Luminous Landscape Forum
Equipment & Techniques => Cameras, Lenses and Shooting gear => Topic started by: dimapant on October 15, 2013, 11:45:30 pm
-
I have Nikon gear, 3 cameras and a number of lenses, both primes and f 2,8 zooms.
After a double test performed last year both on D 800 and D 800 E I decided to not purchase them, following , for my standard, poor image quality for sharpness, mostly with F 2,8 zooms ( 14 – 24, 24 – 70 and 70 – 200 VRII), the lenses that I use most, but also with some primes.
I do consider DxO a serious company and the data of their measured parameters on lenses and cameras to be reliable data.
Some months ago they changed the lens resolution metering system, from the relative lpm to the Perceptual Mpx, the resolution transferred in the final image by that peculiar lens associated with the camera.
Going at DxO site to perform a check of the different lenses fitted on the D 800, the final resolution of the 36 Mpx camera fitted by the best lenses today on the marker never exceed 22 Mpx ( 85 F 1,4 AFS G), having a medium value with the majority of the lenses by far lower than 22 Mpx.
If the lens is not providing the needed resolution to resolve the sensor resolution, the final image, at the best, has the resolution of the lens and not the resolution of the camera.
Canon, with its 5 D MKIII 22 Mpx camera, instead of 36 Mpx, is confirmed to acquire just the same Perceptual Mpx number, 22 Mpx (300 f 2,8), exactly the same, no difference and with 22 mpx camera is anyway able to fully use the performance in resolution of the lenses.
In other words those numbers confirm that , in 35 mm, with today available lenses, a 36 Mpx camera, compared to a 22 – 24 Mpx camera fitted by the same lenses, it is just a waste of money, providing no advantage in final image resolution, and if we put in the bill also a smaller size of the pixel, with 36 Mpx camera we are also kindly gifted by an increase of negative side effects, early diffraction and minor camera shake made visible.
Many thanks in advance for your comments.
Best regards.
Alessandro
-
I have thought long and hard about switching to D800 from 5DMIII for the dynamic range mostly, but also for the resolution. I do not plan on doing so, because, like you point out, the lenses are the bottle neck.
What about the new Zeiss 55mm? That lens should be a good contender to get the very best out of the 36MP.
-
Thats why I got the D600.
-
Hi,
The Zeiss 55/1.4 probably has it greatest advantage at large apertures.
I checked Lloyd Chambers article on Distagon 55/1.4 and according to MTF data it is very sharp at f/1.4. It reaches optimum performance at f/4 and drops beyond it (due to diffraction). I also compared MTF curves with Zeiss Macro Planar 50/20. The Apo Distagon is slightly better at f/4. I don't have figures for the Macro Planar at f/8, but I guess that the two lenses are a close match around f/8.
If you compare the Macro Planar 50/2.8 on the Nikon D800 and Canon 5DIII there is a slight advantage to Nikon D800.
Roger Ciala at LensRentals made an interesting comparison : http://www.lensrentals.com/blog/2013/01/a-24-70mm-system-comparison
(He finally bought Canon 6D.)
Best regards
Erik
I have thought long and hard about switching to D800 from 5DMIII for the dynamic range mostly, but also for the resolution. I do not plan on doing so, because, like you point out, the lenses are the bottle neck.
What about the new Zeiss 55mm? That lens should be a good contender to get the very best out of the 36MP.
-
If the lens is not providing the needed resolution to resolve the sensor resolution, the final image, at the best, has the resolution of the lens and not the resolution of the camera.
Hi Alessandro,
That is a common misunderstanding. While the lesser component will pull down total performance, the performance is a combination of both. That means that a better lens will always improve total performance, even if the camera is the lesser component, but also a better camera will also improve total performance if the lens is the lesser component. So, the lesser component will not create a fixed maximum level. They always go hand in hand, improving one (regardless whether it's the better or lesser component) will improve the total performance, and lowering the quality of one (regardless whether it's the better or lesser component) will likewise lower total performance.
