Canon 5DmkIII

Status
Not open for further replies.
F

Flake

Guest
Don't forget that a camera is more than its sensor, and the reason the D3 appears to produce sharper images, could be down to a number of different reasons.
For a start Nikon sharpen the image something which cannot be turned off, not every photographer is happy about this.
A weaker anti aliasing filter will produce sharper looking images and a weaker bayer matrix will do the same at the expense of saturation and faithful colour depth.
The D3 is intended to be a fast sports & action camera, it's a pro spec model, whereas the 40D is the next step up from the entry model, it's not really fair to compare the two.
Yes I know it's a bit of an odd similie but I think it does explain it nicely! Again don't forget that isolation of pixels from each other is very important as leakage from one to another is a big cause of noise.
 
Upvote 0
Part of it was that the D3 was simply better; Canon's competitor for the "pro sports shooter" camera was APS-H, and not counting the fact that they opted against gapless lenses to avoid vignetting with wide angle lenses, they had inferior noise performance.

The image processor I believe adds read noise to the system, with the rule being lower read noise is better. However, I do believe that read noise is most important in dynamic range at low ISOs, whereas at high sensitivities sensor noise overwhelms the impact of read noise, which remains relatively constant relative to ISO.

As far as the D3 goes, vs the 1D4 (what kind of Canon fanboy are you? D4 was the rumored successor to the D3 and is now the rumored successor to the D3s), part of the issue was that the 1D3 attempted an innovative autofocus system that failed on the field. Another part of the issue was that the D3 was just a comprehensively better camera than the 1D3; not only did it have superior noise performance it also had better dynamic range at low ISOs, which remains BETTER than the current D3s.

The noise advantage really does matter here, however. The D3 is $4000, as is the 1D4. The D3 is just a lot more capable than the 1D4 at photography because of the high ISO feature; it can take shots where the 1D4 cannot, and when you're paying $4000 for a camera I think the versatility is quite important.


Flake: the easy explanation for why the D3 outperforms the 40D is just that one is full-frame, one is APS-C. All the other factors don't add a lot more to that; the D3 is a lot sharper with various lenses because its requirement for lens sharpness is a lot less than the 40D due to the difference in pixel sizes; the 40D only uses the center half of the image circle, and it demands a lot more from its center than the D3 does of the entire lens.

I was recently looking at used 5D1s, and I think that they might be a bargain right now, compared to say, the D90. The high-ISO performance is antiquated, sure, it's about equivalent to the nearly obsolescent D90, has a poorer AF system, and can't do video, but you're looking at something like twice the lens sharpness of the D90 due to the larger sensor format.
 
Upvote 0
N

NotABunny

Guest
fyngyrz said:
8...10 MP. Why? Because large sensels produce larger signals and this will improve the signal to noise ratio, hence allow higher ISOs, a'la the D3s *except* that Canon handles black better than Nikon and also doesn't foul up my images with noise reduction I can't turn off.

At http://www.juzaphoto.com/article.php?l=en&article=83 ( old linkhttp://www.juzaphoto.com/eng/articles/canon_1d_mark4_review_comparisons.htm ) there are 2 images (JPEGs converted from RAW) of a scene, taken in the SAME light, one with Canon 1D4, one with Nikon D3s, at ISO 12800.

Normalize them for sensor size (you have to cut the inner part of the image taken with D3s to fit the area of the 1D4 sensor size - crop 1.3), then normalize them for resolution (because the output resolution would be the same either on screen or on print), and finally normalize them for black level (Nikon applies some in order get rid of the noise in the shadows).

At the end, you'll have two images whose quality is indistinguishable. So, at least at this technological level, pixel size is not a relevant factor for high ISO and low light photography; however, the sensor size is.

----------------------------------------

This is how to scale the resolutions:

1D4 sensor: 27.9 x 18.6 mm, 4896 x 3264 photosites
D3s sensor: 36.0 x 23.9 mm, 4256 x 2832 photosites



Crop the center part of the D3s sensor to the sensor size of the 1D4:

(4256 x 2832 photosites) * (27.9 x 18.6 mm) / (36.0 x 23.9 mm)

=> 3298 x 2204 photosites (on the D3s sensor correspond to a physical area equal with the area of the 1D4 sensor)



To scale to the same resolution, you just need to scale the larger resolution to the smaller one, so scale the image taken with the 1D4 (4896 x 3264 photosites) to 3298 x 2204 photosites. Now you can compare the images either either in full size, or pixel by pixel.
 
