Opticallimits

Full Version: LL & DxO open letter : wide aperture lenses & light transmission
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
Pages: 1 2 3 4 5

Sylvain

Fellow Zoners,



I don't think it has been posted here before and I did use the search function so forgive me if it has. It's rather fresh so...



Reading a french website, I found this article :

http://www.luminous-landscape.com/essays...rers.shtml

&

http://www.dxomark.com/index.php/en/Our-...stop-blues



It's stating very interesting findings.

Among which:

-widest aperture setting provokes a light loss that can render its light collection as "bad" as closing the aperture down

-high pixel pitch sensor are negatively impacted by too wide an aperture (e.g. 7D @1.4 -0.65 EV vs 7D @1.2 -0.95 EV. This would make the use of a lens set at 1.2 rather pointless beyond the DOF aspect.)

-That DOF aspect might itself be questioned by the light ray lost due to above mentioned issue (DxO labs studying it).

-some camera makers are compensating through ISO boosting

...



Some questions :

-I'm not knowledgeable enough to assume but can we consider the light loss from a canon ef 50mm f/1.2L @ 1.4 comparable with a ef 50mm f/1.4 @1.4? My guess would be no, considering that the lens design is different.

-Is focal length of any relevance (for zooms, beyond the scope of this article)?

-Shouldn't they provide "absolute" light transmission data (not sensor based) for a thorough analysis of what's really going on? or is this not measurable?

-How can we interpret the difference between Canon 550D & 7D that are supposed to have a similar sensor?



I'm eager to read more about this and know how it could possibly prove that certain higher specced lenses at certain settings will not give you any advantage over a lesser model in light gathering abilities (sharpness, CA control, etc put aside).



Hope you find it interesting,

S.
Well, there is a reason why it is no longer listed on the main page of LL.



Also, check the forum for the thread on this.



IMO, it wasn't a well written article, with quite a few mistakes, and personally I found it rather disappointing that a well known and respected author actually wrote this piece in the way it was done.



'nuff said.



Kind regards, Wim

Sylvain

[quote name='wim' timestamp='1291156423' post='4615']

Well, there is a reason why it is no longer listed on the main page of LL.



Also, check the forum for the thread on this.



IMO, it wasn't a well written article, with quite a few mistakes, and personally I found it rather disappointing that a well known and respected author actually wrote this piece in the way it was done.



'nuff said.



Kind regards, Wim

[/quote]



Wim, even though you wrote "'nuff said", I'm afraid you did not say anything :/. I did find the article was not well put together (no explanation whatsoever on graph data collection, etc...) but that what it's trying to say could be rather interesting. Or did I misunderstand everything?

DxO still inks to this article and what I could find in the LL forum wasn't really saying anything else than "no explanation". Care to enlighten me? Doesn't have to be a Wim Encyclopedic answer <img src='http://forum.photozone.de/public/style_emoticons/<#EMO_DIR#>/smile.gif' class='bbc_emoticon' alt='Big Grin' />



TIA,

S.

genotypewriter

[quote name='Sylvain' timestamp='1291158324' post='4616']

but that what it's trying to say could be rather interesting. Or did I misunderstand everything?

[/quote]

It's interesting... just like any other piece of science fiction <img src='http://forum.photozone.de/public/style_emoticons/<#EMO_DIR#>/smile.gif' class='bbc_emoticon' alt='Big Grin' />





[quote name='Sylvain' timestamp='1291158324' post='4616']

DxO still inks to this article and what I could find in the LL forum wasn't really saying anything else than "no explanation". Care to enlighten me? [/quote]

The main problem with this article is they're trying to make big claims with limited and possibly badly conducted tests.



There's no doubt that there's light loss due to light entering the photosites from an angle but they're trying to prove this through a methodology that they're hiding. It's a really difficult thing to measure, especially because stopping a lens down reduces optical vignetting that reduces the transmission difference between center and borders. This effect interferes with what they're trying to measure at the sensor level, simply put.



At any rate, a slower aperture doesn't mean there's more (total) light gathered. In fact, I think this whole manufacturers boosting the ISO thing at large apertures is pure fairytale-grade speculation. That can be easily tested for though... I might actually do that sometime soon, now that I think about it.



GTW

genotypewriter

[quote name='LL: Mark Dubovoy']

So the next question is: If the light loss is so significant, how come we do not have great numbers of photographers complaining about severe underexposure at large apertures?[/quote]

What a dumb thing to say! Doesn't this i***t understand TTL metering or vignetting?





[quote name='LL: Mark Dubovoy']This graph clearly shows that camera manufacturers “game the system” by increasing the ISO without the photographer's knowledge.[/quote]

Another mega dumb claim... this light loss problem is in the corners. Increasing the sensor gain would lead the centers to overexpose.



GTW

PuxaVida

AFAIR, they were claiming that the manifacturers compensate the lightloss with adjusted ISO values. And LL leans on DxO's (top secret) test reports. Regardless of the fact that they're not yet technically verified, I'm curious about if there exist any responses from the mentioned manifactuers. I mean it must not be so hard for these companies to publish a disclaimer.