Canon, with its 5 D MKIII 22 Mpx camera, instead of 36 Mpx, is confirmed to acquire just the same Perceptual Mpx number, 22 Mpx (300 f 2,8), exactly the same, no difference and with 22 mpx camera is anyway able to fully use the performance in resolution of the lenses.
In other words those numbers confirm that , in 35 mm, with today available lenses, a 36 Mpx camera, compared to a 22 – 24 Mpx camera fitted by the same lenses, it is just a waste of money, providing no advantage in final image resolution, and if we put in the bill also a smaller size of the pixel, with 36 Mpx camera we are also kindly gifted by an increase of negative side effects, early diffraction and minor camera shake made visible.
That conclusion is not correct. Do not confuse averaged perceptual (as defined by DxOmark.com) with perceivable/actual. The actual resolution, which is different between camera/lens combinations, may not look much different when observed under certain (standardized test) conditions, but they may look very different after post-processing, and under different viewing conditions.
Besides, resolution is not the only criterion for selecting a lens or camera body.
Cheers,
Bart
-
A key motivation for the manufacturers in introducing ever-higher resolution sensors is that this immediately creates a requirement for ever better-performing lenses. It's all about marketing and keeping the churn going. Whilst it's a truism that a few megapixels more - or even a lot of megapixels more - is never going to make average pix into great pix, it's a truism worth repeating regularly.
Of course these websites and forums are part of the mechanism. Without them it would be almost impossible for the manufacturers to generate sufficient demand to keep the churn going - imagine the levels of conventional advertising that would be needed in their absence. Personally I'm very weary of feeling manipulated in this way.
It's interesting to watch the latest feeding frenzy beginning. I'm referring to the Sony 7 / 7r cameras. Anything that hastens the demise of the stage-weight FF DSLR is to be welcomed of course. Carrying these damn things around, unless you're earning a living with them - and maybe even if you are - is absurd.
Roy
-
In a word, "No".
The premise that the 36Mp sensors of the Nikon D800 and D800e are capable of "outresolving" decent quality lenses is absurd.
I use both a D800 and a D800e with the Nikkor "Holy Trinity" of 14-24mm, 24-70mm and 70-200mm f/2.8 lenses and can give an assurance that with those lenses, the full 36Mp can be utilised to huge advantage in those situations that actually do take advantage of it. (Think about that last phrase - it's not as Irish as it sounds.)
Look at it this way. I "downgraded" to the 800 series 18 months ago from a D3s.
For a bog-standard application, such as producing an A3+ print from the full frame of an image, there is absolutely no discernible difference in image quality that can be detected by the naked eye between the 12Mp D3s and the 36Mp D800. Only a numptie would expect any.
But move out to the extremes of performance - such as producing that A3+ print from 10% of the image area - and the differences do start to show.
So, yes, if you are a pedestrian photographer (I mean that in the nicest possible way), any camera above a 10Mp compact is arguably a waste of money. But if you want to continuously push at the extremes of performance, then you must be prepared to pay for it.
I probably have no rational "need" for a 200Mp camera - but when an affordable one comes along (as indeed it must), then you can bet your bottom dollar that I am going to "want" one and will thoroughly enjoy the new, currently unimagined, possibilities it will give me.
-
Far from a waste of money, the D800 family will work very well with the lenses you mention, I have all of them and have not seen any issues.
I would strongly suggest you do either of these:
1. Locate a local Nikon dealer and spend a few hours with a demo, I did this.
2. Rent the body from a group like lensrentals.com They have both the 800 and 800e, and you have the lenses, you can then make the decision based on your results.
Paul Caldwell
-
There's one major defect to most lens tests - sample number. I suspect most people have a sample number of ONE.
As for the demise of SLR format: I prefer the bright prism view, I prefer being able to steady the camera on my face, and I prefer having enough real estate for my hands to comfortably grip. I have size 6 hands, but I like having a non-cramped grip, well spaced controls, and so on, purely from an ergonomic point of view. Guys with big hands, are you ready to ditch your SLRs for compacts? Sports and bird-in-flight and other action specialists, do you feel as comfortable with compacts as with SLRs?
