Is f/1.4 at ISO 200 really f/1.4 at ISO 200?

Luminous-landscape has a very interesting article on how camera manufacturers are silently increasing the ISO in order to compensate for light loss from the lens. The whole analysis is based on data from DxOMark:

click on graph for larger view

The above graph show the light loss at the sensor level for different Canon and Nikon cameras when using a f/1.4 lens wide open. For example, the D3s has -0.30 EV light loss, the D300s has -0.52, the D40 has -0.80, etc. In order to compensate for this light loss and produce correctly exposed images, camera manufacturers are compensating by "secretly" increasing the ISO. The author Mark Dubovoy goes even further by saying that you may get better results with smaller aperture lenses, since they have less light loss and that this phenomenon is also destroying the bokeh:

"One might be better off purchasing smaller aperture lenses and increasing the ISO.  Since these lenses have much less light loss at the sensor, one may well end up with virtually indistinguishable results. 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."

Update - related links from the Web:

This entry was posted in Other Nikon stuff. Bookmark the permalink. Trackbacks are closed, but you can post a comment.
  • santela


    • GlobalGuy

      Hey Admin — do you have backup information on this? I think we have ALL seen direct comparisons of 1.4 vs. 1.8 vs. 2.8 and the 1.4 wins almost every time (in terms of bokeh). LL is a respected site, but they didn’t post any examples, and a chart does not count. Is there study correct?

      The idea that it doesn’t affect DOF is almost absurd considering countless online photos showing otherwise. While noone has isolated this particular issue to test and compare in a VISUAL way, we have all deeply reviewed lenses and compared to their slower counterparts under fixed conditions. This issue MUST receive a follow-up.

      We want to see the real-world phenomenon in examples. And ISO vs. Light Loss needs to be un-confounded, since, currently they are suggesting automatic adaptation. They need to isolate the issues very clearly with visual evidence (this is photography afterall).

      • GlobalGuy

        By the way, its not that I don’t believe them — its that I want to know where does this start and stop (end and begin). That is, is the same true of f/1.4 vs. f/1.8? Or of f/1.4 vs. f/2.0? What of f/1.8 vs. f/2.0 vs. f/2.8?

        Something is MISSING in this report — and that’s the narrow focus on 1.4. The study needs to suggest WHERE IS the ideal point for each camera — not simply knock 1.4. Is the idea point 1.8? Or 2.0? Or 2.8?

        And is the camera increasing EXPOSURE or is it increasing ISO?

        Since both readings stay fixed in the camera data, how can we be sure which was modified?

        • Light loss at the sensor is nothing people should be surprised of. There is no digital imaging sensor design available that manages to capture all the light. Honestly, I think a loss of only -0.3EV as on the D3s’ sensor is quite an achievement. It’s also worth pointing out that some of the initial information (light hitting the sensor’s surface) is also lost due to subsampling (Nyquist’s theorem) since it is impossible to “look at” the image before the A/D conversion.

          A t-stop loss does not mean it affects the DoF. Due to the loss of one third a stop at the sensor a t/1.4 lens will produce about the same transmission as a t/1.6 lens, yet still have the DoF characteristics of t/1.4. More technically speaking — the information lost (blurred) in the optical system won’t be regained due to transmission lost at the imaging sensor. What affects the DoF way more than that is the actual pixel pitch (circle of confusion).

          What does it mean for the end user?
          Since the manufacturers are aware of the transmission loss and have fine-tuned their cameras there is nothing to worry about. If you take a picture at ISO 200 your camera will deliver a result as close to ISO 200 as possible. Better worry about getting the picture right rather than what sensitivity the camera actually uses.

          So film was perfect?
          Yeah, right. NOT

          • Banned

            A very interesting point is, could we actually attribute the D3s better low-light performance compared to D3 solely on the fact that its sensor captures more of the marginal light? In this case, the main difference with a D3 would the design of the micro-lenses since they are responsible for directing the light onto the sensor.

          • Not Banned

            I’m not sure I follow the context in which you bring up Nyquest.

            Nyquest tells us that everything which makes it past the AA filter is reconstructable. Would you mind clarifying what you were trying to say?

      • No, I do not have any backup info on that. I tried to test this by disconnecting my lens as someone suggested, but it did not work for me. DxOMark is a big/legit business and it seems that they are behind this whole thing, LL is just spreading the word. I am curious how this how thing will develop in the future, since DxOMark promises to do more research in this area.

        • R!

          Yes It’s known since day one,for example :don’t enter the lens information in the camera settings you get actually better bokeh and cyan cast for skys deep as film does…
          Figure out the rest by yourself, I’m no teller;lol…

        • The invisible man

          LOL !!
          I just received an email from B&H, the 24mm f/1.4 is now in stock !

          They probably heard about this post and now try to get rid of the f/1.4 lenses !!
          Admin, could you please make a post saying that the D700 is over exposing with all VR lenses ?

        • iamlucky13

          I’m curious why this is coming out now.

          The post is signed “Mark Dubovoy – October, 2010.”

          However, I absolutely swear I read this same essay about two years ago, and I swear it was from Luminous Landscape. I can’t prove this, but does anyone else remember this?

          That’s necessarily an accusation, but if this is just an update of an old plea, Luminous Landscape really should clarify that.

      • Manuel

        “… destroying the bokeh.” This statement is simply ridiculous and it doesn’t make a good impression regarding the seriousness of the rest.

        • Yeah, this whole post is dumb. Testing for yourself proves it. People in the know, who are aware of still cam lenses difference between T and F stops don’t care. It’s irrelevant info in everyday use.

          And why is anybody paying attention to DXO anymore? Their software is amateur automatic crap. And their test results are bad. They intentionally point out wonky results like this to create controversy and grab attention for their real product, which is a crappy piece of software designed to install itself like a virus. Just google it:

          It’s a dishonest company with dishonest software. Really–AVOID. And ignore their testing. Maybe if we just ignore them they’ll go away. The only thing they accomplished is getting Adobe to step up their game. Now that Adobe has, DXO is obsolete and they’re grasping at straws to keep in the game.

          • Trend

            Yep, dishonest company with dishonest software that uses PACE, just like Adobe. So we should stop using LR and PS too right?
            Oh wait.

            • Since when does PS use Pace?!

    • PJS

      Much ado about nothing, imho.

  • i wonder how this would affect my d3000 when using old 1970s nikkors, since the camera has no idea what lens is on the body… lolololol?!?

    • Hmm, yea, good question. Maybe thats how they started noticing the problem.
      Or…THEY just know it!
      They know everything! 😀

    • Worminator

      I’ve noticed that 35mm F2 lenses are considerably darker at F2 than at F2.8 on my D40, where there is no correction as the camera is not aware of the lens aperture. I’ve also noticed the same thing on the D200, where the lens aperture for the Ai lens is set correctly. So while it does seem that light falloff at large apertures is a real, I can’t confirm that the D200 corrects for it by boosting the ISO.

      Also the LL letter makes out like CMOS has light falloff but CCD does not, but several of the cameras listed as having large light falloff are in fact CCD based. (D30X, D60, D200 etc.)

  • I have no idea why the alignment in this post is messed up.

    • Nikkorz

      Looks fine to me.

  • Interesting how the D80 is much worse than the D200 when I thought they had the same sensor.

    • And in real world tests (I used the D80 and D200 together for a while for weddings and location shooting), the D80 actually produced less noisy pictures at larger apertures..

    • TaoTeJared

      They have different sensors. Same MP but are different.

      • WoutK89

        What about same silicon but different electronics? That explains better I would say. Similar for D300-D90-D5000-D300s.

  • Neosxm

    Those are very serious acusations! I need more data and explanation to actually believe it. 🙁

  • John


    • Not really. When camera tweak ISO you loose, in example, dynamic range.

      • GlobalGuy

        I think he means its irrelevant if it cannot be controlled. If it cant be controlled, then the user must rely simply on aesthetics. The artist can choose. And the information is irrelevant compared to experience.

      • John

        Ok guys, so, today, I was shooting an hdr of my dog, and like, the camera tweaked my ISO and I lost like a zillion of dynamic range.

        “This measurbation done by a group of old gear heads has no bearing or impact in reality on anything relevant AT ALL.
        YAAAAAAWN totally and absolutely irrelevant for people who shoot real photos, not photos of walls and resolution charts.”

        • You mean auto-ISO? That’s not what this post is about. It’s about tweaking senor gain without reporting it in EXIF. You’re talking about user error. Please try to be smarter than your gear…

  • I made the test 1 week ago.

    Using 50mm AF-D and the results was that luminous-landscape was right, my D90 underexposed at 1.4 when didn’t know what was the lens attached (just turn the lens until the camera doesn’t recognize it) about 1 stop. It’s easy to test.

