There seems to be considerable discussion on the validity of the topic? How about just cuz it's fun? It's certainly a nice break from "which lens should I buy?".
Anyway the solution (I think) to this problem is coming soon in the form of hemispherically shaped sensors. And maybe even flexible matrix sensors sometime off in the future?
Wow! now it all makes sense. I have always wondered why when using manual non-Canon lenses I always expose 2/3 or more under what the light meter tells me so I get a proper exposure.
I think Mr. Frédéric Guichard's statement questioning the benefit of faster lenses is his personal opinion. The data indicate that the problem is with the camera, not the lens.
gdedwards: You can also try this:
With a fully EF-coupled lens (the 50/1.4 or whatever you have that's faster than F/2.0...) and the camera at "M" setting, shoot a solid surface at shutterspeed "x" and maximum aperture (F/1.2? F/1.4?). Preferably with the camera at a tabletop, or from a tripod.
Then release the lens (without moving or rotating the camera body, or accidentally nudging the focusing ring!), and turn it just a few degrees in the mount. Not enough to make it fall off the camera - just enough to make it lose electrical contact. Usually 2mm is enough. Take another shot with the same settings.
Compare brightness of the shots. :-)
With an 85L mounted at the 60D I lose very close to 1 full stop of light (0.9-0.95?). The REALLY stupid thing with Canon cameras is that in stead of using the real ISO to compensate for this, they do it on the raw-file afterwards - with integer math. This has two effects:
1) you lose 50% of the values in the raw file, only 1/2 of the steps are used.
2) you lose noise performance. Canon cameras typically lower their read-noise to signal ratio all the way up to ISO1600... Under-exposure is bad for Canon cameras.
No, the "problem" is not the lens. T-stop is always the same. It's a physical property of the lens.
How much of the transmitted light the sensor can assimilate is another question. And it's not related to the lens' T-stop, it's related to the physical angle (light cone tip angle + principal ray angle) that the exit-pupil and a point on the sensor makes. And this is more directly related to the F-stop than to the T-stop.
You might call the effect a "system" T-stop.
Then there is the mirror box issue on some cams--probably FT but I wonder about some of that issue with MFT. Early FT lenses were advertized as more "telocentric"
with less divergent rays even in wide angles. I think advances in sensors and data processing pre-RAW file write may ameliorate or at least hide the issue to some degree.
cogitech wrote:
Apparently you, like the scientist at DxO, assume that increased light is the only benefit of large aperture lenses. I know this is not the case, therefore I disagree with any conclusions based on that assumption.
Interestingly, the LL implies that dof may not be affected as much as commonly thought, either, so this issue isn't just about light loss.
cogitech wrote:
That is not what is happening. It is the light loss at the sensor that is the issue, and my point is that this is no reason to question the benefits of large aperture lenses or call larger aperture lenses a "gimmick".
Since when did I ever question the benefits of large aperture lenses?
You seems to put a lot of words in my statements that I never wrote.
You need to read my very first post on this subject, I specifically stated that for those who used the AIS 50mm f1.2 (still own a couple of those) on film cameras is a different story.
You see on film cameras it is not as easy to push/boost the iso unless you exposed the whole roll of film with the same number of stops and tell the developer to compensate for it. Or owns the overhead projector like I used to do and selectively over expose the area with hands or cards over the area you do not want to be exposed.
On digital it is either done within the camera without your knowledge as HerbChong had indicated or post process with softwares with just a few simple key strokes. Remember we are talking about 1 stop or less which can easily be done.
Anyway a lot of additional good information were also provided by Bifurcator, theSuede, Micheal Gordon and others.
Definitely is a better topic to discuss than which lens should I buy as Bifurcator had already mentioned?
Fascinating topic! So the state of the current sensor technology is not advanced enough to fully use the wide aperture lenses, and the issue is secretly dealt with by the manufacturers instead of disclosing the facts to the public. I wonder how long Canikon and the rest have been doing this.