-
Guys with big hands, are you ready to ditch your SLRs for compacts? Sports and bird-in-flight and other action specialists, do you feel as comfortable with compacts as with SLRs?
I used to use OM film cameras which are minute by DSLR standards and they had lovely handling and I need Large/Extra Large gloves. Had a play with the two digital OM cameras the other day and was very taken with them.
I'd do a lot more shots like this if I didn't have to carry a hulking FF body around when biking and the only reason I had my big camera with me last Saturday was because it was meant to be a trail maintenance day. So I was going to document that, but rain cancelled digging and I decided to have a play in woods instead.
-
As for the demise of SLR format: I prefer the bright prism view, I prefer being able to steady the camera on my face, and I prefer having enough real estate for my hands to comfortably grip.
You can still use mirrorless cameras at eye level and for fast lenses they have a big advantage over a traditional viewfinder as you get to see the actual depth of field and can therefore focus more accurately. Use an f1.4 lenses wide open and you see a DoF more like an F2.8 on the ground glass on my Canons. The only way to accurately focus is live view which is not much use for non tripod work.
-
It's interesting to watch the latest feeding frenzy beginning. I'm referring to the Sony 7 / 7r cameras. Anything that hastens the demise of the stage-weight FF DSLR is to be welcomed of course. Carrying these damn things around, unless you're earning a living with them - and maybe even if you are - is absurd.
I'm very intrigued by the new Sony cameras as it would also mean I'd do more high quality landscape work but I not too enthused to mountain bike with a really heavy/bulky backpack, getting up hills around here is hard enough. Not to mention packing for planes would be a lot easier. It can be quite restrictive on which airlines one can use due to the pathetic cabin allowance weight some have. Virgin Atlantic's was particularly mean with one cabin bag at 6kg.
-
Very interesting: thank you for the attention and the link!
Have a nice day and best regards.
Alessandro
-
The resolution certainly is not the only parameter to take into consideration in the selection of a camera and it is not the only parameter to judge an image, even on a large print, but it is very a important one for the natural looking of the image.
I did tests with my D 700 Vs D 800 and even if I limit the test to the zoom lenses, the difference in resolution is huge on monitor, but cannot basically seen on A3 print whilst is visible on A2 print, but the soft borders can be seen some time on D 800 and not seen (they are there but they are masked) by the lower resolution of the D 700.
Still with the zoom, in wxample the 70 - 200 VRII, the difference can be seen between D 800 vs D 800 E but I do not think that it can be seen on the paper unless you really print very large size, more than A2, and you are really very, very carefull with the sharpening work, a really minor unacuracy can spoil the difference.
Referring to Mr. BartvanderWolf, and I thank him for his post, DxO is performing a very practical measurement, providing the average perceptual resolution of the combination Camera + lens: I am not convinced that the perceivable/actual resolution of normal picture is very different from the perceptual resolution of DxO mesured in lab tests.
I agree that DxO is performing lab tests, and a picture can be quite different, but if at DxO they do their job properly, the lab test work normally is arranged to provide a reliable simulation of the real conditions, at least this is what we were doing us, in my company, performing tests on electooptical systems, including lenses, otherwise the lab tests is not fulfilling the proper job, which is to simulate as much as possible the real standard conditions, and becomes just an added cost, and moreover, making measurements of risolution is not a difficult job at all today.
I agree that if you squeese the camera and the lens with important crops and large prints from that crops you probably get some advantage from 36 Mpx, but, at least for me, this is an uncommon way to gat a picture, I am using mostly zooms to take on camera the proper picture, try to avoid crops ( .... D 700 was teching me along the years!).
What make me thinking that DxO is probably right, is the fact that DxO, basically, confirmas what I found on the field, with no tests on optical patterns, but just with many pictures, normal pictures: in few words the fact that many lenses does not performs well enough for that camera, mostly the zooms, the same lenses that in DxO measurements have a average perceptual resolution rather low, but not only the zooms, even primes.
Anyway, I thank you very much all of you for your attention, for your time and for the clear answers and comments: I will think about more.
Have a nice day and best regards.
