ChrisDM wrote:
Weather sealing, built-in IS, ultralight ultracool Limited Pancake prime lenses, greater resolution, DNG file format and in-camera RAW development, LCD color calibration, dust mapping, and last but certainly not least: real Auto ISO... The K20D just simply spanks anything that Canon has to offer in this category now. Hopefully the 50D will have half of the incredible features of the K20D, but I doubt it.
Chris M
www.imagineimagery.com
Fair enough. Have to say I'm very happy with the 450D and it does have auto-ISO
> They don't say that pixel count or pixel size is a disadvantage, they say pixel density is a disadvantage. And it is... However it isn't the only factor in determining a camera's noise characteristics. It is the most reliable first indicator however
No! The whole point I am trying to make is that pixel density makes little to no difference. It is the fifth indicator, after sensor size, date of release, sensor type (CMOS, CCD etc.), and designer/foundry (Canon, Sony etc.). The whole point of this thread is that when pixel density increases, so the space occupied in a print by each pixel decreases, and so little difference is seen. So little space is wasted between sensor sites due to micro-lens technology that pixel density has little effect on the light gathering surface area of the sensor.
If you could get a Canon G9 sensor in full frame dimensions, it would crap on all known DSLR's for high iso performance. It would have a gazillion pixels, and the noise of each would consequently be of much lower significance. Signal-to-noise ratios of compact camera sensors are a generation ahead of DSLRs.
So the point of this thread is exactly the opposite of what you have stated. 1Ds3 high iso performance is being underestimated by reviewers because they are concentrating on per pixel noise, and mistakenly think that pixel density is an important factor in high iso performance when it isn't nearly as important as other factors. Hence your surprise, and other people's, that your 1Ds3 fares so well against your 1D3 at iso 3200. You have your own refuting data in front of you, but you're still clinging to the fallacy it disproves.
dhphoto wrote:
Fair enough. Have to say I'm very happy with the 450D and it does have auto-ISO
David
It's good to hear Canon has finally thought about this. The K20D thinks of ISO as just another exposure setting, like aperture and shutter speed. There's a Sv mode, or "sensitivity mode", and the best feature, a TAv mode, or "Aperture and Shutter Priority Mode". You set the desired Aperture and Shutter speed, and the camera chooses the proper ISO to expose the subject. This feature is simply invaluable for events like concerts, sports etc, where you want to control depth of field and motion. That's just smart, thoughtful camera design.
I do agree with you--but at a practical level, do you bin your pixels by printing at 720 ppi or something? 21mpix is more than I though a standard printer could reasonably print for an 8x10 (or something in that range). Obviously, if you're going bigger, there's no question to the advantage of the 1DsIII.
Daniel Heineck wrote:
...at a practical level, do you bin your pixels by printing at 720 ppi or something? 21mpix is more than I though a standard printer could reasonably print for an 8x10 (or something in that range). Obviously, if you're going bigger, there's no question to the advantage of the 1DsIII.
I don't find printing 21 megapixels a problem, but I did upgrade my Mac when I bought the 1Ds3, and frankly it's still not quick. But sRaw and JPEG are perfectly viable tools. For printing to my Epson I quite often downrez to 360 ppi. I do find that 360 ppi is noticeably sharper in print than 180 ppi. I generally use the full 21 megapixel, and I ignore the fact that 3200, 6400 and 12800 iso files look like fireworks at 100%, because I know that any normal print size will do quite a nice job of ironing it out. In the old days I used to like the quality of grain some films gave, and I went to great lengths, even mounting my condenser enlarger on the wall, to ensure that the grain pattern was resolved. For that reason I would have no problems printing a 12800 iso 1Ds3 file really big, with highly visible noise. It's the new charm of the medium. My 12800 iso crops at the beginning of this thread show that sensor noise can look just like grain, and contributes to the texture of the image. I know that many printers add noise to low iso 5D files because they are so smooth. There is less need for that with the 1Ds3 than with the 5D: it's greater detail and better tones come with an attractive grittiness.
brainiac wrote:
So the point of this thread is exactly the opposite of what you have stated. 1Ds3 high iso performance is being underestimated by reviewers because they are concentrating on per pixel noise, and mistakenly think that pixel density is an important factor in high iso performance when it isn't nearly as important as other factors. Hence your surprise, and other people's, that your 1Ds3 fares so well against your 1D3 at iso 3200. You have your own refuting data in front of you, but you're still clinging to the fallacy it disproves.
You're correct in that I was surprised how well the 1Ds3 fared against the 1D3 in the noise department. But what did not surprise me was that the 1D3 file is still cleaner in a srandard print size, although by less of a margin than I would have expected. Once again, comes back to choosing the right tool for the job. If I'm making standard print sizes at high ISO, the 1D3 is the tool of choice, but if I'm making gallery sized prints at any ISO, the 1Ds3 is of course the better choice. Back to the practical argument for me once again, choosing the right tool for the job.
