-
Hm, sounds a bit unfinished to me. Ok, manfacturers increase the "real" iso without the photographers knowledge, which possibly results in more noise. But hey at f/1.2 its about half a stop only. So the gain of a f/1.2 lens over an f/2 lens is still there. As canon doesnt offer an 85mm f/1.4, which according to Mr Dubovoys argument would be the same noise wise as an f/1.2, I see no alternative to that very lens (unless you want to buy a missfocussing Sigma).
Regarding the depth of filed /maximum blur/bokeh thing: It would have been nice if Mr Dubovoy had facts. This way he is just creating a big stir on the basis of a few assumptions. Better do the tests first before publlishing some wild guessing.
In the end I fail to see the issue. The results of my 85/1.8 are different from the 85/1.2. I can see that in real life, without doing measurements. This "revalation" will certainly not make the f/1.2 less usefull or replacable by an f/1.8 in practice. So yeah, for the sake of a technical dsicussion, why not ask the manufacturers to let us in on the topic. But if they dont, its no big deal, because there is no alternative right now than the technology which is available.
-
I would like to know if this tactic is used on cpu lens only or on any optic. I use an Olympus E-p1, primarily with non-cpu Zukios. During live view the camera adjusts the aperture on the system lenses as the brightness in the scene changes but cannot when a non-cpu lens is attached ;D If one where to repeat these test with non-cpu lenses, would the camera still adjust iso? Does entering the lens data as on a Nikon change this?
This oculd be a work around for critical work ;) but would drive up the prices on old lenses :(
Bill
-
But if they dont, its no big deal, because there is no alternative right now than the technology which is available.
Which is *precisely* the point of his article.
-
Which is *precisely* the point of his article.
Agreed but this was also a key sentence: This graph clearly shows that camera manufacturers “game the system” by increasing the ISO without the photographer's knowledge.
For those of us that like to know what is happening behind the scene, I found the article highly enlightening! And I agree with Mark’s idea of an open letter to resolve this if possible. But like world peace, unlimited free and clean energy and cameras that produce an open, non proprietary raw file, I think it unlikely we users will be heard. Still, I’m thankful for the article, I learned something.
-
It doesn't look like the author knows how to interpret the very data he is quoting in his own article.
Bottom line: Due to the complexity of design and manufacture (let alone the high cost and weight) of large aperture lenses, one may actually end up with better results at virtually the same ISO and depth of field using lenses with more modest maximum apertures.
So many glaring flaws in that logic. The data clearly shows that ultra large aperture lenses still collect more additional light than is lost by the sensor. Going from f/2.0 to f/1.4 may not yield the full stop of light we might expect.... but it still yields a net benefit of 2/3 stops of light. The data is explicitly showing that. An f/2.0 lens does not yield "virtually the same" results a f/1.4.
When you look at the structure of CMOS sensors, each pixel as basically a tube with the sensing element at the bottom. If a light ray that is not parallel to the tube hits the photo site, chances are the light ray will not get to the bottom of the tube and will not hit the sensing element. Therefore, the light coming from that light ray will be lost. It appears from this graph that when using large aperture lenses on Canon cameras, there is a substantial amount of light loss at the sensor due to this effect. In other words, the "marginal" light rays coming in at a large angle from near the edges of the large aperture are completely lost.
The article fails to mention the important fact that there are microlenses at the top of these "tubes" to funnel light into the sensing elements at the bottom. And no, "marginal" light rays coming at a large angle are not completely lost. As microlenses become more efficient at collecting light, so will the sensor.
The article fails to point out that the amount of light lost at the sensor is decreasing with each camera generation. It's right there in the charts, if the author had bothered to look. The 1Ds3 sensor has the same pixel pitch as a an EOS 20D sensor, but the 1Ds3 loses only 1/3 stop of light @ f/1.4 compared to 2/3 stops by the 20D. The pixel pitch of a 1D4 sensor is nearly identical to that of a 450D sensor, but the 1D4 loses only 0.4 stops of light @ f1.4 instead of 0.7 stops. There is progress is being made in minimizing the light loss at the sensor level, and so there is no reason for companies to stop producing ultra large aperture lenses.
I wish to re-emphasize that these issues apply only to the current crop of cameras with CMOS sensors
Odd that the article would be constantly laying the blame on the structure of CMOS sensors, when the very data quoted clearly indicates that the problem is far more severe with CCD sensors. The graphs in the article clearly shows that CCD sensors were losing significantly more light than CMOS sensors from the same generation. The author fails to mention that important fact, and instead misinterprets the data as a reason for the medium format backs and Leica to opt for CCD sensors.
I do not think the author even realized that a lot of the sensors he plotted on those charts are, in fact, CCD sensors. He keeps emphasizing that his article refers to CMOS sensors, when the charts he is using is showing a good mix of CMOS and CCD sensors. Is he not aware that the sensor for the A350, D200, D80, D70, D60, D50, and D40x are all CCD? Is he not aware that the bottom feeders on his charts are populated by these very same CCD sensors? And that the charts are clearly showing a greater efficiency among CMOS sensors?
-
Agreed but this was also a key sentence: This graph clearly shows that camera manufacturers “game the system” by increasing the ISO without the photographer's knowledge.
For those of us that like to know what is happening behind the scene, I found the article highly enlightening! And I agree with Mark’s idea of an open letter to resolve this if possible. But like world peace, unlimited free and clean energy and cameras that produce an open, non proprietary raw file, I think it unlikely we users will be heard. Still, I’m thankful for the article, I learned something.
Me too. Thanks Mark.
Andrew, the camera companies may not respond to this item - for some of them the culture is in the bones not to respond in public to very much at all. But behind the scenes they may well have a look at how they can address the basic issue with improved lens and sensor designs - as indeed some have been doing; then they will market these new designs with all the associated hype about the nature of the improvements, which is fine - gets us where we now know we want to go. I think the more important aspect of the contribution is that we - the customers and users - now know more about this than we did yesterday, and it helps us think about what to buy and how to set our cameras and lenses more intelligently than we knew yesterday.
-
I wonder how much Olympus' telecentric design reduces (or eliminates) the "lost light" effect with its fast (f2) 4/3 format lenses?
-
yup, to the point. Interesting as the article is, it is impossible to understand what the problem/issue is. The author seems to have found a slight discrepancy between the "real" and the "apparent" ISo and completly blown the causes out of proportion. Why the hell is he asking the manufacturers to answer him? The whole article sounds like the camera makers are somehow cheating on us, by not providing us with the last fraction of a stop of "real" iso. Come on Mr Dubovoy, a good journalist is not the journalist who always finds something to critizise. An objektive article simply describing the technical background would have been enough. No assumptions, no complaining about irrelvant performnce differences. Really, these days I dont want to be a camera manfacturer with all those experts on the web constantly complaining about issues which dont exist.
-
I would like to know if this tactic is used on cpu lens only or on any optic. I use an Olympus E-p1, primarily with non-cpu Zukios. During live view the camera adjusts the aperture on the system lenses as the brightness in the scene changes but cannot when a non-cpu lens is attached ;D If one where to repeat these test with non-cpu lenses, would the camera still adjust iso? Does entering the lens data as on a Nikon change this?
I'd like to know this too and also whether the attenuation is measured in the center, averaged over the whole frame or via a center-weighted average? I would say it has to be in the center, otherwise the attenuation would also have to be lens/focal length dependent.
-
The meaning of ISO or EV sensitivity of digital sensors is a bit murky. Do you define proper exposure as filling the sensor charge capacity to a certain level? Then a sensor with a lower full-well capacity would have a higher ISO rating, which seems wrong. Do you base ISO rating on sensor sensitivity, ignoring full-well capacity? Then you cannot properly equate exposures (shutter speed and f/stop and ISO) across different sensors. Given this murkiness, comparing EV settings across cameras introduces a lot of questions, unrelated to lens light loss at large apertures.
If the subject of interest is light loss on digital sensors with large aperture lenses, why not plot sensor response (image brightness corrected for any gain tricks) as a function of lens aperture? Do this for a variety of lenses to see the effect of lens design, and do this for a variety of sensors to see the effect of sensor architecture.
-
Both vertical and horizontal axis titles for the last graph were lost in transit. Anyone care to decipher those?
While it was an interesting article, it reminds me of the definition of kilo in computer terms: 1 kilo is defined as 1,024 by everyone in IT, except hard disk manufacturers who define it as 1,000. The marketing benefit of that is an exercise left to the reader.
-
It doesn't look like the author knows how to interpret the very data he is quoting in his own article.
Hey, no reason to get worked up, it is well-known that we should not trust this site on technical details, all the points you point out are so obvious that it almost feels like Mark and Michael intentionally play this game, as if they wanted to say, we do not care about the details, that is below our attention level or as if they lived in alternative universe where intent and basic idea is everything and execution irrelevant.
What we have learned from this article is that on digital sensors the effects that cause them to vignette more than film (ie, acceptance angle) also cause some measurable 'vignetting' in the center of the lens (I am more and more convinced that these data have to come from the center or at the very least from a central area) and the that the camera manufacturers 'compensate' for this by increasing the ISO as a function of f-stop (and so few cameras are shown in this last graph that it has little practical value and merely generally proves the point).
This is good to know and we should simply ignore all the other obfuscation and declarifications in the article.
-
The meaning of ISO or EV sensitivity of digital sensors is a bit murky. Do you define proper exposure as filling the sensor charge capacity to a certain level? Then a sensor with a lower full-well capacity would have a higher ISO rating, which seems wrong. Do you base ISO rating on sensor sensitivity, ignoring full-well capacity?
If you want to compare cameras (in a somewhat rigorous way as DxO does), you have to use some definition. Whether it is a certain percentage of the full well capacity or the median between a certain noise level and full well capacity does not matter, you just have to do it consistently.
And any ISO amplification factor defacto reduces the full well capacity, thus any ISO higher than base ISO is only higher because you have 'reduced' the full well capacity.
-
Both vertical and horizontal axis titles for the last graph were lost in transit. Anyone care to decipher those?
Same as the other graphs, x-axis is f-stop, y-axis is delta EV stops.
-
If the subject of interest is light loss on digital sensors with large aperture lenses, why not plot sensor response (image brightness corrected for any gain tricks) as a function of lens aperture? Do this for a variety of lenses to see the effect of lens design, and do this for a variety of sensors to see the effect of sensor architecture.
I also have this concern. All of the graphs presented in the open letter are based on a single lens/aperture. I would like to see comparisons among lenses with similar focal lengths/focal length ranges to determine if smaller-aperture lenses do, in fact, lose less light. I'm not convinced that they do. The only way in which a smaller-aperture lens, at the same focal length, could lose less light at the sensor is if the light is more collumnated in the smaller-aperture lens. One could argue for this, by arguing that a smaller pupil is going to cut off light coming in from more extreme angles. However, this loss of off-axis light is going to be balanced by light loss due to increased diffraction from a smaller aperture. I don't know the balance between the two sources of light loss, but it's an easy empirical question to answer.
-
Same as the other graphs, x-axis is f-stop, y-axis is delta EV stops.
Horizontal is pixel pitch on others, but thanks :)
Pretty modest increases, but ~0.4 EV increase in ISO from f/1.8 to f/1.2 might make a significant difference in IQ at very high ISOs. Probably a non-issue below 1600 ISO with most modern DSLRs.
-
Maybe I'm missing the point of this article (it is a little thick to get through), but it seems to me he's missing the point entirely.
My reference points that seem in line with his theme are the 85mm f/1.2, and f/1.8, which I have had both. Saying that "the f/1.2 offers one more stop" is not doing it justice.
That is like saying "this Mercedes has only 40 more horsepower than this Ford, so it isn't worth it."
It's not just about f/stops. The 1.2 is a completely different lens aimed at a different market (i.e. one willing to spend $2k for a lens). It seems to me the 1.2 is the finest 85mm Canon knows how to make, regardless of price. The 1.8 is a great lens at the price point.
I have never heard anyone say that Canon/Nikon/Leica rip off their customers. Seems to me that this article is saying you'll achieve just as good results with the 1.8 as the 1.2, so why bother (i.e. you are being ripped off)?
Compare images with both lenses and you'll easily see why one would bother, assuming you have the $$ required.
-
What was the old saying we had in the film days? Something like f8 and be there. While there may be something to all this, does it really have any practical implications over 99% of the images we capture? As long as the histogram on the camera is a fair representation of what is going on and the lenses do not have any focus aberrations, I suspect all is well in the world. We already know about the diffraction problems at small f stops (in contrast to what we used to do with film to get the great depth of field, f64 anyone?) and now there are some relatively minor issues at the other end. I guess we should go back to the aphorism I already stated, set your camera to f8 and be happy (with some exceptions).
Alan
-
Well, "interesting" article, but very difficult (for me anyway) to evaluate as written. Is it possible to get more detail on how these measurements were performed? Notably:
- Light loss at the sensor
- The "sensor gain function"
A search on the DxO website gave me no hits, and without actually knowing how the measurements were measured, the article is difficult to place in context. E.g., on the face of it, the Nikon data seems to contradict the whole CMOS versus CCD argument; the D50/D70 series (CCD sensor, if I recall correctly) is worse than the D300 series (CMOS sensor). In fact, if I look at the Nikon data, I'd say the cameras with CMOS sensors show only a small difference (about .2 EV) versus pitch. Now maybe that discrepancy is due to other effects, but then whatever the other effects are would have to be taken into account in the Canon comparison.
Sandy
-
In fact, is not even clear that large aperture lenses will deliver a shallower depth of field as intended.
Can someone explain why the circle of confusion of a large aperture lens is impacted by the apparent higher iso of the sensor.
The first item (loss of 0.5 EV due to T stop) is probably true, nice to know but not significant for most shooting. However the DOF story in my mind is unfounded and FUD.
-
As long as the histogram on the camera is a fair representation of what is going on ...
Its not, by a lot (unless you are only concerned with the JPEG and not the raw data itself).
-
Its not, by a lot (unless you are only concerned with the JPEG and not the raw data itself).
Actually on my Nikon D300 it's pretty darn good and useful in determining whether the correct exposure has been obtained. I realize that it's from the JPEG but it is the one tool we have to gauge the exposure without bracketing every shot. When I've done bracketing, it's usually the first exposure that's correct, but maybe I have the one camera in a 1000 that gives this result.
-
Actually on my Nikon D300 it's pretty darn good and useful in determining whether the correct exposure has been obtained. I realize that it's from the JPEG but it is the one tool we have to gauge the exposure without bracketing every shot. When I've done bracketing, it's usually the first exposure that's correct, but maybe I have the one camera in a 1000 that gives this result.
How are you determining its correct?
-
What I want from the exposure index (so-called ISO settings) on a camera is that when I choose f-stop, shutter speed on the basis of the ISO speed and my light measurements, I get correct exposure levels. What Canon and Nikon seem to be doing is ensuring exactly that, correcting for a well-known problem of most sensors with micro-lenses: their sensitivity is distinctly less to light that comes in at a highly off-perpendicular angle, which happens with the light at the outer edges of the very broad light cone from a very low f-stop. It was discusses years ago that microlenses limit the speed advantage of f-stop reductions once one goes much below f/2, and certainly below f/1.4.
To put it another way, the camera is simply adjusting the amplification needed to adjust for the fact that the sensitivity of the sensor actually declines as the f-stop gets very low, and so maintaing constant true sensitivity/exposure index. DxO has a controversial way of measuring "true ISO exposure index", as has been discussed in other threads.
So on the one hand, Canon and Nikon are doing exactly what should be wanted by anyone who wants to make precise manual exposure decisions but knows only the f-stop, not the effective T-stop of a particular lens+sensor combination.
On the other hand, this does show that somewhere between f/2 and f/1.4, further speed gains from reducing f-stop are mostly lost to this "microlens vignetting".
And to respond to a recent post: the DOF reduction is partly or mostly lost for exactly the same reason: it is the extra light coming in near the edges of the larger light cone of a lower f-stop that increases OOF effects, so if the sensor is not detecting much of that extra light when you open up beyond about f/1.8 or f/1.4, the DOF reduction is also mostly lost. By the way, OVF image brightness stops improving even earlier, about f/2 or above, for other reasons, to do with the ground glass of an OVF.
-
without actually knowing how the measurements were measured, the article is difficult to place in context. E.g., on the face of it, the Nikon data seems to contradict the whole CMOS versus CCD argument; the D50/D70 series (CCD sensor, if I recall correctly) is worse than the D300 series (CMOS sensor).
Well, how do you think they would measure it? You provide a target of known illumination, take pictures at various f-stops and compare the raw data, if no effects are there, opening the f-stop by one stop should give you twice the value in the raw data.
And of-course this CCD vs. CMOS argument is very dubious, as the data contradict it. But any argument LL makes that makes MF appear in a more positive light has a high potential to be dubious. Not to say also irrelevant in this case as there is no MF lens faster than f/2 anyway.
-
Can someone explain why the circle of confusion of a large aperture lens is impacted by the apparent higher iso of the sensor.
The first item (loss of 0.5 EV due to T stop) is probably true, nice to know but not significant for most shooting. However the DOF story in my mind is unfounded and FUD.
The long tube effect of the individual sensel has the same effect as a physical aperture (not in a perfect manner as it is in wrong position), thus in effect what Marc describes here is actually a smaller f-stop (except that is in the wrong position and thus behaves not really the same as a normal aperture but it behaves to some extent similarly).
-
What I want from the exposure index (so-called ISO settings) on a camera is that when I choose f-stop, shutter speed on the basis of the ISO speed and my light measurements, I get correct exposure levels. What Canon and Nikon seem to be doing is ensuring exactly that, correcting for a well-known problem of most sensors with micro-lenses: their sensitivity is distinctly less to light that comes in at a highly off-perpendicular angle, which happens with the light at the outer edges of the very broad light cone from a very low f-stop. It was discusses years ago that microlenses limit the speed advantage of f-stop reductions once one goes much below f/2, and certainly below f/1.4.
To put it another way, the camera is simply adjusting the amplification needed to adjust for the fact that the sensitivity of the sensor actually declines as the f-stop gets very low, and so maintaing constant true sensitivity/exposure index. DxO has a controversial way of measuring "true ISO exposure index", as has been discussed in other threads.
So on the one hand, Canon and Nikon are doing exactly what should be wanted by anyone who wants to make precise manual exposure decisions but knows only the f-stop, not the effective T-stop of a particular lens+sensor combination.
On the other hand, this does show that somewhere between f/2 and f/1.4, further speed gains from reducing f-stop are mostly lost to this "microlens vignetting".
I don't think that reducing the effect of one full stop (when going from f/2 to f/1.4) by a third to two-thirds of a stop (for the FF cameras) could be cold as 'mostly', 'mostly' would mean by more than half. But for some of the older and crop-cameras cameras it is can be called mostly as the effect is more than half of a stop.
And the proper thing to do would be to adjust the displayed f-stop, ie, you set your lens to f/1.4 and it reads f/1.5 or f/1.8 (but that would confuse a lot of people, as the same lens would show different values on different cameras).
And the interesting question is what the cameras do with non-CPU lenses.
-
How are you determining its correct?
Andrew,
Two ways. First was to use a standard print designed for printer testing coupled with the X-Rite Passport (I used the print that Jack Flesher created that posted on the Outback Photo site (http://www.outbackprint.com/printinginsights/pi049/essay.html). Exposures were done under daylight conditions (no direct sun on the target as it was a slightly overcast day). I used a 60mm macro lens and kept the aperture at 5.6 and bracketed by changing the exposure time (ISO kept at 200). I had a series of images that could be evaluated both visually and via the histogram rendered in Lightroom and compare it to what was rendered in the camera. In this case, it was the standard (unbracketed) that was closest. Second test was to take bracketed pictures of a rose in bloom. Same thing here, the initial camera reading was the best image both visually and via the histogram.
Alan
-
I had a series of images that could be evaluated both visually and via the histogram rendered in Lightroom and compare it to what was rendered in the camera. In this case, it was the standard (unbracketed) that was closest. Second test was to take bracketed pictures of a rose in bloom. Same thing here, the initial camera reading was the best image both visually and via the histogram.
So you matched (as closely as possible) the histogram on the camera to the histogram in LR based on various settings (I would assume mostly Exposure) on the various brackets until the two histograms matched?
In terms of the “over exposed” but “normalized* images in LR, were they able to be matched?
*Normalized based on the article here on LL: http://www.luminous-landscape.com/tutorials/expose-right.shtml
-
And the interesting question is what the cameras do with non-CPU lenses.
I have three 'old' Nikkor lenses that I had adapted to fit my D300. The f stop registers fine and is properly shown, but of course cannot be altered by the camera during exposure. Whether anything else is going on is anyone's guess.
-
So you matched (as closely as possible) the histogram on the camera to the histogram in LR based on various settings (I would assume mostly Exposure) on the various brackets until the two histograms matched?
In terms of the “over exposed” but “normalized* images in LR, were they able to be matched?
*Normalized based on the article here on LL: http://www.luminous-landscape.com/tutorials/expose-right.shtml
The match is between the normal exposure and 1/3 over exposure which is why I find that only under strange light conditions that I need to expose to the right more than 1/3 of a stop. I would estimate that 80% of the time I do not need to change from the original reading. Again, this is with my camera and others may have different experiences.
-
Assuming the cameras are making these amplification decisions based on the reported aperture of the lens, a simple way to test it would to be to take an adapted non-OEM lens with a programmable EXIF chip and compare images taken under identical circumstances with the chip set to report F1.0, F1,2, F1.4, etc. - the images should get progressively darker down to a plateau, right?
-
Hello everybody.
Thank you for reading the open letter and thank you so much for all your comments.
