DXOMark: Sigma 24-105 f/4 DG OS Reviewed

[Random Orbits]

An f/4 zoom lens is not one you'd expect to have optimal light transmission anyway.

Isn't the point of lens design to have optimal light transmission? And if the f-stop is four and the t-stop half a stop worse, doesn't that say something about Canon's glass elements and coatings? The Canon 24-105 is a good lens but it should not have been branded with the red ring.

Yup, it implies that current coatings are better than something that came out in 2005. Canon's 24-70 f/4 IS has a t-stop = 4. The 24-70 II has a t-stop of 3 while the version I has a t-stop of 3.4.

 

[goldencode]

It's heavier than the Canon 24-70 F2.8 II and its an F4 lens.
82mm filter
Optically very similar to the Canon 24-105 F4 L.

So, what is the advantage of this lens ? I don't think it really adds any " Wow " factor.

 

[Random Orbits]

Yup, it implies that current coatings are better than something that came out in 2005. Canon's 24-70 f/4 IS has a t-stop = 4. The 24-70 II has a t-stop of 3 while the version I has a t-stop of 3.4.

It's not just coatings. The newer lenses have bigger front elements and that counts for a lot (more surface area on the front of the lens = more light gathering ability.)
[/quote]

Maybe... although it would affect some and not all a zoom's focal range. But it can also be done without changing the lens diameter. For example, the 24 f/2.8 and the f/2.8 IS both use 58mm filters. The old one has a t-stop of 3.2 while the new one is 2.8.

 

[unfocused]

Okay, all this debate over the "true" f-stop of a lens has me scratching my head.

Since metering is through the lens and presumably the exposure selected by the camera is going to reflect the actual light hitting the meter, what's the real world impact here?

Yes, I get that a mis-marked lens means that your f4 lens won't have as great of a light gathering power if it is really an f4.5 lens, but unless you are shooting wide open, what's the practical effect. Your images will still be properly exposed.

Now, obviously if you bought a lens thinking it was an f2.8 and it was really an f4.5 that would be some serious fraud going on.

While were on the topic, aren't there also some standards that must be met. After all, ISO stands for the International Order for Standardization.

 

[iowapipe]

Yup, it implies that current coatings are better than something that came out in 2005. Canon's 24-70 f/4 IS has a t-stop = 4. The 24-70 II has a t-stop of 3 while the version I has a t-stop of 3.4.

It's not just coatings. The newer lenses have bigger front elements and that counts for a lot (more surface area on the front of the lens = more light gathering ability.)

Maybe... although it would affect some and not all a zoom's focal range. But it can also be done without changing the lens diameter. For example, the 24 f/2.8 and the f/2.8 IS both use 58mm filters. The old one has a t-stop of 3.2 while the new one is 2.8.
[/quote]

Just as a quick question: while the filter size is the same, is the size of the front element also the same?
I tried a quick lookup to find specs of the element size and couldn't find anything. So I can't answer my own question.

 

[traveller]

There's a lot of Sigma hate going on in this post. I can understand that quite a few people have been burned by third party lenses in the past, I've also had mixed results, but I think that's a reason to be careful with what you buy, not to dismiss an option outright. I don't currently own any Sigma lenses, but I had the 10-20mm f/4-5.6 before moving to full frame, I enjoyed using it very much and got good results. Unlike the argument that is often put forward, I don't think that non-Canon lenses necessarily have poorer resale values, so long as you buy decent quality lenses that appeal to the (often more savvy) second hand market; the hit that I took selling my Sigma was no worse in percentage terms than the hit I would have taken selling the Canon EF-S 10-22mm.

I currently own the Canon 24-105mm f/4L IS and while I am not in a hurry to rush out and buy this new Sigma to replace it, I am glad that people looking for a lens in this class now have another option (other than going for a 24-70 f/2.8 ). If the DXO Mark data is correct (and I am looking at the measurement data, not the silly "single number" metric), this Sigma looks quite a bit better wide open at the long end of the zoom range. This is great news, I'm often torn between the 24-105 and the 70-200 when I get to 70mm (maybe I need a second body, but I can't afford another 5D MkIII -the first one was painful enough on my bank balance!).

