EF-S lens outperforming EF L-lens?

[shinobi]

For the heck of it, I tested my EF-S 17-85mm f/4.0-5.6 IS USM kit lens on a 400D with a borrowed EF 24-105mm f/4.0L IS USM lens recently.

I mounted both lenses in turn on my camera, mounted on a tripod, and aimed it at a controlled indoor scene with both lenses set to 50mm focal length and stopped down to f/5.6. Then I metered the scenes, obtaining a shutter speed of 1/8 for the 24-105 and 1/10 for the 17-85.

What the? If anything I would have expected the much larger and more expensive L lens to be more light efficient and meter a faster shutter speed! I don't know how to explain this - sure, the 24-105 is older than my kit lens... but when we compare lenses we usually say the f/2 is brighter or faster than the f/4 etc... That's true in one sense, i.e. the f/2 can theoretically capture 4 times as much light as the f/4, but ONLY if you open it up. Are we missing some quantity associated with efficiency of light transmission through the entire lens?

To me, it seems that for some reason the L lens is less light efficient than the non-L lens. Under low-light of course it would win because it has the extra stop or so to play with if I used it wide open, but what if I was hand-holding, in relatively low-light, and I cared about maximising my DOF and didn't want to use it wide-open? Wouldn't that mean that the non-L lens has an advantage?

Correct me if I'm wrong, but f-number is simply based on the ratio of the physical dimensions of aperture width to focal length, right? Is there another quantity which gives us a measure of the light efficiency of a lens? It seems to me this could be useful as a basis for comparison across lenses.

Any thoughts? I'm not too experienced as a photographer, and this issue is really bugging me, especially since it seems the L lens really ought to have nailed the non-L lens, and not vice versa, in this case.

 

[Fragnatic]

1. the 50mm on both lenses might not have covered the same field of view, thus the metering might be off.
2. there may be slight ambient light changes.
3. when fired with a tripod, did u cover the VF when u shoot? light entering the VF can throw the metering off.
4. may we know what scene u took for the tests? 1/8 and 1/10 is just a slight difference. many factors can affect this.

 

[shinobi]

Hmmm, I know where you're going with this.

On all but your first point, I'm quite sure the experiment was well-controlled. I switched the lenses back and forth several times and had consistent results each time. Ambient lighting was entirely from indoor lights, and within human muscle error I was in the same position each time although the VF was indeed open and some light could have gone in the back. Unfortunately because I deleted the shots and returned the lens, I can't tell you with certainty that the FOV was exactly the same.

Perhaps you can try with a few lenses that overlap zoom range and see if you get the same discrepancy?

Anyway, my main point is that it seems that the f-number isn't the full story. Is there then a way of measuring and benchmarking the light-efficiency of a lens, all other factors being equal?

 

[foxguy]

please.

1/8 and 1/10 is the metering of your camera body. notthing to do with the lens.

 

[foxguy]

Hmmm, I know where you're going with this.

On all but your first point, I'm quite sure the experiment was well-controlled. I switched the lenses back and forth several times and had consistent results each time. Ambient lighting was entirely from indoor lights, and within human muscle error I was in the same position each time although the VF was indeed open and some light could have gone in the back. Unfortunately because I deleted the shots and returned the lens, I can't tell you with certainty that the FOV was exactly the same.

Perhaps you can try with a few lenses that overlap zoom range and see if you get the same discrepancy?

Anyway, my main point is that it seems that the f-number isn't the full story. Is there then a way of measuring and benchmarking the light-efficiency of a lens, all other factors being equal?

what were you aiming at and what is your metering?
use center weight or spot if you have
aim on same point and same position.