    • Edgar

      If you turn the lens, then the camera cannot properly operate the lever that controls the aperture on the lens.

      • Does not matter.

        You must fix aperture at 1.4 in aperture ring.

        • Edgar

          Ah, if you are using the aperture ring then it’s OK. But beware of the metering issue. The camera is metering through the focusing screen, and modern focusing screen do not see past 2.8 or so. Thus it is normal for the camera to compensate the metering.

          • Not Banned

            No. No. No.
            The camera is metering through the meter.
            The transmission loss of the focus screen is taken into account.

            You are confusing the fact modern AF focus screens are ground (etched) for slow AF zoom lenses are are little help in manually focusing bright lenses with how the meter works.

            • Edgar

              The meter sees the light through the focusing screen. The transmission loss depends on the aperture of the lens. It is taken into account if the camera knows the max aperture of the lens.

  • lolcatmaster FTW

    Who cares? it is totally and absolutely irrelevant if they do it, mostly because the point of using fast lenses is A) depth of field B) poor ambient light conditions.

    This measurbation done by a group of old gear heads has no bearing or impact in reality on anything relevant AT ALL.

    Some of the stuff these geezers aren´t taking into account are: variation on the real value of ISO sensitivity (for years camera makers have told us that the value is a ball park not something etched in stone) and the sensor to sensor variation.

    YAAAAAAWN totally and absolutely irrelevant for people who shoot real photos, not photos of walls and resolution charts.

    • Kendall Curtis

      I’m sorry but your objection is ridiculous and no very well thought out at all. Just because most things of this nature seem similar to pixel peeping and therefore deserve the “real photographers don’t care about this stuff” response, does not mean that any piece of information involving something scientific is inherently irrelevant to “real” photographers.

      You claim that no one would care, but if these claims are legitimate then the cost of lenses of much larger apertures is a complete waste of money. Money is something even those who shoot “real” photos care about. The second relevant claim is that the falloff actually essentially erases light data that creates blur in shallow depth of field photographs which is valuable in the overall impact of a photo which is why “real” photographers pay extra cash for these lenses.

      You even mention that one of the reasons fast lenses are bought is for poor lighting conditions. What this article claims is that it is essentially not delivering any increased performance for this purpose beyond increasing the ISO which you can do with any lens making it, again, irrelevant to spend the extra money.

      I would have to think that you simply do not understand what this article is claiming to rebut in this way.

      • lolcatmaster FTW

        No you numbskull because you can´t replicate the DOF of an 1.2 lens with a an 1.8 lens!!!!! DUH that´s why you pay 3,000 frigging dollars for a frigging f/1.2 lens!

        If the ISO is increased WHO THE FRIGGING EFF CARES! only dumb people who shoot walls!!!! there have been THOUSANDS of THOUSANDS of shooters perfectly happy working and delivering files to clients who are nitpicky beyond what you could imagine and they DON´T CARE BECAUSE IT ISN´T SOMETHING THAT IS NOTICEABLE AT ALL!

        I understand the article but as one of many who has been earning their money in the field for more than 20 years and almost half of that with professional digital photography I can tell you: I DON´T CARE AT ALL because it doesn´t affects image quality in ANY SIGNIFICATIVE WAY.

        This is only interesting to cat shooters.

        • Kendall Curtis

          Wow. You not only misunderstood the fundamentals discussed in the article but you don’t even remotely understand what I wrote in response to your initial rant. I would try to clarify, but obviously even the most simplistic of explanations would likely be completely lost on you. I could say “the sky is blue” and you would respond angrily “bananas do not taste like raisins”. I find it hard to believe you have more than 20 years being anything professional other than a professional moron. Good luck doing whatever it is you do in constant attempt to hide your utter ignorance from everyone around you.

          • PHB

            I have a degree in nuclear physics from Oxford University and I don’t understand the point the piece is attempting to make.

            I think that the authors are simply confused and wrong.

            T-stops are a property of a lens and not the sensor. Cinematography uses t-stops because they care much more about the actual amount of light entering the camera than depth of field. Still photography uses f/stops because they correspond to depth of field.

            Everyone who was paying attention knows that different lenses have different transmission characteristics. Some light is absorbed in every lens. Some are better than others. But none of it has any impact on depth of field which is determined by the focal length and aperture alone.

            • PHB

              OK, I have worked out what the twit is waffling on about.

              The transmission properties of lenses actually change with aperture. That is the reason that the cine world uses t/stops. If you are shooting $10,000 worth of film a day plus the cost of actors and crew, you don’t want to find that it is under exposed.

              Modern lenses have chips that contain profiles of the characteristics of the lens that are used to compensate. That will either mean increasing the amplifier gain or changing the exposure time.

              ‘Increasing ISO’ is meaningless on a digital camera. ISO is a calculated quantity, there is no reference point. The cine folk compensate for lens characteristics by using compensated f/stop numbers (t/stops). It makes just as much sense to use compensated ISO.

              Using compensated ISO numbers is not going to affect noise in the picture. The only thing that will affect noise is changing the gain setting on the camera. And that should not happen unless you have the camera set to use the ISO setting for exposure.

              We have to be very careful here in attributing this to DxO. This is the result of someone getting confused interpreting the DxO data. I did not see this interpretation endorsed by DxO on their site.

            • Not Banned

              Degree in nuclear physics or not, you appear to be missing the point that amplification IS being changed to compensate for the non-linear relationship between f-stop and t-stop.

              This is the least obstructive way to solve the age-old problem. If not the camera would need to lie to manual shooters about the shutter speed or the used f-stop – both of which have much more serious consequences regarding the collected image than a 30% lie in ISO.

            • PHB

              I was posting in response to the people calling people idiots for not accepting the premise of the paper. In my view the problem is with the lack of clarity in the paper.

              I am not sure that the camera is changing the gain. They may just be adjusting the ISO figure without changing the gain.

              It is not clear what the figures mean and certainly not clear how they are being measured.

              I have no idea how DxO purport to be measuring the light reaching the sensor in different cameras. It should be pretty obvious that different cameras have different sensitivities to light. That is why ISO performance is different from camera to camera in the first place. But the variation is way too small for that to be the issue.

            • Not Banned

              The variation in sensitivity being discussed is primarily variation in sensitivity to off-angle light. This is a function of the filters and microlenses in front of the lightwells.

              And, yes, gain is being tweaked. The camera is reporting ISO X when actually it is using ISO 1.3X.

          • Discontinued

            “And, yes, gain is being tweaked. The camera is reporting ISO X when actually it is using ISO 1.3X.”

            And this is an ISSUE to me with consequences, because I still do some light reading with handheld meters (of studio flashes for instance) and set the time of exposure plus aperture manually.

            It already happened to me that I got some overexposures, when using Ais glass and didn’t tell the camera which one I used (Exif 50mm f1.4, 1/250 when I shot 180mm f2.8 for real). SECRETLY raised ISO for f1.4 could be the perfect explanation for the overexposure I got.

            I really hate it if a tool takes control

            • Not Banned

              This FIXES the issue those of us who use an external meter would otherwise have.

              This undocumented change in gain is to correct for the off-angle light rejection issues digital sensors have when compared to film.

            • Discontinued

              @Not Banned,

              I realize that. I am not talking about an issue I would have had, actually shooting at 1.4 (which is compensated via higher ISO, as we learn from the post).

              I am talking about an issue I have had indeed, when the camera thought I was shooting at 1.4 but I was not (due to not changing lens data and my unawareness of this issue).

              I am talking about an issue that stems from unwanted and unnecessary compensation that probably secretly happened while I was using 2.8 glass. IMO that sucks.

            • PHB

              Not Banned is right. If they are making this compensation, it is for good reason. This is not some conspiracy going on.

              It is not at all clear whether anything is happening though.

            • Discontinued


              “conspiracy” ? ? ? What have you been smoking?

              I was just pointing towards an awareness issue. And I am glad I know now (thanx admin for posting it).

              And I’ve tested it. Yes, if you use wrong lens data settings
              (1.4 for slower glas and round the other way) you do get wrong exposures indeed, due to raised ISO (or round the other way, due to uncompensated light loss).

              “If they are making this compensation, it is for good reason.”

              Yes, of course! And if I want to know about it, it is for a very good reason too.

              Until yesterday I thought setting ISO, aperture and shutter speed manually would mean controlling it all. Now I know that I have to give the camera the right lens data too, in order to avoid unwanted ISO compensation and in order to get it when truly needed.