Upvote 0
F

Flake

Guest
Sensor size has very little influence on image quality when comparing sensors of APS H and FF, in fact it is possible to get a better image out of an APS-H because it's not using the weak corners and borders of lenses. If you want sharp images then buy a camera with a weak AA filter, (like the original 5D) , that is going to have more of a bearing on sharpness than the size of the sensor.

Comparing the 7D and the 5D MkII reviewers say it's difficult to tell a difference, and that's when people are aware there is a difference & are looking to find it!

I would suggest to anyone who believes this is a fact, that you have a read of this http://www.luminous-landscape.com/reviews/kidding.shtml A little experiment where industry professionals were asked to tell the difference between two large prints. The fact that they could not distinuguish between one & the other tells a story, but when you consider that one image was taken with a Hasselblad P2 with phase one P45 back, and the other was a Canon G10, that's when you begin to realise that there's an awful lot of BS being talked around the forums.

Truth is as with so many things, photography is being affected by a steeper diminishing returns curve, you spend more & more money to obtain less & less of an improvement.
 
Upvote 0
Flake said:
Sensor size has very little influence on image quality when comparing sensors of APS H and FF, in fact it is possible to get a better image out of an APS-H because it's not using the weak corners and borders of lenses.

It depends on the lens. If the central 1.3x of FF is more than 1.3x sharper than the outer 1.3x of FF, then yes, otherwise no. Same thing for APS-H, but then with the factor 1.6x.

Flake said:
The fact that they could not distinuguish between one & the other tells a story, but when you consider that one image was taken with a Hasselblad P2 with phase one P45 back, and the other was a Canon G10, that's when you begin to realise that there's an awful lot of BS being talked around the forums.

I think that comparison makes a good point, but at the same time it's a bit misleading. Sure there are situations where you can do just as well with a G10 as a FF or MF camera. The comparison you link to proves that. But there are also situations when the differences between the cameras become evident - in particular low-light or shallow-depth photography, but also in those instances when you really need 39 MP or higher dynamic range.
 
Upvote 0
The comparison was based off full prints, not sectional prints. If you dropped the image down to a 25% crop, the difference in image quality would have been a lot more apparent.

Rescaled to 100x150, there is no apparent difference between a picture from a cameraphone and a picture from a 5D2.

Re 5D vs D90; well, actually I have JPEG comparisons where the 5D has comparable performance to the D300, but I'm still hunting down a RAW comparison.

http://www.bobatkins.com/photography/digital/size_matters.html

why larger sensors have an inherent advantage over smaller sensors
 
Upvote 0
B

Bob Howland

Guest
Flake said:
I would suggest to anyone who believes this is a fact, that you have a read of this http://www.luminous-landscape.com/reviews/kidding.shtml A little experiment where industry professionals were asked to tell the difference between two large prints. The fact that they could not distinuguish between one & the other tells a story, but when you consider that one image was taken with a Hasselblad P2 with phase one P45 back, and the other was a Canon G10, that's when you begin to realise that there's an awful lot of BS being talked around the forums.

Both of those images were taken at base ISO, which for the G10 is 80. I own a G10 and it's a great p&S, but above ISO 200, the differences in noise between it and my 5D become readily apparent. At ISO 1600, most of the uneducated, unwashed photographic masses don't even see noise in the 5D image but do see it in the G10 image. Most don't find it objectionable however. What they do like about G10 images is that everything is sharp, because of the G10's huge depth of field.

I will agree with one thing...there's an awful lot of BS being talked around the forums.
 
Upvote 0
E

Edwin Herdman

Guest
NotABunny said:
Normalize them for sensor size (you have to cut the inner part of the image taken with D3s to fit the area of the 1D4 sensor size - crop 1.3), then normalize them for resolution (because the output resolution would be the same either on screen or on print), and finally normalize them for black level (Nikon applies some in order get rid of the noise in the shadows).
This strikes me as pretty reasonable. However, I wonder how many users are actually normalizing their photographs in everyday usage - hopefully somewhere along the 'image supply chain' when an image is shot for media it gets downsampled reasonably. But many shooters are dependent on images close to the maximum resolution their cameras can output (demanded for posterity if nothing else) and probably the vast majority of shooters simply keep JPEGs straight from the camera, or at most crop RAWs in DPP or ViewNX.

In any case it's an extra step that, while reasonable to expect to be done by pros, is something most folks would just as soon not worry about.

