The Sigma 14mm f/1.8 DG HSM ART Lens is one of the most exciting lenses for astrophotography enthusiasts in recent years and here are some of the first real world Milky Way images taken with the lens that also answer the big question: How's the coma?
For photographers that focus on sky images, one of the biggest concerns for any lens is how the stars look in the corners when shooting wide open, and I have to say that the new Sigma 14mm Art lens is looking incredible. Coma or comatic aberration occurs when off axis points of light, like stars, are not sharp due to the optical formula of the lens. Coma tends to look like each star has wings instead of being a point of light like the stars in the center of the image.
The image in this article was taken by Tony Liu with a Canon 5D Mark III with the Sigma 14mm f/1.8 DG HSM ART Lens at f/1.8, 2500 ISO and 10 second exposure.
The image above is zoomed into the top left corner of the main image.
The image above is zoomed into the top right corner of the main image.
My biggest concern when this lens was announced was if Sigma would keep things the same with coma in the corners of the lens as it had with the 24mm f/1.4 DG HSM ART and 20mm f/1.4 DG HSM ART lenses or if it would return to the optical formula that was closer to the 35mm f/1.4 DG HSM ART lens. The answer is that Sigma seems to have landed oh so short of the 35mm Art perfection, but it is much better than the 20mm and 24mm Art lenses. I say this because we are talking about a lens that is a full 1 1/3 stops wider than its competitors, and it looks like Sigma is finding a balance in its optical formula that allows such a wide aperture with fairly well controlled coma.
Is the Sigma 14mm f/1.8 DG HSM ART Lens perfect? No, but it's close and maybe worth another look if you shoot the stars.
Images by Tony Liu and used with permission. You can see more of his work on his Flickr and Instagram.
The astrophotography market has been dominated by the Rokinon manual lenses so far. It would be very interesting to make a side by side comparison of this new Sigma lens with the Rokinon.
I was surprised by the small size of the stars and coma/astigmatism tails in the corners, in comparison with the test results I obtained with my own Sigma 14/1.8, so I made some verifications...if I'm not mistaken, the "zooms" are not 100% crops but less than 50% views, probably about 45%. That should be mentioned because people may suppose that it's 100% crops and even if the lens is very good, 45% vs 100% makes a big difference!
hi Thierry, i simply cropped into the same FOV as it appears on my screen when viewed at 1:1 in lightroom, cheers
Huh? "the same FOV as it appears...at 1:1" ...doesn't really make sense.
Simply put, the images being displayed here are ~995 pixels wide. Is that the pixel width that you originally cropped them at in Lightroom, and exported them at?
what i did was zoom in lightroom at 1:1 and what showed up on the window in lightroom i simply cropped to that exact same crop then exported as jpeg if that makes more sense
Let me try another way: Lightroom has an option to display the actual cropped pixel dimensions. With that image info turned on, are you cropping to 996 pixels long?
Alternately, just throw a CR2 file on a server somewhere, and let us inspect for ourselves.
i've realised something...i am guessing JT would've just took screenshot of what was showing on my flickr page because I had not sent him my files before the article was already posted, now that i've actually clicked on the posted thumbnails it looked like a smaller down sampled version of what i originally cropped, that might explain why
well Tony, I think that we understand now the reason of the discrepancy in scales! :-)
haha hi Theirry yea, i've just posted an 888 pixel wide crop down comment below but here it is again, hope it clears things up a bit more, top left corner
perfect, thank you! :-)
i might also add (just to help put things in context), my copy looks sharp wide open even basically to the border, IQ only really suffers at extreme corners where it noticeably softens but it does it quite uniformly across the 4 corners, feels like to me light field curvature (which incidentally was what Roger at Lensrental reported as well, mostly flat field and just slight field curvature)
I agree Tony, field curvature seems slight, as well as vigneting for such a fast lens (much better than Canon and Samyang 14mm)
what has really impressed me so far is distortion control, the cityscape pic i posted below had no distortion correction, i feel like everyone's so concerned about coma and astigmatism (and fair enough rightfully so) they're overlooking how wonderfully the lens is performing in all other areas like sharpness and distortion etc
Hi Tony, not sure to understand 100% what you mean, but again your crops are not 100% views, they are a small 50%
regards
hi Thierry please see my reply to Matt directly above
Coma ppssh! The marvel is that its a freaking 1.8 and going off your settings used. You could have gone to 30 secs, 1.8 iso 800. That's just insane. Also using image averaging where you can. You're talking about some ridiculously clean astro shots. I don't see a reason NOT to get this less for astro.