  • It sounds as what they’re saying is that with some sensors, you’ll get little noticeable difference when moving from (say) f/2.0 to f/1.4. This should be easy to test…. us all manual settings and take a properly exposed image at f/2.0 of a scene that has a lot of detail across the depth of field, then without changing the ISO or shutter speed, take an image at f/1.4. The image will be overexposed by a stop either because that much extra light was let in, or because the camera secretly boosted the ISO as the article claims, but look to the depth of field: you’ll either notice a difference or you won’t. If not, the claim is correct. If you notice a pronounced change in the depth of field, the claim is wrong.

    Best to test with cameras listed in the lower part of the graph, as they’re the ones with the most claimed compensation.

    • Eric Duminil


      Their point:
      A real-life f/1.4 is not as good as a perfect f/1.4, and a real-life f/1.4 is maybe as good as a perfect f/1.8
      So why not get the perfect f/1.8?

      • I’ve been saying this for years, ever since learning that Canon is selling a fallacy by claiming that you can get perfectly sharp images at F/1.2 (at least as well as with 1.4 or 1.8 lenses). It just isn’t true. Of course it should be remembered that lenses vary by copy, and this will effect sharpness even in the same model. However, it’s irresponsible for companies like Canon to sell these 1.2 lenses as the pinnacle of perfection. The reality is that shooting wide open with the 1.2 lens only produces acceptable sharpness at certain focus distances, while making it impossible to get sharp shots at others. In order to get perfectly sharp images at all distances, one needs to stop down. This in effect removes the value of the lens. In fact, it may even mean it’s a lower value lens, since the bokeh shape at wide open is more appealing due to the perfectly circular pattern, rather than the polygon effect of the aperture blades.

        Too many photographers I see carrying around this trophy glass say the same garbage – namely that it’s somehow their fault that they didn’t nail the focus on a host of shots simply because the depth of field for this glass is so razor thin that the subject moved in the time it took their finger to travel 1/4mm distance. Rubbish.

        Even Nikon’s 50mm 1.4d lens is pushing the envelope. People who pooh-pooh the 1.8 haven’t shot with it, as they obviously don’t realize it produces some incredible results. Personally, I shoot with the 1.4, and I often wonder if it’s really worth the several hundred dollar increase in price over the 1.8.

        Now feel free to flame me.

        • taotejared

          Love it!

          • Jose


  • lolcatmaster FTW

    And the bokeh claims are idiotic at best I have prints done with either film (astia, neopan, t-max, Kodakchrome, Ilford Delta or HP5) and digital too (1Ds MKII and D3x + D700 done with both my EF 50mm f/1.0 L, EF 85mm f/1.2 L, EF 135mm f/2.0 L, Noct Nikkor and 135 and 105mm f/2.0 DC.)

    There´s NO difference in the bokeh at all… what a load of BS from measurbators and cat shooters!!!!

  • Huggs

    Is this why they put an exposure compensation button on my cam? 😀 I kinda get where they’re coming from, though. $1700 for a lens, then another grande or more for a body? I better not have to compensate for anything. mofos.

    • ZoetMB

      >>>I better not have to compensate for anything. mofos.

      Why not? Film cameras had exposure compensation as well. Why? Because the metering wasn’t always exact (lightly lit model wearing black on a black background will over-expose, even with modern sophisticated metering systems) and because the ASA/ISO of film was never exact either.

      I never got exactly the same exposures with Tri-X 400 and let’s say Kodacolor 400 or whatever 400 ASA color film was around at the time.

      I wonder how many people who are screaming about this leave their cameras on auto-ISO anyways?

      There’s only one lens that doesn’t have light loss through the lens: it’s a pin-hole lens (and yes, I know, it’s the equivalent of f64 or whatever.) And whether it’s a sensor or film, light sensitivity is/was never linear.

      Personally, I think this is much ado about nothing. Something is a quarter stop off? ISO400 becomes ISO500 when shooting wide open in certain conditions? So what?

  • They are increasing the amplifier gain to get proper exposure at the given settings, which will give an increase in noise. I don’t see a problem with it. It means that when I expose for f/1.4 1/500th at ISO 400 I get the expected exposure and it is not under exposed by the amount indicated on the chart. Changing to f/2.8 1/250th at ISO 400 will also give proper exposure. I don’t want to have to think about the complicated compensations I have to make for the sensitivity lost at the sensor at wide aperture when adjusting my settings. When I change my aperture, I what a 1 stop change to be equal to 1 stop change in exposure.

    • Kuv

      actually, it won’t. you lost 2 stops at the aperture and gained only one on ss. just sayin’.

      • WoutK89

        He explained he got the same result shooting f/1.4 and f/2.8 but with 1 stop decrease of shutter speed. At least, that is how I read it.

  • Spotpuff

    I think the technical details went over a lot of people’s heads so they’re attacking the authors.

    If bokeh requires out of focus rays to actually be detected by the sensor and show up and they aren’t reaching the sensor then you aren’t getting the bokeh you should.

    Losing a stop of light is a big deal; the “point” of wide aperture can be manifold and while DOF is one of them, using available light is also one of them. Losing almost a stop at the sensor completely negates the point of using a wide lens. If you paid an extra $500 for f/1.4 over f/2 to get DOF, you might be happy, but if you did it to get low-ISO available light photos you are being cheated.

    If ISO didn’t matter people wouldn’t use low ISO to minimize noise. Bumping ISO a stop to compensate for lens or sensor inefficiency is dishonest.

    I don’t understand why camera makers are getting a free pass here and DXO mark is getting attacked. They’re simply pointing out the dishonesty of camera makers.

    • GlobalGuy

      “If bokeh requires out of focus rays to actually be detected by the sensor and show up and they aren’t reaching the sensor then you aren’t getting the bokeh you should.”

      I think the reason they are resisting is because of this bokeh point — when you change from 1.4 to 1.8 people EASILY can tell that the bokeh has changed. Every single time. So let’s not focus on this point.

      As you mentioned, the loss in light is the more important point with respect to exposure and ISO issues. And so for low-light shooters this is a VERY serious study. For bokeh enthusiests, it is a non-issue.

      So let’s separate the two issues quite clearly.

      By the way, daylight shooters might even be pleased their 1.4 setting isn’t letting in so much light. We’ve all taken the “bright white” shots. And they don’t need to be any brighter. 😉 But yeah, low-light shooters should be crying and hollering right now.

      • svane

        Well, actually the article DOESN’T state that you DON’T get better DOF with a 1.4 than with a 1.8. I think what they are trying to say is that since you loose some out of focus rays you don’t get as good a DOF on the 1.4 as you should expect. It’s still better than 1.8 but not as good as it should have been if you were not loosing any light.

        If you enterpret the article like that is really does suggest that all of us that buy expensive large apperture lenses actually do get, at least partially, screwed by the manufacturers for both low light capabilities as well as DOF.

        Again, the article never states that 1.4 doesn’t have better DOF than 1.8, just saying it should have been even better.

        • Sloaah

          The article has nothing to do with DoF… it’s talking about light transmission – how much light that gets through the lens gets to the sensor.

          I’m a little doubtful though – in some cases the difference in light transmission is almost 1 stop, and that would be clearly visible in terms of noise and dynamic range, but I can’t see it.

    • lolcatmaster FTW

      It is easy grab some film: Astia, Ilford, T-max, etc. and Rent a couple of full frame bodies (preferably full frame to keep things equal) like a D3X and an F6, prepare a still life and lit it with continuous lights and come back with the results.

      You will feel pretty dumb for wasting your time after you develop and print the photos from film and print the photos of the digital files.

    • > I think the technical details went over a lot of people’s heads so they’re attacking the authors.
      I think some of the technical details went over the author’s heads, too ;~)

      About eight years ago I pointed out to everyone that what happens “at the sensor” amounts to an optical system. The AA filter, the microlenses, the light sensing area, the Bayer filtration all conspire against the notion that the “film plane” is a perfect plane and that all light reaching that plane is treated equally. It isn’t. Never has been, even with film (multiple layers).

      Personally, I’m impressed that some camera companies have engineered things down to about a third of a stop practical difference across f/1.4 to f/2.8. The problem used to be far more pronounced.

      However, many of conclusions people are reaching from the undocumented DxO charts are, I believe, very incorrect. Moreover, what LL and DxO have failed to achieve is to support any assertion of practical implications.

      As I noted in my 2011 predictions, the coming year will probably see more of this nonsense, as people try to pick apart the low level engineering that’s been going on in our cameras for some time. For example, we have another DxO controversy brewing with the assignment of the Pentax K-5 “sensor” results. Note the change in the data curve that occurs about at ISO 1600 in DxO’s data. Unexplained, but clearly an indication of a change in strategy in the electronics of the camera. So people will argue about that without understanding what the change is or what the practical impacts are. Given what I know about low-level sensor design–remember, I designed more than one digital camera in the 90’s and have tried to keep current with the scholarly literature–we’re going to see a series of these “discoveries” about what our sensors are doing or not doing. To me, the amusing thing is that the camera industry is far ahead of the measurebators here: they’ve been seriously trying to engineer out real issues in the optical aspects of the light capture for 30+ years now.