Let me start with an OOOOPS, I apologize! When the issue that prompted this open letter came about, it was because of a discussion of the structure of CMOS sensors. The whole discussion focused on these structures, and once I received the final data it escaped me that some of the cameras on the chart had CCD sensors. I apologize for any confusion this oversight might have caused.
The whole point of the letter though, remains intact, which is that the light loss depends on the structure of the sensor, and that light loss at the sensor is a very real phenomena.
On a related topic, and to respond directly to one of the comments on the forum: The mention of Medium Format has nothing to do with trying to put MF in a better light or to try to imply that somehow CCD sensors are better than CMOS. Quite the contrary, the mention of MF was meant to point out that the MF back manufacturers face an even bigger problem, which has forced them to produce structures that are quite different. My understanding from talking to the CTO's of the major MF companies is that at this point in time they cannot produce a CMOS sensor that will work for MF backs. While the current CCD structures work well with camera movements, there is no free lunch and they require compromises in other areas, for example, MF CCD backs have a huge negative, which is that Live View is not possible. Personally, I wish I could have a MF back with Live view today! There are other drawbacks to CCD's, but let us not digress.
Let me also respond directly to the comment that the data cannot be found on the DxO site. There is a new release of the DxO site coming in a few weeks. It will contain a lot of new data, including the data that support this open letter.
While there has been some discussion on the depth of field issue, l would like to repeat what I said in the open letter, which is that DxO is in the process of performing thorough focus measurements. Once the final focus measurement data is available, we will be able to put this issue to rest.
Finally, I hope that all of you understand that the reason this was written as an open letter as opposed to an essay is that there are a number of open questions. I asked some of them, and the participants in the forum are asking many other valid questions. We at The Luminous Landscape would like to have an open and constructive dialog with the camera manufacturers to better understand what is going on "under the hood", and to better understand what we can expect from our tools as we use them.
It is my hope that we will receive some interesting feedback from some of these companies. If we do, we will definitely share it with you.
Mark Dubovoy
-
Well, how do you think they would measure it? You provide a target of known illumination, take pictures at various f-stops and compare the raw data, if no effects are there, opening the f-stop by one stop should give you twice the value in the raw data.
Well, no. That doesn't tell you (a) what impact the lens only has independent of the sensor and (b) what impact data processing the camera has. Be aware that the "raw" data in images files, especially for CMOS sensors, is nowhere near the actual sensor data. It's already been been processed to remove noise, dark current, etc, etc
Hopefully the additional info that Mark says in his post above will soon be available will clarify this.
Sandy
-
The long tube effect of the individual sensel has the same effect as a physical aperture (not in a perfect manner as it is in wrong position), thus in effect what Marc describes here is actually a smaller f-stop (except that is in the wrong position and thus behaves not really the same as a normal aperture but it behaves to some extent similarly).
Since the optical circle of confusion is spread over multiple sensels I still do not understand how the long tube effect of one individual sensel inpacts this. So I can understand how the effect results in less light being recorded (hence larget difference between T and F stop) but not how it impacts the DOF.
-
Since the optical circle of confusion is spread over multiple sensels I still do not understand how the long tube effect of one individual sensel inpacts this. So I can understand how the effect results in less light being recorded (hence larget difference between T and F stop) but not how it impacts the DOF.
I understood it so, that light beams coming from the outer part of the lens get more diminished at the sensor than the light beams coming from the central part, because of the angle towards the sensel tube. Therefore adding an F-Stop speed from a fast lens, lets say from f2.0 to f1.4 doesn't add proportionally more light. So - to get a correct +1EV exposure you'd need either longer exposure or ISO gain, since the additional F-Stop doesn't deliver the full amount of light. Shortly: the light from the center of the lens contributes dysproportionally more to the exposure than the light from the "outer rim" ... This appears similar to the leaf shutter effect at short exposure times. This changes the behavior of the transition from sharp to unsharp zones of the image and is one of the explanations, why digital images have a much harder transition from sharp to unsharp zones, as opposed to film. Modern lens design is another factor.
Cheers
~Chris
ADDENDUM: I made a mistake up there, since actually the described effect should make the sharpness transition smoother, not harder. I just don't want to edit above for accuracy of the record. Sorry for the confusion ... Now I myself have some more questions....
-
Hello,
There are so many glaring flaws in that logic. The data clearly shows that ultra large aperture lenses still collect more additional light than is lost by the sensor.Going from f/2.0 to f/1.4 may not yield the full stop of light we might expect but it still yields a net benefit of 2/3 stops of light.
The data is explicitly showing that.An f/2.0 lens does not yield "virtually the same" results a f/1.4.
Regards,
Ali.
-
[...] MF CCD backs have a huge negative [...]
you mean, they shoot film too?! minted!
sorry, couldn't resist...
I'd like to know [...] whether the attenuation is measured in the center, averaged over the whole frame or via a center-weighted average?
i wondered about this too, details from mr dubovoy would be appreciated.
-
Hello everybody.
Thank you for reading the open letter and thank you so much for all your comments.
Let me start with an OOOOPS, I apologize! When the issue that prompted this open letter came about, it was because of a discussion of the structure of CMOS sensors. The whole discussion focused on these structures, and once I received the final data it escaped me that some of the cameras on the chart had CCD sensors. I apologize for any confusion this oversight might have caused.
Hi Mark,
Honestly though, that was not the only oversight. The most glaring omissions are the fact that microlenses reduce the sensitivity to angular rays being blocked by the exposure gates. In fact the microlenses are so effective that the light fall-off is less than with film!, What's more, CMOS devices currently are often used with lenses of a retrofocus design (for shorter focal lengths), and the mirrorbox dimensions/depth limits the maximum angle of incidence.
BTW, you also didn't mention in your introduction that the difference between F-stop and T-stop is caught by the internal exposure meter, and that wider aperture lenses exhibit more vignetting when used wide open (which will be compensated for by the internal exposure meter). I admit that it something that needs to be compensated for when using manual exposure settings, just like one needs to compensate for shorter focus distances (= larger magnification factors, bellows factor).
There are special considerations when the physical sensor array gets larger, and especially when lens shift is introduced. That makes fixed microlens designs (where the lens flange or rather the exit pupil is assumed to be in a fixed position and microlenses can be offset towards the corners) for these scenarios suboptimal, and as such introduces the angular masking issue (including color cast).
The only difference between CCDs and CMOS devices in the context of masking, is that a CMOS device has relatively more surface area that's opaque to incident light, but that doesn't say much about the sensitivity to angular effects. In fact as the presented data shows, CCDs seem to be more sensitive to angular effects. There does seem to be a trend that smaller sensel pitches suffer more from exposure gate masking effects than the larger pitches do, which seems logical from a geometric standpoint (smaller surface area means larger percentage shadow at a given depth).
The whole point of the letter though, remains intact, which is that the light loss depends on the structure of the sensor, and that light loss at the sensor is a very real phenomena.
Well, other than reducing the quantum efficiency of the detector which is a fact of life, it can only be improved by using back-illuminated sensors and improved exposure gate designs. An interesting observation is that apparently, according to DxO, there is a gain compensation based on aperture value setting. That's something that will need to be verified.
Cheers,
Bart
-
Sandymc (one of the correspondents above, and the developer of Cornerfix, which corrects both luminance and color-shift vignetting) surely understands what is missing from this article. But he is tactfully suggesting that we wait for more information from DXO. Fast lenses vignette, so it is important to know how the intensity that is received at the sensor varies as a function of distance from the center of the image. Manufacturers who have announced that they correct for vignetting (only Leica, AFAIK) apply a correction which increases as you move from the center of the image to the outside. In Leica's case, they appear to provide this correction for an aperture value about two stops below wide open, and let the corners darken at wider apertures. (My observation from a few experiments -- Leica doesn't comment on this.) This is a sensible approach, since the color shift is annoying but aperture independent, while luminance vignetting often improves a picture by solidifying the edges and is strongest at wide apertures. Anyway, summarizing the effects as a T-stop to F-stop difference leaves a lot of questions unanswered. It is worth digging into what corrections manufacturers slip in but the full answer requires a curve for each lens and maybe for each of several apertures, and can't be reduced to a single data point.
scott
www.pbase.com/skirkp
www.cs.huji.ac.il/~kirk
-
Until I read this, I thought that I was crazy for noticing that the shutter speed for a given photo was faster when using a prime than a zoom (at the same focal length.) i.e. using a 50mm f/1.4 at f/4 resulted in a faster shutter than a zoom at 50mm that was stopped down to f/4.
Now I know that I'm not quite as crazy as I thought I was :)
-
Until I read this, I thought that I was crazy for noticing that the shutter speed for a given photo was faster when using a prime than a zoom (at the same focal length.) i.e. using a 50mm f/1.4 at f/4 resulted in a faster shutter than a zoom at 50mm that was stopped down to f/4.
I would say that the effect you observed is due to the transmission losses of the zoom lens, since it has more elements than a prime. Both lenses at F/4 will have different T stops. I don´t think the sensor issue explained here is doing anything in your case
-
Leaving aside the technical details and the logic lapses in the main article, isn't this a storm in a teacup? There are dozens of things that are in a way or another normalized for the comfort of photographers and to support legacy hardware. ISO really isn't anything more than a convenient abstraction and we already know different manufacturers ISOs are not really equivalent. Incidence issues are well understood: microlens geometry and lens design have already evolved.
Manufacturers automagically play with the whole process in order to extract the best out of their sensors and stay as close as possible to the legacy photographer's assumptions. So what? At another level, this isn't different from "discovering" they are "cheating" on real pixels counts by using bayer matrices.
AFAIC, I wouldn't mind having a slider for aperture and a slider for gain and some presets.
-
Well, I was going to spout off in detail here but it appears that a large number of readers have already nailed this down for what it is. A tempest in a tea-cup. The article is a bit long winded, uninformative and poorly explains the fact that a digital camera sensor is not a piece of film and as such doesn't behave the same way film does, and intimates that the camera manufacturers are all in a conspiracy to keep important information from the consumer. Knowing the gain curve for a particular camera isn't high on my list of priorities in photography.
-
The whole point of the letter though, remains intact, which is that the light loss depends on the structure of the sensor, and that light loss at the sensor is a very real phenomena
Mark,
Thank you for posting a very informative article. Your use of T-stops actually impaired my comprehension of the article, since T-stops deal with the transmission of light by the lens and do not vary with the aperture, whereas the tunneling effect deals with loss of light at the level of the sensor, and this loss is unlike vignetting, since it also occurs in the central part of the field of view. I was not aware of this effect.
Finally, I hope that all of you understand that the reason this was written as an open letter as opposed to an essay is that there are a number of open questions. I asked some of them, and the participants in the forum are asking many other valid questions. We at The Luminous Landscape would like to have an open and constructive dialog with the camera manufacturers to better understand what is going on "under the hood", and to better understand what we can expect from our tools as we use them.
It is my hope that we will receive some interesting feedback from some of these companies. If we do, we will definitely share it with you.
Yes, right now your essay raises more questions than it answers, but it is a good starting point for additional study and discussion. I would be interested in getting more information on the geometry of CCD and CMOS sensors. As Bart pointed out, CCDs also suffer from this problem, which was particularly acute with the Leica M9 in which the flange to sensor distance is small since there is no mirror box requiring the use of retrofocus lens designs necessary for SLR cameras.
Back illuminated CCDs are available in specialized sensors used for science and space applications (as from Fairchild (http://www.fairchildimaging.com/products/fpa/ccd/area/ccd_486.htm)) and these would greatly reduce the tunneling effect, but I am not aware of their use in general purpose dSLRs or MFDBs. Sony (http://www.sony.net/SonyInfo/technology/technology/theme/exmor_r_01.html) has introduced a back illuminated CMOS for compact cameras, and the graphics in the link demonstrate the geometry of both front and back illuminated CMOS sensors.
-
On a side note: Not being an optical technician or a physisist I cannnot comment on the viability and reliability of the measurment techniques used by DXO. I do have serious doubts about their reliability, however. There are many instances (far more than with other test sites) where their measurements have absoltly nothing to do with practical results. In fact, they are so far off, that its almost ridiculous. Just take the DXo resolution measurements of the Canon 16-35 mm lens, for example. At 16mm the DXO graph shows that the lens' edge and corner reslution decreases when stopping it down. DUDE! Just shoot the lens in real life and report back. Its certainly not a perfect lens and the edges/corner are far from optimal even stopped down. But the corner resolution at f/8 is certainly better than at f/2.8. No question. What a joke.
Same goes for their camera sensor maesurments. Lookin at real life pics and prints from differnt camera models I often find that their rankings are just upside down.
I am therfore very critical about the findings in the current article.
-
Read this interesting comment on another forum http://www.dvxuser.com/V6/showthread.php?227234-Small-bomb-from-LL-and-DxO&p=2142565&viewfull=1#post2142565 (http://www.dvxuser.com/V6/showthread.php?227234-Small-bomb-from-LL-and-DxO&p=2142565&viewfull=1#post2142565)
I've wondered for years (even before digital) why we still use the f/stops and shutter speed combos that we do, figuring that it was just historically convenient.
-
Reading the other forum comments,
The author seems to ignore the autoISO function in some cameras that allow you to set a fixed aperture and a shutter speed and the camera automatically adjust the ISO.
Somewhere down that thread, another poster makes a wrong claim about how the White Balance works, suggesting that there is a gain adjustment in each channel (RGB). White Balance does not work that way, it scales the numerical values of each channel in post processing.
I think that an important issue with varying ISO is the effect on Noise and Dynamic Range. Increasing ISO on any sensor will increase noise and reduce DR.
-
An interesting observation is that apparently, according to DxO, there is a gain compensation based on aperture value setting. That's something that will need to be verified.
An quick evaluation (EOS 1Ds3 + EF 85mm f/1.2 L II), using IRIS software to study the Raw data before demosaicing in linear gamma space, indeed suggests a minute increase in read-noise at apertures wider than f/2. Whether this is also true for other lenses remains to be seen.
The increase in read-noise is a function of gain, and appears to be in the order of a fraction of 1/3rd stop boost. More rigorous testing is needed, but I wanted to share the initial/preliminary finding to allow the discussion to progress on a more fact based second opinion, rather than on speculation or hearsay alone.
Cheers,
Bart
-
Well, no. That doesn't tell you (a) what impact the lens only has independent of the sensor and (b) what impact data processing the camera has. Be aware that the "raw" data in images files, especially for CMOS sensors, is nowhere near the actual sensor data. It's already been been processed to remove noise, dark current, etc, etc
The fact that the Canon data are shown for f/1.2 and the Nikon data for f/1.4 clearly shows that actual lenses were used. And of course, the results will vary from lens to lens, there is nothing you can do about it. Unless there is an agreed ideal model how a f/1.4 lens should look like (and somebody builds such a lens) you always have to deal with real lenses. As long as the difference between several real f/1.4 lenses is noticeably smaller than the difference between f/2 and f/1.4, you can conclude that the average f/1.4 data are somehow representative of the 'f/1.4 effect'.
-
Agreed but this was also a key sentence: This graph clearly shows that camera manufacturers “game the system” by increasing the ISO without the photographer's knowledge.
For those of us that like to know what is happening behind the scene, I found the article highly enlightening! And I agree with Mark’s idea of an open letter to resolve this if possible. But like world peace, unlimited free and clean energy and cameras that produce an open, non proprietary raw file, I think it unlikely we users will be heard. Still, I’m thankful for the article, I learned something.
Me too. Thanks for the info, Mark!
Mike.
-
Folks:
Since there has been some discussion on vignetting, and some postings about microlenses near the edges of the frame, I would like to point out that the DxO measurements were performed at the center of the frame.
There have also been some questions about how DxO performed the measurements, so I think that the following direct quote from the technical experts at DxO should clarify things for everybody:
First, we measure the ISO sensitivity of a sensor without optics, then since all sensors we currently measure are linear (or can be linearized with a simple “Dark value” subtraction) we can use the sensor as a photometer with an optics between the sensor and very stable luminous surface which luminance is precisely measured with a calibrated luminance-meter. It is then easy to compute the actual T-Stop of the lens and sensor combination: one part of the loss comes from then lens itself, another part is due to the sensor inability to capture light from some angles. That measurement is performed at image center, what happens in the field is a different measurement, available on www.dxomark.com as “vignetting”.
Mark Dubovoy
-
Folks:
Since there has been some discussion on vignetting, and some postings about microlenses near the edges of the frame, I would like to point out that the DxO measurements were performed at the center of the frame.
There have also been some questions about how DxO performed the measurements, so I think that the following direct quote from the technical experts at DxO should clarify things for everybody:
First, we measure the ISO sensitivity of a sensor without optics, then since all sensors we currently measure are linear (or can be linearized with a simple “Dark value” subtraction) we can use the sensor as a photometer with an optics between the sensor and very stable luminous surface which luminance is precisely measured with a calibrated luminance-meter. It is then easy to compute the actual T-Stop of the lens and sensor combination: one part of the loss comes from then lens itself, another part is due to the sensor inability to capture light from some angles. That measurement is performed at image center, what happens in the field is a different measurement, available on www.dxomark.com as “vignetting”.
Mark Dubovoy
Mark,
I'd have at least two concerns with that:
- Firstly, using that procedure, DxO would have to have tested a large number of samples of each individual camera, else all you're looking at is sample to sample variation.
- Secondly, I think there's a logical inconsistency between the measurement process and the conclusions made. If you're using the camera as a photometer, to believe you're getting valid data you have to assume that the in-camera processing isn't changing your measurement from exposure to exposure. But the conclusion of the article is that the in-camera processing does change from exposure to exposure, specifically with aperture! Now perhaps you can argue that processing only changes with aperture, and design the test round that assumption, if you were confident enough that aperture is all that's driving the changes. But if there is in-camera processing occurring as claimed in the conclusions, the initial assumption required for the data to be valid starts to look really shaky.
So going from an interesting factoid (the first graph in the article) to the conclusion as to CMOS sensor design starts to feel to me like a huge leap....
Sandy
-
I would like to address the last set of comments on this thread:
I am not sure where the concept of using "the camera with in-camera processing" came from. Please read the DxO statement carefully, they measure THE SENSOR, and they use THE SENSOR as a photometer.
Similarly, regarding the comment about sample to sample variation: Professional testing labs always work using enough samples and scientifically proven statistical methods before they publish any measurements or conclusions.
I believe that we have reached the point where we have beaten this horse to death, so this will be my last post on this thread.
I sincerely hope that all of you have learned something useful from this discussion. I know I have.
May the photons smile on all of you.
:)
Mark Dubovoy
-
Hi Folks,
having read both Mark's essay and the responses with interest, it strikes me that there are two areas of contention: the "attittude" side, (i.e. does it matter?), and the technical side (i.e. is the data analysed and interpreted correctly?).
The first area, the "attitude," will always be personal, and people have to judge for themselves whether it matters to them. Hence I prefer not to comment on that.
However, regarding the technical issues, I would like to point out that Mark is not the first to describe the phenomenon that is subject of his open letter. With that I do not wish to take any credit away from Mark; he is the first (as far as I am aware) to describe the light loss backed by current and more extensive data.
I in no way endorse their system, or make any intimations regarding the overall quality of the results from their system, but the four-thirds consortium were making a lot of noise about light/ sensor issues. In fact, it was put forward as one of the main rationales behind the four-thirds design. They have gone a bit more quiet about it since advances in micro-lens design, but here is the still relevant version of the issue from them (half-way down the page): http://www.four-thirds.org/en/fourthirds/index.html
Hope this helps,
Chris
-
For me the light loss and possible hidden ISO change is not that much a possible issue,
but more the effect on the sharp-unsharp transitions of images, means the bokeh.
What I'd really like to see is an experiment of
- a test image of a motive with enough depth,
- taken with an open aperture, like F1.4
- with 2 sets of lenses: One digitars/modern lens, one old lens/designed for film
- and with film and sensor as media.
and this in a FF/35 mm format and in MF/MFDB format.
2/3 of an F-Stop more or less is a thing, one might care about or not,
but the rendering of the image / bokeh characteristics is something which will be seen, I believe.
A side by side test would be great for that.
Just my $ 0.02
-
I am not sure where the concept of using "the camera with in-camera processing" came from. Please read the DxO statement carefully, they measure THE SENSOR, and they use THE SENSOR as a photometer.
Mark,
Sorry, but I think you misinterpret what DxO are saying; unless both Canon and Nikon gave DxO cameras modified to allow direct access to sensor, DxO are looking at processed data. The best you can get with an unmodified camera is the so called "raw" image data, but that's already been processed extensively. And if by some chance Canon and Nikon did both provide all those models of cameras with special firmware or test connections, they sure didn't do so in sufficient numbers for a statistically valid sample; 20 or 30 samples per model and you'd be dealing with thousands of specially modified cameras! Canon and Nikon may well be super nice guys, but I have trouble believing they're that nice.
Sandy
-
Shortly: the light from the center of the lens contributes dysproportionally more to the exposure than the light from the "outer rim" ... This appears similar to the leaf shutter effect at short exposure times. This changes the behavior of the transition from sharp to unsharp zones of the image and is one of the explanations, why digital images have a much harder transition from sharp to unsharp zones, as opposed to film. Modern lens design is another factor.
Cheers
~Chris
Thanks for taking the time to explain the effect. I now understand it better, but also understand it has more to do with the shape of the bokeh, rather than the "exact dof" (I know "exact dof" is an oximoron).
However if the outer parts of the lens contribute less light vs. the center it reminds me of what the heritage Minolta (now Sony) 135 STF does, they built in a filter close to the aperture which is clear in the center and slowly gets denser as you move outward. This has the effect to make the bokeh really buttery smooth, so if a digital sensor used with a large aperture lens does the same thing (less contribution from the outer rims of the lens) why would the bokeh become harder rather than softer as happens with this STF lens.