hi Kyle, i am the guy that took the images, it was taken in my light polluted backyard, reason for only 10 seconds was to miminise star trails as rule of 600/500 don't always work pending on what part of the sky (which is why you often see sharp as tack stars in the centre frame using UWA and start to trail around the edges
but yea the lens is a marvel for sure, for me personally the coma is very acceptable, lens is sharp across the frame, and the distortion control for such a wide lens is phenomenal, it seems to deal with flare quite well too, at least my copy:
the coma performance of this lens has been heavily scrutinised (and rightfully so) but i can see this lens being used many other ways like architecture, concerts/events etc just for how impressive the other mertrics of the lens are
zero distortion correction:
Do you thing there's anything better to buy instead of this lens when it comes to Astro and landscapes? i have a Sony a7s and Sigma MC-11.. what is your advice?
Lenses like this are a one-trick pony, Mostafa. You've got one focal length, and the lens is heavy as a brick or two, no front filters, etc. There are much more versatile lenses out there, when you want to consider more than "just" astro. The Nikon 14-24 is a fantastic choice, as is the Tamron 15-30. A Canon 16-35 mk3, with front filter threads, lighter weight, and downright incredible image quality wide open, is an even more versatile astro-landscape lens IMO.
I would only buy the Sigma 14 if you already know that you LOVE 14mm, otherwise it's just too risky of an investment. The question is, how much have you shot at 14mm yet?
The thing people say is, well if you wind up LOVING 14mm, then buying the best lens right off the bat is a good idea. To that, I would reply, that if you're that obsessed with a single focal length, you might want to have a backup of it. Especially if you ever get into timelapse photography.
Therefore, it would be a win-win if you bought a Rokinon 14mm 2.8. I think my logic is pretty sound on why, but I'm sure many others will chime in with a more simple yes-or-no reply.
Thanks bro, sigma 14mm could be a good deal since my photography is about 75% astro and landscape right now which means i need something Wide with a good quality control and wide enough aperture to be useful at night and useful at landscapes when i get down to f11 or lower i believe it will give me maximum sharpness, all of these benefits i don't think i will get with any other lens , specially when i have a good offer here in my country which allows me to buy it with a good discount as low as a 40% off from it's original price, i guess it still win-win.
If you have a Sony look at the soon to be announced Laowa 15mm f/2 rectilinear, its more compact, far lighter, engineered for Sony and going by samples from the pre-production model it outperforms the Sigma 14mm f1.8 - Full disclosure, I'm from the Australian distributor of Laowa, but the samples I've seen have been taken with our pre-production copy by people far more skilled than myself.
Adam, I'm especially waiting on the vigneting measurements on the Laowa, all the more that the front lens of the Laowa is much smaller than the Sigma's one. Vigneting is very well managed on the Sigma and that's very important for astrophotography, since software vigneting corrections only hides it (but the underexposition and the bad SNR remain!)
I don't agree with this at all and surprised JT would write the article based on someone else's images. I was very excited about getting the Sigma 14mm and rushed out to take test shots on the 1st day I received it. Sadly I returned it the next day. The coma was terrible and all the way round the edge, not just in the corners, and didn't improve even at f/2.8. Attached is from Top Left side of one of my RAWs as a 2:1 zoom. Taken on a Canon 5DIV, ISO1600, 10s, f/1.8.
hi Jay that looks noticeably worse than my copy...i'd definitely try another if you're able :(
There's a few people I've spoken to who have the same issues. Unlikely I'll go back given the other much cheaper alternatives on the market unless Sigma can prove to me that any future copy supplied is rock solid and as per the test shots they posted that attracted everyone into the idea in the 1st place.
hey that's fair enough, i might've lucked out with a unicorn copy
Well you are showing a 2:1 magnification, so i expect it would be twice as bad at least.
what would be super helpful - if either of you guys are up to it - is access to the original raw files so we can see them firsthand. thanks for considering
i missed the 2:1 comment haha
anywhere i can post the 25mb dng?