So the real issue in my opinion is the lack of ethics from the manufacturers not that the effective Tstop is lower or not. I'm sure as te technology advances it will be dealt with but until when are we going to be treated like idiots, that is the real question.
Did you notice similar behaviour on your A850? I'm not really sure the A900 does it the same way it is described in the article. As a matter of fact the DXO measurments show a loss of 0.4 EV at f/1.4 on the D3X, and this is about the same loss I have noticed with my A900 and a f/1.4 lens, as I usually would always use a + 1/3 to +2/3 EV compensation when shooting wide open or close to. This leads me to believe there is no iso compensation happening.
Perhaps the Minolta engineers who designed the A900 were not as cunning as their Canikon counterparts
theSuede wrote:
gdedwards: You can also try this:
With a fully EF-coupled lens (the 50/1.4 or whatever you have that's faster than F/2.0...) and the camera at "M" setting, shoot a solid surface at shutterspeed "x" and maximum aperture (F/1.2? F/1.4?). Preferably with the camera at a tabletop, or from a tripod.
Then release the lens (without moving or rotating the camera body, or accidentally nudging the focusing ring!), and turn it just a few degrees in the mount. Not enough to make it fall off the camera - just enough to make it lose electrical contact. Usually 2mm is enough. Take another shot with the same settings.
Compare brightness of the shots. :-)
With an 85L mounted at the 60D I lose very close to 1 full stop of light (0.9-0.95?). The REALLY stupid thing with Canon cameras is that in stead of using the real ISO to compensate for this, they do it on the raw-file afterwards - with integer math. This has two effects:
1) you lose 50% of the values in the raw file, only 1/2 of the steps are used.
2) you lose noise performance. Canon cameras typically lower their read-noise to signal ratio all the way up to ISO1600... Under-exposure is bad for Canon cameras.
No, the "problem" is not the lens. T-stop is always the same. It's a physical property of the lens.
How much of the transmitted light the sensor can assimilate is another question. And it's not related to the lens' T-stop, it's related to the physical angle (light cone tip angle + principal ray angle) that the exit-pupil and a point on the sensor makes. And this is more directly related to the F-stop than to the T-stop.
You might call the effect a "system" T-stop....Show more →
cogitech wrote:
Use a camera with a sensor that is better at gathering that extra light? The benefits are obvious.
I know what my choice is.
To use a camera with a sensor that is better at gathering that extra light is of course a sound option, but take care.
Looking at the loss only, one may conclude that the 5D is a better camera for this than the 5DII, but since the 5DII collects more light to start with, the tables are turned again.
At f/1.2, the 5D loses about 0.42 Ev and the 5DII loses about 0.65 Ev, but since the 5DII is about 0.4 Ev more efficient to start with, the 5DII is still the better option if the goal is light gathering alone.
I checked how a Panasonic Summilux-D 25/1.4 on a µ4/3 Panasonic G1 camera worked out with all the electronics engaged and with no communication between the camera and the lens respectively. There was no difference between the resulting images.
Michael Gordon wrote:
Then there is the mirror box issue on some cams--probably FT but I wonder about some of that issue with MFT. (...)
Hope the link works.
Hi Mike,
The link worked. My brain did not.
The Zone10 author Ken Norton claims: On a FourThirds camera, the maximum aperture is around F1.7 depending on the specific camera body it is installed on. (And also mentions APS-C cameras as possible problems.)
Does Ken Norton think of a 50mm lens when saying max aperture is "F1.7" on a 4/3 camera? Surely the rear lens element size differs with the focal length as well as with the lens construction?
When reading the part about the loss of contrast with 4/3 cameras not seen on FF cameras when using super fast lenses I can't help to wonder if Ken Norton is to take seriously at all. In my experience (no scientific measurements though) all the super fast lenses I have tried on FF cameras and 4/3 and µ4/3 cameras has lower contrast wide open compared to stopped down. I don't think there was a real difference between the cameras here.
Unless you are a personal good friend of Ken's I would like to hear your comments on the topic.
