[joze]

Hi,

 

here's my personal experience with the new Sigma 50mm f1.4 ART lens which is hyped all over the place. I tested with a Canon full frame body and mainly wide open (f1.4).

 

To summarize:

• The autofocus was correct in less than 50% of the shots, even for still objects. If shooting the same still object several times, some of the shots are sharp, some aren't. This was true for objects at any distances.
  
• The AF stopped working completely once -- the AF drive just didn't move at all any more. Turning off and on the camera made it running again.
  
• The image quality was not what I had expected from all the reviews to say it politely, see below in more detail.
  

Some initial wide-open shots were sharp (if the AF found the right focus), but not tack-sharp. It was just a subjective feeling that this can't be the thing which is hyped in all the reviews.

 

Here are some image details of a simple "lawn sprinkler" test. The drops from the sprinkler act as many little almost point-like light sources. The lawn sprinkler curtain of drops was shot from the side. The focus was set so that the drops in the center of the image (region 1) are about in focus. The drops on the left (region 2) are in front of the focus and and the drops on the right (region3) are behind the focus. The photo was made with the aperture fully open (f1.4) on a full frame body.

 

Below are an overwiew of the image and three close ups into interesting regions.

 

Region 2 (left side of image)

 

In region 2 on the left side of the image, I observed very weird diamond-shaped bokeh. The drops in this region are in front of the focus and therefore are expected to generate out-of-focus blur (bokeh). The effect increases with field size (== distance from center) and the rotation of the diamond-shapes obviously depends of the azimuthal position of the field point "the diamonds rotate with the field".

 

I would have expected circular out-of-focus blur with part of the out-of-focus disc beeing cut away because of vignetting.

 

The effect of the diamond-like bokeh doesn't seem to be present on the right side of the image (region 3) where the drops are behind the focus.

 

The diamonds don't look coma-shaped as from decentered or tilted lenses (and in this case the effect should be field independant anyway). My gut feeling is more that this could be something like field-dependant astigmatism but in this case, the field-dependency should be linear across the field (beeing zero in the center) and the diamonds should appear also on the right side of the image (with an astigmatism of opposite sign).

 

Any better interpretation of the diamond shape?

 

Center (region 1):

 

The drops are about focused in the center of the image. There are magenta and greenish drops. The only interpretation I can imagine is that the greenish drops are in front and the magenta drops behind the focus (or vice versa) and the color comes from strong longitudinal color such that in one focus position, the green is already defocused (while the magenta beeing still sharp) and in the other case magenta is defocused while green is still sharp.

 

The significance of this effect makes me thinking that this lens has A LOT of longitudinal color (or whatever generates this color split).

 

Any better interpretation of this effect?

 

Right (region 3):

 

This is the region where the out-of-focus blur comes at least closest to what I would have expected. The out-of-focus blobs are systematically colored (center magenta, edge green) which could come from a mixture of longitudinal color, spherical aberration and spherochromatizm.

 

 

My conclusions:

1. I didn't observe unexpected results like this (both image quality and AF) with any other lens I had before (eg. Canon and Zeiss). This clearly doesn't have any statistical significance, as I only had a few lenses and just might have been unlucky with this sigma lens. It makes me wonder though, if sigma sacrifices quality control in favour for beeing competitive. I had a Zeiss Otus for loan some time ago and it delivered exceptional image quality. Again -- to be fair, this is just a comparison of two samples but in this case Zeiss delivered high-end as-built image quality and sigma didn't. To be even more fair, I've to say that with the Zeiss lens not only 50% (like for the sigma) but rather something like 90% of my shots were defocused because my kids move faster than my manual focus capabilities ;-)
   
2. All the single sample reviews are of limited worth because they don't tell anything about the distribution of the as-built image quality but only give results from a single sample. If the sample is provided by the manufacturer, there's an additional risk that the manufacturer cherry-picks the sample to get an excellent review.
   
3. unfortunately many reviews don't evaluate AF capabilities but these might be important for the real world usability
   

Any comments are welcome!

 

 

 

 

Image regions in detail:

[popo]

Your lens is a bad one. Not just a bit off, a lot off. One to return.