Tony you can use dropbox, or if you prefer you can email it to me and i'll put on my dropbox. bvalente at gmail dot com
hi Brian, you've got mail :)
My point in posting this is, its bad! For those wanting a crisp clear lens capable of capturing stars on images that can be used for print, the Sigma falls ways short, especially when there are other contenders out there that might not be f/1.8 or AF, but are still fast enough for astro work and have coma well and truly under control to the point it doesn't exist. Don't forget, the Samyang XP 14mm f/2.4, the Irix 15mm f/2.4 and the Laowa 12mm f/2.8. IMHO, Sigma have some very stiff competition, but for the price point (and lets add the weight difference as well) they didn't deliver on the quality they spruiked. Happy to drop off a DNG somewhere but can't post links in here.
compare it at 2.8 - i think you'll find it smokes the rokinons. not only for coma but also for resolution
btw i own all those lenses. They are all mediocre to bad when it comes to coma. the laowa 12mm distortion has much worse coma, the irix is just as bad, rokinon is bad and has all sorts of problems. But with sigma, you can stop down to 2.8 and have the same speed lens as the others and totally acceptable/near zero coma. It's also true the other lenses could be stopped down to fix it, but then we're talking an f4 or f5.6 lens, a non starter for astro work
that to me is what makes it worth it.
I don't think there's an issue of good copy or bad copy at all.
Jay,I'm not sure it's so different from Tony's pictures, considering that his "zooms" are about 50% views (I calculated 45%), not 1:1.
Are you sure about focus? At f/1.8 it's extremely critical, and I experienced strong coma on a large part of the picture with the Sony 7S when focus was shifted by a very small amount. I also depends where is the star you focus on: the result may be different if the star is in the center or on the edge, due to field curvature (which is small on this lens, thought).
regards
Focus was achieved via Live View method on a 5DIV with jewellers loupe. Focus was checked at top, left and right before image was captured.
i had it look at 2:1 in lightroom and mine does still look better i have to say
Those images are almost completely coma/astigmatism-free, and VERY different from all the other sample images we've been seeing so far. I'm not sure if I feel bold enough to declare BS, but I'm inclined to.
BTW, coma is NOT the "wings", that is astigmatism. Coma is the "stretching" or trailing in the on-axis direction, so, away from the center of the lens. Almost everybody gets this wrong, though. (I got it wrong until recently too, so I won't throw stones.)
hi Matt, i originally posted my findings on POTN because i am obviously one of the earlier folks who has gotten access to this lens and i figured this may help others that are looking at this lens because there is obviously very little info on astro, JT must've saw my post and asked if he could use it as sample
i purchased this lens retail, i have no affiliation with sigma whatsoever, so when you say you are inclined to declare BS ie suggesting i lied or fudged the results, well i can't do much more than to say you're wrong, maybe as i said earlier i had gotten an unicorn copy and is not indicative of rest of the samples, but i have zero gain from posting this, i simply wasted my own time trying to help others
i have to say this whole experience has left a sour taste in my mouth too, you'd think you take time out posting samples helping others the least you'd get is a thanks, but instead you get people attacking and questioning on what you posted, so yea doubt i'd ever do it again i think, just going to go and enjoy my new bit of kit instead of squabbling with people on the internet who you trying to help, cheers
If this experience has left a sour taste in your mouth, then that's understandable. Posting sample images is not for thin-skinned folks. Posting 100% crops of a HIGHLY exotic, expensive lens is risky business, and if your samples dramatically depart from all the other samples being posted by other folks, then yeah you're gonna get a lot of questions, even accusations. That's the nature of the beast.
Enjoy your unicorn of a lens. Or, you could simply post some CR2 files and put all the misunderstanding about what a 100% crop is to rest.