Alessandro
-
"After a double test performed last year both on D 800 and D 800 E I decided to not purchase them, following , for my standard, poor image quality for sharpness, mostly with F 2,8 zooms ( 14 – 24, 24 – 70 and 70 – 200 VRII), the lenses that I use most, but also with some primes."
How did you conduct your tests?
Did you use the cameras' autofocus micro adjustment feature to tune the camera bodies response to your lenses? Every lens and every camera body is unique and if you don't do this it' like playing an out of tune Steinway and not knowing it.
-
In other words those numbers confirm that , in 35 mm, with today available lenses, a 36 Mpx camera, compared to a 22 – 24 Mpx camera fitted by the same lenses, it is just a waste of money, providing no advantage in final image resolution, and if we put in the bill also a smaller size of the pixel, with 36 Mpx camera we are also kindly gifted by an increase of negative side effects, early diffraction and minor camera shake made visible.
Many thanks in advance for your comments.
Best regards.
Alessandro
Well I don't believe there will be negative side effects from diffraction or camera shake if the image from each camera is identically sized. You will not get a lower IQ from more pixels.
Secondly, DXO does not test the 800e AFAIK so there is no perceived pixel rating for that camera on DXO last time I looked but has that changed? (I have not looked for a while)
Thirdly, I don't think you are comparing each sensor using identical lenses. Rather, you compare D800 and Nikon 300/2.8 with 5dIII and Canon 300/2.8. I suspect that this Canon lens is better than Nikon's equivalent as is the 24-70 and the 70-200 F2.8 lenses. You also have an electronic first curtain on the 5dIII which can make a big difference and we do not know what shutter setting DXO use.
So I think it is difficult to draw your conclusion from the DXO tests, particularly as far as the 800e is concerned.
Having said all that, I have the 800e and am also pretty familiar with the Canon system and I print on a Epson 7880 (A1 size). I would say from experience (rather than any scientific testing ) that you would be unlikely to see any resolution difference between a 5dIII and 800e when using the 2.8 zooms. However with a very good prime lens (my best is an APO Leica R) I feel there would be a (just) discernible difference in an A1 print but not much difference in any print smaller. Certainly no difference in a 2mp screen rendering.
For this reason I find the new Sony A7r very interesting to use with some M lenses. I just hope they have dealt with the colour shift issues on WA lenses.
-
It's important to keep in mind some of the things Bart said. Think of the optical "frequency response" as a curve that rolls-off in the higher frequencies. How steeply that curve rolls off, up to the nyquist limit, gives you a good picture of how much information in the higher frequencies is preserved. None of the camera/lens combinations being discussed is actually limited strictly to the "perceptual resolution". This is a suggestively named metric that stretches semantic bounds a bit. In reality, it's a matter of degree.
-
Lens deconvolution is hard. It is made harder by the non-linear/spatially variant nature of sampling. By increasing the sampling rate (pixel count) of the sensor, deconvolving any given lens should be easier.
Even though a given lens might appear to have minimal improvement in detail when switching from 22MP to 36MP when using standard processing, the "optimally" processed image might still have significant improvements.
-h
-
It's important to keep in mind some of the things Bart said. Think of the optical "frequency response" as a curve that rolls-off in the higher frequencies. How steeply that curve rolls off, up to the nyquist limit, gives you a good picture of how much information in the higher frequencies is preserved. None of the camera/lens combinations being discussed is actually limited strictly to the "perceptual resolution". This is a suggestively named metric that stretches semantic bounds a bit. In reality, it's a matter of degree.
Hi Luke,
I agree, semantics can also confuse. Perhaps it's useful for the discussion to repeat a plot (attached) I shared earlier in one of the other threads about this subject.
It shows that, with everything else being kept the same, the MTF curve response for a denser sampling device (4.88 micron versus 6.4 micron sensel pitch) will be higher for the same spatial frequencies, and additionally also even higher spatial frequencies can be resolved (if subject contrast is high enough). Lens resolution often exceeds sensor resolution, so the combination will still benefit significantly from denser sampling. It would even benefit a crappy lens, although the combination would be kind of crappy as well, but less crappy.
