Fast wide-angle lens request

[Peter Connan]

Thanks Jan

Actually, I wish there was an AI-S lens in this class... I love these older lenses.

[Shane]

I thought I would mention the Rokinon 24/1.4.

I have had 4 of these go through my hands before returning them. All of them had misalignment issues (except one) in which one corner or edge of the frame was not in focus compared to the other areas. This was particularly noticeable on star fields where the sharpness of the stars is easily inspected. Wide open, one of them was beautiful in all 4 corners both in sharpness and coma correction and exceeded the performance of MY 14-24/2.8 set at 24mm. I regret returning it BUT it had such bad focus slop it was driving me insane to get critical focus on the stars.

If Rokinon would just work on their QC I would pay double for that "perfect" lens.

[Peter Connan]

Thank you Shane

I am still torn between the 20mm f1.8 and the 15-30mm f2.8.

I think the Nikkor must probably be slightly better for the primary purpose, and has the advantage that it can gather a bit more light if you are prepared to accept lower IQ, but it has some concerning issues (such as severe vignetting exactly at the apertures and focus distances I want to use it at and difficult-to-use manual focus), while the Tamron appears to be a more versatile lens, and here in SA it is cheaper than the 20mm (unlike elsewhere in the world).

Weight and size are not an issue, i have a big car. More to the point, the places I go to mostly don't really allow walking, particularly not after dark.

[Ilkka Nissilä]

I wouldn't say the vignetting is "severe". I don't currently have other 20mm  reference lenses to compare with but according to DXOMark (using the D750 as test camera), the 21mm f/2.8 Zeiss has 1.9 stops of vignetting at f/2.8 whereas the 20mm f/1.8 Nikkor has 0.9 stops at f/2.8 and 1.9 stops at f/1.8. The 14-24 at 20mm has 1.5 stops at 20mm at f/2.8. So at f/2.8 the 20/1.8 would seem to have less vignetting than the "gold standard" lenses. I wasn't able to find a dxomark test for the  Tamron lens yet.

[Bjørn Rørslett]

Plus the various figures for corner fall-off are dubious at best. One can arrive at almost any value simply by massaging image contrast. Since the better RAW converters have vignetting tools, this inevitable fact of optics is not the big deal-breaker it once was, as correction if required can be conducted in the first processing stage anyway. For current digital cameras with their high dynamic range, any quality loss from lifting corners is negligible.

[Ilkka Nissilä]

Vignetting can be measured e.g. by taking a photograph of a gray patch scale which has measured reflectances and positioning this in the center of the image. The pixel values in the corners can be compared with the gray patches and vignetting in EV is determined. The contrast or tone curve has no effect on such a measurement. An alternative approach (which requires no extra tools) is to take a sequence of exposures and convert them all with fixed settings (and vignetting correction off) in the raw converter, and then compare the resulting RGB values (or luminance if you're not interested in color shift) between frame center and corner across exposures. Once you find the pair of images where the center and corner match luminance values then the (maximal) vignetting is approximately  equal to the difference between the exposures of those two images. However, some sites report something else entirely than vignetting in EV even though they mark the result as EV ...  In such a case it can be that the tone curve affects the outcome which is not at all in units of EV anyway. I bite my lips and grind my teeth when I see "tests" like this.

With some lenses vignetting can be so strong that there is some residual  fluctuations even after correction and there is a significant buildup of noise (i.e. 4-5 EV vignetting would potentially cause this). Also in astrophotography it is common to apply several consequtive S curves to the image to bring out the faint features and I would not say the profile based correction works perfectly at least not with ACR lens profiles. Perhaps some other converters have higher quality profiles and produce better results. However, I have not found this to be at all a problem with the mild vignetting of the 20/1.8 in fact I often leave it in place (but as I said before I normally take another exposure at f/2.8 - not because it is necessary but because it is easy to do and it provides an alternative view).

[Bjørn Rørslett]

There obviously is a limit to what level of correction can be applied and no ill effects detected. However, -2 EV is not the draw back it was in the film age.

I have tried on several occasions to verify the reported vignetting figures found on DXOmark and similar, and been unable to get the same result. I did measure pixel levels corner and centre on captures of uniformly lit white panels (reduced exposure to avoid clipping) with gamma=1. 

[Fanie]

Hi Peter, while upgrading lenses for my D800 from using a DX camera, I bought the Tokina 16-28 as a wide angle, being a good lens from f4 and up for the cost. I did not have any particular discipline for it in mind at the time ie landscape, star trials etc, and got it mostly as an all singing all dancing wide angle for whatever situation may require it. Reasonably happy with it.

My only comparison is a photo I took with Jan Anne's 14-24 Nikkor on my D300, and that just blew me away, but I could not afford the 14-24 as well as the 24-70 at the time of purchase.

The attached photo was taken with the Tokina at 16mm.

[Bjørn J]

Your Tokina seems to being just fine. I have not tried it, but have heard mostly good things about it. The Nikkor 14-24 is brilliant, but one has to be aware of the focus shift. However I think the Zeiss 15mm/2,8 is The King of superwideangles.

[Peter Connan]

Thank you gentlemen

BjornJ, I wish I could afford one of those, but sadly, not at this stage.
Fanie, thanks for that.

Bjorn and Ikka, according to Photographylife's test (hope that's not a swearword around here), the 20mm f1.8 has almost three stops vignetting in the corners at f1.8 and infinity focus (it reduces a lot when focused closer).

If shooting at low ISO's, I am sure this can be pulled back in PP without any problems, but when shooting at ISO3200-ISO6400, I am not convinced one would be able to pull back the vignetting, and stopping down to f2.8 will mean additional noise across the frame. Of course, by f2.8 it is out-performing the Tamron, but it is longer, and therefore the shutter allowable shutter speed is reduced...

I will have to see if I can fiddle with both...

Thanks for all the nowledgeable advice received so far.

[Ilkka Nissilä]

Thanks, I will look into the infinity vignetting and see if I can observe a difference. There are some complications in the experiment with finding a homogeneous, infinite, Lambertian surface.

[Bjørn Rørslett]

In particular with a wide field of view, I surmise.

[Peter Connan]

For better or for worse, I have acquired the Tamron 15-30mm f2.8.

Tested both it and the Nikkor, and although the Nikkor has the advantage of f1.8, it only surpasses the sharpness of the Tamron at around f2.8, so it loses the advantage, and I like the shorter focal length the Tamron delivers.

Hopefully I will get an opportunity to use it for it's intended purpose next weekend.

[simsurace]

Plus the various figures for corner fall-off are dubious at best. One can arrive at almost any value simply by massaging image contrast.

If the measurements are read off from RAW data levels (e.g. in RawDigger), this degree of freedom should be absent.
I see something between 1.5 and 2 stops of vignetting with the Nikkor 20/1.8 at f/1.8 using that method. Nothing too severe.