A Brief Comparison of the Canon 300 f2.8 L and the Sigma 120-300 F2.8

pathfinder · October 25, 2006

Greenjk has a thread in the Sports Gallery ( http://www.dgrin.com/showthread.php?t=45249 ) asking about the Sigma 120-300 f2.8 and john asked if they were really the same focal length. I told him to give me a couple days to answer his questions. So here we go.

First some data

Canon 300mm f2.8 L IS

Minimum Focusing Distance 8.2 feet

13 groups/ 17 elements

weight 6.0 lbs

Length 9.9 inches

Diameter of front objective 107mm measured by me with a metal ruler

Sigma 120-300mm F2.8 EX APO DG HSM

Minimum Focusing Distance 4.9 feet

16 groups/ 18 elements

Weight 5.7 lbs

Length 10.6 inches

Diameter of front objective 98 mm measured by my hand

The Canon calculations

300mm/ 107mm = 2.803 The Canon diameter calculates to exactly f2.8 for a 300mm focal length. I doubt that lens diameter was accidental or random, it is just too precise.

The Sigma calculations

300mm/ 98mm = 3.06 not f2.8 UNLESS the lens focal length is shorter than 300mm. How much shorter?? 2.8 x 98mm = 274.4 mm

Is this possible, did the Sigma folks just "round up" the focal length to 300mm for marketing purposes, or is the lens really not f2.8 but about 1/3 stop slower than that.

To find out the answer to the above questions, I mounted the Canon 300mm and the Sigma 120-300 on my trusty 20D on a tripod and focused on a barn across the way. I tried to keep the film/sensor plane of the camera in the same precise position for each lens, and I am sure I was within 1 cm of that in accuracy.

The images were shot at f5.6 1/125 at ISO 200, with each lens, the RAW files were processed exactly the same settings in Adobe ARC. I accepted the suggested values for the first frame and used those same values for the succeeding frames from the Sigma lens and a Tamron 200-500 that I set to 300mm and used for comparison.

Here is the frame with the Canon 300mm f2.8 IS L

I count about 26.5 units of the green fence posts that I am using for a measuring standard. The black circle above the door is where I sampled the color pixels and read 209, 208, 209. Pretty neutral grey. The original file can be seen by using the URL for the image and changing the L to an O.

Here is the frame from the Sigma 120-300 at 300mm.

I roughly measure the green fence post units as 27.5 - a little wider view than the Canon 300. The white spots over the door outlined in black reads 214, 214, 209. Maybe the slightest bit whiter than the Canon, but not enough different to represent a significant difference in exposure. Slightly redder and bluer than the Canon lens.

Here is the Tamron SP 200-500 DI lens at 300mm for comparison

I measure roughly 26 fence post units here - closer to the Canon than the Sigma; the focal length was an indicated 300mm on the lens barrel. The selected pixels read 212, 209, 206. Slightly redder than the Canon lens again, perhaps. Overall the image seems slightly greyer, darker.

Here are the lenses side by side

And in B&W via Channel Mixer

Here is a picture of the frontal elements and their measurements. DO not try to measure the image itself, I cannot verify that they will be representative

And lastly, I like this shot of the front of the Canon lens, as it demonstrates 9 (nine) reflected images of the flat panel flash used for this shot.

I will be happy to answer any questions about how this post was constructed. Criticisms and Comments welcomed..

gluwater · October 25, 2006

PF I think you need to FedEx me those lenses so I can verify your results

. Thanks for the comparison.

pathfinder · October 25, 2006

Nick,

You know you can borrow them in Florida if you promise to be very careful.

Or ply me with brewski's!

ian408 · October 25, 2006

The 300/f2.8 is one sweet lens.

Manfr3d · October 26, 2006

So what's your final conclusion about the fstop of the
Sigma regarding the ability to gather light? If you say
the Sigma is 1/3rd stop slower you are actually only saying
that the dimensions of the sigma dictate a length/opening
ratio of 1/3 less than the 300/2.8 ratio but thats not
exactly the difference in lightgatherbility. One might have
to shoot with these lenses @ 300mm f2.8 with a fixed
exposure time. Then compare the luminance levels of the
taken images.

pathfinder · October 26, 2006

Manfred - That is exactly what I did.

ALL THREE images I posted were shot precisely at 1/125 F5.6 ISO 200.

They all were the processed in Adobe RAW convertor with exactly the same settings.

Looking at the images then, and the pixel data I included, I can see no significant difference in exposure. Certainly nothing significant to the image. But the image with the Sigma is slightly smaller, or has a slightly shorter focal length.

If I take the Simga image and position it over the Canon image in PSCS2, and then blend in the Difference mode, it is very easy to see that the Sigma image elements are slightly smaller between 5 and 10% maybe.

