AF: EF vs EF-S?

Nikolai · July 10, 2008

I've been using my copy of EF-S 17-55/2.8 IS USM a lot since I got it a year or two ago. And for the last few months I started to deviate to EF 70-200/2.8 IS USM.

What I noticed recently is that when it comes to a very low light situation (e.g. after sunset) 17-55's AF system really starts struggling. I find myself switching to MF since the focus hunt becomes really intolerable. At the same time 70-200 focuses just fine long after 17-55 pretty much gives up.

Technically, the amount of light coming to the APC-size sensor is the same in both case, the extra light from EF should be simply lost by the sensor form factor. AF's decision-making system is in the body, not in the lens. Both lenses are f/2.8 IS USM. I wonder what gives. Glass quality? Shall I get 17-40L? Or do I simply have a bad copy of 17-55? :scratch :dunno :rolleyes

silverstangs · July 10, 2008

Nikolai wrote:

Technically, the amount of light coming to the APC-size sensor is the same in both case, the extra light from EF should be simply lost by the sensor form factor. AF's decision-making system is in the body, not in the lens. Both lenses are f/2.8 IS USM. I wonder what gives. Glass quality? Shall I get 17-40L? Or do I simply have a bad copy of 17-55?

The light that hits the APC sensor is the same, but I would thinking that the amount of light entering the body is not considering that the 70-200 has more glass on the backend than the 17-55 lens. Remember the 70-200 was designed for full frame cameras, the focal point on the 17-55 is much smaller.

Nikolai · July 10, 2008

silverstangs wrote:

The light that hits the APC sensor is the same, but I would thinking that the amount of light entering the body is not considering that the 70-200 has more glass on the backend than the 17-55 lens. Remember the 70-200 was designed for full frame cameras, the focal point on the 17-55 is much smaller.

I understand that, however I was under the impression that size of AF mirror/assembly is comparable to the size of sensor, i.e. it's "cropped" to the same factor... No?

Scott_Quier · July 11, 2008

Nik - I've got the same two lenses. I'll run some tests using my 30D and see if I notice a difference. Stay tuned - my results should be posted either later this morning or early afternoon (EDT).

kdog · July 11, 2008

I just got this lens, but I don't have any other f2.8 glass so I can't compare directly. However, I'm not overly impressed at the low-level light focus performance on my 40D. Using the center focus point placed directly on a high contrast area, and the silly thing still wouldn't focus at dusk. One workaround that seems to work great is to enable all the focus points. I think the act of having multiple focus sensors inputs a lot more light to the AF computer, kind of like trying to read a book with just one shade open in a room versus opening all of them. Sure, you don't have explicit control over where your focus is, but you do get feedback where it is, and if you don't like it, you can simply release and refocus and it'll generally choose different spots. And of course the DOF on landcape shots is usually so great that it doesn't matter anyway. At least for my crappy pictures it doesn't. :giggle

You're joking about swapping for the 17-40 though, right? I think you'll be disappointed in going back to an f4 lens, not to mention I don't think the 17-40 is as reliably sharp as the 17-55. I'm pretty impressed at the water-clear quality of this lens so far.

Nikolai · July 11, 2008

Scott_Quier wrote:

Nik - I've got the same two lenses. I'll run some tests using my 30D and see if I notice a difference. Stay tuned - my results should be posted either later this morning or early afternoon (EDT).

I'm standing by, thanks!

Nikolai · July 11, 2008

kdog wrote:

I just got this lens, but I don't have any other f2.8 glass so I can't compare directly. However, I'm not overly impressed at the low-level light focus performance on my 40D. Using the center focus point placed directly on a high contrast area, and the silly thing still wouldn't focus at dusk. One workaround that seems to work great is to enable all the focus points. I think the act of having multiple focus sensors inputs a lot more light to the AF computer, kind of like trying to read a book with just one shade open in a room versus opening all of them. Sure, you don't have explicit control over where your focus is, but you do get feedback where it is, and if you don't like it, you can simply release and refocus and it'll generally choose different spots. And of course the DOF on landcape shots is usually so great that it doesn't matter anyway. At least for my crappy pictures it doesn't. :giggle

You're joking about swapping for the 17-40 though, right? I think you'll be disappointed in going back to an f4 lens, not to mention I don't think the 17-40 is as reliably sharp as the 17-55. I'm pretty impressed at the water-clear quality of this lens so far.

Yeah, exactly my behavior. Unfortunately, my dof is rather on a shallow side, so I can't really go hyperfocal.

And thanks, I totally forgot 17-40 is only f/4. That, and a missing IS is a total deal killer, so scratch that idea...

Scott_Quier · July 11, 2008

Nikolai wrote:

I'm standing by, thanks!

Testing methods:

Reduce ambient until the 17-55 starts hunting and/or fails to get a good focus lock. See how the 70-200 f/2.8IS behaves. Repeat until confident that observed behavior is consistant.
Repeat with the 70-200 going first.
With dim light, do some informal timings (no, a stopwatch was not used, deal with it:D)
Finally - see how these lenses performed with dim but decent light - like you might find at a night-time wedding reception.