Aside from "out of the camera-croppable" quality, I think the simple price to performance (subjective of course) ratio is the most sensible one when comparing cameras. The D3s does very well versus the 1D Mark IV in that regard, at least at launch when it was $300 cheaper.
 
Upvote 0
Jul 30, 2010
1,060
130
dahmooser said:
I was quite intrigued by the post stating that Nikon has better sensor technology. I currently own a 40D (and I do love it), but my (lucky) girlfriend owns a Nikon D3, and aside from it being a full 35mm sensor, I just think the images that are shot on that seem so much sharper, more in focus, something. It's an intangible that I've been unable to properly express.

Please check your sharpness setting. 40D is well known to be soft at standard setting (sharpness at 3). try to set it at 4. You may change your opinion on the 40D.
 
Upvote 0
N

NotABunny

Guest
Edwin Herdman said:
NotABunny said:
Normalize them for sensor size (you have to cut the inner part of the image taken with D3s to fit the area of the 1D4 sensor size - crop 1.3), then normalize them for resolution (because the output resolution would be the same either on screen or on print), and finally normalize them for black level (Nikon applies some in order get rid of the noise in the shadows).
This strikes me as pretty reasonable. However, I wonder how many users are actually normalizing their photographs in everyday usage

They don't, but the normalization for sensor size only needs to be done in order to see how much the sensor size and photosite size influence the image quality. As visible after all normalizations, the photosite size doesn't (visibly) influence the quality of the full image (since viewed on screen or printed, the images look the same).

But since without sensor size normalization the photo taken with the D3s look better, it becomes clear that the sensor size (visibly) influences the quality of the full image. (Of course, for comparison, the relative subject magnification within the frame, must be the same, which is normally true regardless of the sensor size - this is how people do photography, they just have to get closer to the subject as the sensor size increases, for the same focal length.)


By this I don't mean to say that photosite size never influences image quality. These two images only show that photosite size doesn't influences image quality with the technology used in those sensors / cameras. Perhaps if Canon were to use larger photosites, they might be able to include technology which improves the image quality. (This is just speculation.) I prefer they work to improve the tonal range in low light rather than increase the resolution. Shooting indoors / events with a 40D in natural light is not fun, it's limiting / challenging (even with Fn less than 2).
 
Upvote 0
E

Edwin Herdman

Guest
NotABunny said:
They don't, but the normalization for sensor size only needs to be done in order to see how much the sensor size and photosite size influence the image quality.
But then we come to my point - if people aren't normally doing this, it's less of an advantage for one camera or another.

It's also another issue entirely to talk about cropping images to get a usable image from part of it. Canon always has insisted (so I read all over the place, especially when discussing lens sharpness) that all pixels should be usable for cropping, so if a given Nikon is better than its Canon equivalent in this regard then Canon admit themselves they have some work to do.
 
Upvote 0
N

NotABunny

Guest
Edwin Herdman said:
Canon always has insisted (so I read all over the place, especially when discussing lens sharpness) that all pixels should be usable for cropping, so if a given Nikon is better than its Canon equivalent in this regard then Canon admit themselves they have some work to do.

Canon doesn't have anything to admit here. The key word is "equivalent". One must compare photos taken with cameras with sensors of the same size in order to have equivalence.

So, comparing full photos taken with a Canon 1D4 (crop 1.3) and Nikon D3s is not equivalent. When Canon will release either full frame 1Ds4 or 5D3, comparing full photos with Nikon D3s will be equivalent. (Till that moment, in order to be fair, the sensor size must be normalized from the photos.)

Currently, Nikon D3s produces images with a lower noise level because it has a larger sensor, not larger photosites. (That's what those two images from before prove.)


Consider a photo taken with a D3s and the same taken with a camera with a sensor with a size of 360 * 240 millimeters (with the same photosite size). There is no equivalence in directly comparing the two photos; the one taken with the camera with the larger sensor would have an absolutely stupefying technical quality. Why? Because the larger sensor gathers 100 times more light for the same exposure - almost 7 stops (6.64 to be exact), but the noise level doesn't increase proportionally with the sensor size.
 
Upvote 0
NotABunny said:
Consider a photo taken with a D3s and the same taken with a camera with a sensor with a size of 360 * 240 millimeters (with the same photosite size). There is no equivalence in directly comparing the two photos; the one taken with the camera with the larger sensor would have an absolutely stupefying technical quality. Why? Because the larger sensor gathers 100 times more light for the same exposure - almost 7 stops (6.64 to be exact), but the noise level doesn't increase proportionally with the sensor size.