Manfr3d · October 26, 2006

Oh sorry, I shoudl've looked at the EXIF data

Very nice comparison of the two lenses!

claudermilk · October 26, 2006

Thanks for the comparison. That Canon lens is truly amazing (I have had the opportunity to use one...drool) & hard to beat. It looks like the Sigma gets pretty darn close, especially considering the price difference & that it's a zoom not a prime. This definitely cements the 120-300 on my wishlist.

DoctorIt · October 26, 2006

Good technical review, PF

erich6 · October 26, 2006

This is interesting...how accurately do lensmakers have to be in reporting focal lengths? At what point do you claim false advertisement?

Erich

pathfinder · October 27, 2006

I don't know, Erich,

Actually, I think focal length ( the distance from the center of the lens to the focalplane ) varies depending on the distance of the subject from the focal plane also. As with lenses that extend during focus changes, especially zoom lenses. The focal length of some macro lenses change a fair amount when focusing at very close objects also I believe.

I do not know enough about the variables in complex lens design that determine focal length. Focal length is fairly simple for single lens systems, but not nearly so simplistic for lenses with moving elements.

I would be interested in hearing some expert opinion in this regard, but I am not really interested in hearing folks slam Sigma's marketing. I think we need to be sure about what we are talking about before making any attempt at criticism as none was intended in my post.

I emphasize once again, the actual difference in image size between the Canon and the Sigma is barely discernable without superimposing the images.

Blurmore · October 27, 2006

I once conducted at test of my 85 f1.8 to my 17-85 EF-S IS and found that Canon fudged the FL of the 17-85 by about 4 mm. Besides the obvious closer view the 85 f1.8 (no suprise) rendered the subject more three dimensionally, and compressed. This equates in people photography as having a slimming effect on the subject, but withoutpincushion distortion. The 17-85 while ALMOST as close made the subject look flat by comparison to the 85 at all comparable apertures. I think you can chalk this up to the unique properties of primes vs zooms. So far as companies fudging their measurements I'm sure this is nothing new, a 17-81mm lens is no where near as sexy as a 17-85, and most people would never check or know the difference anyway.

pathfinder · October 27, 2006

The discussion of the image quality of primes versus zooms is a fascinating topic, that doesn't seem to get much press these days. We rarely compare the image of a zoom with that of a prime shot at the same time under the same lighting conditions. Zooms are good enough, and so handy, we tend to favor them for routine use because of their convenience.

Modern zooms have gotten so good that we tend not to see their defects - as Blurmore points out, it may not be resolution at all that a zoom lacks, but other qualities, like barrel distortion, and lower contrast. Or a few scant millimeters of focal length

D:D

rofl

erich6 · October 27, 2006

pathfinder wrote:

I don't know, Erich,

Actually, I think focal length ( the distance from the center of the lens to the focalplane ) varies depending on the distance of the subject from the focal plane also. As with lenses that extend during focus changes, especially zoom lenses. The focal length of some macro lenses change a fair amount when focusing at very close objects also I believe.

I do not know enough about the variables in complex lens design that determine focal length. Focal length is fairly simple for single lens systems, but not nearly so simplistic for lenses with moving elements.

I would be interested in hearing some expert opinion in this regard, but I am not really interested in hearing folks slam Sigma's marketing. I think we need to be sure about what we are talking about before making any attempt at criticism as none was intended in my post.

I emphasize once again, the actual difference in image size between the Canon and the Sigma is barely discernable without superimposing the images.

PF,

Your comments made me think about all the potential variables that could be leading to this. I recall Ziggy making the point once that the f-number designator on lenses isn't always the focal length divided by the aperture diameter because the lenses are calibrated against some exposure standard.

What that means is that the actual f-number will depend on how much light gets through the lens. So, if you assume that the actual f-number of the Sigma is f/3.06 then it's possible they can be "rated" at f/2.8 if each of the Signma's lens elements are ~1% more transmissive than the Canon's. That's certainly entirely possible.

I don't think having multiple optics has much to do here...that just changes the optical design so you have an "effective" focal length instead of a single-lens focal length. And being a zoom vs. a prime just gives you a range of effective focal lengths.

So, from a design standpoint you can get a lens with different apertures but same focal length if the transmission of the optics are different. This is a likely design trade that companies have to sort through for each of their products. Highly transmissive optics are harder to make but don't require to be as large. Depending on the supply chain and processes you may opt for one or the other. From a photographer standpoint I would think higher transmission wins all the time because now you have a smaller lens to lug around with less elements and likely better quality because of reduced flare.

As you pointed out, though, the Sigma does have a bit more angle-of-view which means it's probably not quite 300 mm and the optics are not quite 1% more transmissive than the Canon's.

Erich

pathfinder · October 27, 2006

Erich,
One of the facts I was trying to ascertain, was whether there was any significant difference in exposure with each of the three lenses tested - all shot at ISO 200 f5.6 1/125th of a sec

The pixel data I circled in black, above the doors in each image, measured 209,208,209 with the Canon lens, 214,214,209 with the Sigma lens, and 212, 209, 206 which to my eye is not a significant difference in exposure between the three lenses, which I interpret to mean the the apertures are accurately labeled as to aperture number.. The differences may speak to the slightly different color renditions transmitted by each lens. Might I said.

