jorkata wrote:
Here’s a quick math that I did for fun:
The 40D and 1D3 are both 10mp cameras.
The difference in sensor sizes results in a 5.7µm vs 7.2 µm pixel size difference (see the pixels sizes here).
That’s less than 25%.
So, with ~25% larger pixels, the 1D3 has about 1.5-2 stops ISO/noise advantage over the 40D.
The pixels size of the 7D (18mp) is 4.3µm.
This is ~20% smaller than 5.2µm, which is the pixel size of the 450D (12mp).
What that means is that, in theory, if Cannon applies the 7D same sensor tech to a 12mp camera, this camera will have almost 1.5-2 stops ISO/noise advantage over the 7D.
In other words, if the 7D had ‘only’ 12mp, its ISO/noise performance would have been the same as the 1D3 (almost).
Imagine that !
A lot of people welcome the 18mp sensor of the 7D but not all.
Instead of having a single choice of an 18mp sensor, Canon should also have a 12mp version of the 7D with the incredible ISO of the 1D3.
We won’t be arguing on these forums anymore – everybody will get what they want. ...Show more →
Your math is way off. If the 5D2 is barely pulling those numbers in noise improvement what makes you think a 10MP 1.3x sensor is doing the same?
Image noise correlates to sensor size, not pixel size. This is not debateable. It doesn't matter if the 7D was 30MP, overall image noise would still correlate to sensor size, not pixel size. Of course there would be more pixel level noise. As it has already been mentioned you cannot judge noise with 100% crops, it's just not practical. You must look at overall image quality which includes noise, DR, and resolution. The 7D is very comparable at high ISO to the 1D mkiii with sensor improvements and no you can't compare a FlexNR mkiii image to a 7D image and claim it's better. Use DPP for ISO 3200 images from both systems. Looking at the files at imaging-resource I don't know how you claim ANY improvement for the 1D Mkiii over the 7D at high ISO.
Judging them both at the pixel level is like taking a shot with the mkiii and printing it at medium size and printing the 7D at a much larger size and claiming the 1D mkiii the winner. It's not that cut and dry.
John McLean wrote:
To tell the truth I really wasn't concentraiting on the brass, but it is pretty close. It is not a deep, deep brass bezel and the light is reflecting off of it pretty well.
I've noticed one of the side effects of Canon's chroma noise reduction is that it reduces saturation as it is increased in strength. I think this could be the main cause of the change of color in the brass.
I just picked up a 7d today. Now according 2 custom function C.FN III 8 VF display illumination, when enable, my AF point will illuminate in red.
I'm shooting in AI Servo mode with single point AF point selected and my selected focus point does not illuminate. Did I miss something? Shouldn't it illuminate.
I also notice my Auto Lighting Optimizer is set to off. There are no other choice for me to select. The manuel show you off, plus 3 other settings. Should it be doing that?
jorkata:
I'm saying that without noise reduction, ISOs 400 and 800 are maybe the noisiest of any Canon 1.6x after the 20D. Noisier even than the 50D, which is not known as a clean ISO camera.
To those who do like the camera:
Is it true that the noise in ISO 400-800 range is high? I saw quite a few conclusions like that (even liquidstone mentioned that). Many are saying that ISO 1600-3200 range is great but I am not interested in going higher than ISO 800.
I do not have 7D yet (returned my pre-ordered camera because of a three week trip I am on and trying to make up my mind)
I just picked up a 7d today. Now according 2 custom function C.FN III 8 VF display illumination, when enable, my AF point will illuminate in red.
I'm shooting in AI Servo mode with single point AF point selected and my selected focus point does not illuminate. Did I miss something? Shouldn't it illuminate.
The AF point doesn't illuminate in AI servo, at least for all the EOS cameras I have owned and the focus confirm doesn't operate either.
charlyw:I had the 20D, 30D, 40D - on those ISO-800 was end of the line...The 50D was better ...
I read with interest what you have written but you lost me saying that "50D is better". I have not seen a single test showing that. Start with DxO mark for "print". 50D is basically identical (in print) to 40D.
After reading all the posts here and other sites, I cannot wait to see DxO tests. In contrast to many here, I found them matching my experience. By the way, "per pixel" noise matters if one want to use 18 Mpixels to crop to 10 Mpixels for a "lens converter" effect.
charlyw: Guess what: In terms of ISO/noise performance the 7D trounces the 1D Mk.III.