      • YEAH! What he said!

      • PHB

        I think that what they are noting here is nothing more than a difference in calibration between cameras.

        Changes in the lens transmission function have the same effect as adding a neutral density filter. Anyone think that is going to affect bokeh?

        Only some of the light that hits the sensor ends up as signal. Doesn’t matter what brand of camera you have, they all obey the same laws of physics. If CCD chips were magically better in some way then Nikon, Canon and the rest would not have switched.

        Guess what, the same issues affect a f/1.8, f/2 or f/2.8 lens!

        So even though a f/1.4 lens is only equivalent to t/1.6, an f/1.8 lens is only t/2 and so on.

        No digital camera has perfectly flat response curves. If you want to get completely accurate results you have to be prepared to use electronic compensation.

        The real value of 3D cameras may well turn out to be the ability to perform more interesting compensation.

      • Jose

        Maybe K-5 does not have the note posted as previous Pentax cameras, but if you notice, the chart clearly shows where the “change” happens that the RAW data is smoothed. DxO has documented this behavior since the days of K10D.

  • mshi

    It’s a known issue within the community, and that’s why most pros use 2.8 zooms for both convenience and predictability.

  • Bob

    Panic!!!!! I’m selling all my Nikons and getting a Hello Kitty Point-and-Shoot. Hello Kitty never lies…..

    • Kenny Son


    • LOL; Are they available or are they back ordered?

  • Sean

    I want to know if DXO has properly tested this.

    That is, did they check – for example – the Canon 50/1.2, 50/1.4, and 50/1.8 to see if they all fall along that same plot (as we should expect if they’re right), or is this actually just compensation for design specs? [that is, if you read the patents, you’ll see that a 2.8 lens might really be a 3.1 lens, etc.]

    In all cases, the ‘wayward rays’ argument doesn’t sit right with me. We have microlenses for a reason, and as I read the Canon chart, the plot doesn’t follow a ‘propensity to bloom’ line.

    They make a great argument for using T-values just the same, though, and I do wish we could go to it in the photo community.

    • GlobalGuy

      I agree that something is missing. Bokeh very clearly varies at f settings and improves a lot at 1.4 vs. 2.0 for example. Whether or not low-light is perfect is questionable. If there is variable ISO in the equation — as much as HALF A STOP thats a very serious thing.

      I think if this is true, then Canon and Nikon need to think about providing an option for turning that kind of auto-hidden ISO off. You should have a right to have a longer exposure instead of an increase in ISO. For STILL photography on tripod, we’d all prefer a longer exposure. ISO reserved for handheld and moving subjects.

      The data provided in insufficient to be helpful to the artist.

      I assume that its actually NOT better to move to a higher f-stop if you are going for shallow DOF. But this could empower some people to feel more comfortable at say 1.8 when they need MORE dof and previously were too afraid of losing light. If the auto-ISO is going to make the APPEARANCE the same, then we at least need to be able to wisely choose our DOF if thats all that will be different within the “bright” range of f stops.

      • Sloaah

        What we’re missing is proof that this light loss is not a constant, but a variable that depends on the lens used. Some people seem to miss that bokeh has nothing to do with this subject…

  • Ronan


    This instantly void all Nikon and Canon camera’s!

    QUICK all switch to Pentax!!!!!!!

    The world is now upside down!!!

    Zero has been divided!!!

    • PHB

      Dividing zero is easy 0/2 = 0. 0/3 = 3

      Want more?

      Dividing BY zero, different matter.

      • ZoetMB

        0/3 is not 3, it’s 0.

        • I think it was a typo. I’m pretty sure he knew it was 0 since he got the first one right. lol.

  • GlobalGuy

    I just realized whats missing — the STANDARD.

    Is this chart compared to: What? 1.8? 2.0? 2.8? What are they comparing as a “standard candle”?

    Furthermore, it seems clear that for the higher-end cameras (full-frame) this is a 1/3rd stop issue. IF THE CANDLE STANDARD is f/1.8 then that might be relevant. But if it is 2.0 or 2.8 or some other measure, we really need to consider closely to determine low-light DOF sweet spots.

    This chart says NOTHING about bokeh, by the way — its just their theory, but not illustrated in actual photographic case studies.

  • texasjoe

    This is a bold statement. If this is true then Nikon and Canon are going to be having a lot of lawsuits. And who would be next? Sony? Kodak? Leica? Is this something that can be fixed with a firmware update?

    • Just A Thought

      Leica uses CCD based sensor to not an issue. Kodak sensors are usually CCD so you can scratch them off your list.

      • yes, Leica M9 and medium format cameras are not impacted – this is related only to CMOS sensors

        • No, you have to parse their stuff correctly. What is going on is CMOS circuitry has a much smaller light sensitive area than CCD, these problems exist in CCD sensors also, but the gap is theoretically not as pronounced because there is less circuitry shadowing.

          Having said that, you can see from the graph that the CCDs have the highest T-stop loss in the graph. That is mainly because the CCDs tested were pre certain advances in DSP processing, gapless microlensing, shift-microlensing (which reduces vignetting), etc. CCDs still have a theoretically better performance, but in practice do not (as anyone who owns a Leica M8 or M9 and compares it to a Canon 1D or Nikon D3s can attest to). This is even with the ISO compensation going on.

          A lot of reviewers don’t understand this. They compare two lenses, and wonder why the image is dimmer is one at the same f-stop than another. The reason has to do with these off angle rays.

          BTW, there is more than ISO trickery going on. There is also color halos that are probably being compensated for in the software once the lens ID and f-stop are known.

          As lenses get modernized the manufacturers know these problems go away. It’s a function of the lens more than of the sensor here. That’s a fact.

          • Not Banned

            My God, FINALLY some brains on here!

            • Another reason the Leica is worse is the low-pass filter (anti-aliasing) is very thin on Leica because of the target market. A consequence of using a thin filter is the light gathering capabilities (due to mixing) are greatly diminished on-chip.

              The ISO performance thus is greatly diminished. It’s like you rewound the clock on sensor design (other than offset microlensing) five years. It’s terrible!

              I know this and I own a Leica M8 in addition to my Nikons. 😀

          • Very correct. One critical aspect is “fill factor,” or how much of the surface of a sensor is actually dedicated to light detection (other parts may be devoted to power or data lines or on-sensor electronics or cell isolation). Microlenses evolved partially to try to compensate for fill factor. On the Leica, these microlenses are offset to try to compensate for the angle at which light may be hitting them, but I do not believe that the Leica has continuous microlenses, so it would still be slightly vulnerable to the same issue that DxO is trying to report.

            And, yes, it is correct that is more going on than just a change in gain (which again, is not necessarily ISO change; moreover, the ISO standards actually have a +/- margin in them, and I don’t believe that any of the maker’s are violating them; if anything, they’re exploiting them).

            • Thanks for the support, Thom. Much appreciated coming from you.

              I just have a couple things to add.

              1) While offset microlensing (in the Leica) is a Kodak patent that is not necessarily available to Nikon, there is some secret sauce going on in the microlens for off-angle light. My thinking is while “offsetting” may not be allowed, the lenses are manufactured so that they focus differently away from center. That’s the implication in the article above.

              2) In the sample in the article the f/1.2 Canon lens used gives up over almost ISO, so the gain applied is not within the labeling tolerances of the ISO (however, the effective ISO reported is correct). If that sample is the same focal length as the f/1.4 chart (gives around a half stop electronic gain), then one claim in the article is correct: There may be cases where the lower aperture lens actually outperforms the higher one and that difference is “masked” by the internal amplification.

              Another thing about bokeh. There are two separate things here. One is the the technical meaning of “out-of-focus” behavior and the other is the aesthetic quality of such. Bokeh refers to the latter, not the former. Their discussion of bokeh is totally incorrect in this regard. What they should be claiming is that due to the necessity of some degree of telecentricity in the lens, the bokeh of the same lens on a digital sensor will have a measurably different quality (not better or worse) than that same lens on film. As for whether the bokeh will be the same or “worse” the 1.2 (on digital) vs say a 1.4 (on digital), it can claim nothing of the sort! However, usually, there is some correlation between pleasing bokeh and larger apertures, and to that extent, then this is even possible!

        • TaoTeJared

          D80 was a CCD as well as many others.

        • AH

          Nope – Nikon D40, D50, D80, D200 all uses CCD and loses a lot more light than for instance Nikon D700 and Nikon D3.

          The latter probably using a better micro lens design to lead the light into the photon wells.

      • Where do you get this crap?