You didn't. The DxO head scientist did, in his article (did you read the article?) I never inferred that you said this.
I certainly have not done so. You did a fine job with all your words in your first post.
You need to read my very first post on this subject, I specifically stated that for those who used the AIS 50mm f1.2 (still own a couple of those) on film cameras is a different story.
Your first post is what I take issue with. There are numerous statements in that post which are simply ridiculous. I answered each statement before, so I fail to see the benefit of repeating myself. If you think that large aperture lenses only maintain their benefits on a film camera, you are mistaken, and it is because you still think the only benefit of such lenses is light gathering (which is apparently the same mistake that the DxO scientist is making).
You see on film cameras it is not as easy to push/boost the iso unless you exposed the whole roll of film with the same number of stops and tell the developer to compensate for it. Or owns the overhead projector like I used to do and selectively over expose the area with hands or cards over the area you do not want to be exposed.
And you still think that the only benefit of the larger aperture is light gathering, which makes it much more crucial on film, because it is more difficult to push in post. From this you extrapolate to the conclusion that f1.2 is a gimmick on digital. I disagree.
On digital it is either done within the camera without your knowledge as HerbChong had indicated or post process with softwares with just a few simple key strokes. Remember we are talking about 1 stop or less which can easily be done.
You have effectively explained that the light loss issue on digital with large aperture lenses is irrelevant because it is so easy to correct. I agree 100%, but I encourage you once again to consider your earlier statements.
Perhaps an example will help. The following three photos were taken with three different normal (50mm-ish) prime lenses at their max apertures. Can you see a difference?
The differences are obvious and I am not speaking of anything measurable in a lab. If you find this "invalid" then I really am wasting my time here. If you can admit (even just to yourself) that you do see the difference then I encourage you to return to your first post and reconsider your statements.
If you were to perform the same sort of experiment with the EF 50/1.8, EF 50/1.4 and EF 50/1.2L, the results would be quite similar.
In light of this, I maintain my position that while it is true that the light gathering ability of large aperture lenses is not utilized as well on dSLRs as with film (due to light loss at the sensor), this is not a reason to question the benefits of large aperture lenses on dSLRs (as the DxO scientist did) or to make statements such as you did in your original post. Such statements illustrate only that the viewer is fixated on measurable aspects of lens performance and either cannot (or refuses) to see the other benefits of large aperture lenses, which are just as valid on digital as on film.
cogitech wrote:
Perhaps an example will help. The following three photos were taken with three different normal (50mm-ish) prime lenses at their max apertures. Can you see a difference?
I can't stop laughing after looking at the 3 pictures you posted.
Yes! I do see the difference among the 3 lenses.
The contrast are different, the color are different, the sharpness are different.
And that is no scientific feat knowing very well that different lenses perform differently.
The most interest thing is that the picture taken by the Contax 50mm f1.7 is the best with the best contrast, color and sharpness (of course it is sharper due to the smaller aperture or what we call DOF). And the picture taken by the Minolta MC 58mm f1.2 is the softest around the corners (of course it is softer due to the wider aperture or what we call shallow DOF).
So what did your own test conclude?
An interesting point, though difficult to agree with, from the LL article:
In fact, is not even clear that large aperture lenses will deliver a shallower depth of field as intended. The DxO measurements to date prove that the marginal light rays just don’t hit the sensor. The point regarding depth of field is that these rays are also responsible for a larger blur spot when out of focus. If they are lost, they not only don’t contribute to the light intensity at the sensor, but they also don’t blur the out of focus planes as much as you would expect at wide apertures.
As I mentioned earlier, I am convinced that digital captures have even less DOF than film, as I have observed during several years of simultaneous use of both mediums. I am not convinced by the above claim.
However, it does make sense that by not seeing the marginal rays, digital sensors may actually have less veiling haze, and probably different bokeh characteristics than film, or other sensors with more or less efficient sensels.
Of the 3 images, the one with 58 1.2 is the only one that works for me, because that large aperture is needed to blur the background just enough to make it pleasing to my eye.
