BTW, here are two of Phil's iso 3200 crops shown at the same magnification. First is the 1Ds3 unaltered, second is a little fiddling to try to get the 1Ds3 to match colour/contrast of the D3. Last, and least, is the D3. http://cyberphotographer.com/1ds3/dpr_D3v1Ds3_iso3200.jpg
Here's Phil's quote: "There's no doubt that the D3 is the winner once you get above ISO 1600, and that its output remains usable right to the top of its extended range."
What it boils down to is that even Phil Askey isn't capable of interpreting 100% crops from two cameras which capture different numbers of pixels.
Whenever you assess image quality from two cameras on screen, always uprez the lower density image, or both, to some higher resolution at or above the higher density camera's file size.
Tentacle wrote:
My technical instinct (if such a thing exists, else call it educated guesswork) says that both the Mk III cameras are using the same generation sensor technology. Shaky assumption perhaps, but I'll run with it.
Then you can assume that both sensors have the same per-area sensitivity to light. If so, then the 1Ds3 needs more gain in order to get the same output than the 1D3. More gain = more amplification of noise = bad. Likewise, full photo site saturation depends on the physical size, so bigger pixels = better DR. So there is enough grounds for me to be skeptic about it.
I'd like to see a one on one comparison before I really embrace your idea that the 1D3 is outperformed noise-wise by the 1Ds3. ...Show more →
It's very little to do with pixel size and it's everything to do with sensor size and the generation of sensor technology. The 1D3 is a 1.3 crop. Even the D3 beats that. Game over.
I own and use both a 1D Mark III and a 1Ds Mark III, and there's no question which camera I turn to for low light work - and it ain't the one with more pixels.
cwphoto wrote:
I own and use both a 1D Mark III and a 1Ds Mark III, and there's no question which camera I turn to for low light work - and it ain't the one with more pixels.
...is that because (a) the 1Ds3 has no ISO 6400 setting, and (b) you haven't compared results at iso 6400 in print, with sensible noise reduction applied?
brainiac wrote:
[...] It's very little to do with pixel size and it's everything to do with sensor size and the generation of sensor technology. The 1D3 is a 1.3 crop. Even the D3 beats that. Game over.
Please lets not confuse the topics. The crop vs. FF debate can be held elsewhere, as far as I'm concerned.
Signal to noise ratio (which inherently means per-pixel noise) depends on pixel size. Bigger pixels need less gain. Less gain means less noise. So, at the very basis of the discussion, pixel size does matter. And it does too for DR because DR depends both on the noise floor and the point at which a photo site saturates.
Now, you can take pixel size out of the equation by just looking at print size and either downsizing the high res version (thus eliminating noise) or uprezzing the lower resolution. Okay, that's a valid approach, but that merely bypasses the fundamental influence of pixel size, it doesn't invalidate it.
---
The problem with this discussion is that there are different ways to think about noise. Noise per pixel, noise in print, the subjectivity of the appearance of noise, even how well NR can work on the specific characteristics of a given noise pattern a camera creates. And most of the time ("Mea Culpa", I'm guilty too) we leave our own idea about noise wholly implicit.
32067dlm wrote:
I think that making up data through uprezzing or interpolation is no worse an offense. It's like painting an orange red to compare it to an apple. I don't completely disagree with you, but interpolation introduces errors of it's own.
Not really. But if you feel like that, uprez the D3 file to precisely double the baselength using nearest neighbour. Then you have a file which is mathematically identical to the original, except that each pixel is represented by 4 identical pixels. Uprez the 21 megapixel to 48 megapixel by bicubic. It won't make a shred of difference in the perceived quality of the file. Here's an example of just that. It is easy to spot which file is which and assess detail/noise:
Those were one file interpolated up to 177%, and another at 200% identical to the original, followed by 200% by bicubic interpolation. I don't think uprezzing significantly damages real data in a file.
brainiac wrote:
...is that because (a) the 1Ds3 has no ISO 6400 setting, and (b) you haven't compared results at iso 6400 in print, with sensible noise reduction applied?
I very rarely need to upres images, which appears to be what your argument relies upon: same size images.
So at every ISO my 1D Mark III is cleaner.