This benefits the possibility to make a difference in post-processing, i.e. (deconvolution) sharpening and tonemapping. Such post-processing is not a part of the DxOmark metrics. Some of those differences are potentially lost when down-sampling with a lesser algorithm.
Cheers,
Bart
-
Hi,
The way it works you multiply MTF of lens and MTF of sensor. If MTF of lens drops to zero, no more detail can be resolved, but if combined MTF is like 10% we can still extract some detail by appropriate sharpening.
Having a higher resolution reduces the risk of introducing fake detail.
http://echophoto.dnsalias.net/ekr/index.php/photoarticles/75-aliasing-and-fake-detail
http://echophoto.dnsalias.net/ekr/Articles/MFDJourney/FakeDetail/comparison1.html
Best regards
Erik
Hi Luke,
I agree, semantics can also confuse. Perhaps it's useful for the discussion to repeat a plot (attached) I shared earlier in one of the other threads about this subject.
It shows that, with everything else being kept the same, the MTF curve response for a denser sampling device (4.88 micron versus 6.4 micron sensel pitch) will be higher for the same spatial frequencies, and additionally also even higher spatial frequencies can be resolved (if subject contrast is high enough). Lens resolution often exceeds sensor resolution, so the combination will still benefit significantly from denser sampling. It would even benefit a crappy lens, although the combination would be kind of crappy as well, but less crappy.
This benefits the possibility to make a difference in post-processing, i.e. (deconvolution) sharpening and tonemapping. Such post-processing is not a part of the DxOmark metrics. Some of those differences are potentially lost when down-sampling with a lesser algorithm.
Cheers,
Bart
-
Personally I'd like to see the photosite count go up to 64 million. Then I could, via RAW converter (or even in-camera), generate 16mp files using each RGBG Bayer matrix to create one output pixel. 16mp gives me enough spatial detail...I want finer, richer tonal detail.
-Dave-
-
Hi,
Such a technique has been proposed and there is real math supporting it.
Now, todays lenses have something like 0.5 % flare, that is 0.5% of the incoming light bounces around in the lens. Digital sensors cover 12EV or 14EV (3.6 - 4.2 density steps). I think we need discuss lens flare and put less emphasis on sensor DR.
Best regards
Erik
Personally I'd like to see the photosite count go up to 64 million. Then I could, via RAW converter (or even in-camera), generate 16mp files using each RGBG Bayer matrix to create one output pixel. 16mp gives me enough spatial detail...I want finer, richer tonal detail.
-Dave-
-
with 36 Mpx camera we are also kindly gifted by an increase of negative side effects, early diffraction and minor camera shake made visible.
That is not correct, or just partially true. Early difraction and camera shake are made more visible the higher the pixel count at pixel level (100% pixel peeping), but that does not translate to a given output real image. More Mpx never mean your final images will be more affected by diffraction nor camera shake.
Regards
-
Hi,
Guillermo is right.
The image below shows resolution for MTF=50 on three different cameras, 6MP, 12MP and 24MP - all APS-C.
(http://echophoto.dnsalias.net/ekr/images/DMBFigures/EffectsOfDiffraction1.png)
You see essentially that diffraction starts reducing sharpness somewhat on all three systems at f/8 but the higher resolving systems hold advantage up to f/22.
The Alpha 77 used here is 24 MP APS-C, corresponding to 54MP on full frame.
That is not correct, or just partially true. Early difraction and camera shake are made more visible the higher the pixel count at pixel level (100% pixel peeping), but that does not translate to a given output real image. More Mpx never mean your final images will be more affected by diffraction nor camera shake.
Regards
-
Personally I'd like to see the photosite count go up to 64 million. Then I could, via RAW converter (or even in-camera), generate 16mp files using each RGBG Bayer matrix to create one output pixel. 16mp gives me enough spatial detail...I want finer, richer tonal detail.
-Dave-
The question is, if you are going to pay for a 64 MP sensor and the supporting electronics to digitize and (partially) process that data.... Would you not rather store the data with full precision to your memory card, leaving the possibility open to maximize the fidelity in a current (or future) raw converter?