For those that absolutely will not countenance owning anything but Canon branded lenses, perhaps the appearance of this new 24-105 will motivate them to produce a 24-105 f/4L IS II...

 

[LetTheRightLensIn]

Yeah, it only undercuts the Canon because DxO still has the 24-105 costing 1250, which is far above its current street price. It may be a slightly better than Canon's 24-105, but with the Canon version selling at 600-700, the Sigma isn't quite the bargain as when the Canon sold at 1250.

Canon's MSRP for the lens is $1149.

Who cares? Even years ago before the price collapsed I didn't pay close to $1149 for it. You can get them new, for $750 EASILY now, EASILY and for $600 with just a bit of effort. Real world is what counts.

$1149 is nuts! Who pays that? Heck I got a new, from a major camera store, full US warranty, 24-70 f/4 IS for $125 LESS than that, so who on Earth would actually pay $1149 for the 24-105 these days?

 

[LetTheRightLensIn]

Reading the trend in this thread regarding light transmission reminded me of something that I read in Bryan Carnethan's review of the 35mm f/2 IS.

"The above images were identically exposed with exception of the Sigma 35 that needed a 1/3 stop longer exposure to produce a histogram equal to the other lenses in this comparison. The Canon 35 f/2 is about 1/6 stop brighter in comparison, but its exposure was not adjusted in this comparison. "

Just to provide some balance. Most manufacturers fudge a little on light transition and focal length to market lenses at certain acceptable standards. While it's not a huge deal, it is worthy to note lenses that deliver better light transmission because you may be getting slightly more bang for your buck.

Yeah it can vary. I found the sigma 120-300 2.8 non-OS to be about f/3 and 280mm if we take the 300 2.8 IS to be f/2.8 and 300mm.

And I found the 100L f/2.8 to be getting towards 1/3 stop brighter and 1/3 stop less DOF than the 100 non-L macro from Canon both at 100mm f/2.8 (and no I didn't have one in macro zone and the other out).

I forget the details but in a few cases the 70-300L, I believe, seemed to be a bit brighter than the 70-200 f/4 IS at the same aperture and focal length.

Yeah the 24-105 is kinda dim, another reason why the 24-70 f/4 IS and 24-70 II are even better in comparison than a direct aperture comparison makes it appear to be.

 

[LetTheRightLensIn]

It's heavier than the Canon 24-70 F2.8 II and its an F4 lens.
82mm filter
Optically very similar to the Canon 24-105 F4 L.

So, what is the advantage of this lens ? I don't think it really adds any " Wow " factor.

exactly. I'd rather pay $125 more and get the much smaller and lighter 24-70 f/4 IS (with macro mode) or pay less and get by with the 24-105L (or pay more and get a better Tamron 24-70 VC or really pay more and get the best 24-70 II myself since it's not like those add any size or bulk compared to the sigma, although the 24-70 iI does lack the IS, it does offer f/2.8 though).

I think this lens may look pretty good in the Nikon world though.

 

[candc]

I have always been a bit confused about the whole f stop vs. t stop comparison. If I understand it right then a 100 mm f/2 lens has a 50mm max aperture, 100/2 = 50. The lens may pass less light, that is the t stop, an actual transmission or equivelant rating. Does the camera recognize this and make exposure calculations on a lenses light transmission capabilities or does it default to the embedded code of the lens? Is slight underexposure only an issue at max aperture?

 

[Zv]

I have always been a bit confused about the whole f stop vs. t stop comparison. If I understand it right then a 100 mm f/2 lens has a 50mm max aperture, 100/2 = 50. The lens may pass less light, that is the t stop, an actual transmission or equivelant rating. Does the camera recognize this and make exposure calculations on a lenses light transmission capabilities or does it default to the embedded code of the lens? Is slight underexposure only an issue at max aperture?