In fact, the only practical example I can think of which would support your case is if you want to print beyond the 'standard' dpi for a given print viewing distance relationship - which for me is never.
Notwithstanding the fact that not all of my images are presented in hard-copy either.
This has nothing to do with uprezzing for the sake of printing big. The reason I suggest uprezzing is so that you can compare images at 100% on screen AND AT THE SAME PHYSICAL SIZE/MAGNIFICATION. There is no point comparing a square inch from two prints of different sizes. If you allow that, then a Minox is a sharper camera than a Hasselblad. Just print the Hasselblad shot at 50x50 feet, and the Minox at 50x50 cm and hey presto! the Minox wins every time.
That is the sole reason why people think the high iso image quality order is D3>1D3>1Ds3 when really it's 1Ds3>D3>1D3.
brainiac wrote:
This has nothing to do with uprezzing for the sake of printing big. The reason I suggest uprezzing is so that you can compare images at 100% on screen AND AT THE SAME PHYSICAL SIZE/MAGNIFICATION. There is no point comparing a square inch from two prints of different sizes. If you allow that, then a Minox is a sharper camera than a Hasselblad. Just print the Hasselblad shot at 50x50 feet, and the Minox at 50x50 cm and hey presto! the Minox wins every time.
That is the sole reason why people think the high iso image quality order is D3>1D3>1Ds3 when really it's 1Ds3>D3>1D3....Show more →
OK, I'll make it simple: when I print an 8x12" from both a 1Ds Mark III and a 1D Mark III taken of the same subject and same exposure etc my 1D Mark III print looks less noisy.
Tentacle wrote:
[...] It's very little to do with pixel size and it's everything to do with sensor size and the generation of sensor technology. The 1D3 is a 1.3 crop. Even the D3 beats that. Game over.
Please lets not confuse the topics. The crop vs. FF debate can be held elsewhere, as far as I'm concerned.
It is not a confusion. Sensor area matters to high iso. That's why compacts are crap at iso 3200. A full frame camera can provide less noisy _images_ than a 1.3 crop camera of the same generation EVEN IF IT HAS SMALLER PIXELS. That's because enlargement, magnification, whatever you want to call it, is crucial to the intrusion of noise in images. Noise averages out. If pixels are very small, and there are lots of them, then they don't need to be enlarged so much, and therefore they can afford to be noisier.
brainiac wrote:
It is not a confusion. Sensor area matters to high iso. That's why compacts are crap at iso 3200. A full frame camera can provide less noisy _images_ than a 1.3 crop camera of the same generation EVEN IF IT HAS SMALLER PIXELS. That's because enlargement, magnification, whatever you want to call it, is crucial to the intrusion of noise in images. Noise averages out. If pixels are very small, and there are lots of them, then they don't need to be enlarged so much, and therefore they can afford to be noisier.
I understand what you are saying, but I don't see it in practice - sorry.
Like I said, maybe it's because I don't 'enlarge' as you term it.
cwphoto wrote:
OK, I'll make it simple: when I print an 8x12" from both a 1Ds Mark III and a 1D Mark III taken of the same subject and same exposure etc my 1D Mark III print looks less noisy.
Can you show us? Have you taken the same picture, at the same shutter speed and aperture with both cameras and printed them to exactly the same size, after applying appropriate noise reduction tailored to each file?
brainiac wrote:
Can you show us? Have you taken the same picture, at the same shutter speed and aperture with both cameras and printed them to exactly the same size, after applying appropriate noise reduction tailored to each file?
Sure. I'll grab them over the weekend if I get a chance.
brainiac wrote:
[...] Noise averages out. If pixels are very small, and there are lots of them, then they don't need to be enlarged so much, and therefore they can afford to be noisier.
See? And that's why I was specific this time and mentioned per-pixel noise. You either missed that or blatantly ignored it. There is no averaging out if you are talking about noise per pixel.
Whether that pertains to your style of shooting and publishing (print? web? cropped or not?) is another question. You, however, keep insistently projecting your own way of photography on this discussion about noise as if that's the One Path. While you do have some very valid points, that doesn't mean that you can push your view onto everyone else.