Maybe I'm missing a trivial point, so pls. bear with me, I'm just trying to understand this whole story better.
-
Light coming from the lens and forming a point (ideally) has the shape of a cone with its base at the lens.
That is why wide angles are so sensible against defocusing errors.
The greater the angle at the tip of that cone is, the faster and intenser defocusing leads to blurring.
Light coming from the "outer rim", which is only revealed at open apertures, forms a cone with a greater angler than light coming from the central parts, which are always used, even at high F-Stops=small apertures.
The light forming a point comes from the central beam, beams from the outer rim and all between.
But: Due to the different angles, the light from the "outer rim", which is only there if the aperture is wide open,
is more sensible to the blurring effects of defocusing.
The light from the center is not.
Therefore - the larger the amount of light from the outer rim is, the faster defocusing effects become visible, and we see that as low DOF at open apertures.
Theoretically you could put a center filter with varying density (transparent to grey or grey to transparent) at the plane of the aperture and use this to manipulate the behavior of the bokeh. You could make the transition from sharp to blurred areas softer or harsher. I guess, the more light from the outer rim is there, the harsher the transition would be, the more from the center the softer.
To Pegelli:
ADDENDUM: Erm ... I just realized my error above ... Yup - the bokeh should be softer. But generally digital files seem to have a harder transition - Maybe its the thickness of the image plane film vs. sensor which comes into play ? Or is it the construction of the modern digital lenses ... ? Or is it "just" the generally smaller formats/sensor sizes?
Sorry for the mistake ... I'd really like to see a test and have some enlightenment from the tech nerds here ...
What remains for me is, that from experience there is a clear difference between the sharpness behavior of digital vs. film imaging and I thought (which I had to revise) the described problem could be an explanation. Now it appears the opposite is true and the light loss at sensor level is probably working against the otherwise harsh sharpness of digital systems. So - if I look at it from this side, this could actually be something good ...
-
They are testing the sensors in the camera. If they wanted to test the sensor outside the body, they would have to provide custom electronics anyway, and their results wouldn't be relevant to the camera anymore. What manufacturer extract from their camera depends a lot on the electronics behind the sensor (see for example the different behavior of the KAF-3200ME in the QSI and SBIG astro cameras).
Here an interesting link on light transmission.
http://www.cctv-information.co.uk/i/Light_Transmission_Through_Lenses
Not the final word, for sure, but I think some of us will be surprised by the numbers.
-
When the camera determines shutter speed and aperture, does it simply choose the nearest "standard" shutter speed and aperture or does it set them very precisely? E.g. if it determines that the correct aperture is f2.95 does it use f2.95 or does it use f2.8 ? My pentax cameras (film and digital) only ever display "normal" values for aperture and shutter speed. In which case you'll always be slightly under or over exposed anyway?
Anthony
-
I think Canons are limited to 1/3 stop changes. That was a point in the marketing material for new Sony lenses. I don't know if continuous changes of aperture are very good for videos though. Better adjust the gain automagically I think.
-
I'd like to say just one thing :
The author claims the light reduction is a result of the inclination of light rays, which gets higher with wide apertures.
Then grab a "usual" lens, meaning no telephoto, retrofocus or pseudo-telecentric one. A classic one. Let's take a 40mm or 50mm lens on a 24x36 sensor, maximum aperture 1.4.
I happen to have a 40/1.4 at home, i'll calculate on that basis, and I can even give you a graph (in French, but the only thing to know is that's real scale) :
(http://as.bz.free.fr//alpha/20101029_15h33m56s_00773.jpg)
Compute the angle of incidence of rays coming from the border of the cone for a pixel in the center. Approcimately 0.3 radians.
Compute the angle of incidence of rays coming to the border of the sensor when you are at f/22 (meaning the ray from the center of the cone). Approximately 0.55 radians.
Meaning IF there are some effects from inclination of rays on the center of the frame with large aperture lenses, those are really weak in comparison to effets on the border of the sensor whatever aperture you use.
(Don't imagine that sensors are made in a way which corrects that, apart from Leica M9 that's not the case).
And this is a very gentle inclination, try with some ultra-wide angle lens that isn't pseudo-telecentric and you'll see CRAZY inclination that even a f/0.5 lense won't provide.
-
Taking into account that every camera ends in a different exposure in the RAW file (in some cases differing by more than 1 or even 2 stops), for the same scene, metering, lens, shutter, aperture and ISO, I wonder why should be interesting finding out what camera vendors are doing to compensate for loss of light in the lenses by raising their ISO.
It's no problem at all, just take your sensor+lens, compare it to any other candidate, and choose the one performing best (noise) for your application. It's the only thing that really matters, not how accurate are the ISO values reported by the manufacturer, that are just a reference. We could in fact call them: first ISO, second ISO, third ISO,... without referring to any particular numbering, and nothing would change.
Much more interesting IMO is the DOF question. If a lens of a given maximum aperture is NOT producing the desired shallow DOF according to its f-number, then those manufacturers selling these lenses are cheating the buyer.
Regards
-
... grab a "usual" lens, meaning no telephoto, retrofocus or pseudo-telecentric one. A classic one. Let's take a 40mm or 50mm lens on a 24x36 sensor, maximum aperture 1.4.
Compute the angle of incidence of rays coming from the border of the cone for a pixel in the center. Approximately 0.3 radians.
Compute the angle of incidence of rays coming to the border of the sensor when you are at f/22 (meaning the ray from the center of the cone). Approximately 0.55 radians.
Meaning IF there are some effects from inclination of rays on the center of the frame with large aperture lenses, those are really weak in comparison to effets on the border of the sensor whatever aperture you use.
(Don't imagine that sensors are made in a way which corrects that, apart from Leica M9 that's not the case).
True: the jargon is that even the chief ray (middle of the cone) hits the corners of the frame at a highly off-perpendicular angle if the exit pupil is too close to the sensor, as it is with "classic" wide to normal lens designs.
... and avoiding this "microlens vignetting" towards the edges of the frame is why lens design for digital cameras that have microlenses on their sensors favors keeping the exit pupil high, meaning "near-telecentric" so that wide to normal lens designs are somewhat retro-focal, not "classic" designs. And this is why wide-angle Leica rangefinder lenses in particular have troubles near the edges: the classic range-finder "symmetric" wide-angle designs have lower exit pupils than SLR lenses, which are forced by the presence of the mirror box to have a higher exit pupil. And this is turn is why Leica first kept to a smaller sensor in the M8 (avoiding the corner problem) and then adopted off-set microlenses for the M9. Meanwhile, DSLR's also started with the "smaller sensor" solution, and according to Thom Hogan, Nikon at least had been moving to more telecentric lens designs from early in the digital era, even before it launched its first full 35mm format DSLR.
And software correction for this microlens vignetting is available: it basically does for the edges and corners what Canon and Nikon are doing across the entire frame: bringing the luminosity of to the level to be expected from the chosen combination of f-stop, shutter speed, and exposure index ("ISO").
By the way:
(Don't imagine that sensors are made in a way which corrects that, apart from Leica M9 that's not the case).
That is true in most cases, but the few medium format sensors I know of that have micro-lenses, all 44x33mm ones from Kodak, also use offset micro-lenses. See the note to Figure 6 on page 15 of http://www.kodak.com/global/plugins/acrobat/en/business/ISS/datasheet/fullframe/KAF-31600LongSpec.pdf So it seems that all medium format sensors are designed to avoid this microlens vignetting in order to accommodate all the "classic", not so telecentric medium format lenses in use: they do it by either omitting microlenses or offsetting them.
Actually, this might have changed recently, or be due to change soon: Dalsa has described a new type of microlenses that have a far wider acceptance angle and so avoid this vignetting problem, but I am not sure if any MF back on the market is using them.
-
Taking into account that every camera ends in a different exposure in the RAW file (in some cases differing by more than 1 or even 2 stops), for the same scene, metering, lens, shutter, aperture and ISO, I wonder why should be interesting finding out what camera vendors are doing to compensate for loss of light in the lenses by raising their ISO.
It's no problem at all, just take your sensor+lens, compare it to any other candidate, and choose the one performing best (noise) for your application. It's the only thing that really matters, not how accurate are the ISO values reported by the manufacturer, that are just a reference. We could in fact call them: first ISO, second ISO, third ISO,... without referring to any particular numbering, and nothing would change.
I agree, and I'd like to add that ISO have a definition which is independant from sensor sensibility... 100 ISO is what allows you to have a good exposition given aperture (well, T-number, more exactly) and given a certain amout of light.
Doesn't matter at all if the sensor sensibility is not linear with its ISO settings.
Much more interesting IMO is the DOF question. If a lens of a given maximum aperture is NOT producing the desired shallow DOF according to its f-number, then those manufacturers selling these lenses are cheating the buyer.
Regards
That's not an issue at all.
Take a lense that your reflex doesn't know (put some adhesive tape on your lens contacts ?), take a picture with a digital reflex and then with a film reflex, at f/8 -at least 2 stops smaller aperture than wide open to avoid inherent vigneting from inside obstruction-.
You'll see the amount of vignetting is virtually the same.
Yes, there might be some difference, maybe 10% because of the effect the author presents, or because film doesn't absorb light as easily if rays are highly inclined. But it will be nowhere near 1 EV.
And then remember what I said earlier : inclination of rays coming from the border of the cone with a large aperture lens is very small compared to inclination of rays that hit the border of the sensor, even at f/8.
So the effect on DoF is just negligible.
True: the jargon is that even the chief ray (middle of the cone) hits the corners of the frame at a highly off-perpendicular angle if the exit pupil is too close to the sensor, as it is with "classic" wide to normal lens designs.
... and avoiding this "microlens vignetting" towards the edges of the frame is why lens design for digital cameras that have microlenses on their sensors favors keeping the exit pupil high, meaning "near-telecentric" so that wide to normal lens designs are somewhat retro-focal, not "classic" designs. And this is why wide-angle Leica rangefinder lenses in particular have troubles near the edges: the classic range-finder "symmetric" wide-angle designs have lower exit pupils than SLR lenses, which are forced by the presence of the mirror box to have a higher exit pupil. And this is turn is why Leica first kept to a smaller sensor in the M8 (avoiding the corner problem) and then adopted off-set microlenses for the M9. Meanwhile, DSLR's also started with the "smaller sensor" solution, and according to Thom Hogan, Nikon at least had been moving to more telecentric lens designs from early in the digital era, even before it launched its first full 35mm format DSLR.
And software correction for this microlens vignetting is available: it basically does for the edges and corners what Canon and Nikon are doing across the entire frame: bringing the luminosity of to the level to be expected from the chosen combination of f-stop, shutter speed, and exposure index ("ISO").
Thanks for the jargon :)
As I said above, software correction is not that strong. I use RF lenses on my Nex -no software correction-, and I don't get awful vignetting, even with 0.35 radians inclination of chief ray at corners (and 0.6 radians angle from extreme ray of the cone) with a 40/1.4. It doesn't have any noticeable vignetting at f/8 -I'd dare to say I'll notice a 1/3 EV vignetting.
So, even with a cone whose chief ray inclination is 0.35 radians (higher than any f/1.4 lens can provide - apart from anti-telecentric ones ? ^^), the effect is small.
Software correction is useful for lenses vignetting, not "sensor vignetting", apart from ultra-wide non-telecentric lenses, at least.
By the way:That is true in most cases, but the few medium format sensors I know of that have micro-lenses, all 44x33mm ones from Kodak, also use offset micro-lenses. See the note to Figure 6 on page 15 of http://www.kodak.com/global/plugins/acrobat/en/business/ISS/datasheet/fullframe/KAF-31600LongSpec.pdf So it seems that all medium format sensors are designed to avoid this microlens vignetting in order to accommodate all the "classic", not so telecentric medium format lenses in use: they do it by either omitting microlenses or offsetting them.
Actually, this might have changed recently, or be due to change soon: Dalsa has described a new type of microlenses that have a far wider acceptance angle and so avoid this vignetting problem, but I am not sure if any MF back on the market is using them.
Thanks for the medium format insight, I was just speaking of 24x36 and smaller, I think was right for these ? :)
I do not disagree that very high incidence angles are a huge problem with wide angle non telecentric lenses, the point I wanted to raise is that inclination from the border of the cone with wide aperture lense -subject of this article- is really low compared to inclination on the corners of the sensor even with (non telecentric) normal focal at whatever aperture you want.
Anyone can see that a normal non-telecentric lense doesn't have a dramatic vignetting in the corners at say f/8 -even without software correction-, so the issue cannot be that important with fast lenses...
Edit : we (they - the manufacturers) began to think a lot about ray inclination in corners with 24 or wider mm lenses on M9, or wide angle and extreme shift/tilt/whatever with view cameras.
24mm on a M9 means corners receive rays at a 0.75 radian inclination... 15mm (the lenses which makes the real big problem rise with the M9) mean 0.97 radians !!!
To get a 0.97 radians incidence angle from a ray hitting the center of the sensor, you'll have to use a f/0.33 aperture lense... :) (f/0.53 for a 0.75 radian angle).
-
Much more interesting IMO is the DOF question. If a lens of a given maximum aperture is NOT producing the desired shallow DOF according to its f-number, then those manufacturers selling these lenses are cheating the buyer.
Regards
Sorry, I don't think anybody is cheating anybody else. No lens manufacturer sells lenses by dof, but by max aperture, which is a optical constant that is simply defined by the focal length devided by the pupil opening. You can't blame lens manufacturers that these lenses might have (in my mind very slightly different) dof or bokeh depending if they're used on film or a digital sensor.
-
To make it shorter :
Caring about rays inclination on f/1.2 lenses means caring about 0.35 radian angles.
Corner rays from a typical 40mm lense at f/11 have a 0.35 radians inclination on an APSC sensor.
Do you see any vignetting at f/11 with a 40mm lense on an APSC sensor ? No. Even without software correction.
So you won't see any effect on the border of your f/1.2 lense light cone.
NB : the 40mm lense is supposed to be non-telecentric.
Edit : we (they - the manufacturers) began to think a lot about ray inclination in corners with 24 or wider mm lenses on M9, or wide angle and extreme shift/tilt/whatever with view cameras.
24mm on a M9 means corners receive rays at a 0.75 radian inclination... 15mm (the lenses which makes the real big problem rise with the M9) mean 0.97 radians !!!
To get a 0.97 radians incidence angle from a ray hitting the center of the sensor, you'll have to use a f/0.33 aperture lense... :) (f/0.53 for a 0.75 radian angle).
-
I think some people have observed the secret ISO boosting discussed in this article first hand:
http://forums.dpreview.com/forums/read.asp?forum=1037&message=36770709
-
I think some people have observed the secret ISO boosting discussed in this article first hand:
http://forums.dpreview.com/forums/read.asp?forum=1037&message=36770709
In my opinion the assumption made here is a lot better than the one on the luminous landscape article.
I believe this exposure compensation specific to fast lenses only compensates for their own huge vignetting issues (from inner lense ray obstruction, the same that causes swirling bokeh), not for some sensor vignetting issue.
Could very well explain the exposure compensation, but not those first two DxO graphs. I'd like to have them with various lenses and various apertures.
-
Meaning IF there are some effects from inclination of rays on the center of the frame with large aperture lenses, those are really weak in comparison to effets on the border of the sensor whatever aperture you use.
(Don't imagine that sensors are made in a way which corrects that, apart from Leica M9 that's not the case).
And this is a very gentle inclination, try with some ultra-wide angle lens that isn't pseudo-telecentric and you'll see CRAZY inclination that even a f/0.5 lense won't provide.
Of course, nobody is saying this center 'sensel-vignetting' is nearly as bad as the wide-angle vignetting. The former is according to the data about between 1/3 and 2/3 of a stop. Vignetting numbers for some wide-angle lenses has been measured as high as 3 EV:
http://www.photozone.de/canon_eos_ff/506-zeiss18f35eosff?start=1
-
Caring about rays inclination on f/1.2 lenses means caring about 0.35 radian angles.
Corner rays from a typical 40mm lense at f/11 have a 0.35 radians inclination on an APSC sensor.
Do you see any vignetting at f/11 with a 40mm lense on an APSC sensor ? No. Even without software correction.
NB : the 40mm lense is supposed to be non-telecentric.
You might have a point, but I think you are making a false assumption which overestimates the corner "pixel vignetting": that there are "totally non-telecentric" 40mm SLR lenses with exit pupil height equal to that focal length of 40mm. That is not the way 35mm format SLR lenses work; certainly not Nikon F-mount ones because the flange is about 47mm from the focal plane, and so the exit pupil is higher than that: 55mm and up usually. See this table from 2003, and note that there has been a trend since then toward higher exit pupils: http://www.swissarmyfork.com/lens_table_1.htm
-
Bart said
An quick evaluation (EOS 1Ds3 + EF 85mm f/1.2 L II), using IRIS software to study the Raw data before demosaicing in linear gamma space, indeed suggests a minute increase in read-noise at apertures wider than f/2. Whether this is also true for other lenses remains to be seen.
I am a sucker for useless tests ;-0: same result here on the 85 1.2 and 50 1.4 on the 5DMKII
(4 secs dark frames at 400 ISO, separated by 20 seconds to let the sensor cool, in a dark room to avoid light leaks, analyzed in Iris with bgnoise, average of 2 runs, confirmed with Maxim)
aperture exposure 85mm 50mm
8 4 5,62 5,61
5,6 4 5,64 5,62
4 4 5,65 5,61
2,8 4 5,63 5,63
2 4 5,68 5,69
1,6 4 5,91 5,93
1,4 4 6,15 6,15
1,2 4 7,5
-
Bart said
I am a sucker for useless tests ;-0: same result here on the 85 1.2 and 50 1.4 on the 5DMKII
(4 secs dark frames at 400 ISO, separated by 20 seconds to let the sensor cool, in a dark room to avoid light leaks, analyzed in Iris with bgnoise, average of 2 runs, confirmed with Maxim)
aperture exposure 85mm 50mm
8 4 5,62 5,61
5,6 4 5,64 5,62
4 4 5,65 5,61
2,8 4 5,63 5,63
2 4 5,68 5,69
1,6 4 5,91 5,93
1,4 4 6,15 6,15
1,2 4 7,5
Hi Pierre,
Only useless if we fail to learn something from it ... ;)
Your noise statistics for the 5D2 show similar values as for my 1Ds3, although you took 4 second dark frames at ISO 400, while I took 1/8000 second black frames at ISO 100. That only confirms the potentially improved output from the 5D2 (more recent design), if it were not for the higher pattern noise probability.
You also experienced the slight increase of gain generated noise at apertures wider than f/2.0. So that kind of confirms that something is going on, although the increase of noise is very slight. Because the gain boost is so small, and assuming it is to compensate for exposure gate shading (QED), I do wonder why the manufacturers even do it, because the effects of vignetting plus light fall-off towards the corners of the image are much more pronounced, and spatially variant. The larger issue at wide aperture, as far as exposure is concerned, is the significant under exposure of the corners of the image, which leeds to visible differences in signal to noise after vignetting correction in postprocessing.
Another, somewhat related, issue is 'color cast'. That's a shift in color balance which also depends on the angle of incidence, but that is also not addressed by an overal minor gain boost.
The reason for the gain boost remains slightly puzzling, but given it's minor effect on noise (and only in scenarios that create larger problems), it's not a big issue. For those who like to understand their tools, it's intriguing though ...
Cheers,
Bart
-
Of course, nobody is saying this center 'sensel-vignetting' is nearly as bad as the wide-angle vignetting. The former is according to the data about between 1/3 and 2/3 of a stop. Vignetting numbers for some wide-angle lenses has been measured as high as 3 EV:
http://www.photozone.de/canon_eos_ff/506-zeiss18f35eosff?start=1
http://en.wikipedia.org/wiki/Vignetting
You're confusing mechanical and optical vignetting -ray obstruction inside the lens, mainly- and sensor vignetting -ray inclination on sensor and cos^4 law-.
The first two are totally irrelevant to our problem -thus my indication to use f/8-, since they have nothing to do with ray inclination. And the -3ev you give is of this kind.
Besides, this lens is of retrofocus and telecentric design, its rays are no more inclinated than any old 50mm lens.
You might have a point, but I think you are making a false assumption which overestimates the corner "pixel vignetting": that there are "totally non-telecentric" 40mm SLR lenses with exit pupil height equal to that focal length of 40mm. That is not the way 35mm format SLR lenses work; certainly not Nikon F-mount ones because the flange is about 47mm from the focal plane, and so the exit pupil is higher than that: 55mm and up usually. See this table from 2003, and note that there has been a trend since then toward higher exit pupils: http://www.swissarmyfork.com/lens_table_1.htm
Unless I have rangefinder lenses -not telecentric- and I use them on my APSC Nex ? Which is exactly what I do :-) And I happen to have a 40mm I measured myself yesterday.
I also have a 50/2.8 Sony macro lense, whose exit pupil is somewhere like 5mm from flange. Sony has a 45mm-ish flange.
And DxO says its lens have a mere 1/3 EV sensor vignetting on a full frame sensor. So basically DxO says that 0.40 radians inclination doesn't mean more than 1/3EV sensor vignetting (and it might even be the cos^4 law talking and not pixel vignetting).
You can test it yourself by grabbing any lens in the list you gave (thanks, it's really intersting by the way :) ) that have a 55mm (from sensor) exit pupil : you'll have a 0.37 radians angle which correspond to a 1.2 aperture.