You are correct than one lens could possibly transmit more light per unit area than another. But how much more could a lens transmit per unit area??

I calculate the area of the front element of the Canon lens as (107/2)(107/2) * Pi = 8991.8 sq mm. The areas of the Sigma lens is (98/2)(98/2) * Pi = 7542.7 sq mm.

The Canon lens has about 1449.1 more sq mm of frontal surface area. That is about 16% smaller lens area for the Sigma. Lens designers strive very hard to maximize light transmission in lenses to minimize reflections and scatter, so I find it hard to believe Sigma could get 16% more light through their smaller frontal optic. The largest difference in light transmission is easily less than 1% with the best modern lenses.

I still think the front element in the Sigma lens is smaller because the focal length is slightly shorter. Not enough to be upset about , but shorter nonetheless.

erich6 · October 29, 2006

pathfinder wrote:

Erich,
One of the facts I was trying to ascertain, was whether there was any significant difference in exposure with each of the three lenses tested - all shot at ISO 200 f5.6 1/125th of a sec

The pixel data I circled in black, above the doors in each image, measured 209,208,209 with the Canon lens, 214,214,209 with the Sigma lens, and 212, 209, 206 which to my eye is not a significant difference in exposure between the three lenses, which I interpret to mean the the apertures are accurately labeled as to aperture number.. The differences may speak to the slightly different color renditions transmitted by each lens. Might I said.

You are correct than one lens could possibly transmit more light per unit area than another. But how much more could a lens transmit per unit area??

I calculate the area of the front element of the Canon lens as (107/2)(107/2) * Pi = 8991.8 sq mm. The areas of the Sigma lens is (98/2)(98/2) * Pi = 7542.7 sq mm.

The Canon lens has about 1449.1 more sq mm of frontal surface area. That is about 16% smaller lens area for the Sigma. Lens designers strive very hard to maximize light transmission in lenses to minimize reflections and scatter, so I find it hard to believe Sigma could get 16% more light through their smaller frontal optic. The largest difference in light transmission is easily less than 1% with the best modern lenses.

I still think the front element in the Sigma lens is smaller because the focal length is slightly shorter. Not enough to be upset about , but shorter nonetheless.

PF,

The 16% is exactly the difference and the point I was making but it only requires each optical element to be better by 1%.

Based on the data you provided the Sigma has 18 elements. If you consider each element to have 0.98 transmission then the total transmission is 0.98^18=0.6951. The Canon has 17 elements. If each element is 1% less transmissive or 0.97 then the total transmission is 0.97^17=0.5958. That means the Sigma would have 0.6951/0.5958=17% more transmission. That would make the Sigma's exposure about the same as the Canon's even though its aperture is smaller and collects 16% less light.

Like you said the focal length is probably a bit shorter than claimed which would make the optics not quite 1% better. Even so, it seems this is not that far off.

Erich

pathfinder · October 29, 2006

Interesting idea, Erich.

I wonder how much more light is lost due to the IS mechanism - there must be two air-glass interfaces for the mobile lens unit that is activated by the IS mechanism, so is it possible that IS introduces light loss that a non-IS lens does not have? Or not?

erich6 · October 29, 2006

pathfinder wrote:

Interesting idea, Erich.

I wonder how much more light is lost due to the IS mechanism - there must be two air-glass interfaces for the mobile lens unit that is activated by the IS mechanism, so is it possible that IS introduces light loss that a non-IS lens does not have? Or not?

I'm not sure.... I don't really know the details of how they implement IS. I would guess it involves some for of piezoelectric actuator to tip/tilt the selected elements inside the lens. This may require a more hefty mount than a standard lens and this could encroach in the clear aperture of each lens. I always thought that the actual aperture stop for all SLR lenses is actually inside the lens as opposed to the front element. You want that big lens to be it but ultimately you have to trade off stray light and flare against max aperture.

This is the optical layout from the Canon Camera Museum:

They really don't show you much...

I think if you really want to know the focal length you have to measure the magnification. One easy thing you can do is compare the relative width of the fence posts between lenses and at least you'll figure out how much shorter is the Sigma from the Canon.

Erich

pathfinder · October 30, 2006

I did use the fence as a ruler, Erich and it was roughly 26.5 or so for the Canon image and 27.5 give or take for the Sigma image.

A better way to evaluate the relative image size is to put one image on top of the other in the Layers Pallete and look at them in the Difference Blending mode - This mode only shows the pixels that do not match up, and the Sigma image is very slightly wider than the Canon, but I can't give a number for the difference in size.

As for the size of the frontal objective, larger, faster lenses have larger frontal elements to gather more incident light because light begins diverging from the subjects surface immeditely. This is true for all telescopes unless they have multiple front lenses as an array.