It is becoming even more interesting. Many considered 1D mkIII better, as far as S/N is concerned, than 5D. Am I to assume that you are saying that "7D double trounces 5D" in S/N category?
Is it true that the noise in ISO 400-800 range is high? I saw quite a few conclusions like that (even liquidstone mentioned that). Many are saying that ISO 1600-3200 range is great but I am not interested in going higher than ISO 800.
I do not have 7D yet (returned my pre-ordered camera because of a three week trip I am on and trying to make up my mind)
That is not true, generally and scientifically, I have measured it. The noise relative to midtone signal increases with about 3 dB per ISO stop from 100 to 12800. There is nothing special about 400 and 800. I cannot rule out copy variations, but for photon shot noise that is only dependent on the sensitivity, and as long as the camera exposes properly, there is no chance at all that the photon shot noise will be different. The read noise may vary though.
On a subjective basis, some people may form that opinion about 400-800 still. That can happen if they are accustomed to the per-pixel noise level from the 40D at ISO 400, and find it to be so good that it doesn't degrade image quality in their eyes. Now, if the per-pixel noise on the 7D is higher, they feel it is above their subjective limit at 400. Even if the per-pixel noise at ISO 200 is higher by the same amount, it still falls below their subjective limit.
Some people buy into the extra resolution to be able to make larger prints, with the same per-area quality. If they then expect to use the same ISO setting, they either don't know or don't respect the laws of physics. The same goes for cropping.
In addition to Alundeb's analysis, there's (for example) a pretty big 7D image pool on Flickr now, at every ISO from 100 to 12800.
I've been through every last one of those images - at the biggest size I could find as well as at web presentation size - and while there's the odd one where you go "hmmm...", the vast majority are perfectly fine or better: the high ISO ones are very impressive of course, but there are plenty of 320, 400, 800 and 1600 ISO files that look great.
So much for only having acceptable IQ at 100 and 200 ISO.
And, as I and others have pointed out, once our favourite converters catch up with the 7D, things will get better again.
Still, why let the facts and the evidence get in the way of a bit of selectively-quoted flamebait trolling, eh?
That there may be odd cameras that are misbehaving is not in doubt: but there's no consensus here or anywhere else that the 7D is a noisy camera at any given ISO.
jorkata wrote:
Here’s a quick math that I did for fun:
The 40D and 1D3 are both 10mp cameras.
The difference in sensor sizes results in a 5.7µm vs 7.2 µm pixel size difference (see the pixels sizes here).
That’s less than 25%.
So, with ~25% larger pixels, the 1D3 has about 1.5-2 stops ISO/noise advantage over the 40D.
The pixels size of the 7D (18mp) is 4.3µm.
This is ~20% smaller than 5.2µm, which is the pixel size of the 450D (12mp).
What that means is that, in theory, if Cannon applies the 7D same sensor tech to a 12mp camera, this camera will have almost 1.5-2 stops ISO/noise advantage over the 7D.
In other words, if the 7D had ‘only’ 12mp, its ISO/noise performance would have been the same as the 1D3 (almost).
Imagine that !
A lot of people welcome the 18mp sensor of the 7D but not all.
Instead of having a single choice of an 18mp sensor, Canon should also have a 12mp version of the 7D with the incredible ISO of the 1D3.
We won’t be arguing on these forums anymore – everybody will get what they want. ...Show more →
However, your determination to ignore the clear evidence that has already been put in front of you on this thread suggests that you may never understand that per pixel noise isn't an absolute measure of how well a camera takes a picture. All I can say is that you are wrong in your assertion that the 7D is worse than previous crop cameras in terms of noise at isos above 200. There are examples on this very thread of it smacking the 50D, and I think it is highly likely that it is a better camera at isos 200 and above than the 1Dmk3. Of course, to ascertain the truth of that, a per pixel analysis like yours will not do, as it will be completely misleading to compare 100% crops from 10 and 18 Mpixel cameras. The 1D3 file will need to be uprezzed to 18 Mpixels to make a fair comparison on screen, at which point I think we will see that the greater technological maturity of the 7D will overcome the 1D3's sensor size advantage, perhaps even before appropriate chroma noise reduction is applied. It's clear to me, jorkata, that you simply haven't understood these issues at all, and the opinion that you are proffering here isn't based on experience of comparison of real output.