        On the chart itself, the Nikon D70, D40 D50. D40x, D60 D80, D200 are all CCD based. And they’re some of the worst performers.

        By big gripe with the graph is it is a function of the lens type. Clearly the Nikon 1.4 lens they used has more off-angle rays than the Canon 1.4 they used. They don’t publish what lenses these are. It’s obvious because how poor performing the CCDs are in this (those were the days before gapless microlenses, however CCDs have a much larger image area with no circuitry backlighting). Overlaying the two graphs are apples-oranges.

        Me wonders what would happen if I compare a nano-coated lens like the 24mm f1.4G to the pre-digital 24mm f/1.4L. Hmmm…

        • Edgar

          DxO is measuring the effect at the center of the field (that’s what Mr. Dubovoy told me). Therefore the angle of incidence of marginal light rays does not depend on the specific lens design. It’s about 20 deg for a 1.4 lens.

          • Really? All lens designs bring light rays in at the same angle? Nonsense. Nikon, in anticipation of the change to digital, actually started making significant design changes to their wide angle lenses, partly to make the angle into the sensor more telecentric.

            • Edgar

              Yes really, at the center of the field. The changes you are talking about (telecentric design) are intended to improve the uniformity across the field, which is quite another issue.

          • No, not quite correct. Even at the center of field, light must be focused through the outer edges of a large aperture, those rays are not telecentric from that perspective. If this weren’t the case, then the center of the field would have an ISO brightness independent of aperture size (think about it)—which is clearly not the case.

            In film, light hitting at any angle is equally sensitive. In the case of digital sensors, there is a “stack” and off angle rays hitting this “stack” may be a bit off.

            • What I’m saying is draw a ray from the center of the field, to the edge of the outer element of the lens and back through the aperture (edge) so that it still focuses on the film plane in the center. These are the sort of light that is “gathered” even at the center in large aperture lenses. You will see that this ray does not necessarily hit the film plane perpendicular (telecentric).

              Mark Dubovoy is incorrect, or you misheard him. There is no way the engineers at DxO aren’t aware of this as it doesn’t take a degree to understand how lenses work 😀

            • Correct. And what I was trying to point out is that where you place the aperture in the design will have an impact on the telecentricity of the light, too. Nikon narrowed the implied angle on most wide lens designs starting in the late 90’s.

          • Edgar

            Maybe I was not clear enough. At the center of the field, the angle of incidence of marginal light rays does of course depend on the f-number. But for a given f-number, it does not depend on the lens design. According to the Abbe sine condition, it’s something like asin(1/2N), where N is the f-number.

            The telecentricity of the lens (or more generally, the position of the exit pupil) makes a difference only when you move of of the field’s center.

      • The Nikon D40 uses a DDC and is also effected. So the usage of a CCD alone is no indication for not seeing the described effect.

  • Just A Thought

    Apparently only an issue if camera has a CMOS sensor – FX about 1/3 of a stop, so a 1.4 becomes like a 1.8. On some DX it is a bigger deal where a 1.4 becomes say a 2.5.

    On a DSLR just don’t use lenses faster than f2.8 and it becomes a non-issue.

    If you want to use faster glass, then grab your Nikon F4 or Canon 1N and you’re good to go (I knew there was a good reason not to sell them on eBay years ago). Anyone know of a good place to pickup a few rolls of Fuji ISO 6400 Velvia????

    • AH

      Many of the cameras in the test chart are using a CCD. Nikon D40, D50, D80 and D200 and they all lose more light than for instance the Nikon D700 that uses a CMOS 🙂

      • Just A Thought

        You’re right. Thanks for the correction.

  • Fascinating. Sounds like a class action suit in the making. I just wish they would put a dang ISO setting of 25 on these cameras to really use these fast lenses–now that Nikon has seen fit to get back into the fast lens market. Why would I buy a f/1.4 lens for $2k when I have to use it at ISO 100 or 200?

    • imheretosetyoufree

      Axel… You mean a “glass” action suit. 🙂

  • I’ve made test last year. Took tripod, D700, different leses, and shoot the wall at same settings, changing aperture.

    Shoot RAW.

    Here is the results:

    As you can see, several lenses have better T-stop than another (85mm f1.8 and 135mm f2 are 1 stop brigther!).

    Also, 135mm f2 gives same exposure results on f2 and f2.8. This is very very weird.

    I’ve create post here

    • FINALLY someone with some sense.

      Your chart demonstrates quite clearly what the article is describing (variability). Note, however, that it happens ACROSS THE BOARD and is different for EACH lens.

      LL and DxOMark are making an issue of a total and complete non-issue. Suck it up, stop measuring your gear and go out and USE IT.

      • Yeah im using it instead of meausring.

        But, understanding T-stops is critical if you really shooting in low-light situations. Or if you are relying on flash power output. For example, this year i did not bought 80-200 f2.8 lens because its 2.8 have same t-stop as 4 on my primes.

  • lolcatmaster FTW

    People are so dumb, why do you care? Luminous Landscape is sponsored by DXO and they are only whoring the “test” to promote sales of the DXO software, this is beyond stupid, sometimes the internet is the nest of *diots.

    Grab films of the same ISO rating and a friggin full frame camera and do the “test” again and then smack your head on the wall for wasting so much time in such an idiotic way.

    • Man, you are so agressive here, but I agree with you at all points. This issue doesn’t affect my image in any way, cause when I capturing the frame, I think about moment, light, shadow, pose, composition, not about this technical shit for brick-wall shooters.

  • Film.



    • Film had layers. When light comes in at an angle on film, it registers in a different X,Y position in each layer.

  • Nikon Tandoori

    sometimes with Nikons like the D90 you can clearly see shooting by night that the ISO200 coming out is as grainy as ISo800 at least.

    • How long exposure might be when you shoot at night? I used to capture still images ~32 minutes long at f16 during fullmoon night and noticed that my D5000 (which uses the same matrix as D90) had noise level at ISO200 not much more than standart photo, taken at shiny day same setups.

    • This is because the D90 has a CCD sensor. In a CCD sensor, digitization is performed in a separate DSP chip (instead of on-sensor). While at first you might thing this would produce less noise (and you would be correct!) than CMOS, you must realize that the electons have to be gained and exported to the separate chip for processing. This chip therefore runs at a higher voltage (and thus emits heat).

      The net result is the heat from the DSP chip ends up contaminating (accumulating false electron charge unrelated to light) in the light sensor itself. This is evident in long exposures and is cancelled partly by the photographing of a dark image and digital subtraction.

      Even if you don’t do the dark image subtraction. Nikon still does a baseline subtraction anyway because the amount of heat on long exposures contaminates the image that much.

      And even with the subtraction, the statistics just wears you down and is evident as more noise.

      This is only for long exposures, btw. Now what is being discussed above.

      And overall CCD is way better. It’s just when you are talking about astronomy applications (like the Hubble or something), the DSP is put further away or in a cryogenic bath so there isn’t heat contamination. Unfortunately, you can’t have a little LN2 tank in your D90 😀

      That’s why the manufacturers have switched to CMOS (with few exceptions like Leica).

  • M!

    Luminous Landscape: “1. When I select a specific ISO, I have a reason to do so. I find it unacceptable for the camera to change it without my knowledge.”

    I don’t see the problem when let say, you selected ISO 200 and the camera knows that it is under by -0.3EV, and compensates it back to what should be ISO 200 for you, so that your f1.4 is still f1.4 and your shutter speed is still your shutter speed.

    Can LL’s author tell in a picture the difference between a picture that is taken at ISO 200 and ISO 200 +0.8EV ? (every thing being equal)

    • Michael

      I do not see the problem either. No, you can’t. But if you know, of yeah, than you will see it. 🙂

    • Just A Thought

      You commented about base ISO 200 +0.8 EV. You can buy a fast lens for DOF and also for Low Light. What if the Low Light Shooter uses an f.14 at ISO 6400 and then the camera adds in +0.8EV. Think that would affect the image quality?

      Can you tell the difference in sale price between say a 85mm f1.4 and a 85mm f2.0 and a 85mm f2.8? My Bank Manager sure can.

      • M!

        the test run by DXO is not based on the parameters you are describing.
        they are testing the problem at ISO 200.
        and yes, if you have 3 pictures, each taken at everything equal, at the widest aperture, you can tell the difference btwn 85mm 1.4, 2.0 and 2.8.

    • They should be evident in the increased noise performance and less accurate color rendition/gamut. Realize if it’s as much as a stop, then if you are on the edge of what is considered acceptable (that depends on your level of pixel peeping, for me and a Nikon D3, I’d put it at somewhere between HI+1 and HI+2, but for these guys it’s anything above 200 ;-)), it may (may may may!) push it into the unacceptable.