About the G9 comments? My G9 has greater absolute resolution than my 1D3, more megapixels. But can it make a large print anywhere in the same universe as my 1D3? Absolutely not. Simple reason? Tiny little pixels at high density. Once again, talking in practical terms, looking at actual print sizes or actual cameras.
ChrisDM wrote:
About the G9 comments? My G9 has greater absolute resolution than my 1D3, more megapixels. But can it make a large print anywhere in the same universe as my 1D3? Absolutely not. Simple reason? Tiny little pixels at high density. Once again, talking in practical terms, looking at actual print sizes or actual cameras.
My point was about extending the G9 sensor to 24x36mm while maintaining pixel pitch. I am not talking about your G9, or anything like it. I am talking about a 239.3 megapixel camera using the same sensor technology as the G9. Such a camera would undoubtedly be less noisy at iso 12800 than your 1Ds3 and the D3. To get an idea of how good the G9 sensor is compared to your 1Ds3, crop an area of 1186 pixels by 890 pixels from a 1Ds3 file, and uprez it to 4000 by 3000. Then compare it to a G9 file. Pick any iso you like.
> Simple reason? Tiny little pixels at high density.
No. Simple reason? Too small a sensor.
By your argument, putting a 1 megapixel sensor in the G9 would make it as good as the 1D3.
Edited by brainiac on Aug 18, 2008 at 05:57 PM GMT
ChrisDM wrote:
Once again, comes back to choosing the right tool for the job. If I'm making standard print sizes at high ISO, the 1D3 is the tool of choice, but if I'm making gallery sized prints at any ISO, the 1Ds3 is of course the better choice. Back to the practical argument for me once again, choosing the right tool for the job.
Like your test shows, there isn't much visible difference between 1D3 and 1Ds3 files when you equalize them (wether uprezzing or downrezzing). And this is on screen. I doubt any difference will be there in the final print at these standard to medium sizes. When it comes to printing really big, obviously the 1Ds3 has the advantage. So in the end, the 1Ds3 will always be the best tool for the job if you only take the quality of output into consideration
Too small a sensor creates the high pixel density... It's the same thing. The Fuji F30 was the best example of this. They limited the compact sensor to 6 megapixels and produced the best high ISO compact ever seen, by a long shot. Then they tried putting 12 megapixels in the same camera, and of course it didn't work. Pixel density matters, no matter the size of the sensor.
ChrisDM wrote:
> Too small a sensor creates the high pixel density... It's the same thing.
A D3 and a 1Ds3 have significantly different pixel densities, but the same sensor size. Density is independent of sensor size. The D3 and 1Ds3 also have similar sensor technology, both being current micro-lens CMOS designs.
Image for image, cameras of the same sensor size and generation of sensor technology will deliver similar high iso performance, regardless of the pixel density. If their pixel densities differ widely, the per pixel noise rate will likely differ widely too, but the noise per square inch of print will not.
For this reason, noise expectations should generally ignore pixel density, and instead consider much more important factors like sensor size, release date, manufacturer, and sensor technology. Pixel density should have almost no place in a discussion of a camera's high iso performance.
So let's keep this simple: by your reasoning, pixel density matters a lot to noise, and therefore the D3 ought to be much better at iso 3200 than the much denser 1Ds3. By my reasoning, pixel density doesn't make much difference, and therefore the two cameras should produce similar results, except that one has more detail. Once again, here is the correctly magnified iso 3200 DPR crop (1Ds3, 1Ds3, D3): http://cyberphotographer.com/1ds3/dpr_D3v1Ds3_iso3200.jpg
Bearing in mind the Nikon D3's characteristic chroma-blending noise reduction which is pretty obvious here, these crops suggest to me that chroma noise is probably LESS intrusive on the 1Ds3 at iso 3200. I say that because we are looking at less noise-reduced data from the 1Ds3, and yet it still seems to have no worse chroma noise, and the chroma noise is on a smaller scale in the image than in the D3 file, which has questionable large scale areas of colour in the tree's foliage. The large green and purple chroma errors in the tree foliage will be visible at most print sizes, whereas the 1Ds3's smaller scale chroma noise probably won't. Certainly the results are very similar, but if pixel density was the primary factor, the Canon without NR surely couldn't be better than the D3 with NR.
brainiac wrote:
A D3 and a 1Ds3 have significantly different pixel densities, but the same sensor size. Density is independent of sensor size.
I
Pixel density is not "independent of sensor size". In fact, just the opposite: Pixel density is total number of pixels divided by the sensor size. Completely dependent on sensor size... In other words, for any given number of pixels, as sensor size goes up, pixel density goes down. And conversely, as sensor size goes down, pixel density goes up. It's a simple equation.