Only my 7D shows the maze/basket-weaving patter to any noticeable degree when using ACR. While 7D ACR is BETA, in the past, the final version tended to usually perform the same in terms of maze artifacts.
If you just use DPP and in cam jpgs or perhaps certain other programs (?) it might not be much of an issue at all for anyone but it is kind of a mess to leave it to converters to recognize and try to counteract it since if you end up with a body that is not perfectly balance and you don't like DPP or in-cam jpgs too much....
Before I level on just this patch alone, when the orange was brighter and the levels tool didn't have the shadow slider brought to high, the patch looked totally smooth on the 5D2 but still had some basket/maze texture on the 7D which adds the appearance of a lot more noise than it should have and it can also obscure certain fine details like actual fabric weaves or dimpling on walls, etc.
Also note I am not at all saying that all or even most or even more than a handful here and there have this. With the 5D2, it was very rare.
(the 20D and 50D were taken outdoor under direct sunlight, the others indoors, i didn't carefully match brightness here, etc. don't compare general noise between them, especially not the 20D/50D to the others here)
mfurman wrote:
I read with interest what you have written but you lost me saying that "50D is better". I have not seen a single test showing that. Start with DxO mark for "print". 50D is basically identical (in print) to 40D.
The 50D's bad rep is probably based on the error of 100% comparison with the 40D, i.e. the standard magnification error. The 50D seems to perform as well as the 40D at high iso once you normalise magnification, and what is more, at lower isos with adequate lenses it probably captures more detail. Of course there are plenty of online samples of the 50D capturing no more detail than the 40D, but there are also plenty of samples showing a 5D2 capturing no more detail than a 5D. The sensor isn't the only bottleneck for detail.
After reading all the posts here and other sites, I cannot wait to see DxO tests. In contrast to many here, I found them matching my experience.
DxO tests tell you little about the quality of noise patterns, and evenly distributed random noise is far less intrusive and far more responsive to NR than stripy or hatched large scale noise like that found in the 50D or 5D2. Furthermore, DxO presents per pixel data in comparisons between cameras with different numbers of pixels, which is a pointless and thoroughly misleading way of pretending to compare camera performance.
By the way, "per pixel" noise matters if one want to use 18 Mpixels to crop to 10 Mpixels for a "lens converter" effect.
The advantage of the higher rez and lower noise camera lies EITHER in cropping more OR in use of higher isos. You can't expect to get away with no degradation when doing both things at once, as that is trying to count the advantage twice.
To brainiac (you actually answered this question in your post but I did not see it when posting):
I completely understand your analysis but I have a question regarding cropping. Let's assume that one wants to use 18 Mpixel camera to crop further. Let say that we are interested in 3300x2200 pixel image so anything above that is going to be cropped (the number of "pixels per duck" factor). We are not downrezzing or uprezzing, we may not be even printing - we want this bit map file.
What do you (and others) think about 7D to 5d mkII and 40 D noise comparison in this context? I am not trying to be controversial. This is the scenario I am genuinly interested in.
mfurman wrote:
To brainiac (you actually answered this question in your post but I did not see it when posting):
I completely understand your analysis but I have a question regarding cropping. Let's assume that one wants to use 18 Mpixel camera to crop further. Let say that we are interested in 3300x2200 pixel image so anything above that is going to be cropped (the number of "pixels per duck" factor). We are not downrezzing or uprezzing, we may not be even printing - we want this bit map file.
What do you (and others) think about 7D to 5d mkII and 40 D noise comparison in this context? I am not trying to be controversial. This is the scenario I am genuinly interested in....Show more →
I think you would do better to choose a different lens.
mfurman wrote:
What do you (and others) think about 7D to 5d mkII and 40 D noise comparison in this context? I am not trying to be controversial. This is the scenario I am genuinly interested in.
I only owned a 40D briefly, but my impression is that at native res it would go like this: 5D2 > 7D > 40D.
But in this scenario, you could do even better in both the 5D2 and 7D by shooting sRAW1 or MRAW respectively, and still obtain an image large enough to crop to your specified area.