Do you have awful vignetting in corners (on FF sensors) with those lenses once closed to f/11 ? Probably not.
I don't have it on my Nex with a 40mm non-telecentric lens, DxO doesn't measure it stronger than 1/3 EV with a non-telecentric 50mm on FF sensor...
... and friends of mine start to have that kind of vignetting and magenta shift with 20mm -non telecentric- lenses and shorter : angles like from f/0.7 lenses...
Another, somewhat related, issue is 'color cast'. That's a shift in color balance which also depends on the angle of incidence, but that is also not addressed by an overal minor gain boost.
The reason for the gain boost remains slightly puzzling, but given it's minor effect on noise (and only in scenarios that create larger problems), it's not a big issue. For those who like to understand their tools, it's intriguing though ...
Color cast happens on APSC sensors with 20mm exit pupils and shorter, meaning angles like 0.6 radians, such as with f/0.7 lenses... Nothing to worry about at f/1.2 or on corners with DSLR lenses.
I firmly believe this gain boost is here to compensate for LENS vigneting (mainly optical vigneting) which is often severe with wide aperture lenses all over the frame apart from center -look at that swirling bokeh-, and give images an overall under-exposed look.
If you know your f/1.4 lens has a -2.5 ev vignetting in corners, -1.5 ev on horizontal borders, and -1 EV overall... A 1 EV overexposure will over-expose the center -what DxO measured, I think-, but the whole image will look better...
-
I firmly believe this gain boost is here to compensate for LENS vigneting (mainly optical vigneting) which is often severe with wide aperture lenses all over the frame apart from center -look at that swirling bokeh-, and give images an overall under-exposed look.
If you know your f/1.4 lens has a -2.5 ev vignetting in corners, -1.5 ev on horizontal borders, and -1 EV overall... A 1 EV overexposure will over-expose the center -what DxO measured, I think-, but the whole image will look better...
I'm not so sure. The gain boost I'm getting seems to be in the order of a fraction of 1/3rd of a stop (I need to do a more exact determination, but IRIS is not cooperating). No way is that going to improve the loss of 1 or more stops in the corners in any significant manner. When using the internal exposure meter, the automatic correction for corner underexposure may well turn out to be more by adding actual photons, but overexposure of the image center (assuming I'm 'exposing to the right' there, e.g. with manual exposure) is not going to help image quality either. The puzzle remains.
Cheers,
Bart
-
The reason for the gain boost remains slightly puzzling, but given it's minor effect on noise (and only in scenarios that create larger problems), it's not a big issue. For those who like to understand their tools, it's intriguing though ...
I agree, Bart. The effect is interesting and we should thank Mark Dubovoy for raising the issue. But I wonder just how useful in a pratical sense in the field or studio such knowledge would be if the manufacturer were to specify and reveal any current secret boosting of ISO when a lens is used at very wide apertures.
We really need some comparison shots demonstrating that an F1.2 shot with a particular lens really has no shallower DoF, and/or has noticeably greater noise, than an F1.4 shot of the identical scene using the same lens.
If the increase in noise (or lack of reduction in DoF) is not noticeable, except at the extreme pixel-peeping level, then the issue becomes academic and perhaps only of concern to organizations like DXO Labs who need extremely accurate information on the performance of sensors and lenses in order to improve their RAW converter.
Another issue with such a comparison is the resolution at the point of focus. We would expect an F1.2 lens to be slightly sharper at F1.4 and slight sharper again at F1.8.
If we have a situation whereby a shot at F1.4 has a noticeably blurrier background than the same scene shot at F1.8, but the plane of focus in the F1.8 shot is noticeably sharper, then at a certain print size which brings out that greater sharpness, the differences in the perceived DoF of the two shots will be reduced.
What is perhaps of greater concern to the practical photographer is the variance in 'light transmission' efficiency of different lenses used at the same f stop. Do we need T-stop markings on lenses? Is the variance great enough to warrant that?
I recall some confusion a couple of years ago when I compared my Nikkor 14-24/2.8 with my Sigma 15-30, using the Mark Welsh adapter on my Canon 5D.
For those who are interested, the thread is at http://www.luminous-landscape.com/forum/index.php?topic=28926.0
The outcome was, I learned that my Canon 50/1.8 has about 2/3rds of a stop greater light-transmission efficiency than my Canon 24-105/F4 zoom at 50mm. I also bought a Nikon D700 and sold my Mark Welsh adapter.
The practical benefit of such knowledge might be of some significance in certain circumstances when, for example, I need a 50mm lens and a fast shutter speed for a moving subject. I might not bother changing lenses from the zoom to the 50mm prime for the sake of the marginal increase in resolution that the 50mm prime might provide.
However, knowing that I can also use a faster shutter speed at the same ISO and f stop (up to F4) with the 50/1.8 (a 320th as opposed to a 200th) for the same ETTR exposure, I might just take the trouble to change lenses and get a better quality shot as a consequence.
By the way, the Nikkor 14-24 on the D700 seems to have the same T-stop as the Sigma 15-30 on the 5D. The variance is very slight and appears to match the very slight differences in ISO sensitivity between these two cameras as shown on the DXOmark site.
-
You guyes are barking up the wrong tree.
The issue is NOT the way the sensor is built. This is easy to prove by using the tables Mark provided in his argument. Cameras with essentially identical sensors, but different sensor sizes have completely different results when the lens is used at F1.2.
The reason is NOT the sensor!
It's the mirror box. If you look at the exit pupil of the lens in question you will see that it is actually larger than the mirror box itself. The mirror box itself is vignetting the light path. The tighter the mirror box, the more vignetting you get.
Those of us who have a bit of experience with Perspective Control lenses have run into this problem for ages. We used to blame the problem on lens falloff, but it was actually light-loss due to the walls of the mirror-box impinging on the light path.
Ken N.
www.zone-10.com
-
I think some posters (BartvanderWolf, pegelli, b_z and image66 among others) haven't understood Marc Dubovoy's excellent article because they are confused about what microlenses accomplish.
Microlenses improve the angular response of a sensel, but only up to a limit.
The marginal rays of the light cone projected by a fast — e.g. f/1.2 — lens are quite tilted.
A pont source of light, e.g. a small LED photographed from a fairly large distance, stars in the night sky etc. form, when in focus, a point on the imaging sensor or film.
When out of focus, the intersection between the lens' light cone and the imaging plane forms a disc, not a point.
Depending on the degree of defocus, that disc can obviously have various diameters.
If the degree of defocus is small, the disc diameter might be smaller than the diameter of the circle of confusion which figures in depth of field calculations.
Regardless of the OOF disc's diameter, the angle of the marginal rays relative to the sensor doesn't change. This obviously implies, as Mark Dubovoy points out, that the diameter of the circle of confusion recorded by an imaging sensor is also affected by the sensel's acceptance angle, and that depth of field, in turn, must be affected by the sensel's angular response.
How can we estimate that acceptance angle with any degree of reliability ?
If, despite the microlens' best efforts, the marginal rays are too tilted to reach the photodiode at the bottom of thick sensel structure, these light rays can be considered to be non-existent for imaging purposes.
If these marginal rays cannot reach the photodiodes, the recorded diameter of the disc formed by an out of focus point light source will be reduced.
A simple way to assess the critical incidence angle — that is, the acceptance angle — above which the rays cannot be recorded anymore is thus to measure the diameter of the out of focus disc formed by a point source.
Film doesn't have any problems recording even very tilted light rays. Due to the very geometry of the lens' light cone, the diameter of the OOF disc must be directly proportional to the lens' f-stop a.k.a. aperture setting. On film, doubling the aperture thus necessarily doubles the recorded OOF disc diameter.
With some digital sensor designs, acceptance angles can be quite limited. At some point, the linear relationship between the lens aperture and the OOF disc diameter recorded on the picture must then break down, as the tilted marginal rays cannot reach the photodiodes anymore.
A quick measurement of the dimensions of the OOF discs visible in the test picture (http://www.photozone.de/images/8Reviews/lenses/canon_50_12_5d/bokeh.jpg) of the Canon EF50mm F/1.2L lens on Photozone.de (http://www.photozone.de/canon_eos_ff/472-canon_50_12_5d?start=1) show that with the Canon 5D Mark II's CMOS sensor used in the test, the diameter of the OOF disc increases, as expected, when the lens is opened up from f/4 to f/2, but then plateaus at a diameter geometrically corresponding to about f/1.5, even when the lens is opened up to f/1.2
This indicates that the acceptance angle of the Canon 5D Mark II's sensor is limited to about arctan(1/(1.5*2)) i.e. about 18.4 degrees from the perpendicular, and that a f/1.2 lens thus performs on the 5D2 essentially as a f/1.5 lens as far as actual lens speed, DoF and bokeh are concerned.
To fully record the bokeh and "draw" of fast lenses like the Canon EF50mmF1.0L, EF50mmF1.2L and EF85mmF1.2L, ir thus seems that one will have to use a film-based Canon EOS body, instead of a Canon DSLR.
An interesting pint of comparison to bring up here might be the — presumably identical — sensels used in the CCD sensors equipping the Leica M8 and M9.
CCDs don't need the multiple transistors surrounding each photodiode of a CMOS sensor architecture. As such, CCDs don't require the multiple metal layers of a CMOS needed for the transistor's signal lines. The resulting CCD pixel stack is typically much more shallow than a CMOS sensor's tunnel-like architecture.
With a CCD, the distance between the microlens and the photodiode can thus be made closer than with a CMOS sensor, and the microlens' acceptance angles can thus be much wider.
The marginal ray of the light cone of a Leica Noctilux f/1.0 lens is about arctan(1/2) = 26.6 degrees. According to its datasheet (http://www.kodak.com/global/plugins/acrobat/en/business/ISS/datasheet/fullframe/KAF-10500LongSpec.pdf), the Kodak KAF-10500 CCD used in the Leica M8 still has an angular response, at an angle of 27 degrees, of about 70% of the peak response.
Unlike the DSLRs, the CCD-based digital Leicas thus seem entirely able to record the light rays of the wide light cone emanating from Leica's super-fast Noctiluxes.
As for b_z's assertions about acceptance angles, using his NEX-5 and a 40mm F/1.4 lens as examples, they are flawed for several reasons:
- b_z doesn't properly consider the distance between the sensor and the lens' exit pupil, from which the light rays can be considered to emanate. That exit pupil distance is unknown. If it is 40mm, the chief ray's tilt angle, even in the extreme corner of an APS-C sensor, would be arctan(14mm/40mm) i.e. about 19.2 degrees, that is, quite close to the Canon 5D2's f/1.5-equivalent 18.5 degrees. As such, given the incertitude about the exit pupil distance of the 40mm lens used by b_z, his test doesn't bring any useful piece of information about the angular response of f/1.2 lenses.
- The minimal exit pupil distance relative to the image plane of most Leica-mount rangefinder lenses, including ultra wides, is about 28mm. I know of no actual instances of Leica-mount lenses with a 15mm exit pupil distance hypothesized by b_z.
- The Sony-brand APS sensor used in Fuji's X100 fixed-lens digital rangefinder camera uses offset microlenses. There is no publicly available information indicating whether Sony uses offset microlenses for the NEX series of cameras. This means that it is impossible for us to state whether the NEX-5 uses, or does not use offset microlenses. If the NEX-5 uses offset microlenses — a plausible scenario, given the very short flangeback of the NEX mount and the compactness of the NEX 16mm pancake lens — the acceptance angle of the corner sensels could be even larger than the back-of-the envelope calculations done above, rendering b_z's NEX-based analysis even more meaningless.
Let us now consider image66's assertion that the vignetting is not caused by the sensor, but by the mirror box.
If the mirror box were to cause vignetting at the center of the image where DxO performed their mesurements, then, the OOF disc of an f/1.2 or f/1.4 lens at the center of the image would not appear as a perfect circle, but as a circle visibly truncated by the straight edges of the mirror box.
Given that anybody with an APS or full-size DSLR and a f/1.2 or f/1.4 lens can verify that, even with the lens wide open, an OOF point lioght source appears at the center of the image as a perfect circle and not as a truncated one, image66's assertion can be dismissed out of hand.
As a camera geek, I applaud Mark Dubovoy for his very insightful article, and look forward to DxO publising their detailed findings as to how they assessed the amplitude of the ISO correction camera manufacturers automatically apply to compensate for the non-recording of the marginal light rays emitted by fast lenses.
-
I think some posters (BartvanderWolf, pegelli, b_z and image66 among others) haven't understood Marc Dubovoy's excellent article because they are confused about what microlenses accomplish.
Mark's article is good because it's a relevant question to ask (regarding false ISO setting), but it's not excellent because his ideas of a cause and his whole "depth of field" idea is totally irrelevant here.
Microlenses improve the angular response of a sensel, but only up to a limit.
Yes.
The marginal rays of the light cone projected by a fast — e.g. f/1.2 — lens are quite tilted.
A LOT less than rays coming to corners of sensor with a telemetric (non telecentric) wide angle. And still less than a ray coming to corner of a FF sensor with a 50mm non telecentric macro lense. See my calculation above, appearently you can understand them.
A pont source of light, e.g. a small LED photographed from a fairly large distance, stars in the night sky etc. form, when in focus, a point on the imaging sensor or film.
When out of focus, the intersection between the lens' light cone and the imaging plane forms a disc, not a point.
Depending on the degree of defocus, that disc can obviously have various diameters.
If the degree of defocus is small, the disc diameter might be smaller than the diameter of the circle of confusion which figures in depth of field calculations.
Regardless of the OOF disc's diameter, the angle of the marginal rays relative to the sensor doesn't change.
True
This obviously implies, as Mark Dubovoy points out, that the diameter of the circle of confusion recorded by an imaging sensor is also affected by the sensel's acceptance angle, and that depth of field, in turn, must be affected by the sensel's angular response.
Yes, ONLY IN THE MIDDLE OF THE SENSOR, borders of the circle consists of inclinated rays. But as soon as you're not in the center anymore, that's NOT TRUE. And it does not appear in reality.
How can we estimate that acceptance angle with any degree of reliability ?
Reading my calculations above would be a good start...
If, despite the microlens' best efforts, the marginal rays are too tilted to reach the photodiode at the bottom of thick sensel structure, these light rays can be considered to be non-existent for imaging purposes.
Not, the effect is progressive.
If these marginal rays cannot reach the photodiodes, the recorded diameter of the disc formed by an out of focus point light source will be reduced.
No, the effect is progressive. At most you should see border of the circle in the middle of the frame fade a little, nothing more.
A simple way to assess the critical incidence angle — that is, the acceptance angle — above which the rays cannot be recorded anymore is thus to measure the diameter of the out of focus disc formed by a point source.
Not true, see above.
Film doesn't have any problems recording even very tilted light rays. Due to the very geometry of the lens' light cone, the diameter of the OOF disc must be directly proportional to the lens' f-stop a.k.a. aperture setting. On film, doubling the aperture thus necessarily doubles the recorded OOF disc diameter.
True
With some digital sensor designs, acceptance angles can be quite limited. At some point, the linear relationship between the lens aperture and the OOF disc diameter recorded on the picture must then break down, as the tilted marginal rays cannot reach the photodiodes anymore.
False, the effect is PROGRESSIVE.
A quick measurement of the dimensions of the OOF discs visible in the test picture (http://www.photozone.de/images/8Reviews/lenses/canon_50_12_5d/bokeh.jpg) of the Canon EF50mm F/1.2L lens on Photozone.de (http://www.photozone.de/canon_eos_ff/472-canon_50_12_5d?start=1) show that with the Canon 5D Mark II's CMOS sensor used in the test, the diameter of the OOF disc increases, as expected, when the lens is opened up from f/4 to f/2, but then plateaus at a diameter geometrically corresponding to about f/1.5, even when the lens is opened up to f/1.2
Absolutely irrelevant example, what you're seeing there is that because the disc is NOT IN THE MIDDLE OF THE FRAME, the lens show mechanical vignetting because of the conception of the lens. It's related to vigneting, and bokeh discs into sort-of-elliptic shapes.
You're just seing that, nothing else.
This indicates that the acceptance angle of the Canon 5D Mark II's sensor is limited to about arctan(1/(1.5*2)) i.e. about 18.4 degrees from the perpendicular, and that a f/1.2 lens thus performs on the 5D2 essentially as a f/1.5 lens as far as actual lens speed, DoF and bokeh are concerned.
To fully record the bokeh and "draw" of fast lenses like the Canon EF50mmF1.0L, EF50mmF1.2L and EF85mmF1.2L, ir thus seems that one will have to use a film-based Canon EOS body, instead of a Canon DSLR.
Not at all, try on film you'll see exactly the same shape. It's due to lens design, not sensor.
An interesting pint of comparison to bring up here might be the — presumably identical — sensels used in the CCD sensors equipping the Leica M8 and M9.
M8 is APSC and have no telecentric microlenses. The sensors are TOTALLY different in regard to their response to rays inclination.
CCDs don't need the multiple transistors surrounding each photodiode of a CMOS sensor architecture. As such, CCDs don't require the multiple metal layers of a CMOS needed for the transistor's signal lines. The resulting CCD pixel stack is typically much more shallow than a CMOS sensor's tunnel-like architecture.
With a CCD, the distance between the microlens and the photodiode can thus be made closer than with a CMOS sensor, and the microlens' acceptance angles can thus be much wider.
Well, I'm not sure about that, and the fact that DxO datas show that CCDs are WORSE in regard to this "unknown" effect makes this assumption quite weird.
The marginal ray of the light cone of a Leica Noctilux f/1.0 lens is about arctan(1/2) = 26.6 degrees. According to its datasheet (http://www.kodak.com/global/plugins/acrobat/en/business/ISS/datasheet/fullframe/KAF-10500LongSpec.pdf), the Kodak KAF-10500 CCD used in the Leica M8 still has an angular response, at an angle of 27 degrees, of about 70% of the peak response.
Unlike the DSLRs, the CCD-based digital Leicas thus seem entirely able to record the light rays of the wide light cone emanating from Leica's super-fast Noctiluxes.
What's the angular response for say the A900 CMOS ? You don't know ? Then how could you compare ???????????????
As for b_z's assertions about acceptance angles, using his NEX-5 and a 40mm F/1.4 lens as examples, they are flawed for several reasons:
It's rather funny you call my assumptions flawed when in fact you just didn't understand them, or considered I hadn't thought of testing some things.
b_z doesn't properly consider the distance between the sensor and the lens' exit pupil, from which the light rays can be considered to emanate. That exit pupil distance is unknown.
As a matter of fact, I consider lens's exit pupil.
If it is 40mm, the chief ray's tilt angle, even in the extreme corner of an APS-C sensor, would be arctan(14mm/40mm) i.e. about 19.2 degrees, that is, quite close to the Canon 5D2's f/1.5-equivalent 18.5 degrees. As such, given the incertitude about the exit pupil distance of the 40mm lens used by b_z, his test doesn't bring any useful piece of information about the angular response of f/1.2 lenses.
Exit pupil of my 40mm lens is somwhere between 35 and 45mm. I measured that. Now, you see why my corner rays are relevant ? They are more tilted than this f/1.2 lens rays. And still I got no real vignetting when stopped down a little.
The minimal exit pupil distance relative to the image plane of most Leica-mount rangefinder lenses, including ultra wides, is about 28mm. I know of no actual instances of Leica-mount lenses with a 15mm exit pupil distance hypothesized by b_z.
Look at voigtlander 15mm. Just look.
The Sony-brand APS sensor used in Fuji's X100 fixed-lens digital rangefinder camera uses offset microlenses. There is no publicly available information indicating whether Sony uses offset microlenses for the NEX series of cameras. This means that it is impossible for us to state whether the NEX-5 uses, or does not use offset microlenses. If the NEX-5 uses offset microlenses — a plausible scenario, given the very short flangeback of the NEX mount and the compactness of the NEX 16mm pancake lens — the acceptance angle of the corner sensels could be even larger than the back-of-the envelope calculations done above, rendering b_z's NEX-based analysis even more meaningless.
You don't know if Nex sensor has offset microlenses. I do know it doesn't have any. Saying my analysis is meaningless just because you don't understand and don't even try to search on your own to check my sayings is quite rude.
Let us now consider image66's assertion that the vignetting is not caused by the sensor, but by the mirror box.
If the mirror box were to cause vignetting at the center of the image where DxO performed their mesurements, then, the OOF disc of an f/1.2 or f/1.4 lens at the center of the image would not appear as a perfect circle, but as a circle visibly truncated by the straight edges of the mirror box.
Given that anybody with an APS or full-size DSLR and a f/1.2 or f/1.4 lens can verify that, even with the lens wide open, an OOF point lioght source appears at the center of the image as a perfect circle and not as a truncated one, image66's assertion can be dismissed out of hand.
ONLY if DxO test is made on center of the frame (which seems to be the case, I agree). And it's a good concern because it happens sometimes, see there : http://forums.dpreview.com/forums/read.asp?forum=1042&message=36846243
As a camera geek, I applaud Mark Dubovoy for his very insightful article, and look forward to DxO publising their detailed findings as to how they assessed the amplitude of the ISO correction camera manufacturers automatically apply to compensate for the non-recording of the marginal light rays emitted by fast lenses.
Mark's questions are good, but his (and your) assumptions about the causes are wrong.
This ISO correction could be made only to compensate the huge mechanical vignetting every fast lens showand make the whole image look better exposed.
-
Mark's article is good because it's a relevant question to ask (regarding false ISO setting), but it's not excellent because his ideas of a cause and his whole "depth of field" idea is totally irrelevant here.
Depth of field appears "irrelevant" to you because you don't understand the notion of "Circle of Confusion" used in DoF calculations.