      I understand the argument, but I don’t buy it. Really this is not an argument for the camera so much as an argument for the lens. Lenses with older designs are going to perform worse than they are rated for.

      As time goes on, this will nearly be eliminated. After all…

      Do you think that if Nikon had a NEW f/1.4 lens (they’ve introduced two this year), and they perform on digital worse than the f/1.8 in overall ISO (meaning the 1.4 needs to be gained more than a full stop and the 1.8 only 1/3 of a stop), how would that fly? The answer is someone like Bjorn Roslette would see it instantly and smash that lens into the ground. I guarantee it.

      The only company I see getting away with any fiasco would be Canon and only for making nominal 1.2L lenses that perform worse than 1.4L lenses. That’s it! There is no way they’d get away with it between the 1.4 and 1.8 or within different iterations of the 2.8L line or any such thing.

      That’s why this thing is a non-issue. The reason this is going on is the manufacturers are trying to create the best experience for the customer and one of the things they have to do is make lenses designed for film work correctly on digital sensors. They do it using this. That’s a good thing™.

  • Michael

    I am not sure whether this is a real issue or an Internet hype.
    Everyone knows (should know) that every lens loses light, this is not something ‘new’.

    In the ‘old’ days of film and mechanical cameras ISO could not be tweaked. But even in those days there had to be a compensation somewhere because otherwise we would have seen underexposed photos since 1870. In the ‘old’ tutorials one could read: ‘Make some test pictures and check the exposure so you know how and when to compensate’. Compensate for errors of your meter and camera and compensate for loss of light.

    You can read in test reports already for dozens of years ‘Test results of lens don’t match specification, 1.8 is in reality 2.1’ And the next line ‘This is no problem as the camera adjusts for this automatically.’ By exposure of course. And nowadays it adjusts by changing the ISO a little. To get a correct exposure. What is the big deal? And, the camera adjusts not only for loss of light but does a lot more tricks with its SW. Are we going to fight these alterations as well?

    A good photo stays a good photo regardless of this or any other technical issue. And a good photo, that’s what is all about.
    Or am I missing something?

    • Bryan

      thats very true, a good photo IS what its all about. but its also true that lousy photographers are always looking for an excuse.

      • Just A Thought

        Yes, but I can pay XXXX amount of dollars for af f1.2 or f1.4 lens or I can pay a heck of a lot less for a f2.8 lens.The resulting good photo could essentially be the same with either lens (depending on the DSLR it is used on). Your bank account, on the other hand, could be quit different after said lens purchase.

        • Michael

          But you see a difference nów, do you? Between a 1.4/85 and a 2.8/85mm eg? Well the good news is that they both suffer from loss of light, that’s the way optics work.
          So your 1.4 is may be in reality a 1.8 (at the end of the lens). But the 2.8 is may be in reality a 3.1. So the difference you see right now is NOT the difference between 1.4 and 2.8 but between 1.8 and 3.1. The 1.8 is still a lot faster than the 3.1. So you and I all the rest have always seen a ‘net’ 1.8 result and never a 1.4. And paid for 1.4 at the front and 1.8 at the back. Or paid for 2.8 at the front and 3.1 at the back. The standard notation is F-stops and not T-stops, that is may be confusing but this LL-story is really nothing new at all.

    • Be careful here. And this is where the article is trash.

      Lenses actually report T-stop not F-stop (to within a third stop accuracy… something manufacturers fudge constantly and have since the dawn of time). Don’t believe me, just measure the physical aperture of your telephoto and do the math. You’ll see that say… the 200mm f/2 lens does not have a physical aperture that’s 10 cm wide 😀

      Multicoatings have reduced the light loss in the lens to small levels allowing designs with tons of elements and groups. We’re talking about early 80’s technology here. What really exploded things in modern times is improvements in manufacturing and CAD.

      What is going on is the F stop on your lens is just the T stop for your lens and lens alone. But the REAL T-stop is the lens/camera combo and when we mean camera we mean digital sensor. There is real variational differences going on.

      All that stuff about how the movie industry uses T-stop and the photography industry uses F-stop? Hogwash. They’re talking about crap that fell in the dustbin of history around the time I was born. The author is clearly confused. He didn’t understand what the DxO engineers were talking about.

      • >You’ll see that say… the 200mm f/2 lens does not have a physical aperture that’s 10 cm wide

        Wrong: F number is entrance pupil between focal length, and entrance pupil exists in object space and can be bigger than the diaphragm…

  • brave new world

    It tells us: check for the results and adapt the settings as required … that’s it. BTW 135mm at 2.0, 85 at 1.4 and 24 at 1.4 perform nicely on a D700. Just another reason using the combination. It shows also: use the RAW format to compensate for any unintentional changes by the camera built-in software.

    • Just A Thought

      During RAW conversion you cannot reduce the ISO which the camera used to take said image file.

  • Phil

    I’m not exactly understanding this. Are they saying modern optics have crap T-Stops, or are the sensors somehow not properly responding to the light values the larger lenses are transmitting?

  • TaoTeJared

    Here is my take –

    The data shows that there is less light at F1.4 then at say F4.
    Take a flashlight – the narrower the beam (smaller app) the more intense the light. The world is a light bulb and the lens is the top of my flashlight.

    Second part – The larger, newer, less dense sensors perform better than one’s from 5 years ago. Once again – Duh! Techenology is improving, design, electricity flow, materials, everything is constantly changing.

    I don’t think there is any conspiratorial agenda here. If they round a number or two, who cares. I don’t care if it says iso 100 or iso 110 – I assume they do. I think someone is digging too deep in the numbers and forgetting some real life knowledge.

    Come up for Air guys – Really the world is more than charts and numbers.

  • Edgar

    This issue has already been discussed at Check my answer, especially the section Update 2 about my correspondence with Mr. Dubovoy.

    Bottom line: it’s quite easy to prove him wrong.

    • Sean

      I see what you’re saying about the optical path, but I wonder, shouldn’t we also be comparing SNR across apertures so as to verify that the ADC isn’t (as DXO claims) boosting the signal to keep that brightness constant across images [that it doesn’t change across the field is something I don’t disagree with]?

      • Edgar

        Well, it may very well be that with fast lenses, the sensor is slightly less sensitive to marginal light rays, and that the camera compensates with the ISO. We will know more when DxO publishes their full data, but even if it is true, it is nothing to worry about.

        My issue with Mr. Dubovoy’s article is that he claims that marginal light rays do not hit the sensor at all! And therefore fast lenses do not provide any advantage over, say, f/1.8 lenses. This is quite a strong claim! That’s why I took the pain to prove him wrong, at least for my sensor.

        • Sean

          I don’t dispute that DXO’s explanation was awful – short on facts, and the rationale there just wasn’t a good one. But I do agree with them and a few others here inasmuch that, if indeed the sensors are losing some % of highly off-axis photons as both you and their data (appear to) suggest, it should be a user choice to compensate for that either with shutter or with ISO. Locking everyone into ISO directly impacts IQ, when we might be able to afford that 1/3-1 stop of shutter speed in order to get the better result.

          Everyone wants the best quality which they can get for the situation.

          Thanks, btw, for testing the aperture issue. That was an important point for clarity about what is really at play here.

        • LB

          It is so easy to test. Long before DxO people had noted that if you take black frames with a fast lens mounted to the camera instead of a slow lens/body cap and then measure the SNR of the black frame that you consistently get a different result when the fast lens is mounted. The amount of difference varies by camera body. Often it’s not too much of a big deal, especially unless the lens is very, very fast. It’s not going to be an instant cut-off, but you slowly get just a little bit less of everything that you expected from the fast lens the faster it is.

  • G

    So why do film cameras expose accurately at the desired ISO speed? There’re too many anomalies in this article.

  • Strange results on DXOmark lens compare. Try these three Nikon lenses: 70-200 F/2.8, 18-70 f/3.5-4.5 and 18-200 f3.5-4.5 on a D300.

    The cheap D70 kit-lens (18-70) scores a 27 Lens Peak Score. The mighty 70-200 is just a lousy 2 ponts better with 29???!!!!

    Must be a big mistake or these comparissons are totally useless and therefor this “ISO-question” is totally bogus too???

  • Arthur

    Interesting article about the sensors. Frankly I don’t know how Nikon does it, but Canon has these (gapless) micro lenses in front of the sensor. And it looks to me like like pretty much every photon gets to the sensor in the right angle. Of course, the border between this micro lenses can’t be 100% accurate, but still I would think that these micro lenses would eliminate the “wrong angle the photons are falling on the pixels” the above mentioned article is talking about.

    • Gapless microlensing is a Nikon technology that’s licensed to Canon (sort of like Optical Image Stabilization in the reverse).