Serkan

Sylvain

[quote name='genotypewriter' timestamp='1291162647' post='4620']

What a dumb thing to say! Doesn't this i***t understand TTL metering or vignetting?







Another mega dumb claim... this light loss problem is in the corners. Increasing the sensor gain would lead the centers to overexpose.



GTW

[/quote]



Thanks for your answer GTW. I understand a lot of information is missing in these articles and that Mark Dubovoy might be telling bullcr@p but I'm still giving *some* credit to DxO lab people, not much though. I assume they should be rather educated people, at the very least on the software engineering side, and that they do have some scientific education on proper test procedures. I'm guessing they're just not very good at reporting (which is also a part of one's scientific education I agree) and that even though we're left clueless with one guy poor analysis, there must be some things to investigate (caused by light angle or not). The practical consequences.

Now one could argue about the relevance of such findings IRL, but then, one could discuss the relevance of the slimmest form of pixel peeping. Who sets the limit?

Also, a much more constructive move would be, IMO, to structure an open reply to highlight the shortcomings of their "draft" and request a of their methods.



Greetings,

S.

Guest

[quote name='genotypewriter' timestamp='1291160485' post='4618']

It's a really difficult thing to measure, especially because stopping a lens down reduces optical vignetting that reduces the transmission difference between center and borders. This effect interferes with what they're trying to measure at the sensor level, simply put.

[/quote]



Exactly the point, gtw. I think that the light loss they measured is indeed the increase of vignetting at larger apertures. A vignetting would decrease the overall sum of light if one does not distinguish the distribution of light within the optical cross section. Optically, if the light in the outer regions of the lens would be lost at wide open aperture, then it must be similar to an additional aperture directly in front of the pixel, something they seem to assume by bringing in the angle at which the light hits the photon buckets. But in this case, the out of focus blur circles wouldn't show an increase in diameter as well when using wide open aperture.



And I guess that the increase in ISO for compensation was not measured but back calculated: they assumed that the amount of recorded photons has to be constant for the photo not to be underexposed, then brought in the measured light loss and calculated, what the "real" ISO must have been for the equation to be valid.



Christian



PS: one last idea: if one uses an old f/1.4 lens and does not program the data into the camera (or uses false data), the camera should not be aware of when to compensate and for how much. I wonder if the images would show the amount of under exposure that has been predicted.
[quote name='TheChris' timestamp='1291200372' post='4638']

Exactly the point, gtw. I think that the light loss they measured is indeed the increase of vignetting at larger apertures. A vignetting would decrease the overall sum of light if one does not distinguish the distribution of light within the optical cross section. Optically, if the light in the outer regions of the lens would be lost at wide open aperture, then it must be similar to an additional aperture directly in front of the pixel, something they seem to assume by bringing in the angle at which the light hits the photon buckets. But in this case, the out of focus blur circles wouldn't show an increase in diameter as well when using wide open aperture.



And I guess that the increase in ISO for compensation was not measured but back calculated: they assumed that the amount of recorded photons has to be constant for the photo not to be underexposed, then brought in the measured light loss and calculated, what the "real" ISO must have been for the equation to be valid.



Christian



PS: one last idea: if one uses an old f/1.4 lens and does not program the data into the camera (or uses false data), the camera should not be aware of when to compensate and for how much. I wonder if the images would show the amount of under exposure that has been predicted.

[/quote]

The test is WAY more simple. I would do it if I would have a fast enough prime!



Attach the prime, wide open, to the camera (Canon EOS DSLR). Make a photo. Camera is aware of the lens, and the used aperture.



Press the lens unlock button and turn the lens a bit. Now the electrical contacts are not used anymore. Camera sees NO lens, and has no idea of used aperture. Make a photo.



Compare both photos. My guess: Both photos will be exposed exactly the same.



If my guess turns out to be correct: Articles are bullshit.



Simple enough test.

genotypewriter

[quote name='Sylvain' timestamp='1291197218' post='4633']

Now one could argue about the relevance of such findings IRL, but then, one could discuss the relevance of the slimmest form of pixel peeping. Who sets the limit?

[/quote]

Yes... the light loss at the corners is real but I just have serious doubts about their claims... especially since their methodology needs to be very carefully constructed to measure such things but they're so secretive about it.



Just for the record, I too think DxO is one of the better sites out there for sensor testing. For example the results from a test that I did and what DxO published on the same topic were similar (and both were published almost simultaneously):



November 8, 2008: http://www.flickr.com/photos/genotypewriter/3012247230

November 16, 2008: http://www.dxomark.com/index.php/en/Our-...omparisons



But I don't think they're doing everything the right way.





[quote name='Brightcolours' timestamp='1291210980' post='4652']

Attach the prime, wide open, to the camera (Canon EOS DSLR). Make a photo. Camera is aware of the lens, and the used aperture.



Press the lens unlock button and turn the lens a bit. Now the electrical contacts are not used anymore. Camera sees NO lens, and has no idea of used aperture. Make a photo.



Compare both photos. My guess: Both photos will be exposed exactly the same.

[/quote]

This is what I was going to do... except I was thinking about putting something like a thin film over the contacts. Just haven't had time to do it.



GTW
Pages: 1 2 3 4 5