The intersection of the lens' light cone and the image plane is a disc. The light rays near the periphery of the cone are more tilted than the central light ray (the chief ray).
If the tilt exceeds a critical angle, the rays cannot reach, and therefore are not recorded by the photodiodes. Ergo, the recorded diameter of that disc is affected by the sensel's acceptance angle, and that recorded diameter can therefore be smaller than the geometric diameter of the light cone at the imaging plane intersection.
The CoC is also a disc, and is conceptually and physically identical to the disc formed by the intersection — discussed above — of the light cone and the imaging plane.
Ergo, a sensel with a limited acceptance angle will record a disc that is smaller than the disc recorded by a medium like film that hasn't any acceptance angle limits.
A smaller disc, in CoC terms, is a sharper, less blurred disc. If the constitutive points of an image are sharper, it means that the sharpness zone in the object field, i.e. the DoF, must be deeper. Ergo, a sensel with a limited acceptance angle necessarily results — as Mark Dubovoy correctly points out — in a deeper DoF than you'd get with film.
The photozone.de OOF disc samples, while not ideal as they haven't been taken at the center of the imaging field, provide a good measure of the plateauing of the disc's dimensions at a dimension corresponding to about f/1.5
Considering the typical price, size and weight differentials between a f/1.2 and a f/1.4 lens, the fact that — when used on a DSLR like the Canon 5D2 — the effective speed of an f/1.2 lens, as well as its DoF and bokeh might be equivalent only to that of a f/1.5 lens is a quite relevant observation on a photo forum.
The apparent deformation affecting the exit pupil at larger image heights — that is, at points far from the image center — is essentially an elliptical one on photozone.de's pictures. Elliptical deformations affect the dimensions of a circular exit pupil only along one axis, and quite valid dimensional measurements can thus be made along the axis unaffected by the elliptical squeezing.
These photozone.de measurements are tangible and totally consistent with the physics of light cone propagation and microlens acceptance angle limitation models.
All your assertions, OTOH, are unsubstantiated, including the one that a test with the NEX-5 and a 40mm lens of unknown exit pupil distance is somehow relevant to judge a DSLR's sensel's response to a f/1.2 cone of light.
The diagonal of an APS-C sensor like the one used in the NEX-5 is 28mm. If we assume, as you do, that the lens' exit pupil distance from the imaging plane is 40mm, the chief ray's angle in the corner has a tilt of arctan(14mm/40mm) — i.e. corresponding to the tilt of the marginal rays of a light cone of a 40mm/28mm = f/1.4 lens.
So, even if the NEX-5 had no offset microlenses — something that is impossible for me and you to know, — your assertion that your f/1.4 NEX-5 thought experiment "proves" that ray tilt cannot be a vignetting factor with a f/1.2 lens is totally irrelevant.
Obviously, if the NEX-5 had offset microlenses, your thought experiment would be even more irrelevant, as the corner sensels' sensitivity to the chief ray's tilt angle would be even less observable.
-
The marginal rays of the light cone projected by a fast — e.g. f/1.2 — lens are quite tilted.
I wouldn't characterize the maximum possible angle as "quite tilted" (whatever that may mean):
(http://www.xs4all.nl/~bvdwolf/temp/OPF/EP2FP.png)
That doesn't seem like a very problematic angle for a microlens to me, especially when we consider the modest light fall-off in the corners of the image when stopped down where the central rays strike at an even more oblique angle of 29.65 degrees ...
Cheers,
Bart
-
I wouldn't characterize the maximum possible angle as "quite tilted" (whatever that may mean):
An incidence angle that is larger than the acceptance angle of common DSLR sensors is, in my mind, quite tilted.
I also invite you to review what "exit pupil" means, optically speaking.
Considering that most SLR lenses with a 50mm focal length have an exit pupil to image plane distance of about 50mm, the notion that a short tele 85mm Canon lens designed to be mounted on an SLR could possibly have an exit pupil distance of 38mm is, well, quite amusing.
Furthermore, any assertion that the sensel's acceptance angles are not a factor should also be accompanied with a physically coherent explanation as to why the OOF discs recorded with a DSLR happen to be clipped at a radius corresponding to around f/1.5 even with f/1.2 lenses while, with film cameras, the recorded OOF disc diameter maintains a strict proportionality with the lens' selected aperture.
-
@ Antithenes, interesting discussion but in my mind still inconclusive.
One thing that might clear up some doubt for me is if the light fall off proportionally (or gradually) drops down as the inclination angle increases, or if there is a very steep fall off from near 100% to near 0% once you exceed a certain threshold inclination angle. What's your understanding how that works?
-
Depth of field appears "irrelevant" to you because you don't understand the notion of "Circle of Confusion" used in DoF calculations.
You make me laugh... Seriously. I perfectly understand circle of confusion and depth of field calculation. They are irrelevant here because you will never EVER see any modification of DoF because of ray inclination.
The intersection of the lens' light cone and the image plane is a disc. The light rays near the periphery of the cone are more tilted than the central light ray (the chief ray).
Which cone ? In fact it's a lot more complicated than that. A point in object field send a light cone which intersect sensor=image plane in a disc (provided there is no mechanical vignetting, thing you seem unable to understand), yes.
But, as I already said, the chief ray is the less inclined of the disc ONLY in the middle of the sensor. At the border, the chief ray is MORE tilted than some of the border rays.
If the tilt exceeds a critical angle, the rays cannot reach, and therefore are not recorded by the photodiodes.
That's utterly wrong. The transition is PROGRESSIVE (I said that earlier, you did not understand ?), there's no "critical angle after which NO light is recorded".
Ergo, the recorded diameter of that disc is affected by the sensel's acceptance angle, and that recorded diameter can therefore be smaller than the geometric diameter of the light cone at the imaging plane intersection.
Again, this would only be true right at the middle of the sensor, if there was such an "acceptance angle" (which is NOT the case at any angle we're speaking of).
The CoC is also a disc, and is conceptually and physically identical to the disc formed by the intersection — discussed above — of the light cone and the imaging plane.
Not true. CoC is the diameter a disc can have on the sensor and yet be considered ponctual by an observer placed at the right distance of the printing. When the light cone of an object point intersect the sensor in a circle which is smaller than CoC, then it is in focus.
Ergo, a sensel with a limited acceptance angle will record a disc that is smaller than the disc recorded by a medium like film that hasn't any acceptance angle limits.
Not true. At most you'll have more vignetting, but appearently this "pixel/sensor vignetting" is always weaker than the cos4 law.
A smaller disc, in CoC terms, is a sharper, less blurred disc. If the constitutive points of an image are sharper, it means that the sharpness zone in the object field, i.e. the DoF, must be deeper. Ergo, a sensel with a limited acceptance angle necessarily results — as Mark Dubovoy correctly points out — in a deeper DoF than you'd get with film.
Not true. You're totally confused here, it makes absolutely no sense.
The photozone.de OOF disc samples, while not ideal as they haven't been taken at the center of the imaging field, provide a good measure of the plateauing of the disc's dimensions at a dimension corresponding to about f/1.5
Wrong. Their vertical size is approximately the same than at f/1.5, BECAUSE OF LENS MECHANICAL VIGNETTING AND NOTHING ELSE, but if you were honest, you would point out that their horizontal size is really what's expected for f/1.2 lens. And it proves you're wrong.
And had this crop been taken at center of the frame, you'd see a perfezct circle, the size you except to have for f/1.2 (unless ther is mechanical vignetting due to the chamber, it sometimes happens as I said earlier).
Considering the typical price, size and weight differentials between a f/1.2 and a f/1.4 lens, the fact that — when used on a DSLR like the Canon 5D2 — the effective speed of an f/1.2 lens, as well as its DoF and bokeh might be equivalent only to that of a f/1.5 lens is a quite relevant observation on a photo forum.
Except this "might be" is in fact a "is not".
The apparent deformation affecting the exit pupil at larger image heights — that is, at points far from the image center — is essentially an elliptical one on photozone.de's pictures. Elliptical deformations affect the dimensions of a circular exit pupil only along one axis, and quite valid dimensional measurements can thus be made along the axis unaffected by the elliptical squeezing.
Yes, and had you been honest you would have seen only vertical size is smaller than expected. Horizontal size is fine.
These photozone.de measurements are tangible and totally consistent with the physics of light cone propagation and microlens acceptance angle limitation models.
Yes, and measuring these disks along horizontal axis prove that it's really a f/1.2 bokeh.
I'm laughing at you saying things about "physics of light cone propagation" when in fact you fail to understand even the simplier arguments. "microlens acceptance angle limitation models" is just b***s*** when you talk of 30° angles. It's an issue only for a lot higher angles, and even so, cos4 law is a waaaay bigger issue.
All your assertions, OTOH, are unsubstantiated, including the one that a test with the NEX-5 and a 40mm lens of unknown exit pupil distance is somehow relevant to judge a DSLR's sensel's response to a f/1.2 cone of light.
I'm laughing. Really. "unsubstantiated" describes very well your own posts. Mine consist of facts I've checked, or when it's not I say it.
The exit pupil is not unkown, you failed to read it was between 35 and 45mm ? Yes, I measured it, because I know how to do...
The diagonal of an APS-C sensor like the one used in the NEX-5 is 28mm. If we assume, as you do, that the lens' exit pupil distance from the imaging plane is 40mm, the chief ray's angle in the corner has a tilt of arctan(14mm/40mm) — i.e. corresponding to the tilt of the marginal rays of a light cone of a 40mm/28mm = f/1.4 lens.
Yes, you're right, for once ! But I'm not "assuming", I don't "assume" things or I say I do. I measured it.
So, even if the NEX-5 had no offset microlenses — something that is impossible for me and you to know, — your assertion that your f/1.4 NEX-5 thought experiment "proves" that ray tilt cannot be a vignetting factor with a f/1.2 lens is totally irrelevant.
The fact you don't know that doesn't mean nobody knoes. I do know Nex has no offset microlenses. And I proved that tilted rays are a totally irrelevant issue for f/1.4 lenses with that lens/sensor combination.
And read one of my above post again : I also measured a 50~55mm exit pupil for m 50/2.8 macro Sony lens, and on FF sensor (Alpha 850/900) it has no vignetting. Meaning for f/1.2 lenses (the angle is almost the same), this issue is irrelevant too.
Obviously, if the NEX-5 had offset microlenses, your thought experiment would be even more irrelevant, as the corner sensels' sensitivity to the chief ray's tilt angle would be even less observable.
Except it doesn't have. And I'm not guessing, it's a fact.
-
An incidence angle that is larger than the acceptance angle of common DSLR sensors is, in my mind, quite tilted.
Define acceptance angle.
I also invite you to review what "exit pupil" means, optically speaking.
Considering that most SLR lenses with a 50mm focal length have an exit pupil to image plane distance of about 50mm, the notion that a short tele 85mm Canon lens designed to be mounted on an SLR could possibly have an exit pupil distance of 38mm is, well, quite amusing.
"Exist pupil" on this scheme is in fact "rear lens element".
Furthermore, any assertion that the sensel's acceptance angles are not a factor should also be accompanied with a physically coherent explanation as to why the OOF discs recorded with a DSLR happen to be clipped at a radius corresponding to around f/1.5 even with f/1.2 lenses while, with film cameras, the recorded OOF disc diameter maintains a strict proportionality with the lens' selected aperture.
Define acceptance angle before even thinking of going on with this.
The bold sentence is completely false. There is no such OOS disc clipping APART FROM MECHANICAL VIGNETTING. But you seem unable to understand that.
Try with film and sensor, you'll get the exact same OOF discs shape. Exactly. Not even a 1% difference in size. Maybe a 1~10% luminosity decrease at borders with sensor compared to film, but no shape difference. At all.
-
An incidence angle that is larger than the acceptance angle of common DSLR sensors is, in my mind, quite tilted.
That is quite a bit of circular reasoning. What is the acceptance angle of common DSLRs? You seem to conclude that it's exceeded, so you suggest you know.
I also invite you to review what "exit pupil" means, optically speaking.
And I invite you to review the difference between exit pupil (http://en.wikipedia.org/wiki/Exit_pupil) (especially in tele lenses) and (secondary) principle plane (http://en.wikipedia.org/wiki/Cardinal_point_(optics)) ...
In fact, the diameter of the exit pupil of the EF 85mm f/1.2 at 44mm distance from the flange is almost exactly the same at the rear lens element and the virtual image of the front lens aperture. True, the exact distance of the exit pupil may be at a different position, but it's visible diameter isn't different (as viewed from the fixed distance at the center of the sensor). When moving off optical axis, one can see (from the position of the sensor through the rear lens element) the front of the lens barrel, which explains the vignetting at full aperture, the aperture is no longer exactly circular. All of the above is when focused at infinity.
Considering that most SLR lenses with a 50mm focal length have an exit pupil to image plane distance of about 50mm, the notion that a short tele 85mm Canon lens designed to be mounted on an SLR could possibly have an exit pupil distance of 38mm is, well, quite amusing.
Unless you are confusing exit pupil of a telelens (where the rear element acts as an aperture) with principle plane, I fail to grasp your misplaced amusement.
Cheers,
Bart
-
@ Antithenes, interesting discussion but in my mind still inconclusive.
One thing that might clear up some doubt for me is if the light fall off proportionally (or gradually) drops down as the inclination angle increases, or if there is a very steep fall off from near 100% to near 0% once you exceed a certain threshold inclination angle. What's your understanding how that works?
http://en.wikipedia.org/wiki/Vignetting very good article on vignetting.
Two kinds of light fall-off gradually increase depending on the inclination angle.
The biggest fall-off in regard of inclination angle is the cos4 law, which is the same on film and sensor. It's a physical law, you can only correct that by making your lenses telecentric (meaning rays are less inclined in the borders of the frame).
The second is the pixel/sensor vignetting, which is Mark's concern. And in fact it's known to appear in some bizarre kind of way (purple or magenta zones because of selective refraction on microlenses, I read somewhere, and added vignetting) only at very tilted angles you can get with a 15~20mm non telecentric lens with APSC or bigger sensor, EVEN with offset microlenses (Leica strongly advise not to use old design (non telecentric) ultra wide angles on M9 because of that).
The angle concerned is approximately what you'd get with a f/0.7 lens. With smaller angles, no noticeable sign of pixel/sensor vignetting.
-
@b_z, BartvanderWolf, also thanks for all the insights and links you're providing, should have included you in my earlier posted question At least I can follow your logic since it's clear concepts and data (I'm a ChemE by training, so have some appreciation for that ;) ). I'll continue to follow, but might not be able to contribute a lot to the insights other than asking a stupid question now and then ::)
-
As far as mechanical vignetting is concerned for fast optics, I can confirm it seems to exist in Canon cameras
http://dl.dropbox.com/u/2445259/boxeffect.jpg
I have consistently observed the above vignetting pattern (with slight variations) from my Canon cameras on a few of my instruments and lenses (on 10D, 20D, 40D, 1DMKII, 5D, 5D MKII) - it is usually not perceptible when looking at short exposures full size, barely noticeable on long exposures full size, but quite noticeable on processed images that haven't been properly calibrated with a flat field. It's orientation is always the same relative to the mirror box: there is always a darker linear area at the bottom of the frame.
The only other explanation I haven't been able to rule out with absolute certainty is that there would be a sensitivity issue at the bottom of all Canon sensors I have and does not show up with the astro CCDs I own. I haven't looked at the issue in depth with canon lenses as they are often not ideally suited for astro-photography (not to say that you can't get very good results, but they don't compare with the 2.7 FFC).
Last but not least, the whole subject is extremely complex, involves advanced lens design, advanced sensor design, some murky (at least in our minds) definitions, the inability to test components individually, etc. The consequences are relatively marginal in both senses of the term, and there are glaring holes in most of the complex arguments developed here (as they would be in my arguments if I went into them).
Let's be realistic here, we can't really second guess entire teams of competent engineers, each with their own domains of expertise, who have at their disposal the full resources of large manufacturers who happen to be locked in a fierce competitive war.
No point in getting into complex point by point rebukes or flame wars imho.
-
@ Antithenes, interesting discussion but in my mind still inconclusive.
One thing that might clear up some doubt for me is if the light fall off proportionally (or gradually) drops down as the inclination angle increases, or if there is a very steep fall off from near 100% to near 0% once you exceed a certain threshold inclination angle. What's your understanding how that works?
The photodiode at the bottom of the CMOS pixel stack has a finite size. Provided the light ray lands somewhere, anywhere in the photodiode's active area, it will presumably generate electrical charges, indicating that light has been detected.
As the ray's tilt increases, at some point, the ray won't fall anymore within the photodiode's active area. Beyond that critical angle, the light ray can be considered, for imaging purposes, as lost.
The photodiode's active area is not "fuzzy". It is created with IC manufacturing-class precision — perimeter variations of a few tens of nanometers at most against total sensel longitudinal dimensions of several thousand nanometers.
The angle transition between "detectable" and "not detectable" will thus presumably be quite sharp.
And indeed, the clipped f/1.5-equivalent OOF disc visible in photozone.de's picture has an clearly recognizable boundary — i.e. its perimeter is emphatically not progressively dissolving with, say, a Gaussian kind of blur where no definite boundary exists.
-
Unless you are confusing exit pupil of a telelens (where the rear element acts as an aperture) with principle plane, I fail to grasp your misplaced amusement.
I found it amusing because you're obviously conflating the rear element with the "exit pupil".
The exit pupil is the virtual image of the lens aperture — or diaphragm blades, if you prefer — as seen from the back of the lens.
Your drawing showing an SLR tele lens with an exit pupil distance of 38mm is therefore utterly unrealistic.
The chief ray tilt calculations based on that unrealistic placement are thus, of course, totally meaningless.
Let F be the f-number of a lens. The half-angle of the light cone emitted by that lens will necessarily have as value arctan(1/(2*F))
For an f/1.2 lens, the half-angle must thus be arctan(1/(2*1.2)) ~ 22.62 degrees.
The fact that you indicated 27.76-degree half-angle for a 85mm f/1.2 lens is also evidence that your calculations of chief ray tilt angles at various image heights are, I regret to say, absolutely unrealistic.
-
I perfectly understand circle of confusion and depth of field calculation. They are irrelevant here because you will never EVER see any modification of DoF because of ray inclination.
Yawn. Will you ever be able to understand that it is not sufficient to make assertions, but that you must also provide a modicum of evidence or physics- or math-based reasoning supporting your assertions ?
So, do please explain, preferably using sound optics principles, why "one will never EVER see any modification of DoF because of ray inclination"
But, as I already said, the chief ray is the less inclined of the disc ONLY in the middle of the sensor. At the border, the chief ray is MORE tilted than some of the border rays.
Yawn.
APS-C sensor diagonal: about 28mm.
In your thought experiment, your lens has an exit pupil at a distance of 40mm from the imaging plane.
This happens to correspond to the half-angle of the light cone of a f/1.4 lens.
Problem is, we're talking about the light loss affecting the light cone of a f/1.2 lens, not f/1.4.
Last time I checked, f/1.2 lenses had wider light cones than f/1.4 lenses, meaning that the f/1.2 cone's marginal rays are MORE, not LESS tilted than the f/1.4 light cone's marginal rays.
If the tilt exceeds a critical angle, the rays cannot reach, and therefore are not recorded by the photodiodes.
That's utterly wrong. The transition is PROGRESSIVE (I said that earlier, you did not understand ?), there's no "critical angle after which NO light is recorded".
Yawn. Where is the substantiation of your gratuitous assertion that the transition is PROGRESSIVE ? Are the depletion zones of the photodiodes at the bottom of a CMOS pixel stack drawn using some kind of fuzzy semiconductor technology ?
Ergo, the recorded diameter of that disc is affected by the sensel's acceptance angle, and that recorded diameter can therefore be smaller than the geometric diameter of the light cone at the imaging plane intersection.
Again, this would only be true right at the middle of the sensor, if there was such an "acceptance angle" (which is NOT the case at any angle we're speaking of).
Yawn. Where is the substantiation of your gratuitous assertion that there can't possibly be acceptance angle effects with large aperture lenses ?
The CoC is also a disc, and is conceptually and physically identical to the disc formed by the intersection — discussed above — of the light cone and the imaging plane.
Not true. CoC is the diameter a disc can have on the sensor and yet be considered ponctual by an observer placed at the right distance of the printing. When the light cone of an object point intersect the sensor in a circle which is smaller than CoC, then it is in focus.
Yawn. English language reading comprehension isn't your forte, is it ?
Ergo, a sensel with a limited acceptance angle will record a disc that is smaller than the disc recorded by a medium like film that hasn't any acceptance angle limits.
Not true. At most you'll have more vignetting, but appearently this "pixel/sensor vignetting" is always weaker than the cos4 law.
Yawn. Where is the substantiation of your gratuitous assertion that "pixel/sensor vignetting" is always weaker than the cos4 law ?
A smaller disc, in CoC terms, is a sharper, less blurred disc. If the constitutive points of an image are sharper, it means that the sharpness zone in the object field, i.e. the DoF, must be deeper. Ergo, a sensel with a limited acceptance angle necessarily results — as Mark Dubovoy correctly points out — in a deeper DoF than you'd get with film.
Not true. You're totally confused here, it makes absolutely no sense.
Yawn. Has it occurred to you that it might not make sense to you because:
1) your English-language reading comprehension skills are pretty low
and
2) your understanding of geometric optics concepts like CoC is basically non-existent ?
The photozone.de OOF disc samples, while not ideal as they haven't been taken at the center of the imaging field, provide a good measure of the plateauing of the disc's dimensions at a dimension corresponding to about f/1.5
Wrong. Their vertical size is approximately the same than at f/1.5, BECAUSE OF LENS MECHANICAL VIGNETTING AND NOTHING ELSE, but if you were honest, you would point out that their horizontal size is really what's expected for f/1.2 lens. And it proves you're wrong.