 

I just tried mine on a simulated starfield (painted some white dots on black background on computer screen). Using AF (previously micro-adjusted) I took an in focus shot wide open using it on a 5D2. Dots remained mostly circular, with minor signs of astigmatism in corners.

 

Switching to MF, I focused in front of the monitor plane. The bokehballs remained circular with a slight green fringe. This may be stronger if the contrast was higher. Cats eye effect started to happen away from centre. There was also a horizontal cut in the bokeh on the side closest to the centre, presumably due to the rectangular mirror box in camera. The rear element of the lens is bigger than the short side of the mirror box!

 

I repeated this, now focusing behind the monitor. The bokehballs remained circular with a magenta fringe. There was also the cats eye and chopped effect, but this time on the outside of the balls.

 

I don't know what element(s) need to be out of place to create the diamond effects but that lens is not functioning as designed while it does so.

[joze]

popo / thanks for confirming this. I already sent back the lens to the dealer.

[JJ_SO]

Your "lawn sprinkler test" has two serious problems. fast moving drops and different light situations - some drops are in bright light, others in shadow. The third problem is only mine. Shooting a sprinkler at f/1.4 is a kind of testing which doesn't say much about the quality of a lens. Especially with a narrow DoF. Also we don't know about the shutter speed, I couldn't read the EXIF.

 

But maybe the Sigma is just not the best for shooting your sprinkler wide open, maybe you just need to get a very decent sprinkler lens  ^_^ I'm sorry, I don't belong to the sprinkler-fanclub, I was just taking this one because I found the water surface nice. But even at 1/4000 the drops are not perfectly round. And I wouldn't expect from a lens which costs 23% (!) of the reference to be equal wide open. This is what the hype's about: it is close in some aspects, to me close enough, but it is not apochromatic. I do find it a bit weird: First get the top of the top notch lens and then compare any competitor. Do you need reasons to buy the Otus? 

 

 

<p style="margin:0px 0px 0px 0px;">

<p style="margin:0px 0px 0px 0px;"> 

<p style="margin:0px 0px 0px 0px;">For drop-peepers larger version

<p style="margin:0px 0px 0px 0px;"> 

<p style="margin:0px 0px 0px 0px;">Could you show us the drops of the Otus, please?

<p style="margin:0px 0px 0px 0px;"> 

<p style="margin:0px 0px 0px 0px;">Here's another one from the same series showing the same phenomenon as yours. If the drops are very close to the camera, they show a rhomboid - so what? Important thing is, what does the lens show in the proper focus plane.

<p style="margin:0px 0px 0px 0px;"> 

<p style="margin:0px 0px 0px 0px;">I think the rhomboids are easy to explain. Same effect like in the old days when we found it creative to turn a zoom ring while the shutter was open. The drops out of the sprinkler come in a ballistic curve and are moving in space, not in a plane. So there are different "blur sources": the speed, the curve shape of the movement and the approach towards the camera which by itself increases the bokeh not constantly but increasingly faster. Most of us have seen the picture with the race car with oval wheels because of the shutter curtains movement, a bit of that we see here, too. Next time I visit those sprinklers here I'll take some different shutter speeds to see if my guess is completely off or not.

[joze]

Hmm Joju, you're right, the lawn sprinkler test isn't rigorously scientific ;-)

 

Was the lens for your images also the sigma 50mm ART lens?

 

I've no drops shot with the Otus as I had the Otus in last winter (no sprinklers then) and the shots I have are too personal to share here. The sharpness of the Otus fully open (in the shallow depth of focus where it's sharp) is a bit like if you would have stopped another high-end lens down to something like 5.6.

[Klaus]

As mentioned in the review, the foreground bokeh is harsh.

The diamonds - this is probably a combination of a variety of factors (astigmatism, velocity of the drops).

As far as focus accuracy is concerned - welcome to the world of DSLRs.

Just take the Canon 50mm f/1.2 USM L as another example.