If you have two lenses of similar design - one T 2.8 and other T 3 and put them both on the same camera at the same settings you would notice one is slightly brighter. With the camera in Av mode the metering system would compensate via shutter speed so they look similar. The downside being you lost speed. How important that is depends on what you're shooting - action vs landscape for example.

Since the 24-105L isn't really known for it's action freezing abilities it's not even an issue. For low level light situations it could make a slight difference though. But then again for low light you'd use a faster lens.

However, more light is always welcome so in the grand scheme of things better T stop values are desirable.

 

[Zv]

So you attach an f/2.8 lens to the camera, the max aperture reading f2.8 you see is based on what the lens tells the camera, not what the actual light reading the camera see's based on the focal length?

Yes exactly. F stops only refer to the physical size of the aperture. It does not represent the actual amount of light the lens can transmit.

The maximum amount of light that gets through when the lens is at f/2.8 is what the T stop refers to. By the time that light has gone through the front element, all the elements in the middle (could be several) and out the back and onto the sensor it is no longer at 100% because some of it got reflected along the way. How much got reflected depends on the quality ofthe glass and the coatings.

Easy way to think of it. F stop is just how big the hole is. T stop is how much light is allowed to pass through the hole. If that hole is covered in spider webs not much light gonna get thru! Haha!

Also when you stick a filter in front of the lens it also affects the T stop since the filter is essentially adding another element in the light path.

It makes sense why prime lenses have better image quality than zooms. The zooms tend to have more glass elements thus blocking more light.

No lens is perfect, you'll always lose a bit of light.

 

[candc]

Okay, I think I understand how the whole lens, camera metering works. The lens tells the camera its max aperture, the camera takes a reading based on that and the light it's seeing, it's actually basing the calculation on the t stop it's seeing relative to the f stop it's given?

 

[privatebydesign]

Okay, I think I understand how the whole lens, camera metering works. The lens tells the camera its max aperture, the camera takes a reading based on that and the light it's seeing, it's actually basing the calculation on the t stop it's seeing relative to the f stop it's given?

Close but not really. Forget T stop. The difference between f stop and t stop does not result in slight underexposure even at max aperture.

The camera is basing the exposure on the light it sees. The T value is one factor in that but a very minor one, of magnitudes greater importance is the luminosity of the scene. It doesn't matter to the camera if you have a low transmission efficiency, or the scene is darker, or if you put on a ND filter etc, it adds up to the same thing, longer shutter speeds.

The camera is not basing the calculation on the t stop it sees per se, it never knows what the t stop is and doesn't care, it is basing the calculation on the light it sees, that is all. The difference between the light there is (which an ETTL meter never has a way of knowing) and the light that comes through the lens (which it does), is impacted by the t stop and filters etc, but that difference is not a factor in the exposure calculation.

 

[Zv]

Okay, I think I understand how the whole lens, camera metering works. The lens tells the camera its max aperture, the camera takes a reading based on that and the light it's seeing, it's actually basing the calculation on the t stop it's seeing relative to the f stop it's given?

T stop is just a theoretical (measured in a lab??) value.

When you half press the shutter the meter is activated and the lens does nothing really, it's already wide open to allow as much light in as possible. The camera then does it's calculations based on the amount of light hitting the sensor.

When you fully press the shutter the aperture then adjusts to the required setting and the shutter moves to achieve a correct exposure.

The only time you might notice T stops is if you were using M mode and switching lenses for the same scene. For example lets say you set up a shot at f/2.8 1/60 ISO 100 and it looks perfect. You then switch lenses and set it to the exact same settings and notice the shot is under exposed slightly. No big deal you just crank up that ISO or adjust shutter speed and you're back on track!

An example where you might encounter this situation is say you're using the 24-105L and shooting at 24mm and say f/8. You decide you're not a fan of the 24-105L at the wide end so you switch to your 24L or whatever and try again. Surprise surprise it looks better now!