And had this crop been taken at center of the frame, you'd see a perfezct circle, the size you except to have for f/1.2 (unless ther is mechanical vignetting due to the chamber, it sometimes happens as I said earlier).
Yawn. Where is the substantiation of your gratuitous assertion that only mechanical vignetting is to blame, and that the horizontal size is what's expected for a f/1.2 lens ?
How about, well, actually calculating yourself what that horizontal size should be, using sound optical principles, and then providing us with your calculated value, and the measured value ?
Considering the typical price, size and weight differentials between a f/1.2 and a f/1.4 lens, the fact that — when used on a DSLR like the Canon 5D2 — the effective speed of an f/1.2 lens, as well as its DoF and bokeh might be equivalent only to that of a f/1.5 lens is a quite relevant observation on a photo forum.
Except this "might be" is in fact a "is not".
Yawn. Where is the substantiation of your gratuitous assertion that it "is not" the case that a f/1.2 lens can behave as a f/1.5 lens with a DSLR like the Canon 5D2 ?
The apparent deformation affecting the exit pupil at larger image heights — that is, at points far from the image center — is essentially an elliptical one on photozone.de's pictures. Elliptical deformations affect the dimensions of a circular exit pupil only along one axis, and quite valid dimensional measurements can thus be made along the axis unaffected by the elliptical squeezing.
Yes, and had you been honest you would have seen only vertical size is smaller than expected. Horizontal size is fine.
Yawn. Where is the substantiation of your gratuitous assertion that the horizontal size is fine ? Are you even able to calculate what that horizontal size should be ?
I'm laughing. Really. "unsubstantiated" describes very well your own posts. Mine consist of facts I've checked, or when it's not I say it.
The exit pupil is not unkown, you failed to read it was between 35 and 45mm ? Yes, I measured it, because I know how to do...
Yawn. You "know how to do" an exit pupil distance measurement of a rangefinder lens with a 40mm focal length, yet you can't even provide that value with an incertitude better than 10mm ?
So, even if the NEX-5 had no offset microlenses — something that is impossible for me and you to know, — your assertion that your f/1.4 NEX-5 thought experiment "proves" that ray tilt cannot be a vignetting factor with a f/1.2 lens is totally irrelevant.
The fact you don't know that doesn't mean nobody knoes. I do know Nex has no offset microlenses.
Yawn. Where is the substantiation of your gratuitous assertion that the NEX has no offset microlenses ?
And I proved that tilted rays are a totally irrelevant issue for f/1.4 lenses with that lens/sensor combination.
Yawn. Where is that "proof" [sic] ?
And read one of my above post again : I also measured a 50~55mm exit pupil for m 50/2.8 macro Sony lens, and on FF sensor (Alpha 850/900) it has no vignetting. Meaning for f/1.2 lenses (the angle is almost the same), this issue is irrelevant too.
Yawn. I suppose you don't even realize that your measurements are meaningless if you cannot provide solid evidence that:
- the Sony 50/2.8 macro lens has an exit pupil distance relative to the imaging plane of 50-55mm
and
- the Sony A900 sensor does not have offset microlenses.
-
For an f/1.2 lens, the half-angle must thus be arctan(1/(2*1.2)) ~ 22.62 degrees.
On a symmetrical lens design. However, we are not dealing with a symmetrical lens design, were're dealing with a telelens. Ever hear about a concept called pupil magnification (http://toothwalker.org/optics/dofderivation.html#pupil)?
Cheers,
Bart
-
Will you ever be able to understand that it is not sufficient to make assertions, but that you must also provide a modicum of evidence or physics- or math-based reasoning supporting your assertions ?
Maybe this discussion might get somewhere if you provided some data and evidence versus your current attitude of "I know it all and you guys don't"
-
For an f/1.2 lens, the half-angle must thus be arctan(1/(2*1.2)) ~ 22.62 degrees.
On a symmetrical lens design. However, we are not dealing with a symmetrical lens design, were're dealing with a telelens. Ever hear about a concept called pupil magnification?
Do you actually read the information on the pages you're linking to ?
I've highlighted an, um, relevant section on that quite interesting web page.
What part of "paraxial approximation" and "sin(theta) ~= tan(theta)", or the relation "theta=arctan(tan(theta))" don't you understand ?
(http://farm5.static.flickr.com/4010/5158359139_b2fa57d07e.jpg)
-
Will you ever be able to understand that it is not sufficient to make assertions, but that you must also provide a modicum of evidence or physics- or math-based reasoning supporting your assertions ?
Maybe this discussion might get somewhere if you provided some data and evidence versus your current attitude of "I know it all and you guys don't"
Maybe this discussion might get somewhere if some people actually had experience with film-based photography and knew that with film, the diameter of an OOF disc is directly proportional to the lens' aperture.
Maybe this discussion might get somewhere if some people had a sufficient intellect to understand that when a digital sensor delivers an OOF disc whose diameter is less than what you'd get on film, that this necessarily implies that light rays are being lost — i.e. not recorded by the photosites.
-
Antisthenes, you're the one providing no evidence at all.
I won't even reply to you, you just proved your uselessness : your only argument is "b_z might have done some error here, so all he said is wrong".
Besides, the fact you need to criticize my English to feel better about your inability to counter my arguments with facts and not assumptions that I'm wrong is just lame.
I strongly advise everyone here not to pay attention to you or answer any more.
-
Antisthenes, you're the one providing no evidence at all.
I won't even reply to you, you just proved your uselessness : your only argument is "b_z might have done some error here, so all he said is wrong".
Besides, the fact you need to criticize my English to feel better about your inability to counter my arguments with facts and not assumptions that I'm wrong is just lame.
I strongly advise everyone here not to pay attention to you or answer any more.
And it would be of great help if the two of you would act in a civil manner, use your own names rather than some pseudonym (as most of us do because of a general request last year), and provide some evidence of your scientific background. There are enough of us with advanced degrees in engineering and the physical sciences that we don't engage in these shouting matches but do, in fact, understand the physics of all this.
Alan
-
On a symmetrical lens design. However, we are not dealing with a symmetrical lens design, were're dealing with a telelens. Ever hear about a concept called pupil magnification?
Do you actually read the information on the pages you're linking to ?
Yes. However, the part you seem to have difficulty with is that the rear element of the 85mm f/1.2 restricts (or is of equal size from the perspective of the sensor) the exiting cone of light. While not at the actual exit pupil distance, it would virtually act as a rear aperture if it wasn't dimensioned exactly right (that's why it's so large).
(http://www.photozone.de/images/8Reviews/lenses/canon_85_12_5d/fatso.jpg)
Photo courtesy of www.photozone.de (http://www.photozone.de/canon_eos_ff/502-canon_85f12ff)
The rear lens element is situated 6mm behind the lens flange, as indicated in my earlier diagram.
Cheers,
Bart
-
And it would be of great help if the two of you would act in a civil manner, use your own names rather than some pseudonym (as most of us do because of a general request last year), and provide some evidence of your scientific background.
A "great help" for whom ?
I happen to be, in real life, in a professional position such that having my name traceable to some random Internet forum would be against my own, and many other people's interests.
I also have the relevant degrees in science and professional experience thst qualify me to give me a reasoned opinion on the topic at hand.
That background also makes me immediately recognize the vacuous factoids and "logic" proferred by some people, who presume that they are sufficiently kowledgeable to dismiss DxO's and Mark Dubovoy's quite competent observations. The presumption of such people can elicit, as you've seen, a fairly critical reaction from myself.
Should you think that you have a sufficient professional qualification to contest or disprove even a single one of the elements in my exposition, please be my guest.
-
Maybe this discussion might get somewhere if you provided some data and evidence versus your current attitude of "I know it all and you guys don't"
Maybe this discussion might get somewhere if some people actually had experience with film-based photography and knew that with film, the diameter of an OOF disc is directly proportional to the lens' aperture.
Check, I do
Maybe this discussion might get somewhere if some people had a sufficient intellect to understand that when a digital sensor delivers an OOF disc whose diameter is less than what you'd get on film, that this necessarily implies that light rays are being lost — i.e. not recorded by the photosites.
Do you have any other evidence than the pictures on photozone.de to substantiate your claim that the oof disk at f1.2 on the 5DMkII is smaller than it would be on film? Because I don't think those pictures prove that.
-
Do you actually read the information on the pages you're linking to ?
Yes. However, the part you seem to have difficulty with is that the rear element of the 85mm f/1.2 restricts (or is of equal size from the perspective of the sensor) the exiting cone of light. While not at the actual exit pupil distance, it would virtually act as a rear aperture if it wasn't dimensioned exactly right (that's why it's so large).
The part you seem to have difficulty with is that the half-angle of the light cone emitted by a lens is determined essentially by its aperture — i.e. the particular f-number setting of that lens, which isn't necessarily the lens' full open aperture.
If you don't believe me, let me quote from the very web page to which you provided the link:
"Indeed, at infinity focus the apical angle of the light cone that impinges upon the film is the same for all lenses at the same F-number"
The diameter of the rearmost element can restrict the light cone's apical angle, but does not determine it.
Furthermore, the position of that last element does not define the distance between the imaging plane and the exit pupil.
An accurate determination of the exit pupil position is required for your chief ray tilt angle calculations to have any semblance of reality.
As an example, a lens that is "telecentric" on the image side is one that has its exit pupil at a large, possibly infinite distance from the image plane.
In such a case, it is obvious that the tilt angle of the chief rays — which, from the sensor's point of view, emanate from the exit pupil — will exhibit almost no variation across the sensor's entire diagonal.
With your simplistic approach, you'd probably assign an entirely fictitious position to the exit pupil, resulting in a totally incorrect chief ray tilt angle behavior model.
To repeat, your strange fixation that the distance or diameter of the lens' last element are somehow relevant pieces of information to determine the exit pupil — a.k.a. virtual aperture — position as well as chief ray tilt angles is totally misguided, and your calculations are therefore meaningless.
-
Do you have any other evidence than the pictures on photozone.de to substantiate your claim that the oof disk at f1.2 on the 5DMkII is smaller than it would be on film? Because I don't think those pictures prove that.
Will you please first provide some evidence that you've bothered to read and understand the contents of my posts — e.g. these sentences:
A simple way to assess the critical incidence angle — that is, the acceptance angle — above which the rays cannot be recorded anymore is thus to measure the diameter of the out of focus disc formed by a point source.
[..]
With some digital sensor designs, acceptance angles can be quite limited. At some point, the linear relationship between the lens aperture and the OOF disc diameter recorded on the picture must then break down, as the tilted marginal rays cannot reach the photodiodes anymore.
Given that anybody with an APS or full-size DSLR and a f/1.2 or f/1.4 lens can verify that, even with the lens wide open, an OOF point lioght source appears at the center of the image as a perfect circle [blah]
That "anybody" in the above sentence also includes you, pegelli.
-
Will you please first provide some evidence that you've bothered to read and understand the contents of my posts — e.g. these sentences:
That "anybody" in the above sentence also includes you, pegelli.
I do read and understand everything you write, it's no more than a qualitative description of what you are failing to quantify. I conclude from your remarks that the answer to my question is "no", which means your whole evidence is on shaky ground.
-
I do read and understand everything you write, it's no more than a qualitative description of what you are failing to quantify. I conclude from your remarks that the answer to my question is "no", which means your whole evidence is on shaky ground.
Nope. You think you understand, but the noetic reality is quite different.
You don't seem to be the sharpest tool in the drawer, so let me try to put you on the right track.
Me:
"If you were to hold an apple some distance above the ground, and let it go, it would fall towards the ground. Anybody with an apple can verify that"
You:
"can you provide across the Internet some evidence that the apple would, as you assert, fall ?"
Me:
"well, anybody with access to an apple could verify it for himself if he/she bothered to read the description of that (trivial to perform) experiment'"
You:
"Your description is purely qualitative, and I conclude that you cannot provide, across the Internet, evidence that the apple would fall. I won't bother to perform the experiment myself, and I conclude that your evidence is on shaky ground"
Me:
::)
-
I refuse to use my real name on internet, sorry for that.
As for the background check, I'm still a student, in one of the best French scientific schools. Namely the Ecole Normale Superieure de Lyon, where I had Cedric Villani (one of this year Fields medalist) as a teacher and tutor. Obviously I'm not in an English major, but mathematics major and physic minor.
Pegelli, you're fighting against windmills.
-
As for the background check, I'm still a student, in one of the best French scientific schools.
A good English word to describe your ilk is sophomoric
Obviously I'm not in an English major, but mathematics major and physic minor.
Tidbit: Besides my science background (my thesis advisor was a Nobel laureate, BTW), I happen to have a French minor, and judging by your prose, my French happens to be miles above your English skills...
-
Antisthenes said...
You don't seem to be the sharpest tool in the drawer, so let me try to put you on the right track.
Cough, cough, cough.
A good English word to describe your ilk is sophomoric
Hmmmm...
Tidbit: Besides my science background (my thesis advisor was a Nobel laureate, BTW), I happen to have a French minor, and judging by your prose, my French happens to be miles above your English skills...
Yeah, sure. Unsubstantiated ad hominem attacks.
Once more (parce que je pense que le franc n'est pas tombe...), I believe this is a place where real people, using their real names, enjoy disagreeing in a respectful way, a skill you have obviously not mastered.
-
Once more (parce que je pense que le franc n'est pas tombe...), I believe this is a place where real people, using their real names, enjoy disagreeing in a respectful way, a skill you have obviously not mastered.
Should you think that you have a sufficient professional qualification to contest or disprove even a single one of the technical elements in my exposition, please be my guest.
Being respectful to others, in my book, includes refraining as much as possible from proffering ill-informed, ignorant opinions and judgments about a technical topic that demonstrate nothing more than the vacuity and presumption of the person behind them.
The problem with such people is that they are so ignorant that:
1) they don't even realize their ignorance about some particular technical topic
2) they can't recognize or appreciate the application of real knowledge.
In particular, a sure sign on photography-oriented forums identifying these conceited types is their criticism and summary judgment of some DxO-related methodology, even though they don't have a shred of an inkling of the qualification to really understand the issue at hand.
I have no connection with DxO whatsoever, other than being an occasional visitor of their site.
Still, I think I know enough about some domains germane to DxO's activities, including applied mathematics, physics and engineering to realize that the odds of some random bloke being right about some arcane technical topic, and DxO wrong, is vanishingly small.
DxO is of course not special, as the same thing can be said for any number of domains, including camera design, IT, law, economy, marketing, history, music, foreign relations, literature etc.
The day conceited ignoramuses become intelligent enough to appreciate the negligible odds of they having a better insight about a particular topic than the people, groups or companies that have cultivated real, specialized knowledge, ability and experience is the day this kind of tedious thread will disappear from the Internet.
Meanwhile, in technical, fact- and science-based topics, I don't think it's an entirely bad thing for some people to occasionally come across a person who doesn't tolerate fools gladly, and makes them realize the value of restraint and of thinking twice about whether their background really equips them to issue informed opinions and judgments.
;)
-
Posted by: Antisthenes
Should you think that you have a sufficient professional qualification to contest or disprove even a single one of the technical elements in my exposition, please be my guest.
Tempting.
Posted by: Antisthenes
Considering the typical price, size and weight differentials between a f/1.2 and a f/1.4 lens, the fact that — when used on a DSLR like the Canon 5D2 — the effective speed of an f/1.2 lens, as well as its DoF and bokeh might be equivalent only to that of a f/1.5 lens is a quite relevant observation on a photo forum.
This is what you call a relevant fact. I happen to have a bunch of lense and can easily check if your facts are grounded in reality.
So here we go.
(http://dl.dropbox.com/u/2445259/12vs.jpg)
Can you spot the image that was shot at 1.2 on a Canon 5D MK II and can you spot the one that was shot at 1.6?
Any difference in depth of field and/or bokeh? Yeah, I though so.
Any photographer who has shot eyes at 1.2 and eyes at 1.6 knows what your reality is worth.
BTW, I shot this at 3200 iso, in poor lighting, because what can be seen is that the 1.2 image is noisier, which seems to indicate that the automatic gain boost that has already been observed has an impact on the quality of the image in extreme conditions (it can be measured on more normal exposures, but is not as obvious).
So that's it: a clear statement of fact in one of your messages doesn't appear to be connected with the real world.
While I don't rebreathe as many Nobel blessed oxygen molecules as you do, you'll easily understand I cast a doubtful eye on your digressions after that divergence.
PS: and yes, I found the text "knowing E we know everything" particularly well suited to the occasion. <G>
-
You don't seem to be the sharpest tool in the drawer
You mean "sharp" as in 'con-artist', yep, there you outdo me but I have no problem with you being very proud of that.
Btw, interesting dissertation on the apple, your first doctorate was under Newton I guess?
Why don't you go here (http://www.ashwooduniversity.net/) or here (http://www.realisticdiplomas.com/ProductInfo.aspx?productid=FAKE-PHD-DOCTORATE-DEGREE-DIPLOMA) and add to your impressive scientific qualifications. Who knows, next time you might be two or three times as right as you are now.
;D
-
To whom it may concern:
The increasing ad hominem attacks in this thread start making reading it increasingly unpleasant ...
Please stop it.
<°)(((o><
-
To whom it may concern:
The increasing ad hominem attacks in this thread start making reading it increasingly unpleasant ...
I agree.
And therefore, in an attempt to further the discussion in a more constructive direction, I've modified my initial diagram on the assumption that the Pupil Magnification effects at near infinity focus should not play a role. Hence I've attempted to determine the position of the exit pupil a bit more accurately as shown in the updated diagram:
(http://www.xs4all.nl/~bvdwolf/temp/OPF/EP2FP_V2.png)
Interestingly, the maximum angle of incidence has reduced from my more pessimistic diagam earlier. So the question becomes even more to the point; How hard is it for a microlens to cope with that? It would be nice to have some answers on that, rather than attempts to side track the discussion.
In fact, I think another issue may be more prominent, and potentially the reason why manufacturers attempt to compensate for such a small loss in the first place, and I've not seen it addressed before if I'm not mistaken:
The effect of reflecting light from the sensor's IR filter / cover-glass. While the protective layer is usually anti-reflection coated, it does still reflect a certain amount of light, and the coating introduces a color cast for more oblique rays of incident light.
Cheers,
Bart
-
So the question becomes even more to the point; How hard is it for a microlens to cope with that? It would be nice to have some answers on that, rather than attempts to side track the discussion.
Here is some data for green light at the center of the field for one Kodak sensor with micro-lenses, the 31MP 44x33 with 6.8 micron pixels used in the H3D-31 etc. See figure 6, page 15:
http://www.kodak.com/global/plugins/acrobat/en/business/ISS/datasheet/fullframe/KAF-31600LongSpec.pdf
It seems that at 27º, the sensitivity is off by a factor of 0.7: a loss of 30% or about 1/2 stop.
For comparison, in another Kodak sensor with the same 6.8 micron pixel design but no microlenses (the 48x36mm 39MP KAF-39000) the loss is only about 10% at 27º and increases to 30% only at 40º: see figure 8, page 15 of
http://www.kodak.com/global/plugins/acrobat/en/business/ISS/datasheet/fullframe/KAF-39000LongSpec.pdf
In fact, I think another issue may be more prominent, and potentially the reason why manufacturers attempt to compensate for such a small loss in the first place, and I've not seen it addressed before if I'm not mistaken:
The effect of reflecting light from the sensor's IR filter / cover-glass.
An interesting thought: the split between reflection and transmission increases as the angle of incidence increases, so that could also be a factor. But I suspect it is a smaller factor than the micro-lens effect: a table on page 28 of the above document suggests that cover glass reflection is less than 2%. Even if that is for the best case of perpendicular incidence, refection losses at 27º seem unlikely to be near the above 30% figure. I suppose that with some reasonable assumptions on the refractive index of the cover glass, one could compute the reflection/transmission fractions as a function of incident angle from basic optics formulas, but I am too lazy to look that stuff up.
-
Here is some data for green light at the center of the field for one Kodak sensor with micro-lenses, the 31MP 44x33 with 6.8 micron pixels used in the H3D-31 etc. See figure 6, page 15:
http://www.kodak.com/global/plugins/acrobat/en/business/ISS/datasheet/fullframe/KAF-31600LongSpec.pdf
It seems that at 27º, the sensitivity is off by a factor of 0.7: a loss of 30% or about 1/2 stop.
For comparison, in another Kodak sensor with the same 6.8 micron pixel design but no microlenses (the 48x36mm 39MP KAF-39000) the loss is only about 10% at 27º and increases to 30% only at 40º: see figure 8, page 15 of
http://www.kodak.com/global/plugins/acrobat/en/business/ISS/datasheet/fullframe/KAF-39000LongSpec.pdf
Kodak wrote up their experiences with those sensors, and specifically addressed addressed the microlens question. The article is here (warning: highly technical): http://wwwcafr.kodak.com/global/plugins/acrobat/en/business/ISS/supportdocs/31Mp_and_39Mp_Full-Frame_CCD_paper.pdf
There's also a less technical Dalsa article here: http://www.dalsa.com/public/dc/documents/Image_Sensor_Architecture_Whitepaper_Digital_Cinema_00218-00_03-70.pdf
If you access to SPIE, there's also the Catryse/Liu/El Gamal article "QE Reduction due to Pixel Vigneting in CMOS Image Sensors"; I don't have a download location for that.