I'm not very fond of this simplified argument, removing the optical system from the question, which you really can't: for your simile to work, you have to assume that both detectors use the same (10xFF) lens. Now what if you were imaging a bird that just fit onto the small FF frame. Would the image of the bird be better with the 10xFF detector? No. Sure, you would capture 100x more photons, but 99% of those photons would come from the boring forest, of no consequence for the image quality of the bird.

A detector doesn't produce an image by itself. It needs optics. The reason a FF camera has an IQ advantage over APS-C is that it is easier to produce suitable optics for FF than the equivalent for APS-C. In your example, if you put a 50/1.2 lens in front of the FF and a 500/12 lens in front of the 10xFF, they would produce equivalent images. They would collect the same number of photons. There would be no difference in IQ. But while a 500mm f/12 lens can be readily produced at home by an amateur astronomer (they actually do a bit better), you need Canon's expensive top-of-the-line L-optics to find a 50mm f/1.2 lens.
 
Upvote 0
E

Edwin Herdman

Guest
NotABunny said:
Canon doesn't have anything to admit here. The key word is "equivalent".
My point is that I'm talking about situations which most people consider relevant, while you're harping on this theme of "in theory." That's nice, and in theory I don't have any substantive disagreement with what you've written. I realize you were writing earlier in response to some earlier stuff, but my point remains that this simply ignores some pretty well-established critical points laid out by users and even the companies themselves in trying to target those users.

If I can use a sort-of similar example: Everybody agrees that you lose a bit of control and ultimately image quality when shooting JPEGs processed internally by the camera, instead of saving RAWs and processing them later. But people do it because it's convenient. You and I might have the time to apply some more sophisticated techniques, but Canon wants to market their systems as being usable, and ultimately competitive, not just when the sophisticated techniques are used, but also when the system is not at its best - like in JPEG shooting or in the case of camera crops.

Basically, the theory is nice but it has to come to grips with the mundane indignities of everday shooting. ;)
 
Upvote 0
N

NotABunny

Guest
epsiloneri said:
I'm not very fond of this simplified argument, removing the optical system from the question, which you really can't: for your simile to work, you have to assume that both detectors use the same (10xFF) lens. Now what if you were imaging a bird that just fit onto the small FF frame. Would the image of the bird be better with the 10xFF detector?

I understand your point, but you're talking about the artistic side. The original post was talking about noise level in the context of photosite size. I was also talking about that. I've used "image quality" not to refer to the artistic side, but to the technical side, that is, also including things like tonal range. It's irrelevant for instance how much motion blur you have when talking about noise level, tonal range, dynamic range.

(Also note that DXO Labs test the camera sensors only, excluding the lens entirely.)

My argument is that photosite size is irrelevant in the context of current technology, and this is not theory - the two images that I gave are proof (but one must first understand the math behind the comparison method; otherwise, people must wait for cameras with same sensor size).
 
Upvote 0
NotABunny said:
I understand your point, but you're talking about the artistic side.

No, I'm not talking about the artistic side, I just gave the bird as an example. I could just as well have said "the ability to image an intensity distribution of a given solid angle that just fits onto the FF sensor".

Of course you can test the sensors only, without an optical system. It's the statement "bigger sensors collect more light" that is misleading. I agree that photosite size is only of second-order importance for a well-sampled image.
 
Upvote 0
N

NotABunny

Guest
epsiloneri said:
I just gave the bird as an example. I could just as well have said "the ability to image an intensity distribution of a given solid angle that just fits onto the FF sensor". It's the statement "bigger sensors collect more light" that is misleading.

Ok, I think I understand. You may have missed when I said that this (= the noise level is lower in photos taken with larger sensor) works only if the relative subject magnification is the same (pointless to specify since it's the norm, but accurate). I presume that this is what you meant with the bird example - that the bird must have the same magnification in the photo.

This is how the usual photographs are taken. For example a closeup portrait is taken from the top to the bottom margins of the frame. Basically the constant relative magnification ensures that the noise level relative to the subject is lower as the sensor gets larger, since the light gathered by the sensor in the subject's area is increased.

In the case of a flying bird, as the sensor becomes larger, the photographer needs to either get closer or use a longer focal length in order to preserve the magnification.
 
Upvote 0
Status
Not open for further replies.