Afternoon Matt, I'd like to first off say that you are correct about my statement of coma vs astigmatism. After shooting for 5 years, I still make mistakes or have just not been specific enough in understanding lens construction and clarity is great to have. I'll reach out to have this post updated. Secondly, lenses are universally created with compromise and some are simply constructed better than others even in the same batch constructed at the same time. We see this regularly and it's why there are always an average for data sets to see where the overall lens may stand. I've seen this from the Rokinon/Samyang lenses to Canon L lenses and it usually comes down to what you as an artist are ok with or willing to work around. I've been shooting the Canon 50L since fall last year for creating panoramics that have nearly zero star issues dealing with coma and astigmatism because I use technique to overcome that flaw. It's also one of my most used pieces of kit for everything else so it has value beyond its issues and I work with it. As stated in the article, is the lens optically perfect? No. Does it allow for higher dynamic range at night (usually the bigger deal and why I'm looking at a tracking mount eventually)? Yes. As with anything, if we focus on just one thing, we may miss what matters, and that's having an experience that fuels us to keep creating. I look forward to more of your commentary and appreciate the correction. Best, JT
Matthew, I did a simple calculation.
The full field image with the Milky way (the first image of this page) has been reduced by a factor of 6, its resolution being 946x631 pix vs 5760x3840 pix for a 5D III raw image. If you take the distance (in pixels, inches, mm...) between two bright stars in a corner, and then measure the distance between the same stars in the "zooms", there must be a factor of 6 between the two figures if those crops are actually at 1:1 scale.
But the ratio of distances is only 2.75 instead of 6, leading to the conclusion that the "zooms" crops are not displayed at 100% scale but only about 45%.
I do believe that Tony made a mistake. It's not a crime, all of us make mistakes, but if we compare his images with other ones displayed at 1:1, we just compare apples and oranges. No "accusation", no "beast"! ;-)
Edit: Tony has just posted the reason of the discrepancy and why his "zooms" in this page are not 1:1.
hi Matt if you have read above comments both me and Jay have offered to post our DNG files somewhere just give us somewhere to upload to, hell i am even happy to email it directly to someone if asked
on side note outright accusing someone lying (then calling someone thin skinned) is very different to simply asking questions about the results, it's amazing the sense of entitlement some people have, the way you responded you make it sound like i am indebted to you somehow to prove myself?
let me ask again, what do you suggest i gain out of this going out of my way to post sample pics of supposedly made up result except for the time i wasted?
and again i have DNG file ready to post somewhere, maybe i can send it to JT directly and he can make it available as a link on this article?
For submissions, there is a requirement on size/ limited size of the image shown. As Tony stated, I had to do a screen grab for the images to put out the article as they were not downloadable from his Flickr. The link is in the comments to take a look at a direct upload of the images. As such, these are not full res as Thierry has figured from the 2:3 ratio at nearly 1000px on the long edge. Again, there is a limit on the size. I wanted to thank Tony for going above and beyond to offer a full res file for anyone to peruse. I can't create a link in the post unfortunately. I wanted to thank Tony for the use of images as there is an enormous amount of interest in this lens as the dialogue has shown. I would stress that any one copy of a piece of equipment, especially newly released, will have a bit more variation when compared singularly to others, and the only way to find a realistic result is with an average of results. If nothing else, the optical formula shows it's possible to innovate lens design and I'm sure larger and smaller manufacturers are taking notice.
oh and until then here is a crop of the top left corner cropped to 888 pixels wide, exif intact
Tony Liu graciously shared his full res raw file (in dng format) and can be downloaded for inspection here:
https://www.dropbox.com/s/6xjt4i4zpsvei17/522A8092.dng?dl=0
my impression after examining this raw is the reviewer is spot on:
"Is the Sigma 14mm f/1.8 DG HSM ART Lens perfect? No, but it's close and maybe worth another look if you shoot the stars."
and
"Sigma seems to have landed oh so short of the 35mm Art perfection, but it is much better than the 20mm and 24mm Art lenses"