 

[foxguy]

one more thing,
If you had the chance to compare, you would have found 24-105 sharper and focusing is faster

 

[Fragnatic]

i believe there might be slight ambient light changes. unless u do it in a studio/test lab with strict control of lighting. what i can say is that 1/10 and 1/8 are really slight differences.

hmmm, i believe foxguy might be right too. the camera's metering might be causing the differences. its not always precise. maybe you should have taken like 10 shots or more (meter the scene, shoot, repeat metering and shooting like 10 times, and finally find the average exposure the metering gives for each lens)

and also, performance of the lens is not only about light efficiency. its about focusing speed, max aperture, image quality, sharpness (esp. in the corners), colours and contrast, flare control, distortion control, chromatic abberations, vignetting, build quality, dust/whether proof, etc etc.. I'm sure most L lenses out-perform non-Ls. (though some non Ls perform equally well.)

for this case, i believe the 24-105L beats 17-85 in most aspects i've mentioned.

 

[shinobi]

Please. I think we're missing the point that I'm trying to focus discussion on, which is light efficiency. I'm certainly not naive enough to suggest that the L lens is getting beat on all fronts by the non-L, or even on any other measure of IQ. All I'm saying is that for some reason (more lens elements?), the L appears to transmit less light. There may therefore be cases where 2 lenses exhibit the exact same IQ characteristics, but one is less light efficient. If so, the more efficient and transparent lens would warrant a price premium right? How do we find objective data to test lenses on this metric?

What I'm saying is that the L lens (or at least that copy) was less light efficient than the non-L. And I've said it before and I will say it again, this admittedly adhoc experiment was repeated several times, i.e. (1) take out 17-85, put in 24-105, set to 50mm/5.6, meter shows 1/8. (2) take out 24-105, put in 17-85, set to 50mm/5.6, meter shows 1/10, (3) repeat from step (1) several times. I consistently got 1/8 and 1/10 every time respectively. Yes, it's a small difference, but certainly not beyond the precision of the camera's metering system to resolve. So the questions in my first para still remain. Appreciate some feedback on those, rather than critique of the "experiment" methodology. If you want to be clinical about it, I suggest you run your own tests. I have a feeling this is something that you'll find in many other pair-wise comparisons of lenses - the question I've repeated is simply: how do you measure this difference?

 

[zerartul]

Mmm light effeciency isnt really a huge performance issue. In this case even if there is, its only 1/3 a stop of difference and can be easily remedied. That said, I believe that its the metering in the camera thats causing the difference. It maybe the way the lens is designed as such the difference. Did you try taking pictures and see if this 1/3 stop actually affects the exposure of the scene? Also lastly, its true that more light due to aperture only happens when you open up, but isnt it better to have the option to do it then no option at all? You will never know when you need that extra stop or effect of blurring. Anyway if you want to have a Ef-S lens nailing an L, try the 17-55 Ef-S F2.8. That lens can best most L Lens of its range.

 

[AReality]

Please. I think we're missing the point that I'm trying to focus discussion on, which is light efficiency. I'm certainly not naive enough to suggest that the L lens is getting beat on all fronts by the non-L, or even on any other measure of IQ. All I'm saying is that for some reason (more lens elements?), the L appears to transmit less light. There may therefore be cases where 2 lenses exhibit the exact same IQ characteristics, but one is less light efficient. If so, the more efficient and transparent lens would warrant a price premium right? How do we find objective data to test lenses on this metric?

What I'm saying is that the L lens (or at least that copy) was less light efficient than the non-L. And I've said it before and I will say it again, this admittedly adhoc experiment was repeated several times, i.e. (1) take out 17-85, put in 24-105, set to 50mm/5.6, meter shows 1/8. (2) take out 24-105, put in 17-85, set to 50mm/5.6, meter shows 1/10, (3) repeat from step (1) several times. I consistently got 1/8 and 1/10 every time respectively. Yes, it's a small difference, but certainly not beyond the precision of the camera's metering system to resolve. So the questions in my first para still remain. Appreciate some feedback on those, rather than critique of the "experiment" methodology. If you want to be clinical about it, I suggest you run your own tests. I have a feeling this is something that you'll find in many other pair-wise comparisons of lenses - the question I've repeated is simply: how do you measure this difference?

U buy a Lamborghini and a SUV, and ask why the Lamborghini so small?

Look at the lenses as a whole, not just how much light gets thru it.