Non-Scientific results:

When the light was reduced to the point that the 17-55 was starting to have problems, the 70-200 had virtually none.
When the light was reduced to the point where the 70-200 was just starting to have problems, the 17-55 never achieved focus lock.
In dim light but bright enough that both lens were able to achieve focus lock, the 17-55 usually hunted a bit (to a lot). The 70-200 might pass the focus point in one direction and then return to get the lock - no issues. As a result, the 70-200 usually achieved focus lock in about 1/2 the time it took the 17-55. Bear in mind, that this was in quite dim ambient (ISO 100, f/2.8, 1sec - you do the math for the EV).
When light was dim but decent, both lens performs wonderfully; there was virtually no difference in focus performance.

Conclusion - It would appear, based on these results, that the 70-200 is the better performer, getting focus lock under adverse conditions much more readily and more accurately than the 17-55.

It would be interesting if more individuals could contribute to this. In this way, we might be able to say something with a little more conviction.

Blurmore · July 11, 2008

My only question for the tester is what focal length were the tests performed at. It would stand to reason that at WA the 17-55 would hunt more as it has more real estate to stimulate the focal points, and it is harder to grab onto one spot and stick with it. I've killed 2 IS units in my 17-55 so I'd say I am a heavy user of the lens, and I have no issues with it during wedding receptions (probably the worst light you can get) yeah it hunts but not more than my 85 f1.8 at twice the aperture. I definitely notice it hunting more at the 55 end (like during a first dance) than at the 17 End like shooting a group dancing. It (or more likely the camera) also doesn't appreciate backlight AT ALL when it comes to focus. I was shooting a tent wedding lit by paper lanterns over the tables but NONE (nada) over the dance floor. Ofcourse I wanted to use the lamps as a background instead of the DJ table or the outside but WOW I couldn't get the 17-55 to stay focused on the action AT ALL. Compounding the problem was that there was no videographer whose hot light always provides a good contrast target. I imagine that the focus MOTOR quality of the 70-200 vs the 17-55 would be higher but I'm not sure how that would effect AF performance unless the motor in the lens was comunicating more information to the AF system. Another consideration of the 70-200 is that it has a focus limiter which probably give it an advantage in that it is ignoring close distance if set from 3m to infinity. Sometimes I wish the 17-55 had a 2 stage focus limiter Macro -> 5ft 5ft -> infinity.

mercphoto · July 11, 2008

Blurmore wrote:

My only question for the tester is what focal length were the tests performed at.

Agreed. There is no focal-length overlap between the two lenses, so your best results will be testing one lens at 55mm (full zoom) and the other at 70mm (full wide). Think of it this way, if you test one lens at 17mm and the other at 200mm, notice that you are greatly magnifying any area of contrast at 200mm. This magnification will make the edges easier to focus on.

kdog · July 11, 2008

Blurmore wrote:

It would stand to reason that at WA the 17-55 would hunt more as it has more real estate to stimulate the focal points, and it is harder to grab onto one spot and stick with it. [...]

That's a really good point.

I definitely notice it hunting more at the 55 end (like during a first dance) than at the 17 End like shooting a group dancing.

Wait a minute. It hunts more on the narrow end? Isn't that exactly the opposite of your previously stated hypothesis that it would hunt more on the wide end?

Scott_Quier · July 11, 2008

Good point about the focal lengths. I thought about it as I was "engineering" my test and just forgot to include that in my method description.

On the 70-200 I used 70 and 55 on the other and used a bit of carbon-based zoom (I walked) to get the target scene the same size (near enough) in the view finder. Oh, and I used only the center focal point. In this manner, I figured I was doing a decent job of compensating for the lack of over-lap in the focal lengths.

Blurmore · July 11, 2008

Well that is what MY lens does. The WA comment was more directed at the difference between the 2 lenses. I don't use focus tracking with the camera in WA and I do use it at full zoom (like shooting a first dance) because I don't often shoot WA at f2.8 I usually shoot my WA dance stuff at f 4-5.6 to give the most DOF and allow for focus error or not having the camera to eye at all. I shoot first dances at f2.8 to isolate the subjects and you NEED focus tracking at this aperture as inches separate in focus from out. But yeah I contridicted myself.

Nikolai · July 11, 2008

Scott,
thank you very much for a great test!

WA vs tele brings an interesting point. WA lens should bring MORE light overall compared to tele simply because it covers a larger part of a light-emitting area. From this POW, WA lens should focus faster than a tele one.

Yet the results show quite the opposite.

kdog · July 11, 2008

That's not the way I see it. The camera focuses on a contrasting edge. The closer you are to the contrasting area, the faster the focus. So zooming in is better. Also, I don't agree that the amount of light on the focus sensors, or image sensor for that matter, is at all related to focal length. The entire field width is focused over the entire sensors either way, so it's completely dependent on the light in the scene.

Cheers,
-joel

AF: EF vs EF-S?

Comments