That's basically what this is all about. The Sigma lens would be slightly brighter and of course better. But the amount is fairly small.

 

[Sporgon]

T stops and f stops all are pretty easy to understand in principle but....

Ever since I first started using the 24-105 in 2005 I found that if you shot from the camera's meter, or suggested exposure in manual, this lens underexposed the frame by about one third of a stop compared with say a fast lens such as the 50/1.4, so the image was slightly more 'dense' and the histogram more dense and to the left.

Now according to DXO the 24-105 has a t stop of 5.1 against an f stop of 4, so it looses two thirds of a stop.

Using the same source the 50/1.4 has a t stop of 1.6 against 1.4 so it is one third of a stop less; so; the difference between the two lenses actual transmission is one third and this is exactly the difference I find in exposure between the two, and this includes using a hand held incident light meter and setting the camera shutter speed and aperture manually.

So far so good; this difference equated exactly to the t stop differences between the lenses.

However there is a problem. The 24-70 f4 IS which by the same source has a t stop of f4 against f4 does exactly the same thing. Compared with a faster prime with less elements it underexposes by one third of a stop. So does complicated element lenses such as the 70-300L. Also I believe Edward Lang (eml58) found that his 200-400L underexposed on the same meter reading compared with the 400/2.8L .

So what's going on ? By the definition of what a 't' stop is, lenses of equal t stop should expose the same, but it seems to me that the more elements that are added the more the lens is likely to underexpose. Yet this is just what t stop is supposed to measure.

And before anyone asks, this isn't just on one copy of the lens; it is universal.

To me it suggests that 't' stop value is not accurate in practice, or at least in this application. Any ideas ?

 

[Zlyden]

Interesting discussion!

Now I think I understand practical purpose of this parameter in DxO stats.

Did anyone see lenses with calculated T-stop equal to F-number on them?

 

[TWI by Dustin Abbott]

Interesting discussion!

Now I think I understand practical purpose of this parameter in DxO stats.

Did anyone see lenses with calculated T-stop equal to F-number on them?

Many of the better primes (particularly the new ones) are much closer. I've already mentioned the new EF 35mm f/2IS, which has both an f-number and t-stop of 2. In the past, I would say that many primes are more likely to be close than zooms (where more compromises are made), and that does speak well of the light transmission of the new Sigma 24-105. because it manages to transmit just about (almost) the amount of light that the f-stop suggests.

 

[Zv]

Interesting discussion!

Now I think I understand practical purpose of this parameter in DxO stats.

Did anyone see lenses with calculated T-stop equal to F-number on them?

Many of the better primes (particularly the new ones) are much closer. I've already mentioned the new EF 35mm f/2IS, which has both an f-number and t-stop of 2. In the past, I would say that many primes are more likely to be close than zooms (where more compromises are made), and that does speak well of the light transmission of the new Sigma 24-105. because it manages to transmit just about (almost) the amount of light that the f-stop suggests.

How can the T value be the same as the f number if the lens clearly vignettes wide open? Is T value only measured in the center?

Just asking, I have no clue how the T value is actually determined.

 

[TWI by Dustin Abbott]

Interesting discussion!

Now I think I understand practical purpose of this parameter in DxO stats.

Did anyone see lenses with calculated T-stop equal to F-number on them?

Many of the better primes (particularly the new ones) are much closer. I've already mentioned the new EF 35mm f/2IS, which has both an f-number and t-stop of 2. In the past, I would say that many primes are more likely to be close than zooms (where more compromises are made), and that does speak well of the light transmission of the new Sigma 24-105. because it manages to transmit just about (almost) the amount of light that the f-stop suggests.

How can the T value be the same as the f number if the lens clearly vignettes wide open? Is T value only measured in the center?

Just asking, I have no clue how the T value is actually determined.

Good question. I can't see it being anything else, as the average from almost every lens towards the periphery would bring the score down