.

 

[AdyH]

Dude y r u testing ur lens in such a manner? Wat is e objective of ur test in e 1st place? To determine which lens have more light passing thru?

I really do not understand y u even need to perform such a test? Cuz at e end of e day, u shoot based on wat u deem e situation calls for. Unless u using aperture priority without wanting to go into manual mode.

 

[Fragnatic]

Please. I think we're missing the point that I'm trying to focus discussion on, which is light efficiency. I'm certainly not naive enough to suggest that the L lens is getting beat on all fronts by the non-L, or even on any other measure of IQ. All I'm saying is that for some reason (more lens elements?), the L appears to transmit less light. There may therefore be cases where 2 lenses exhibit the exact same IQ characteristics, but one is less light efficient. If so, the more efficient and transparent lens would warrant a price premium right? How do we find objective data to test lenses on this metric?

What I'm saying is that the L lens (or at least that copy) was less light efficient than the non-L. And I've said it before and I will say it again, this admittedly adhoc experiment was repeated several times, i.e. (1) take out 17-85, put in 24-105, set to 50mm/5.6, meter shows 1/8. (2) take out 24-105, put in 17-85, set to 50mm/5.6, meter shows 1/10, (3) repeat from step (1) several times. I consistently got 1/8 and 1/10 every time respectively. Yes, it's a small difference, but certainly not beyond the precision of the camera's metering system to resolve. So the questions in my first para still remain. Appreciate some feedback on those, rather than critique of the "experiment" methodology. If you want to be clinical about it, I suggest you run your own tests. I have a feeling this is something that you'll find in many other pair-wise comparisons of lenses - the question I've repeated is simply: how do you measure this difference?

you see, how can we confirm that light transimission thru the lenses are different if the experiment methodology might be causing the fault?

anw, your thread title says "EF-S lens outperforming EF L-lens". which is what you are not refering to.. zzz
more like "light transmission".

how sure are u that the camera's metering is that precise? i do get slightly different exposure values (like diff. in shutter using same aperture) on the same scene when metering it a few times on a single lens sometimes.


hmmm. ok, you might be right that there might be some light loss when light is passing through the lens. but frm your description, i still find that it might be an experimental fault (like not covering the VF when metering, etc..) :dunno:

not to start a war or sth..
maybe someone can try testing out L lenses and EF-S lenses to confirm ur point.

 

[shinobi]

I have to say it's hard to appreciate some of the responses on this thread, especially when they fail to read carefully, miss the point, and then dismiss the issue with a wave of the hand.

But thanks Fragnatic and zerartul. If the final answer is that "it really doesn't matter" then I'd be fine leaving it at that. I still think there ought to be a objective measure for lenses though. Maybe it's overshadowed by other measures of IQ, maybe not... but people rate lenses for build and those things don't affect IQ... so I'm just curious why something like that doesn't seem to be readily documented. Photography is about drawing with light right? Does no one care that the light ain't being fully utilised?

(Anyway, for an example of the importance of light transmission, astronomy buffs can tell you that the quality of the dielectric coatings and prisms/mirrors in their telescopes and binoculars are everything.)

 

[shinobi]

Fragnatic - to your point, I did choose 'outperforming' for a spot of sensationalism. But I think if I had used "more light-efficient than", I may not have gotten as colourful a comparison as lamborghinis and SUVs - really, care to explain this analogy?

But yes, maybe the experiment wasn't as tightly controlled as it could have been. (I really find it hard to explain why I would be standing slightly closer to the tripod every time I had the 24-105 mounted, and farther when the 17-85 was mounted though... *shrug*)

Not having an L lens in my inventory, I wanted to see if all things being equal, the L would have more action stopping potential than my other lenses - it turns out the answer is no (at least for the copies I had at the time). Actually I think there's a simpler answer to this, but I'll need to borrow that same lens again to double-check that hypothesis, so that will have to wait.

 

[Dream Merchant]

Interesting discussion.

Apeture is only a physical force controlled by just that - a physically defined 'hole'.

