed with the filter and rely these are correct.
The reason for this necessary shift is the optical path in any telescope. The larger the aperture and shorter the focal length (high focal ratio) outer beams enter the filter in a slightly slant angle which causes the filter to "recognize" this beam at a longer wavelength. For slower telescopes these outer beams are still well within the bandpass width.
However, for faster telescopes the effective wavelength raises with the radius and slowly leave the bandpass region and therefore will be blocked by the filter. The total transmission is reduced.
For a much too long period I was convinced that this could be compensated with some more integration time, but the situation is slightly more complex.
- First of all this shift only affects the emission lines of interest. The unwanted photons continue through the filter unimpressed, which effectively reduce the SNR (signal-to-noise ratio).
- The transmissions are blocked with increasing radius, which is fatal for two reasons. First of all, due to their larger surface, these outer regions would contribute to the overall transmission most. Much more critical is the reduction of optical resolution, which is basically determined by the effective aperture. With the outer regions blocked by the filter, the overall sharpness will be reduced.
- Most reflector telescopes do have a central obstruction. This blocks just those parts, with the highest transmission.
As a result with those filters my "fast" newtonian required more integration time to compete with the results from the slower refractor. In addition the angular resulution was lower, as the transmission effectively drops below 50% at a diameter of 90mm and larger.
I performed some tests and reduced the aperture with a mask to 100mm, which affected narrowband transmission only slightly, but reduced the unwanted photons by alot, which increased the SNR. But this could not be the final solution.
So I ordered the highspeed version of my existing 3nm Hydrogen-Alpha filter and did some direct comparisons in the same night and identical conditions. My target of choice was the region of the Horsehead Nebula (IC434), as this area contains both faint smooth and sharp, high contrast elements.
Without the stars removed two mini-stacks with only two 5 minute subs each looked like this:
It is pretty obvious that the emission lines are much better captured using the highspeed filter. Barely visible fainter regions are well distinct in the highspeed filter's image.
The differences get much more obvious after removing the stars and use NoiseXTerminator to reduce the noise. An additional STF was applied in addition to maximize the contrast:
For this comparison I also used 4 subs from the regular filter, effectively doubling the integration time, to check if this would help with fainter regions. From my experience with the local light pollution I guess doubling the integration time a second time would still not be enough.
By the way, if you look closely, you'll recognize the remnant of a satellite trace in both images, a clear indicator that I changed the filter after each sub.
More differences within the details and overall sharpness reveal by zooming in. This crop also exposes a typical issue with modern, AI based enhancement tools, which tend to emphasize supposed structures within too noisy data (lack of subs), which simply do not exist. Just compare the obvious bright spot right to the center of this image.
This bright "spot" is a remnant of NGC2023, a fairly bright reflection neubula, which only in part was removed during star extraction. The following noise reduction then interpreted this artefact as signal and did some enhancements.
I probably would remove that spot from the stack, as it is no significant element in H-Alpha, but it probably would not matter in the final image, as it is overlayed by much brighter layers, as shown on the right.
In Orion displays the full color image.
The rest of the short window within the clouds was used to collect 1.6 hours H-Alpha from the Flaming Star Nebula, which was captured exactly one year ago and contains a total of 3.3 hours of this band. The new stack was taken with a clear sky and about 50% Moon, the older stack was taken at full Moon and a more humid air (star halos). The full Moon does not affect the image quality significantly when using the refractor, but it may cause some additional degradation in contrast with the fast telescope.
But the difference in transmission is quite similar to the image above (remember, the image taken with the highspeed filter only has 50% integration time):
With the stars as a brightness reference it is quite obvious, that the overall transmission of H-Alpha is much better with the highspeed filter.
But let's also check the starless and restreched versions:
The speckled regions are primarily remnants from the star halos due to larger humidity, the fainter regions are still much better and more detailed in the image using the highspeed filter. Even with half the integration time.
Of course the story repeats when zooming in, which again reveals an overall loss of sharpness in the image using the regular filter:
As a final test I mounted the image train on my slower refractor to check the expected loss in transmission on this slower telescope. Technically there is about a loss of 15% which can be compensated using a little more integration time without negative side effects.
Sadly IC434 dropped into the haze during these shots, and there were clearly some faint clouds running through. But there is no significant drop of the overall transmission, so I do not have to bother purchasing a second set of filters. If I ever repeat this experiment with better conditions, the images receive an update. But for the moment these have to fit.