Long and the short of it - microlens efficiency depends on a LOT of things :)
Sandy
-
Here is some data for green light at the center of the field for one Kodak sensor with micro-lenses, the 31MP 44x33 with 6.8 micron pixels used in the H3D-31 etc. See figure 6, page 15:
http://www.kodak.com/global/plugins/acrobat/en/business/ISS/datasheet/fullframe/KAF-31600LongSpec.pdf
It seems that at 27º, the sensitivity is off by a factor of 0.7: a loss of 30% or about 1/2 stop.
Thanks for that document. BTW, 70% is 1/3rd of a stop loss if I'm not mistaken. However, the way I interpret that figure 6 on page 15 is that they're talking about full diagonal performance. In the center of the sensor array (where the DxO measurements are taken) there is no significant loss. That means that any automatic correction in gain will lead to relative overexposure in the center. I know we're talking about a small correction, but then why bother at all and confuse matters? Afterall, it's a minor correction compared to the vignetting and light fall-off (some 1.8 stops on a 5D2 at f/1.2)! The distance to the corner of a 24x36mm sensor is onlly 21.6 mm, so the angle of the center rays hit the sensor at an angle of some 20 degrees (if my estimate of the exit pupil position is correct).
I would agree that such a (gain, or exposure) correction is best left to the photographer. Some photographers do know what they're doing, and don't shoot on P-mode only (some even bracket to optimize output ;)).
For comparison, in another Kodak sensor with the same 6.8 micron pixel design but no microlenses (the 48x36mm 39MP KAF-39000) the loss is only about 10% at 27º and increases to 30% only at 40º: see figure 8, page 15 of
http://www.kodak.com/global/plugins/acrobat/en/business/ISS/datasheet/fullframe/KAF-39000LongSpec.pdf
Which demonstrates that different designs (this one doesn't have the surrounding buffer and light shielded pixels around each sensel) achieve different results. There are too many design differences to only attribute it to the microlenses.
An interesting thought: the split between reflection and transmission increases as the angle of incidence increases, so that could also be a factor. But I suspect it is a smaller factor than the micro-lens effect: a table on page 28 of the above document suggests that cover glass reflection is less than 2%. Even if that is for the best case of perpendicular incidence, refection losses at 27º seem unlikely to be near the above 30% figure.
It's hard to say how much of an effect the individual components contribute to the resulting cumulative effect graphs without comparing with a coverglass-less version.
Cheers,
Bart
-
To whom it may concern:
The increasing ad hominem attacks in this thread start making reading it increasingly unpleasant ...
Please stop it.
<°)(((o><
Christoph, you're right and I'm sorry. But after so many personal attacks and insults to several members (incl. myself) it felt good just to vent back. Won't happen again if this remains a respectful technical discussion.
Let me try to make up to you and the other respectful forum members by adding another technical observation. If you look at the diameter ratio of the out of focus circles at the different apertures on photozone.de my conclusion is that even if you assume these are from an infinitely small point light source the lens at f1.2 has an out of focus circle that would correspond with an ~f1.4 image (so already better than the f1.5 that was claimed). Since the out of focus highlight is probably not infinitely small (but a diameter of maybe something like 10-15% of the out of focus circle) the ratio of the circles of confusion points more to something like between ~f1.2 and ~f1.3. So it seems the magnitude and impact is much smaller as some want us to believe, if at all there. Obviously these are estimations based on hand measurement of out of focus circles that were not in the middle of the frame, so it's probably best to wait with definite conclusions until the additional DxO measurements as promised in Mark Dubovoy's article are available. However my calculations are in line with Pierre Vandevenne's test as well as BartvanderWolf's updated diagram, so I'm looking forward to see more from DxO in the future and get to the bottom of this issue.
-
Just to summarize the findings so far, from a practical point of view
- there is indeed a "hidden" gain (ISO) boost at wide apertures, 1.2 and to a lesser extent at 1.4. This is both measurable and visible in some conditions such as low light and high ISO for 1.2.
as far as photography is concerned, using 1.2 and 3200 ISO is rarely a creative choice. If you do it, you need to do it and can live with somewhat suboptimal quality.
- the 1.2 lenses are not as good wide open as they are stopped down. But that's hardly unexpected...
- bokeh and dof appear to behave as expected. Used in the conditions, 1.2 does provide a shallower DoF than 1.4, 1.6 and obviosuly 1.8. The Canon 85mm 1.8 and the 50mm 1.4 are excellent deals from a price/quality point of view, but not to the point that they perform identically to their higher priced siblings.
-
Thanks for that document. BTW, 70% is 1/3rd of a stop loss if I'm not mistaken. However, the way I interpret that figure 6 on page 15 is that they're talking about full diagonal performance.
A few comments.
1. Stops are a logarithmic/power scale, so 70% is half a stop (as 50% is one stop). A one stop decrease in illumination is a factor of 1/2 in light, so a half stop increase is the square root of that, or sqrt(0.5) or about 0.7.
2. The caption says that these are measurements at the center of the sensor:
"Normalized Angle QE at the center of the sensor under green LED illumination."
With low f-stops, there is a reduced sensitivity to light at the outside edges of the broad light cone everywhere, though the problem gets worse towards the edges. So correcting for the part of the sensitivity loss experienced everywhere including at the center while leaving the additional off-center effect to lost-processing correction makes sense.
3. It has been said in a great many places that micro-lenses, in the usual designs, decrease the acceptance angle, and that this is the dominant reason that MF sensors almost never have micro-lenses, despite the penalty in low-light performance from omitting them. Those other design differences you an Kodak mention, like the narrower "windows", are there purely for the sake of the use of micro-lenses, being necessitated by the micro-lenses, to avoid cross talk (light entering through one CFA but then getting counted in a neighboring photosite). This cross-talk is worse with microlenses (though also present without) because some light gets bent even more off-perpendicular by the microlenses. This is discussed nicely in the papers that sandymc mentions above (thanks sandymc!)
But to put it simply: Kokak acknowledges the problems cuased by the worse off-perpendicular sensitivity of its sensors with microlenses (uses strategies like off-set microlesnes to mitigate the effect), and so if Kodak could reduce this disadvantage of it sensors with microlenses by changing other details like the masking to be more like in its sensors without micro-lenses, surely Kodak would do it! Any worsening of off-perpendicular sensitivity in can only be explained as something necessitated by the design needed when micro-lenses are present.
I could dig up a Dalsa reference on microlenses reducing acceptance angle if you want: Dalsa has a new approach where the microlenses are far closer to the wells, which essentially eliminates the off-perpendicular" disadvantage. Or to paraphrase what Kodak says about its sensors without microlenses, the effect is about equal in magnitude to the cos^4 effect in play with any sensor, including film. So the effect is far worse than "cos^4" with Kodak's microlensed sensors!
-
A few comments.
1. Stops are a logarithmic/power scale, so 70% is half a stop (as 50% is one stop). A one stop decrease in illumination is a factor of 1/2 in light, so a half stop increase is the square root of that, or sqrt(0.5) or about 0.7.
You are right. Must have hit the wrong calculator button (Log(0.7) / Log(2) = -0.51 stops).
I could dig up a Dalsa reference on microlenses reducing acceptance angle if you want.
That's not necessary, I can find it if needed. There obviously is a limit to the acceptance angle of a microlens, but micro-lens technology has also advanced (e.g. offset micro-lenses (not good for shifted lenses) and different shapes and refractive indexes). Also, Mark's essay showed DxO charts with maximum size 24x36mm sensors, so the maximum angle of incidence near a corner is more limited than on a MF sensor.
In the center of the sensor array (as in the DxO measurements) most of the exit pupil rays strike the sensels almost perpendicular, and those from the edge of the exit pupil will be refracted to something closer to perpendicular. So whatever DxO measured, it was almost certainly not impacted by shading effects.
Cheers,
Bart
-
There obviously is a limit to the acceptance angle of a microlens, but micro-lens technology has also advanced (e.g. offset micro-lenses (not good for shifted lenses) and different shapes and refractive indexes).
Of course the technology is advancing, but the evidence of quite recent Kodak FF CCDs with microlenses is that they still have far more severe off-perpendicular sensitivity fall-off than sensors without (30% loss or 1/2 stop at 27º with, vs. 10% at 27º and 30% at 40º without.) And off-setting is irrelevant to the problem I am talking about which applies even at the center; off-setting is only relevant to the additional problems with wide-angle (low exit pupil) lenses, an issue that seems of little importance with DSLRs when used with SLR lenses, due to modern mainstream SLR lenses having fairly high exit pupils relative to sensor size.
Perhaps you are thinking that somehow Kodak is behind what Canon, Sony etc. are achieving with their CMOS and ILT CCDs, but the opposite is probably true: CMOS and ILT CCD sensors require stronger microlenses than FF CCD (the wells are a smaller portion of the total photosite) making the problem worse, and it is yet worse for CMOS sensors (front-illuminated ones anyway as all current DSLR CMOS sensors are) as the stack of transistors on top of the photosite forces the microlenses to be further from the well. Some quotes from Dalsa page 7 of http://www.dalsa.com/public/dc/documents/Image_Sensor_Architecture_Whitepaper_Digital_Cinema_00218-00_03-70.pdf
- about ILT CCD in particular, but true in general:
At low f-numbers, microlensed pixels can suffer from vignetting, pixel crosstalk, light scattering, diffraction and reduced MTF—all of which can hurt their resolving power.
- about T3 CMOS in particular, and even more so of T4/T5 CMOS:
The tradeoffs involved with microlenses are more pronounced with CMOS imagers since the microlenses are farther from the photosensitive surface of the pixel due to the “optical stack” of transistors.
In the center of the sensor array (as in the DxO measurements) most of the exit pupil rays strike the sensels almost perpendicular, and those from the edge of the exit pupil will be refracted to something closer to perpendicular.
Dalsa disagrees on that last point, indicating instead that the already oblique rays are bent even more off-perpendicular. See the top-center figure on the same page 7 as cited above. Or just note that microlenses are converging lenses.
This illustrates the potential problem: light from the edge of a low f-stop light cone, at any part of the sensor, can be sent off to the side of the well, and be either not detected or get smeared into an adjacent well. Kodak's masking is to avoid the latter.
-
Highly entertaining! I'm glad we are beginning to see the results of some test images at shallow DoF, but I can't help being reminded of the following amusing parable. ;D
In the year of our Lord 1432, there arose a grievous quarrel among the brethren over the number of teeth in the mouth of a horse. For thirteen days the disputation raged without ceasing.
All the ancient books and chronicles were fetched out, and wonderful and ponderous erudition such as was never before heard of in this region was made manifest.
At the beginning of the fourteenth day, a youthful friar of goodly bearing asked his learned superiors for permission to add a word, and straightway, to the wonderment of the disputants, whose deep wisdom he sore vexed, he beseeched them to unbend in a manner coarse and unheard-of and to look in the open mouth of a horse and find answer to their questionings.
At this, their dignity being grievously hurt, they waxed exceeding wroth; and, joining in a mighty uproar, they flew upon him and smote him, hip and thigh, and cast him out forthwith. For, said they, surely Satan hath tempted this bold neophyte to declare unholy and unheard-of ways of finding truth, contrary to all the teachings of the fathers. After many days more of grievous strife, the dove of peace sat on the assembly, and they as one man declaring the problem to be an everlasting mystery because of a grievous dearth of historical and theological evidence thereof, so ordered the same writ down.
-
Of course the technology is advancing, but the evidence of quite recent Kodak FF CCDs with microlenses is that they still have far more severe off-perpendicular sensitivity fall-off than sensors without (30% loss or 1/2 stop at 27º with, vs. 10% at 27º and 30% at 40º without.)
I don't think that applies in general, while it does in this specific case where one sensor design only uses a small part (< 1/2 of the sensel pitch) of the surface to collect light, and the other half as signal buffer). Besides, compare that performance with the supposed Cos(a)^4 light fall-off (e.g. on film).
Perhaps you are thinking that somehow Kodak is behind what Canon, Sony etc. are achieving with their CMOS and ILT CCDs, but the opposite is probably true: CMOS and ILT CCD sensors require stronger microlenses than FF CCD (the wells are a smaller portion of the total photosite) making the problem worse, and it is yet worse for CMOS sensors (front-illuminated ones anyway as all current DSLR CMOS sensors are) as the stack of transistors on top of the photosite forces the microlenses to be further from the well.
No, I'm not thinking Kodak is behind the curve, there are more Kodak sensors being used in all sorts of equipment than most people know. It's a very mature division within that company.
Some quotes from Dalsa page 7 of http://www.dalsa.com/public/dc/documents/Image_Sensor_Architecture_Whitepaper_Digital_Cinema_00218-00_03-70.pdf
- about ILT CCD in particular, but true in general:
At low f-numbers, microlensed pixels can suffer from vignetting, pixel crosstalk, light scattering, diffraction and reduced MTF—all of which can hurt their resolving power.
- about T3 CMOS in particular, and even more so of T4/T5 CMOS:
The tradeoffs involved with microlenses are more pronounced with CMOS imagers since the microlenses are farther from the photosensitive surface of the pixel due to the “optical stack” of transistors.
Dalsa disagrees on that last point, indicating instead that the already oblique rays are bent even more off-perpendicular. See the top-center figure on the same page 7 as cited above. Or just note that microlenses are converging lenses.
Maybe you misread my remark, I was specifically talking about the center of the sensor array. The only obliquness there is from the marginal rays from the exit pupil, the rest around the chief ray goes in almost perpendicular. Nobody contests that in the corners of the sensor array the task for microlenses is more daunting. Looking at the DxO graph in Mark's essay, CMOS doesn't seem to do that bad in the center of the sensor array, compared to CCD (although there are more variables in play that could cause that).
This illustrates the potential problem: light from the edge of a low f-stop light cone, at any part of the sensor, can be sent off to the side of the well, and be either not detected or get smeared into an adjacent well. Kodak's masking is to avoid the latter.
Yes, manufacturers are very well aware of optical cross-talk, and take precautionary measures, e.g. by varying the thickness/shape/and refractive index (which can also influence internal reflection), and adding exposure masks, to reach a better overall optimum, as demonstrated here:
http://www.lumerical.com/fdtd_microlens/cmos_image_sensor_pixel_microlens.php
And here is another interesting paper (published in 2000, it may be a bit dated, but it shows the considerations nicely):
http://www-isl.stanford.edu/groups/elgamal/abbas_publications/C074.pdf
Although it deals with sensors without microlenses, they do add at the end of the document:
"Microlenses: QE can be increased using microlenses.The design and manufacturing of these
microlenses must, however, take into consideration the geometry of the tunnel. The improvement in
on-axis QE can be considerable. For off-axis pixels, it is not clear how much improvement, ifany, is
achieved without the use of a telecentric imaging lens or an unconventional angle-dependent microlens
design, which may not be practical. Another important consideration is the material needed. To
focus the light onto the photodiode surface, the focal length of the microlens may be too long to be
manufactured using silicon dioxide. In this case, other materials such as polymethyl methacrylate
(PMMA) may be used, which increases the cost of the microlens fabrication.".
Cheers,
Bart
-
I don't have a 5dII and a 85/1.2 L lens, but I do have a Sony A850 and an older Minolta AF 85/1.4 G lens (very nice combo for portraits)
So I decided to test and see if the circle of confusion at f 1.4 became smaller as expected by comparing it to the circle of confusion at f2.8.
Here's the three images:
Sharply rendered highlight:
(http://i227.photobucket.com/albums/dd43/pegelli/DSC03623.jpg)
out-of focus @ f2.8:
(http://i227.photobucket.com/albums/dd43/pegelli/DSC03624.jpg)
out of focus @ f1.4:
(http://i227.photobucket.com/albums/dd43/pegelli/DSC03625.jpg)
Since the Minolta 85/1.4 has some internal focussing I made the out-of-focus shots at the same distance setting (~1 meter) but with a 12 mm extension ring added. So I know for sure all 3 shots are taken at exactly the same focal lenth.
My conclusion is that the diameter ratio of the circles of confusion between the f1.4 shot and the f2.8 shot is exactly 2, so the out-of-focus and dof rendering at f1.4 is exactly as expected. So the dof reduction effect and shielding of marginal rays at large f-stops as hinted at by Mark Dubovoy does not show up with this camera/lens combination.
Unfortunately no f1.2 A-mount lenses were ever produced, so we'll never know how that would work. There are however also a 35/1.4 and 50/1.4 for the A-mount but since I have neither I cannot test those at this moment.
I'll do some tests later to see if I can find any "ISO" gain as Mark is describing with this combo, but since that interests me much less it will have to wait until I make some time for that.
-
I don't think that applies in general, while it does in this specific case where one sensor design only uses a small part (< 1/2 of the sensel pitch) of the surface to collect light, and the other half as signal buffer).
The Kodak sensors compared are both full frame CCD's, so as far as I can tell they have no signal buffer, just a little space at the edges for lateral overflow drains ... and one of the main claims in that Kodak paper is that the design reduces the space needed for the LODs, giving an impressively high 69% of the photosite being electron well (some is then masked off from direct illumination, but can still hold electrons). This loss to LODs is less than with the CMOS or interline CCD designs in most SLRs.
To repeat, the need for a smaller "window" (more masking) over the well in the microlens design is forced by the difficulties like pixel cross-talk that are made worse by microlenses; it is not some dumb stupid thing that Kodak did with the KAF-31000, degrading its performance in one respect, while knowing how to to better in another (the KAF-39000 without microlenses).
Since I am the only one providing actual data about actual sensors and quotes from sensor making companies, can you provide authoritative evidence to the contrary, like spec's a sensor with microlenses than has off-perpendicular sensitivity "at the center of the sensor array" as good as in that non-microlens sensor? You might find one such, a new one from Dalsa using its new "low profile" microlens design, but that approach only works with CCDs, not CMOS, as Dalsa itself explains, so I doubt that any recent mainstream SLR sensor (all CMOS with microlenses) has such performance.
Maybe you misread my remark, I was specifically talking about the center of the sensor array.
The only obliquness there is from the marginal rays from the exit pupil, the rest around the chief ray goes in almost perpendicular.
No, I understand it perfectly: as I have said repeatedly, we are both talking about an effect that applies to marginal rays at low f-stops everywhere on the sensor, including at the center of the sensor where the chief ray is perpendicular but marginal rays are substantially off-perpendicular. And the illustrations that I referred to in the Dalsa document, the middle column, are exactly for this situation.
Yes, manufacturers are very well aware of optical cross-talk, and take precautionary measures, e.g. by varying the thickness/shape/and refractive index (which can also influence internal reflection), and adding exposure masks, to reach a better overall optimum ...
And from here on we agree, and so do your sources, neither of which in the slightest contradict my point that light loss also occurs at the center of the sensor array with low f-stops. The problems increases away from the center, and only produces "vignetting" (variation in detected luminosity detected de to spatial variation in QE) off-center, and so the off-center effect is discussed more but produces somewhat reduced QE everywhere, even at center, with sufficiently low f-stops. Also no MF lens has an f-sop low enough for this to be an issue, whereas many have low exit pupils, and Kodak explicitly says that these designs are optimized for MF, so the low f-stop effect is not worth Kodak's discussing in that document.
Until you come up with superior authority or arguments, I have to go with Kodak's statement that
the KODAK KAF-39000 Image Sensor (39 Mp) is designed without microlenses to maximize incident light-angle response ... the critical crosstalk angle is increased
and of the KAF-31000 that
The primary drawback of this design is reduced incident light angle response compared to a non-microlens design
and Dalsa's that
At low f-numbers, microlensed pixels can suffer from vignetting, pixel crosstalk, light scattering, diffraction and reduced MTF ...
The tradeoffs involved with microlenses are more pronounced with CMOS imagers since the microlenses are farther from the photosensitive surface of the pixel due to the “optical stack” of transistors.
-
The Kodak sensors compared are both full frame CCD's, so as far as I can tell they have no signal buffer, just a little space at the edges for lateral overflow drains ... and one of the main claims in that Kodak paper is that the design reduces the space needed for the LODs, giving an impressively high 69% of the photosite being electron well (some is then masked off from direct illumination, but can still hold electrons). This loss to LODs is less than with the CMOS or interline CCD designs in most SLRs.
To repeat, the need for a smaller "window" (more masking) over the well in the microlens design is forced by the difficulties like pixel cross-talk that are made worse by microlenses; it is not some dumb stupid thing that Kodak did with the KAF-31000, degrading its performance in one respect, while knowing how to to better in another (the KAF-39000 without microlenses).
From the KAF-31600 data sheet, page 4, you linked to:
DESCRIPTION
The KAF-31600 is a dual output, high performance color
array CCD (charge coupled device) image sensor with
[...]
Microlenses are added for improved sensitivity.
The photoactive pixels are surrounded by a border
of buffer and light-shielded pixels.
I interpreted that (enforced by SEM images of the sensel structure I saw somewhere) as all individual photoactive pixels being surrounded by other pixels. Obviously that would reduce the photosensitive area, and microlenses would be needed to refocus the light onto the sensitive areas. On second reading it could also mean (and would be more likely) that there is a row of shielded pixels around the entire array of photosensitive sensels.
On the KAF-39000 sheet description, there is no mention of surrounding pixels/shielding, suggesting a significantly different design, at least that's how I interpreted it.
All I'm saying is that differences in design make comparisons difficult, and certainly not a basis to generalize on, and the (also mentioned in the datasheet, offset) microlenses were instrumental in increasing the overall sensitivity of the (more special ?) design. Despite the identical 6.8 micron sensel pitch, the datasheets mention different sizes for the sensor array, but perhaps they are measuring 2 different things, that also gave me the impression that there is something else going on. Maybe it's just inconsistent measurements, e.g. total area versus effective area. They also mention different numbers of pixels, confusing ...