      And the answer is no, it gathers all the photons but its optics and there is no guarantee that all of them hit the light sensitive portion of the photodiode (not even close, but there are other Nikon technologies that have been licensed). If the majority of the rays come in at a sharp enough angle (edge of the frame), they can actually focus on the neighboring pixels which produce an interesting form of color fringing that Leica testers know a lot about (because of the lack of an AA filter).

      This is why when you look at the graph above, you’ll see the D3 and D300 generation cameras significantly outperform the D200 and D70 generation sensors. A similar jump for the Canon designs (which occur a year and a half later).

  • PHB

    The article does not make a lot of sense to me. The most significant structure in any sensor, regardless of the technology is the microlens array.

    Mention of the medium format makers makes me somewhat suspicious. There is a lot of complete rubbish talked on the forums and much of it originates with the manufacturers trying to spin their product (cf Canon’s ‘Nikon can’t make f/1.2 lenses because of the f-mount’) . This looks like something the Hassleblad marketing dept thought up to explain why they are sticking with CCD.

    f/stops and t/stops are functions of the lens, not the sensor. Different sensors have different sensitivities. There is no absolute value for ISO in digital sensors, gain is a function of the sensor. So what are they measuring here?

    One of the features of a DSLR design is that the rear of the lens has to be quite some way from the sensor. Thus it is geometrically impossible for a light ray to hit the center of the sensor at anything other than an 80-100 degree angle for all but a handful of lenses regardless of how the light is distributed.

    The only sense I can make of the piece is if the author is confused about a peculiarity of the Canon f/1.2 lens which is that allegedly the Canon cameras apparently have to focus it at f/1.4 and open up a stop for the shot. Which does not make sense to me either. I can quite believe that the Canons can’t focus for such a shallow depth of field, but can’t see how trying to focus with a narrower aperture would help. I suspect that is also nonsense.

    • The author is confused. Very confused.

      The engineers at DxO are not at all.

      What they engineers are trying to say is the microlens array inside the sensor forms part of the “lens system” if you will when coupled with the lenses. The two should be considered together, and can diminish the aperture by as much as a full T-stop (depending on the sensor being a very old design and the lens being a pre-digital large aperture prime normal I’ll bet 10-1 (something without a wide angle reverse telescope in it) ;-)).

      The net conclusions are:

      1) If you try to find this yourself you might not see this (unless you cover up the electical contacts of your lens so the camera can’t see what lens is attached).
      2) Hey, if you buy a newer camera, your existing photos might be less noisy and have better color rendition. (If that isn’t f–king obvious).
      3) If you buy an older lens at f/1.4 or f/1.2 it might not perform as well as a newer lens at f/1.4 or f/1.2. (Again, massive stupidly obvious). And don’t trust just comparing the resolution charts either because they are theoretically computed (Canon) or measured in testing on film with a tool from Zeiss, not your camera (Nikon). (Again, you must know nothing about such charts to draw not know this.)
      4) This different will be more noticeable the larger the aperture of the lens you use when shooting at maximum apertures. (Okay, this is novel, it hadn’t occurred to me since I read the article, but honestly… really?)

  • hah

    T stops do not equal fstops. The 70-200 vrI has less transmission (t stop) than the VRII. the vrII looks brigher as a result. yet bokeh is the same at f2.8.

  • Anonymous

    sorry, nothing new here, it’ve been around for many decades. the light loss is caused by heavy vignetting at wide open. just throw in +0.3 or +0.7EV or turn on vignetting control when shooting wide open, problem solved.

    • xjrx

      confusing. not only the original claim, but also the varoius issues people conclude from it.

      ok, so f-stop is the “hardware”-specification of any given lens, focal length/ aperture etc.
      that does not say how much light is “wasted” by the lenses within that hard specs. for sure there is “light-preserving” and “light-wasting” designs.

      in comes t-stop. which makes perfect sense for moviecameras and everybody, who needs to care about 1/3s of f-stops for whatever reason (f.e. filmstudios and pixelpeepers)

      the lensmanufacturers ofc know exactly, how much light is wasted/preserved by their lens-desings, they desinged them, they produce them and they have to market them.

      comparing lenses by their f-stops makes sense as a marketer. telling your lens/camera-CPU specific profiles of these lenses, so the camera can -theoretically- compensate the (small) differences in actual t-stops makes also perfect sense from the manufacturers pov.

      you can easily compare lenses by their f-stop “hardware”-specs, adding the “soft”-specs to that list would be stupid marketing. 90% of all users do not care about 5% or 10%

      why not measure the actual t-stop for all those lenses out there?
      isn´t that

    • xjrx

      confusing. not only the original claim, but also the various issues people conclude from it.

      ok, so f-stop is the “hardware”-specification of any given lens, focal length/ aperture etc. that does not say how much light is “wasted” by the lenses within that hard spec. for sure there is “light-preserving” and “light-wasting” designs.

      that´s nothing new plus these lens-specs have always been approximations, no?

      in comes t-stop. which makes perfect sense for moviecameras and everybody, who needs to care about 1/3s of f-stops for whatever reason (f.e. filmstudios and pixelpeepers).

      the lensmanufacturers ofc know exactly, how much light is wasted/preserved by their lenses, they designed them, they produce them and they have to market them.differentiating lenses by their f-stops makes sense as a marketer. telling your lens/camera-CPU specific profiles of these lenses, so the camera can -theoretically- compensate the (small) differences in actual t-stops makes also perfect sense from the manufacturers pov.

      it happens all the time! a BMW 528 has “only” 2699 ccm, the 70-200 VR is “only” a 135mm at 200mm @ closest focusdistance, there is a lens called f/1.8 but it is f/1.9 etc. there is room for variation everywhere, the technologial culture improves and uncovers those small margins by time.

      what i do not get is what DoF has to do with it? DoF should be determined by the f-stop and has nothing to do with the actual amount of light, that hits the sensor in the end.

      second mixup is the sensor-design-issue. ok, so CCD and CMOS behave differently, still the actual amount of light hitting whatever sensor is determined by the lens´t-stop. ofc it makes the lens-testing more complicated if you want to give people the real t-stops for their lens and sensor-combination.

      so this new advanced second spec t-stop could just be pickud up by the manufacturers as addon-info for each individual lens. grand, a good reason, to once again test all the lenses in the world for a new quality. actually thats what dxo started this year, they catalogued sensorlens-combination-tests.

      if there was a conspiracy, it may be that modern design vs cost engineering makes lenses significantly “darker” than their hard spec “f-stop” should allow. i´m thinking of millions of plastic kitlenses. they are built 3.5-5.6, but maybe the designers know, they only yield 4.0-6.1. nobody buys that lens. engineering found out, that by firmware-patch the ISO can be inceased to recapture those 1 f-stops and nobody ever notices. maybe there is even a toplevel built-in ISO-margin, again: to compensate the various physically/engineeringwise differences between the lensdesigns and make lenses transparent and comparable to a mass of comsumers.

      so a 2011 lens-test looks like this
      1. specs incl f-stop and t-stop. something new like “t-yield”-factor included, which is just a theoretical number. in extreme cases the accusation that a f/3.x lens in really a f/4.x lens in terms of light-yield, not even counting transition to any given sensor.

      2. transfer-factor of “t-yield” to camerasensor-ratings for all those sensors out ther, see dxo.
      after that, we all know, what our lenses (in case they are like the testsamples) with our sensor (in case our sensor is like the test-sample) are doing. millions of hours will be spent chewing on that in the internets. sample-variations will unsharpen all that, beautiful box, pandora…

      thanks NR admin, always good sunday reading here. sorry if this was too long, ppl 😉

      • xjrx

        whoops. another sorry. too stupid to post properly. shutting up now

    • This is where the author is so confused (and you are too).

      The author wants to distinguish this report from the known result that there is heavy vignetting on digital sensors at max apertures (that which you report).

      These measurements are taken at maximum aperture in the center of the field and even that needs to be gained up because light, even there, is not perfectly telecentric. The larger the aperture of the lens (if it’s an old enough design, something the author and the DxO engineers are purposely ignoring) the more pronounced this light loss.

  • Bruce

    Hey folks… nothing on the DXO web site says they endorse the conclusions drawn by the LOONIE TUNES over at LL. The LL guys have got their knickers discombobulated over a non-issue. The ISO doesn’t matter. What matters is the exposure and Depth of Field. The cinema folks use t-stops because they MUST be sure the EXPOSURE is correct. With still photography we have shutter speed/aperture/ISO to play with. The f-stop characteristic of each lens sets the depth of field and an f/1.4 will ALWAYS have a shallower depth of field than an f/2.8 or f/4 lens. Nothing that DXO have come up with changes this fact. The LL folks have run off on an erroneous tangent. LAUGHABLE.