IMO, the f/1.2 lens also looks to be the sharpest one in the plane of focus.
Edward:
The A850 compensates with almost exactly +0.3Ev at F/1.4 - same for all lenses I've been able to try. There doesn't seem to be any corner correction going on, it's a uniform brightening of the picture. Same for the D3x (though it has less loss than the A850/900 due to the better sensor topping, and some corner correction mixed in too). It seems the +0.3Ev makes the D3x go almost perfectly linear, but is slightly too little to fully compensate my A850.
Jonas:
Olympus even went so far as to implicitly specify a maximum incident ray angle conforming to using F/2.0 in their recommendations for lenses for their µ4/3 and FT formats. This means that you can use a F/1.4 construction IF you plan an at least 2:1 telecentricity into the lens - it will result in approximately the same angles. This is pretty normal for a modern WA construction. F/1.4 in a WA telecentric lens like a distagon or similar base (telecentric is a misused word here... "lens with large pupil magnification ratio" is better) behaves like a symmetric design F/2.0 of twice the focal length. Telecentric 24F/1.4 > losses like a 48F/2.0
You can also see a marked difference in reality if you compare (say) the AF-S 24F/1.4 to the AF-S 50F/1.4. The loss at the sensor plane is almost twice as high @ F/1.4 with the 50mm lens, even though it has lower actual internal loss (lower optical T-stop).
The LL quoted text in Edward's post two above this is pure %&"%/. (at the lack for better words). It's not like you "lose" everything outside the F/2 angle at the sensor surface, it might be easier to visualize like if you have a "grid" over the microlens surface.
Only the white parts of the grid "measures" incoming light http://ezpicshare.com/images/active-1292421831.gif
-So your F/1.4 "rim" rays give less effect on the image than intended, but it's not like they disappear. The loss starts already at F/2.8, and the ACCUMULATED loss sum at F/1.4 might be 0.5-0.7Ev depending on sensor construction (effective pixel area x microlens angle efficiency). That's not the same as saying that you loose half of the F/1.4 illuminance effect (though quite close). You lose a little bit of the F/2.4 illuminance, a little bit of the F/2.0, a little bit more of the F/1.7 and so on. The SUM of losses is what you see at any given aperture. That sum gets bigger the bigger the aperture angles you use.
edwardkaraa wrote:
An interesting point, though difficult to agree with, from the LL article:
In fact, is not even clear that large aperture lenses will deliver a shallower depth of field as intended. The DxO measurements to date prove that the marginal light rays just don’t hit the sensor. The point regarding depth of field is that these rays are also responsible for a larger blur spot when out of focus. If they are lost, they not only don’t contribute to the light intensity at the sensor, but they also don’t blur the out of focus planes as much as you would expect at wide apertures.
As I mentioned earlier, I am convinced that digital captures have even less DOF than film, as I have observed during several years of simultaneous use of both mediums. I am not convinced by the above claim.
However, it does make sense that by not seeing the marginal rays, digital sensors may actually have less veiling haze, and probably different bokeh characteristics than film, or other sensors with more or less efficient sensels. ...Show more →
I agree, Edward. I saw the vague discussion of this in the LL article and I had the same thoughts. Perhaps at least some sensors offer the advantage of less DOF and simultaneously better-defined focal plane than (some) films. LL seems to suggest the opposite. Or, at least is seems so.
OTOH, some may in fact consider shallower DOF to be a disadvantage.
theSuede wrote:
Edward:
The A850 compensates with almost exactly +0.3Ev at F/1.4 - same for all lenses I've been able to try. There doesn't seem to be any corner correction going on, it's a uniform brightening of the picture. Same for the D3x (though it has less loss than the A850/900 due to the better sensor topping, and some corner correction mixed in too). It seems the +0.3Ev makes the D3x go almost perfectly linear, but is slightly too little to fully compensate my A850.
Thanks!
Well I can live with an ISO increase of 1/3 a stop. Luckily it's only that much.