I would rather draw general conclusions about microlenses based on identical sensor array designs, only differing in the use of microlenses or not, that's all. That will probably be data that's hard to find.
Since I am the only one providing actual data about actual sensors and quotes from sensor making companies, can you provide authoritative evidence to the contrary, like spec's a sensor with microlenses than has off-perpendicular sensitivity "at the center of the sensor array" as good as in that non-microlens sensor? You might find one such, a new one from Dalsa using its new "low profile" microlens design, but that approach only works with CCDs, not CMOS, as Dalsa itself explains, so I doubt that any recent mainstream SLR sensor (all CMOS with microlenses) has such performance.
We'll see what can be found, but not all manufacturers have such easily accessable datasheets as Kodak.
Until you come up with superior authority or arguments, I have to go with Kodak's statement that
the KODAK KAF-39000 Image Sensor (39 Mp) is designed without microlenses to maximize incident light-angle response ... the critical crosstalk angle is increased
and of the KAF-31000 that
The primary drawback of this design is reduced incident light angle response compared to a non-microlens design
You make it sound like we're in some sort of pissing contest, which we're not (I'm not). We agree for the most part, we just cannot find independent sources that publicly quantify what we'd like to know without additional guesswork.
and Dalsa's that
At low f-numbers, microlensed pixels can suffer from vignetting, pixel crosstalk, light scattering, diffraction and reduced MTF ...
The tradeoffs involved with microlenses are more pronounced with CMOS imagers since the microlenses are farther from the photosensitive surface of the pixel due to the “optical stack” of transistors.
Yes, that's all true in general, "microlensed pixels" can suffer from all that but do so in differing degrees, and changing the fill-factor with microlenses does indeed change the MTF. No discussion needed. Now to find 2 comparable sensor array designs, preferably the same, where one has microlenses and the other one has not, and the same between CCD and CMOS (although we already know that CMOS has more layers in general, and thus sensitivity to tunneling effects). Then we will be able to quantify the differences, without design variables that make such a comparison shaky at best.
Cheers,
Bart
-
Could I maybe ask what maybe a really dumb question? F stop is just length/width, ie a ratio. A relative measure, an indicator of lens mechanics, not an absolute measure of light. Something from history which has suited us well and is still used because, well, it is easily understood. You could have a lens with a piece of cardboard in the middle and it'd still be f 1.2 :). Isn't the real problem here the uninformed belief that f stop is an absolute measure of light and therefore every f1.2 is the same throughout history, rather than just a calculation useful for that particular lens only?
-
Could I maybe ask what maybe a really dumb question? F stop is just length/width, ie a ratio. A relative measure, an indicator of lens mechanics, not an absolute measure of light. Something from history which has suited us well and is still used because, well, it is easily understood. You could have a lens with a piece of cardboard in the middle and it'd still be f 1.2 :). Isn't the real problem here the uninformed belief that f stop is an absolute measure of light and therefore every f1.2 is the same throughout history, rather than just a calculation useful for that particular lens only?
Hey nass, there's no dumb questions, only dumb (or wrong) answers ;). Hope mine doesn't fall in that category :o
Question at hand here in this thread is not so much wether a 1.2 lens meets the definition you state and lets through all the light it's supposed to do, but wether the digital sensor behind it can capture and record this light. Mark Dubovoy's article talks about two effects. First a general darkening due to the rays coming from the outer rims of the lens wide open not being recorded and secondly that due to the same effect on the sensor the depth of field is different, as it might cut out preferentially the outer rims of the circle of confusion that is created when an object is out of focus. Mark claims the general darkening is compensated by a "secret" ISO boost that is happening in the body and if you look at Pierre Vandevenne's pictures there seems to be some evidence for that. On the other hand you can see in my post that I have not been able to reproduce the depth of field/circle of confusion effect with my f 1.4 85 mm lens. Doesn't mean it doesn't exist for other lenses/sensors, but at least I'm "safe" ;D
-
nass said
Could I maybe ask what maybe a really dumb question? F stop is just length/width, ie a ratio. A relative measure, an indicator of lens mechanics, not an absolute measure of light. Something from history which has suited us well and is still used because, well, it is easily understood. You could have a lens with a piece of cardboard in the middle and it'd still be f 1.2 . Isn't the real problem here the uninformed belief that f stop is an absolute measure of light and therefore every f1.2 is the same throughout history, rather than just a calculation useful for that particular lens only?
It's not a dumb question. That type of lens exists, for example in catadioptric telescopes, with a central obstruction. The key point (as you suggest) is that they would not behave as expected for the photographer who uses those ratio to reason about the proper exposure. In very, very broad terms (because each subpoint of that topic could generate longish discussions)
- their geometry would remain F/D 1.2
- their resolving power would remain roughly equivalent to an unobstructed lens (at the extreme, that is the principle behind large base interferometry and interferometry in general)
- their contrast would be "modified" - this is really a complex issue see http://www.astrosurf.com/legault/obstruction.html for more details
- their light gathering ability would be reduced by the obstruction (which is probably the main reason why they wouldn't behave as the F/D 1.2 photographers expect)
And I don't even dream thinking about the analysis of marginal rays in micro-lensed sensors ;-)
-
Mark claims the general darkening is compensated by a "secret" ISO boost that is happening in the body and if you look at Pierre Vandevenne's pictures there seems to be some evidence for that.
Hi Pieter,
Indeed, I have been able to detect that as well, but the weather (and the IRIS software) hasn't been nice enough to do a quantitative test yet. Sofar it seems like a fraction of 1/3rd of a stop gain, but the more accurate test will tell it more accurately (for my 1Ds3+ 2 lenses).
On the other hand you can see in my post that I have not been able to reproduce the depth of field/circle of confusion effect with my f 1.4 85 mm lens. Doesn't mean it doesn't exist for other lenses/sensors, but at least I'm "safe" ;D
I haven't been able to detect that either, on a 85mm f/1.2 lens. The OOF blur 'disks' follow a perfectly predictable path, with very high reliability (R^2=0.9985) in a power function regresssion test. I'll try and post some results tomorrow.
Cheers,
Bart
-
Hi Pieter,
Indeed, I have been able to detect that as well, but the weather (and the IRIS software) hasn't been nice enough to do a quantitative test yet. Sofar it seems like a fraction of 1/3rd of a stop gain, but the more accurate test will tell it more accurately (for my 1Ds3+ 2 lenses).
That would be nice, but to be fair I'm not to worried about this effect for my (amateur) shooting. Not all my exposures are spot on and I sometimes make larger corrections in my raw converter and still get acceptable results. Yes pixel peeping the noise increase, but for an A3 print or downsized web presentation I find changes of up to a 1/2 stop are not significant. However I understand that very serious amateurs/pro's who want to get the absolute maximum quality out of their exposures even less than 1/3 stop can be a big deal.
I haven't been able to detect that either, on a 85mm f/1.2 lens. The OOF blur 'disks' follow a perfectly predictable path, with very high reliability (R^2=0.9985) in a power function regresssion test. I'll try and post some results tomorrow.
LOL, I only tested f 2.8 vs. f1.4 so I guess my R^2 is 1.000 (if you want to find a straight line relationship only do 2 experiments ;D), but you seem to be way more rigorous and scientific which I applaud. I found the DOF story a bit far fetched from the beginning and I'm glad we're starting to see the evidence this effect can be dismissed.
-
I found the DOF story a bit far fetched from the beginning and I'm glad we're starting to see the evidence this effect can be dismissed.
Well, there's not much wrong with a hypothesis being rejected, as long as one gains insight in the proces it still counts as progress ;)
Here's a quick summary of my findings with regards to the sensel tunnel shading effect on DOF.
I took a series of images at 1/3rd stop intervals, keeping the exposure time constant. The subject was a MagLite 'Solitaire' (very small, single AAA battery powered). I focused the lens at infinity, and positioned the tiny bare bulb (just unscrew the front of the torch) at a little more than 1 metre distance in the approximate center of the image. That should create a nice out-of-focus (OOF) highlight. Off-center measurements will suffer from vignetting and light fall off, so I stuck to center image testing.
The images were evaluated in Photoshop with the ruler tool, measuring the approximate (because the 8 segment aperture is not exactly circular) diameter of the OOF highlight images. I tried to measure at the same gradient of darkening at the edge of the OOF highlight 'circle'.
I tabulated the measurements in a spreadsheet, and converted the numbers of pixels across to millimetres by multiplying with the sensel pitch of 6.4 micron. I converted the diameters to approximate area (pi * (diameter/2)^2). The rounded F-numbers were replaced by their more exact mathematical equivalents, and a plot was made in a graph, and a regression trendline was calculated. The regression trendline was compared to the actual observations in the table, and an overall very good fit was found.
The trendline does not follow a perfect power of 2 relation, although it is pretty close. Afterall we are talking about a mechanical aperture and I know I have some issues with mine. The main point is that there is no evidence of a suddenly degrading fit at the widest apertures. On the contrary, the trend is very stable with a low average deviation, and the widest apertures show no sign of a reduced diameter.
Based on this, I tend to rejecting the hypothesis that the DOF at the widest apertures is significantly impacted by sensel tunnel shading.
Cheers,
Bart
-
The photoactive pixels are surrounded by a border
of buffer and light-shielded pixels.
I interpreted that (enforced by SEM images of the sensel structure I saw somewhere) as all individual photoactive pixels being surrounded by other pixels. Obviously that would reduce the photosensitive area, and microlenses would be needed to refocus the light onto the sensitive areas. On second reading it could also mean (and would be more likely) that there is a row of shielded pixels around the entire array of photosensitive sensels.
On the KAF-39000 sheet description, there is no mention of surrounding pixels/shielding, suggesting a significantly different design, at least that's how I interpreted it.
Pretty much all sensors have have an band of masked pixels around the active array; these are used to measure and compensate for dark current, the amount of charge that accumulates in a pixel even if no light falls on it. Usually this band is of the order of 10-20 pixels wide.
No sensor that I know of has any masked pixels in the active array.
As regards microlenses, CMOS and CCD sensors are different as regards their need for microlenses. Generally, a CCD array is just a big array of active cells, so the ratio of active pixel area to "dead" space, the "fill factor" is high. Also, CCD cells tend to be relatively shallow. CMOS cells need a lot of extra circuitry - amplifiers, etc - round each pixel, so the fill factor is lower, down to as low as 20-30%, and the pixels tend to be deeper. The fill factor is the reason for the so called "back lit" sensors. So when Kodak talk about not using microlenses to maximize sensitivity, be aware that's for a CCD sensor with a relatively high fill factor. A CMOS sensor is a very different proposition.
Sandy
-
When I got my 1.2 lenses, I spent a few hours obliquely shooting calipers and rulers at different aperture settings. The progressive reduction in DOF was quite obvious (I haven't kept the images, a few GB of raw images of calipers and rulers just add to the size of the backups). Of course, I can't be sure it wouldn't be even better with film. That's a test that's a bit hard to make in controlled conditions.
Also, I did notice that portraits shot at 1.2 mandated a plane for the eyes that was perfectly parallel to the sensor, whereas 1.6 offered a safety margin.
But I have to say I was a bit surprised by the effect of the gain boost, which I expected to be measurable but not visible. There's also something I noticed and that I will try to investigate a bit later: the images shot at 1.2 and 1.6 seem roughly identically exposed, with the 1.2 gain boost and, somewhat surprisingly the same shutter speed (1/1250) in that case.
Since the conditions weren't really stable (I may have been casting a shadow, the economic bulb could have still been heating up...) I can't be sure of what it means, but keep an eye for this in more controlled tests
-
First, let me note that I am not offering precise quantification of the magnitude of the effect we are discussing; after all, the data id not for any of the Canon, Sony or Nikon (CMOS or ILT CCD) sensors that the original article is dealing with; just with what seems an clear fact, that microlenses cause a significant fall-off in sensitivity as light comes from an off-perpendicular angle, including the light form the outer parts of there light cone from a low f-stop lens (like f/1.2 and maybe f/1.4) even at the center of the frame. But for reasons I have stated before and restate below, it is likely if anything to be greater with current DSLR front-illuminated active pixel CMOS sensors than with Kodak's FF CCDs.
Second, even if one denies that microlenses are the cause, the effect is still there in every single microlenses sensor I have data for (four different FF CCDs of three different pixel spitches from Kodak.)
... On second reading it could also mean (and would be more likely) that there is a row of shielded pixels around the entire array of photosensitive sensels.
The second reading is correct: this is a border around the edge of the entire sensor, not within each photosite. This is made clear in the early pages of the long spec document for the KAF-39000, and in every Kodak FF CCD long spec: see http://www.kodak.com:80/global/en/business/ISS/Products/Fullframe/
Despite the identical 6.8 micron sensel pitch, the datasheets mention different sizes for the sensor array, ... They also mention different numbers of pixels, confusing ...
It is not at all confusing if you read a bit more carefully: these are sensors using the same basic 6.8 micron photosite design but of different sizes (44x33mm vs 49x37mm) and thus of different pixel counts (39MP vs 31MP). But that difference is irrelevant to per pixel performance characteristics.
All I'm saying is that differences in design make comparisons difficult, and certainly not a basis to generalize on, and the (also mentioned in the datasheet, offset) microlenses were instrumental in increasing the overall sensitivity of the (more special ?) design.
I would rather draw general conclusions about microlenses based on identical sensor array designs, only differing in the use of microlenses or not, that's all. That will probably be data that's hard to find.
The difference in overall sensor size and pixel count is irrelevant to the question we are discussing: we are looking at what happens at an individual photosite, anywhere on the sensor, when light strikes it at various angles, and Kodak gives data for those two contemporary sensor designs for what happens at a photosite at the center of the sensor. By the way, here is another earlier one with microlenses but not offset, that KAF-8300 of the Olympus E-300 and E500: http://www.kodak.com:80/global/plugins/acrobat/en/business/ISS/datasheet/fullframe/KAF-8300LongSpec.pdf
and the KAF-18000, an 18MP, 44x33mm sensor with 9 micron pixels and non-offset microlenses:
http://www.kodak.com:80/global/plugins/acrobat/en/business/ISS/datasheet/fullframe/KAF-18000LongSpec.pdf
which you might want to copare to the same generatio 22MP, 49x37mm KAF-22000 with no microlenses and far less off-perpendicular fall-off:
http://www.kodak.com/global/plugins/acrobat/en/business/ISS/datasheet/fullframe/KAF-22000LongSpec.pdf
The angular fall-off is even worse for the older KAF-5101 of the Olympus E-1, with 6.8 micron pixel pitch and non-offset microlenses, but that is no longer at Kodak's site.
It becomes hard to conclude anything that than, consistently over eight years of putting microlenses on FF CCDs, Kodak has been forced to sacrifice off-perpendicular sensitivity fall-off as part of the price of adding microlenses.
The size difference is however relevant in a different sense: despite the substantial advantages in QE that microlenses give when Kodak uses them in FF sensors (about doubling QE and thus adding one stop of sensitivity) and despite Kodak having had this technology since the KAF-5101 from 2002 with this same 6.8 micron pixel spacing, Kodak has never used microlenses on its larger MF sensors, the ones about 49x37mm, while using them on three generations of its smaller 44x33mm FF CCDs for MF (KAF-18000, KAF-31600, KAF-40000). The reason seems obvious, especially since Kodak more or less states it in text I have already quoted: adding microlenses has the unfortunate side-effect of causing greater fall-off in sensitivity to off-perpendicular incident light, causing a kind of vignetting, and this increases as distance from the center of the frame increases. The effect is thus more tolerable with the smaller radius (corner distance) of the 44x33mm MF sensors than with the larger radius of the 49x37mm sensors. Before Kodak had off-set microlenses, it did not use microlenses on any MF sensor.
I ask again: can anyone suggest any plausible reason why Kodak would hamper every one of its microlensed FF CCD sensors in this way if it could combine the QE advantages of microlenses with the greater off-perpendicular sensitivity that all its non micro-lensed FF CCD sensors have, going back many years?
Or to put it another way, why does Kodak continue to hamper the sensitivity of all its largest 49x37mm MF sensors with substantially lower QE by not offering any of them with microlenses, except due to its inability to add microlenses while avoiding the vignetting problem due to sensitivity that declines too fast as the angle of incidence increases?
And to the idea that this is just inability on Kodak's part, I will note again that (a) Dalsa says the same thing about off-perpendicular sensitivity fall-off being a disadvantage of microlenses, and (b) as Dalsa explains, the problem is likely to be worse with all recent DSLR sensors because they are all front-illuminated CMOS sensors, and thus microlenses need to be further from the wells.
-
In case someone is still interested in this, the latest edition of Chasseur d'Image has an article in French that I find to be a more accurate explanation of the issue Mark wrote about.
Regards,
Bernard
-
I'd be most interested in the DoF change suspicion. There's posts in this thread that show this effect could not be demonstrated with f 1.4 and f 1.2 lenses on two different bodies but maybe Mark and/or DxO have investigated this further and can provide some new insights or measurements.
I think the discussion (and subsequent articles) have the artificial ISO gain pretty well understood (despite the lack of response from the manufacturers), but it's been very silent on the DoF issue where I have more doubts about.
-
Lens/camera designers from Fuji are very aware of the subject raised by Mark. They do not seem to follow the road of secretly boosting ISO.
A quote from: http://www.finepix-x100.com/story/
2Road to the F2 aperture value.
* Designing an F1.6 or F1.8 lens is not so difficult; however, in the case of a digital camera, even if an aperture larger than F2 is used, the light receiving elements on the sensor cannot effectively use the brighter portion of the incoming light because of low incident light gathering efficiency.
-
I did a few tests
http://www.datarescue.com/photorescue/freefiles/compare40d5dmkII.jpg
http://www.datarescue.com/photorescue/freefiles/5dmk2.jpg
http://www.datarescue.com/photorescue/freefiles/40d.jpg
40D / 5DMK2 - tripod - remote trigger - 1 meter from target - auto exposure - lighting poor on purpose to better see the noise. Shot in RAW, converted to jpeg at max quality with no adjustment whatsoever, ISO 800
AFAIC
- the gain boost is quite visible, maybe a bit less on the 40D, but the auto-exposure chose different parameters.
- the practical impact, at least at pixel peeping level, of the gain boost is more significant than the raw noise number indicate, probably because noisy pixels have a bigger than expected impact on the RAW conversion.
- the DOF decreases as expected (it is very visible on near tangent images of printed text, as posted earlier) at least on the 5D
- the 40D doesn't seem to exploit that DOF increase fully, but that's just a gut feeling, and DOF is bigger anyway.
-
I did a few tests
http://www.datarescue.com/photorescue/freefiles/compare40d5dmkII.jpg
http://www.datarescue.com/photorescue/freefiles/5dmk2.jpg
http://www.datarescue.com/photorescue/freefiles/40d.jpg
40D / 5DMK2 - tripod - remote trigger - 1 meter from target - auto exposure - lighting poor on purpose to better see the noise. Shot in RAW, converted to jpeg at max quality with no adjustment whatsoever, ISO 800
AFAIC
- the gain boost is quite visible, maybe a bit less on the 40D, but the auto-exposure chose different parameters.
- the practical impact, at least at pixel peeping level, of the gain boost is more significant than the raw noise number indicate, probably because noisy pixels have a bigger than expected impact on the RAW conversion.
- the DOF decreases as expected (it is very visible on near tangent images of printed text, as posted earlier) at least on the 5D
- the 40D doesn't seem to exploit that DOF increase fully, but that's just a gut feeling, and DOF is bigger anyway.
Pierre,
Interesting data, nicely done. Seems indeed that gain (noise) increases at wider apertures. The change of DOF is clearly visible.
Expected to see slightly more gain (noise) increase of the 40D compared to the 5D as per DxO labs figures / smaller 40D photo sites.
Probably the difference of 0.1EV is neglectable / not visible in this test.
Please note that the DOF of the 40D is smaller because you are shooting both camera's from the same distance.
Calculated DOF for a 85mm lens on both camera's focussed on 1 meter distance is as follows:
5Dmk2 40D
f/2.0 1.6cm 1.0cm
f/1.8 1.5cm 0.9cm
f/1.4 1.1cm 0.7cm
f/1.2 1.0cm 0.6cm
Of course these figures are approximate for standard eyesight on printed matter. However close enough for validation of this test I think.
Used the DOF calculator on the Cambridge in Colour site.
http://www.cambridgeincolour.com/tutorials/depth-of-field.htm
These distances are very close according your test shots.
So it looks that gain (and noise) is increasing towards wider apertures and still the influence of aperture on DOF remains visible towards f/1.2.
Herb
-
- the DOF decreases as expected (it is very visible on near tangent images of printed text, as posted earlier) at least on the 5D
- the 40D doesn't seem to exploit that DOF increase fully, but that's just a gut feeling, and DOF is bigger anyway.
Thanks for the test!
The width of DoF doesn't seem that impacted, but I fell that the bokeh changes with the max aperture, see the blue line at left side of the ruler...
-
Let me also respond directly to the comment that the data cannot be found on the DxO site. There is a new release of the DxO site coming in a few weeks. It will contain a lot of new data, including the data that support this open letter.
While there has been some discussion on the depth of field issue, l would like to repeat what I said in the open letter, which is that DxO is in the process of performing thorough focus measurements. Once the final focus measurement data is available, we will be able to put this issue to rest.
Sorry to reopen a very old discussion, but I was reminded of this thread due to a question in another forum.
Has anybody seen the promised further data from DxO, either on the "hidden" iso gain as well as on the "reduced DOF" claim? I haven't, but would be interested to read more on this in case it has been published and I missed it.
-
Has anybody seen the promised further data from DxO, either on the "hidden" iso gain
photonstophotos.net