    • Except for the fact that there’s a nice big DxOMark logo on the LL chart, nothing at all, nosiree!

  • MG

    time to go back to film 🙂

  • Ken Elliott

    Actually, this is a very clever way to compensate for light loss due to sensor design and lens design. I’m GLAD they do this.

    Yes, we would all be better off if we used T-stops (actual transmission values) rather than F-stops (a mechanical ratio that only gives a rough and inaccurate indication of transmission values). A change of 1 F-stop is close, but not exactly a change of 1 stop. T-stops are simply more accurate. With all the confusion about “crop factor”, I doubt the industry will try to make that change.

    A photo taken at 1/250 and f/1.4 on a D3 will have be brighter than a D300 using the exact lens and settings. Do you want to always be thinking about exposure compensation between bodies? I don’t.

    The solution is for the manufacturer to compensate by adjusting the ISO so digital cameras are consistent from one body to another – just like film. This appears to be an extremely well thought out solution, and it works. After all, nobody noticed it until DXO figured it out. If they didn’t do this, I would have to do this via ISO, shutter or aperture. So this is a GOOD THING.

    • Just A Thought

      So paying for an f1.4 lens and getting an approx f2.5 lens (worst case – depending on the camera) is a GOOD THING?

      • fakekenrockwell

        a f/1.4 lens is an f/1.4 lens. You get a bigger physical aperture and all the characteristics, benefits and compromises of that lens. Focal lengths don’t change (sigh), and neither does actual effective EV assuming the camera is doing some adjustments to compensate for light loss as needed.

        But yes, what people don’t seem to understand (something which nikon has been on to for several years now), is if you make sensor sensitivity a non-issue, your kit lens at 18mm f/3.5 might actually out perform someone’s fancy f/2.8 prime on a competitive body.

        Basically, in the very near future, the only reason one should choose a very big aperture lens is for DOF and artistic reasons and not not ‘speed’, i.e. shutter speed.

  • Anton

    Considering the topic is in focus, i have these remarks:

    1)How does a particular lens say get a 1.4 F stop marking? By that i mean, that such a lens is usually usable starting only from 2.0. What’s the point of a 1,4 label then? So is there some (international) standard that states under what circumstances a lens get a certain f stop label. (NR admin do you know of any) Do different manufacturers use the same guideline when labeling their lenses.

    2)As many others mentioned, the main purpose of the f stop is to control the DOF, and not the amount of light that passes through. So, I don’t see anything wrong with the tweaking of the sensor sensitivity. This is how it should be. Some cameras have automatic correction of the distortion, vignetting and chromatic aberration (It’s not like the particular lens does not have any of them). Cameras can also correct (compensate) the colors. That’s how it should be. BTW, what is the definition of ISO 200 for instance?
    I would expect the camera to automatically adjust it’s sensitivity across the sensor to meet the standard. The cameras should do the automatic compensation in such a way that there would not be much difference in brightness between different lenses at the same settings. So if a lens has more lenses ( 🙂 , recursion) in it, the brightness should be the same. Shouldn’t it? Imagine that for each lens a photographer would have to know how much sensitivity he should add to make the brightness similar with some other lens?

    BTW, from my knowledge the Hasselblad camera tweaks EVERYTHING. That’s what you pay for when you buy it.

  • FakeKenRockwell

    Wow. Folks, ISO in digital photography is a made up number. It is not a measurement of anything lol. It is more or less arbitrary anf relative The point of iso is so that the outcome is predictable given our analog frame of reference. Why do you think nikon is spending money to make iso basically irrelevant?

    • taotejared

      Thank you!

  • broxibear

    “If the camera is automatically going to increase the ISO due to a significant light loss at the sensor, does it make sense to buy bigger, heavier and much more expensive large aperture lenses?”
    Now that’s an interesting question… The Nikon AF 85mm f/1.8 is regarded as a great lens and costs £300, the AF-S 85mm f/1.4G costs £1400 ?

  • it happens. using my d200 with nikkor 50 f/1.2 at 1.2 with the exact same iso and shutter settings, the camera will under expose if you don’t input the lens aperture informations. its about a 1/3-1/2 stop difference even tho there is no change in shutter spd and aperture nor STATED iso value. the dof however, is still noticably shallower than with the f/1.4 AIS.

    now i know why my wide open shots always seemed noisier…

    • Michael

      I think the D200 is known for underexposing in dull light; mine does anyway. So in these conditions I always correct 0.3+. May be you use your 1.2 in these dull/dark situations?
      Btw, the D200 is not the top of the bill regarding freedom of noise. But even if the camera would change the iso from 200 to 250 you won’t see it imho. And I do not think you will see a difference between 800 and 850, they both will be noisy.. If your wide open shots are noisier, I wonder, are they may be still underexposed?

  • anonymule

    After reading the article, and the many comments and opinions on the subject, my conclusion is: It doesn’t make much of a difference to me. I shoot with a D90, and I’m pretty satisfied with its dynamic range. The third stop increase in ISO the camera is keeping from me matters little to me in the real world. Bottom line. I don’t see how this changes anything. Some of you people spend too much time peeping at pixels, measuring depth of field and light loss. Who cares? And every reply to this effect has been challenged by the pseudo-intellectual rocket scientist pointing out…..well, I can’t really tell what they are pointing out. When someone says, it doesn’t matter, why is it you people’s jobs to make us believe that there is some kind of conspiracy going on here?

  • Eric Pepin

    This is a moment where tech geeks need to relax, and just be photographers. I couldnt care less if my d300s decides to use iso 240 or 300 even when i bust the aperture down to 1.4.

    • Not Banned

      One may not care, but another may.

      Anyone who has a brain “calibrated” to film behavior likely does quite a bit of mental translation when it comes to digital behavior in trying to produce the image they see in their head.

      A better understanding of how the sensor responds, especially when it differs significantly from film as in this example, is helpful for that translation.

      • anonymule

        I disagree and reaffirm my lack of interest. I shot film 10 years before I picked up a digital. There’s no translation. Film is not magical in how it captures a latent image that makes it differ greatly from digital. However, with digital, I get to see what I did immediately and can translate what I see without waiting an hour or a day to get the film back, thereby making the process of brain-image translation easier and a lot faster. I don’t see how people who are “brain calibrated” to film would have such a tough time, and I fail to see how this subject would change anything if they did. Once again, you fail at making a point, like everyone else going on and on about it. It simply makes no more difference today than having heard nothing about it the day before yesterday. If people spent less time debating the relevance of this and more time practicing technique, their results would become better than what could be expected if we all just sat around beating this horse into the ground.

        • Phil

          …of course film is magical… 🙂

  • jk

    If the standard set in the analogue days has changed then it has changed. Lets just hope that if it is true they will have new “digital standards” so everything stays consistent. There was a time when film was not all exactly iso 200 for example. That is why iso and asa exists today.

    This variation in exposure, light loss or whatever is highly likely but who really cares? especially if it takes such a controlled nerd test to reveal this happening.

    The very fact that there are lenses now that are considered ideal for digital bodies clearly reveals it is not the same as film. One should expect this as technology changes and stop comparing digital to film. There is not a whole lot they share other than the (most important) ability to capture an image.

  • The only real “disadvantage” I can see of this is that if your shooting in EXTREME low light with a 1.2 or 1.4 lens and you already have your ISO cranked, your camera will top out a little before it should. Maybe a half stop.

    All I can say is….thankgod I have a D3, its on the top of the list 🙂

  • Also, I like how DxO just blanketed “you may want to consider buying smaller aperture lenses” without saying “well, we mean, that depends on how shitty your sensor is.”

    Us D3s/D3/D700 users are sitting pretty and can actually benefit from good glass from the looks of those charts… but isn’t that what they’ve been saying about lens resolution for years anyway? They should just say “if your camera is loosing a whole stop at the sensor, buy a new camera.”

    Nothing’s changed. Nothing to see here.

  • Mark Astle

    What I love about digital is that I don’t have to arse about making stupid calculations before I take a picture. I, shoot, look at the back of the camera, check the histograms if I’m being really picky, adjust, and shoot again. If I want a small depth of field, I’ll select a large aperture. If I want to shoot in low light, I’ll up the ISO setting. And I know what those things are going to do to the image. If the f2.8 ISO 200 shot, is in reality an f 3.1 ISO 220 shot, I DONT CARE. I know what kind of image those settings produce on my camera with that lens and I work from that point. Simple as that. What the camera’s doing behind the scenes is of no interest to me and makes no difference to the end product.

    • Not Banned

      Not every subject sits still while you chimp.

      Not every moment comes around again… ever.

      While I’ll admit we’re discussing a corner case – I see no need to be dismissive of understanding what exactly is going on.

  • Back to top