It's independent in the sense that a designer is free to vary sensor size while maintaining density, or vary the density while maintaining sensor size. The fact that you can get full frame cameras with 8.8 micron and 5.2 micron linear density shows that density is independent of sensor size.
However, you are right in saying that pixel density is not independent of both sensor size and pixel number. But that is a different issue, since not all full frame cameras have the same number of pixels. That kind of dependency is irrelevant here, since we are not all bound to use 12 megapixels. For the purpose of discussion of cameras with different numbers of pixels, density is independent of sensor size, like I said.
brainiac wrote:
It's independent in the sense that a designer is free to vary sensor size while maintaining density, or vary the density while maintaining sensor size. The fact that you can get full frame cameras with 8.8 micron and 5.2 micron linear density shows that density is independent of sensor size.
However, you are right in saying that pixel density is not independent of both sensor size and pixel number. But that is a different issue, since not all full frame cameras have the same number of pixels. That kind of dependency is irrelevant here, since we are not all bound to use 12 megapixels. For the purpose of discussion of cameras with different numbers of pixels, density is independent of sensor size, like I said....Show more →
Richard, you should have been a politician! But "for the purpose of discussion", pixel density is a simple formula derived by the total number of pixels divided by the area of the sensor (sensor size). It is a simple formula, there's no way around it.
ChrisDM wrote:
Richard, you should have been a politician! But "for the purpose of discussion", pixel density is a simple formula derived by the total number of pixels divided by the area of the sensor (sensor size). It is a simple formula, there's no way around it.
I think his point is "a 1.3 crop sensor does not have to have a certain density", and he has said that quite explicit, you trying to troll more on this issue makes you look like.. well a troll! So please drop it
ChrisDM wrote:
Richard, you should have been a politician! But "for the purpose of discussion", pixel density is a simple formula derived by the total number of pixels divided by the area of the sensor (sensor size). It is a simple formula, there's no way around it.
This is physics, not politics. A variable is 'independent' of another if it can vary freely while the other remains constant. Pixel density can vary while sensor size remains constant. The consequence is that number of pixels varies too, but that doesn't take away the independence of density from size, UNLESS YOU STATE THAT NUMBER OF PIXELS IS A CONSTANT. Clearly number of pixels is not a constant, since we are discussing cameras with differing numbers of pixels.
The formula is...
number of pixels = density * area
In physics, all three of those quantities are known as 'independent' variables, because each can vary unless you specify the other two. That is why it is correct, in physics, to say that pixel density is 'independent' of sensor size. Fixing sensor size DOES NOT DETERMINE PIXEL DENSITY.
Think of it another way: I have a 56 x 56 mm sensor here. What is the pixel density? That's right: you can't say. Pixel density does not depend (hang) on sensor size alone. It is independent of sensor size until you determine something else. There are no useful formulae which contain variables which have no possible inter-dependence. The sense in which you wish to use the word 'independent' might better be served by the word 'isolated'.
brainiac wrote:
Think of it another way: I have a 56 x 56 mm sensor here. What is the pixel density? That's right: you can't say. Pixel density does not depend (hang) on sensor size alone. It is independent of sensor size until you determine something else. There are no useful formulae which contain variables which have no possible inter-dependence.
You're talking semantics here obviously, but it doesn't change the simple fact that pixel density is a simple formula determined by number of pixels divided by the sensor area (size). Call it whatever you want.
By my tests brainiac is right regarding comparative noise of Canon's MkIIIs-as he expresses it. Again: I mounted a 35L on my 1Ds Mk III and a 28mm F1.8 on my Mk III. This yielded equal coverage (effective FL of a 28mm on a 1.274x sensor is 36mm). Enlargement of these images at equal percentages yielded the same displayed noise (and that's up to 500%).
These comparison are equivalent because Canon has equal sensor pixel technology and, at least, similar if not identical noise and image processing parameters. I also have the D3 which is very effective as my Go-To body for low light dancers. The D3's features fall between the Mk III and 1Ds Mk III. And with the, new and expensive $500, memory buffer, upgrade the D3 equals he buffer size of the Mk III.
Yes, my side by-side test showed this as well, it is very, very difficult to see the difference between the 1Ds3 and 1D3 at 3200 and standard print sizes much to the 1Ds3's credit. But of course the 1Ds3 doesn't shoot past 3200, so the "new high ISO" generation of shooting (6400+) is left to those cameras designed to shoot up there.
These comparison are equivalent because Canon has equal sensor pixel technology and, at least, similar if not identical noise and image processing parameters. I also have the D3 which is very effective as my Go-To body for low light dancers. The D3's features fall between the Mk III and 1Ds Mk III. And with the, new and expensive $500, memory buffer, upgrade the D3 equals he buffer size of the Mk III.
Yes again, every major reviewer who has worked with the D3 agrees, it is the current "go-to for low light", as you say.. Hopefully for us Canon shooters that will change this fall
