This topic is in the 2009 archive. You can read it and like its posts here. It is read only.
Archive 20097D noise vs 50D numbers
p.6 #17D noise vs 50D numbers
Daan B wrote:
Even more confusing why they choose to cramp their sensors with more MP's again and again
For a given sensor size and technology, if the overhead percentage remains constant then more pixels will result in improved image detail and less aliasing artifacts; this is true on a per picture basis (assuming reasonably large prints).
Regards,
Joe Kurkjian
Even more confusing why they choose to cramp their sensors with more MP's again and again
For a given sensor size and technology, if the overhead percentage remains constant then more pixels will result in improved image detail and less aliasing artifacts; this is true on a per picture basis (assuming reasonably large prints).
Regards,
Joe Kurkjian
Sep 05, 2009 at 08:29 AM
p.6 #27D noise vs 50D numbers
jkurkjia wrote:
For a given sensor size and technology, if the overhead percentage remains constant then more pixels will result in improved image detail and less aliasing artifacts; this is true on a per picture basis (assuming reasonably large prints).
Joe, I know
What is worrying me is the pattern/banding noise (at lower ISO's) in cameras like the 5D2 and 50D. I sincerely hope Canon got it right with the 7D sensor. But until I have seen and developed some 7D RAW files, I will remain somewhat sceptic. 18MP's on a 1.6x crop sensor is a lot.
Daan B wrote
Even more confusing why they choose to cramp their sensors with more MP's again and againJoe, I know
What is worrying me is the pattern/banding noise (at lower ISO's) in cameras like the 5D2 and 50D. I sincerely hope Canon got it right with the 7D sensor. But until I have seen and developed some 7D RAW files, I will remain somewhat sceptic. 18MP's on a 1.6x crop sensor is a lot.
Sep 05, 2009 at 08:53 AM
p.6 #37D noise vs 50D numbers
Fred Relaix wrote:
In fact I feel the 7D is really designed to compete with the D400s rather than the D300s. Any MP-upgraded D300s would look like old news now.
Agree.
The 7D will have a longer update cycle for sure. It won't get updated every 12-18 months like the xxD series, so by putting 18mp into the 7D Canon wanted to make it 'future proof', IMO.
This way it can stay on the market for 2-3 years without looking outdated.
Megapixels are such a polarizing topic.
The fact is, detail gives photos depth and 'dimensionality'.
So, on one hand, adding extra megapixels does improve image quality.
Noise suffers, of course, hence the anti-MP sentiment.
But here’s the thing – despite the anti-MP sentiment, the 5DII is a hot selling product and seems to outsell the cheaper and yet better spec’d Nikon D700.
So, it would seem that people like megapixels after all.
In fact I feel the 7D is really designed to compete with the D400s rather than the D300s. Any MP-upgraded D300s would look like old news now.
Agree.
The 7D will have a longer update cycle for sure. It won't get updated every 12-18 months like the xxD series, so by putting 18mp into the 7D Canon wanted to make it 'future proof', IMO.
This way it can stay on the market for 2-3 years without looking outdated.
Megapixels are such a polarizing topic.
The fact is, detail gives photos depth and 'dimensionality'.
So, on one hand, adding extra megapixels does improve image quality.
Noise suffers, of course, hence the anti-MP sentiment.
But here’s the thing – despite the anti-MP sentiment, the 5DII is a hot selling product and seems to outsell the cheaper and yet better spec’d Nikon D700.
So, it would seem that people like megapixels after all.
Sep 05, 2009 at 12:57 PM
p.6 #47D noise vs 50D numbers
ejmartin wrote:
When I tried doing a similar analysis using the ISO 100 image, I noticed that the std devs for the two green subarrays of the RAW could differ wildly (tens of percent) for the same patch. I'm not sure I trust the results at the moment.
Agree.
Happy to see you here, btw
I find the same wild difference in Std.dev for the two green subarrays at low ISO. The relative difference diminishes at higher ISO's.
Using a white wall for the mean-variance method introduces the risk of getting real image texture into the Std.dev. To get the best possible results, we need to balance image texture against read noise. I ran some calculations for the 50D and found the optimal ISO here to be 1600.