Light transmission and the efficiency, on the other hand, are probably subject to physics that are beyond my comprehension like light reflectance, refractance indexes of the glass used, internal reflection, abberations and so on. IIRC, even the different wavelengths of light travel at different speeds. I may be mistaken, but deriving effective optical formulars are so complex it almost seems to make brain surgery look as simple as re-wiring a house.

I'm sure ways to measure how effieicntly a lens allows light through exists, but these would be in the order of hundreds of thousands of dollars of lab equipment.

Have a look at the documents chronicling the development of camera/photographic lenses (almost all German designed, mainly Leitz and Zeiss) and optical abberations. Quite interesting reading if you're into that sort of stuff.

CHEERS!

 

[zerartul]

Light lose through the glass in the Lens is actually quite a complicated issue that unless you have avery controlled situation and advance equipement can you then actually take a correct output. That said as I have wrote above, unless that 1/3 Stop produces a significantly different exposure in the picture, its actually very hard to prove one lens has better light transmission rate then the other.

 

[varf]

All I'm saying is that for some reason (...), the L appears to transmit less light.

how did you arrive at this conclusion? the EF-S should transmit less light at 50mm than the 24-105L.

lenses stay open at their widest aperture while your camera performs metering, except when you use DOF preview. unless i miss my guess (since i don't have either lens), the 17-85 should have a smaller max aperture at 50mm than f/4.

 

[tiduck]

shinobi:
What you're asking about is known as the "T" stop (for transmission or transmissivity stop). Unlike the f-stop, which describes the physical size of the lens opening, the t-stop considers the actual light transmission level through the multiple elements of the lens in question. So yes, what you're asking about is a known issue.

Also remember that f-stop numbers are simply labels. Lens elements and designs don't come in "1/3 f-stop blocks" that you throw together, resulting in nice f-stop upper and lower bounds. The lens designers may end up with a 17-85/3.95-5.55 (t-stop), but label it as a 4-5.6. That's fine, as it's just a 2% discrepancy, right? It happens all the time with focal length labelling too.

Test Methodology Issues
Unfortunately, the above point means that a "quite well-controlled" experiment doesn't cut it here. Fragnatic is absolutely right in saying that slight changes in ambient light levels and FOV will be sufficient to throw your results off from 1/8 (0.125s) to 1/10 (0.100s). That gives a midpoint of 0.1125s. Let's say your metering for the 17-85 actually resulted in 0.110s, and the 24-105L returned 0.115s. That's a difference of 1.05%, yet it would have been enough to flip your shutter speed from one value to the other.

I think your question is a valid one. Of course, I may simply be biased, having wondered the exact same thing 4 years ago It's an interesting piece of information that we don't have to consider now with the advent of AE systems, sans external light meter. That said, zerartul, foxguy, Fragnatic et al have it right when they point out all the good stuff the 24-105L has over the 17-85 (which I still love for its versatility).

varf:
That is correct, EOS cameras meter with the lens wide open. However, the poster is referring to a "transmissivity factor". Even with different max apertures on these 2 lenses, given identical aperture diameters and light transmission losses through the various elements, the internal metering calculations should produce the same shutter speed.

 

[foxguy]

u say the light gathering power of the 24-105 is poorer compare to 17-85.

bro, bring your 17-85, and borrow a 24-105 and go to a evening open concert. first use your 17-85. shoot till you find your lens starts to hunt a lot, then change to 24-105 and try. Others i hope understands what i meant.
i have both 24-105 and 24-70. i tried many lens too. i find that only prime lens would be better. 50 F1.4

 

[foxguy]

for people who are interested in such a light test.
you would need to have a laser, a resever and a stepper power system, fire the laser thru the glass and measure the light loss in terms of dB. i am sure most L are good due to the "Flo****"(dun remember the spelling) element in L lens.
any way, putting that aside, i find lens that can perform in the real to be "good", not lab test reslts or hypes generated by "information".

LENS ARE JUST A TOOL, ENJOY PHOTOGRAPHY.