I am not confident what "tonal detail" really is, but I don't think that in-camera downsampling is the way to get it. Whatever the camera can do (automatically and fixed) to your raw digital data, a software raw developer can do (with better precision, flexibility and user-interaction).
In my view, the "megapixel wars" is moving us further and further into the oversampled region. If the camera manufacturers can increase the number of sensels for a given sensor area without increasing image-level noise (they seem to be able to do the opposite), while lens manufacturers cannot improve at the same pace, then we will have more and more (accurate) samples of a 2-d waveform that gets (relatively) blurrier and blurrier (thus easier to sample and recreate). To me, that is happy-land.
-h
-
That is not correct, or just partially true. Early difraction and camera shake are made more visible the higher the pixel count at pixel level (100% pixel peeping), but that does not translate to a given output real image. More Mpx never mean your final images will be more affected by diffraction nor camera shake.
Regards
Precisely. But it amazing how many of the self-appointed gurus get that wrong. Mark you, it was not helped by the strange Technical Note that Nikon put out just before the D800 launch.
-
That is not correct, or just partially true. Early difraction and camera shake are made more visible the higher the pixel count at pixel level (100% pixel peeping), but that does not translate to a given output real image. More Mpx never mean your final images will be more affected by diffraction nor camera shake.
Yup, absolutely.
Similarly [of sorts] is why comparing different MP cameras at 100% is also a bit daft. Compare them at same print size that's a far more useful metric. If a 40Mp camera has marginally more noise at 100% than a 20MP camera, you are are simply not going to see it as it will be much smaller/less noticable at same enlargement.
-
Amazing
Thought I knew something about photography but little about the functioning of sensors before I came here. I know nothing about sensors and very little photography now
But this seems another, "... no one needs all these pixels and the camera companies are evil and just want us to give them more money their R&D didn't earn..." thread. Of which I've watched numerous spring up over the past year
On one hand you have "pixels I don't need", on the other "it's all about empiric testing (which of course has nothing to do with looking at a picture.) Sort of like road testing a car still in the garage (but you did read an empiric article about the exhaust note)
One thing I really like to do is look at and manipulate images. Too bad I'm too old to actually use it beyond my family. Stopped reading about sensor-metrics. No serious scientist would base a "true statement" on the "n's" used by these publications. But I do know that my RX-1 is an amazingly good camera for taking interesting and enjoyable pictures! And when I have enough light, the resolution and beauty of a DP-2 & 3 Merrill (with some "adjustments" to tifs) is at once manifest and sublime. I see things in these pictures that my eye likes.
Understood that some like to "bench" a bunch of data and call it authoritative, but doesn't one get a bit curious about how it actually looks? When I was much younger audiophiles were the rage and I worked for a time at a high end audio store. Same thing. Look at spec sheet all day and perseverate about the measured minutiae, but once the music came on, all the data was no better than "interesting".
-
... I worked for a time at a high end audio store. Same thing. Look at spec sheet all day and perseverate about the measured minutiae, but once the music came on, all the data was no better than "interesting".
Me too. Those people tended to either:
1) Not care about spec at all, or
2) Care about some single spec (like purity of Cu in cables) without paying much attention to the full picture
I am convinced that measurements and practice _can_ be combined, and that a good combination i better than either in isolation.
-h
-
Enter Blade Runner :D
Eduardo
In a word, "No".
The premise that the 36Mp sensors of the Nikon D800 and D800e are capable of "outresolving" decent quality lenses is absurd.
I probably have no rational "need" for a 200Mp camera - but when an affordable one comes along (as indeed it must), then you can bet your bottom dollar that I am going to "want" one and will thoroughly enjoy the new, currently unimagined, possibilities it will give me.
-
Me too. Those people tended to either:
1) Not care about spec at all, or
2) Care about some single spec (like purity of Cu in cables) without paying much attention to the full picture
I am convinced that measurements and practice _can_ be combined, and that a good combination i better than either in isolation.
-h
+1 ;)