[popo]

Makes me want to try this now, although a bit late at the moment! Thinking more I may have been a bit quick to blame a "bad lens", although in the use I've had so far from my 50 Art I hadn't come across anything like that. The nearest is some effect in the corners, but I've not tried with high contrast (very bright highlights) type shots so that might expose weaknesses I haven't seen yet.

 

I probably should get more sleep...

 

Edit: added the test shots I did earlier. No diamonds here, but as said these are lower contrast. And relative defocus may play a part in it?

[JJ_SO]

Hmm Joju, you're right, the lawn sprinkler test isn't rigorously scientific ;-)

 

Was the lens for your images also the sigma 50mm ART lens?

 

I've no drops shot with the Otus as I had the Otus in last winter (no sprinklers then) and the shots I have are too personal to share here. The sharpness of the Otus fully open (in the shallow depth of focus where it's sharp) is a bit like if you would have stopped another high-end lens down to something like 5.6.

 

Yes, I also had the idea to run across a sprinkler with the Sigma. So different are tastes, I like what it did with the drops. Also, the Sigma might not be as sharp wide open as the Otus, but from f/2 on it's better than all other lenses I own at f/4. I also like very much the way it renders colors and models the third dimension  :wub:

 

I can't judge about Canon, but I'm not overly indulged by Nikon's AF technology and I had lenses with some keeper rate of about 50% wide open (and with some movement). The Sigma doesn't blow it up to 100% but it is at least as reliable as any Nikkor. After 2 years learning with D800 I'm used to take more than one shot, if I have to open up the aperture and not every click is sharp, but the ones which are - always an awesome impression.

 

And now I need to sleep as well 

[davidmanze]

 Probably the hardest test ever conceived for any lens!

 

     There is aways extreme contrast around water droplets when the sun is shining through them, to me it is not the least bit surprising that there are aberrations probably of every sort, it is one hell of a big ask to expect any lens wide open and at 100% to cope with that. Even the ripples on water can produce purple fringing as is seen in Joju's image of the pond. 

 

   Try the same test with the Otus it's likely to be the only lens to better it.

 

 

 

   However, I wouldn't be complaining about it raining diamonds, bring it on!

[JJ_SO]

The thing about the hype around this lens is, some people now expect miracles. "Better or equal to Otus" plus AF plus USB dock plus... pardon me, contrary to Hasselblad, Zeiss doesn't put some nicer knobs and shell around a Sony and is selling it for 5× the price. They made some expensive decisions and they got simply the best lens. Everyone interested in zoom gets a 70-200/2.9 for 20k $ from them which doesn't focus shift because it's designed for movie cameras.

 

If I were in a business photographing crystal chandeliers or sprinkling fountains and all in dim lit places and wide open, the Otus is a bargain. As long as one doesn't need AF, because it saves so much time in post by just being apochromatic corrected and very well made.

 

But the Sigma remains a bargain, too, and is not only 23% of the Otus performance. The bargain and the real part of the hype is the comparison to other AF lenses, usually more expensive than the Sigma and it outperforms them - not in each aspect and not always but in summary - please show me a better one amongst the competitors and under 3000$. 3k is a huge amount of money for the perfect sprinkler shot - do I like sprinklers enough?

 

Here's one with tiny water drops falling down from a defect gutter. So the velocity (and shutter speed, too) was slower. Pushing water upwards needs pressure and increases speed when it's leaving the hose. I didn't correct CA in post, I find it difficult to some pictures, especially when spherical transparent bodies are involved - I never get rid of all colors, some are fringing, others are normal physical behavior of water drops. The diffraction of them will always show colors - that's what we see in each water drop in a macro picture, don't we? Drops are part of rainbows. How could there be a rainbow like the ones we're used to see, if water drops wouldn't behave like prisms? In that shot, I spotted round drops in focus, lots out of focus and some of them showed yellow and blue instead of magenta and green.

 

Today, I won't be shure if Otus would show a different shape (than romboid) in the corners, with the same camera and parameters, while drops are

• travelling at speed towards the lens and
  
• out of the picture center and
  
• rectangular to the shutter curtain
  

Today I like to go back to the sprinkler and see what higher shutter speed will do, maybe some flashlight, too.