I will crop a shot up to 2:1 pixel ratio if that is what it takes to get the composition I couldn't get through the lens. I'd do that with a 1Ds3 and a 1D3. The first would give me more range/reach, but at comparable pixel ratio (and the 1D3 having a typical working distance that has to be closer than with the 1Ds3) the 1D3 will beat the 1Ds3. But that's just for how I photograph! It does not have to apply to everyone else.
cwphoto wrote:
OK, I'll make it simple: when I print an 8x12" from both a 1Ds Mark III and a 1D Mark III taken of the same subject and same exposure etc my 1D Mark III print looks less noisy.
...and here are the crops from Phil Askey's comparison at iso 3200. The 1D3 seems to have more intrusive colour noise, simply, as I have been trying to explain, because it's bigger: http://cyberphotographer.com/1ds3/1Ds3v1D3.jpg
brainiac wrote:
...and here are the crops from Phil Askey's comparison at iso 3200. The 1D3 seems to have more intrusive colour noise, simply, as I have been trying to explain, because it's bigger: http://cyberphotographer.com/1ds3/1Ds3v1D3.jpg
Well that's not my experience, I'll grab the samples when I can over the weekend.
Tentacle wrote:
See? And that's why I was specific this time and mentioned per-pixel noise. You either missed that or blatantly ignored it.
I ignored it because per pixel noise is only relevant where the two cameras have the same number of pixels. It is precisely because people concentrate on per pixel noise, without realising how irrelevant it is in the context of two cameras with different numbers of pixels, that the confusions over this issue arise.
brainiac wrote:
Can you show us? Have you taken the same picture, at the same shutter speed and aperture with both cameras and printed them to exactly the same size, after applying appropriate noise reduction tailored to each file?
Brainiac, I see what you're trying to do, but I personally believe you're contaminating your tests. Take the image straight from the camera, don't uprez or apply noise reduction. Or if you do, apply them equally, but even then, that'd be wrong. If you want to compare cameras, you have to control the variables, not add more. The 1DIII is a 1.3 crop with 10MP, the 1DsIII is FF with 21, the D3 is FF with 12MP. They are what they are; print test prints, compare, choose a camera then begin to doctor your images.
The 1DsIII is a great camera, but it is what it is, and you don't have to justify your choice or findings to us.
32067dlm wrote:
Brainiac, I see what you're trying to do, but I personally believe you're contaminating your tests. Take the image straight from the camera, don't uprez or apply noise reduction. Or if you do, apply them equally, but even then, that'd be wrong. If you want to compare cameras, you have to control the variables, not add more. The 1DIII is a 1.3 crop with 10MP, the 1DsIII is FF with 21, the D3 is FF with 12MP. They are what they are; print test prints, compare, choose a camera then begin to doctor your images.
The 1DsIII is a great camera, but it is what it is, and you don't have to justify your choice or findings to us....Show more →
I can't agree with you I'm afraid. It is precisely in order control the variables that you must magnify the images to the same size and view them fairly. Without printing, that means setting the files to the same number of pixels, since monitors are quite low rez and can favour one interpolation over another.
I am trying to help others to avoid the thoroughly misleading comparisons of high iso noise and detail resolution which take place all over the web, and invariably fail to properly compare camera output. You don't have to agree with the fact that comparison between 100% crops from different cameras is misleading, and you are welcome to compare by your own means. But if you look at Phil's cited comparison, and the properly magnified samples I have posted here, I hope you can see that there is something in what I am saying. If not, refute it, but not with received wisdom and opinion. Show us what's wrong with my explanation and demonstration of how Phil's test fails to properly assess these cameras. You've seen the crops, you've heard the science. Refute it with some tangible argument, evidence and demonstration.
I hasten to add that I have owned 5D, 1D3 and 1Ds3. I don't need to justify anything as I make a living from photography and can use whichever tool does the job best. All these cameras are great. But I find the 1Ds3 has less intrusive noise at high iso's than my 1D3, which matters to me a lot as I shoot evening events. I'm simply looking for the least noisy camera in almost no light, for any given print. I think it's the 1Ds3, and Phil Askey's crops show why, even though he has misinterpreted them himself.