The real system gain for the 50D at IS100 is 2.2 (clarkvision)
The results for system gain normalized to ISO 100, estimated with IR images taken at various ISO's for the 50D:
100: 1.43
200: 1.74
400: 1.95
800: 2.09
1600: 2.19
3200: 2.15
6400: 2.05
At ISO1600, the contributions from image texture and read noise are both so low that the estimate is very good, 2.19 vs. 2.2
The similar table for the 7D is (one green channel):
100: 1.02
200: 1.15
400: 1.32
800: 1.65
1600: 2.00
3200: 2.63
6400: 2.70
12800: 2.56
I would expect more contribution from image texture with this higher resolution camera, and thus worse results at low ISO, but I am not confident that the read noise is so low that ISO 6400 is optimal here. Since these images were taken with High ISO NR OFF, there shouldn't be any "cooking" of the raw files in camera either.
When I tried doing a similar analysis using the ISO 100 image, I noticed that the std devs for the two green subarrays of the RAW could differ wildly (tens of percent) for the same patch. I'm not sure I trust the results at the moment.
Agree.
Happy to see you here, btw
I find the same wild difference in Std.dev for the two green subarrays at low ISO. The relative difference diminishes at higher ISO's.
Using a white wall for the mean-variance method introduces the risk of getting real image texture into the Std.dev. To get the best possible results, we need to balance image texture against read noise. I ran some calculations for the 50D and found the optimal ISO here to be 1600.
The real system gain for the 50D at IS100 is 2.2 (clarkvision)
The results for system gain normalized to ISO 100, estimated with IR images taken at various ISO's for the 50D:
100: 1.43
200: 1.74
400: 1.95
800: 2.09
1600: 2.19
3200: 2.15
6400: 2.05
At ISO1600, the contributions from image texture and read noise are both so low that the estimate is very good, 2.19 vs. 2.2
The similar table for the 7D is (one green channel):
100: 1.02
200: 1.15
400: 1.32
800: 1.65
1600: 2.00
3200: 2.63
6400: 2.70
12800: 2.56
I would expect more contribution from image texture with this higher resolution camera, and thus worse results at low ISO, but I am not confident that the read noise is so low that ISO 6400 is optimal here. Since these images were taken with High ISO NR OFF, there shouldn't be any "cooking" of the raw files in camera either.
Sep 06, 2009 at 01:51 PM
p.6 #57D noise vs 50D numbers
So what does this green channel discrepancy imply, guys?
And re: the 7D figures - what do they suggest about Real World noise at (say) 400 ISO compared to the 50D?
I guess "higher is better", in which case is a 0.63 difference in favour(?) of the 50D a big deal?
And re: the 7D figures - what do they suggest about Real World noise at (say) 400 ISO compared to the 50D?
I guess "higher is better", in which case is a 0.63 difference in favour(?) of the 50D a big deal?
Sep 06, 2009 at 02:24 PM
p.6 #67D noise vs 50D numbers
I have no idea about the discrepancy, Keith.
The numbers will have to be anchored to an exposure reference before we can tell to what extent they reduce noise or increase dynamic headroom at a given ISO. It can mean both.
The numbers will have to be anchored to an exposure reference before we can tell to what extent they reduce noise or increase dynamic headroom at a given ISO. It can mean both.
Sep 06, 2009 at 02:43 PM
p.6 #77D noise vs 50D numbers
keithreeder wrote:
So what does this green channel discrepancy imply, guys?
If you look at the new RAW-sharpness sample at IR, while looking very crisp, it also has lots of artifacts (referenced in a DPR thread). Now DPP is known for that, however, if you look at 200% you can easily start seeing the maze-pattern issue which can also mean a sensor issue. I'm not sure if DPP is that bad. I think a few 5D2 copies had it and they got a new sensor. I vaguely seem to recall someone found that perhaps a green channel imbalance caused it
Maybe the sample are using is a poorly tuned pre-pro model?
So what does this green channel discrepancy imply, guys?
If you look at the new RAW-sharpness sample at IR, while looking very crisp, it also has lots of artifacts (referenced in a DPR thread). Now DPP is known for that, however, if you look at 200% you can easily start seeing the maze-pattern issue which can also mean a sensor issue. I'm not sure if DPP is that bad. I think a few 5D2 copies had it and they got a new sensor. I vaguely seem to recall someone found that perhaps a green channel imbalance caused it
Maybe the sample are using is a poorly tuned pre-pro model?Sep 07, 2009 at 02:04 AM
p.6 #87D noise vs 50D numbers
The 7D has an improved gapless pixel design and so should (might?) gather a bit more light than a 50D's sensor (buckets of photon per unit area). But a crop sensor will have nowhere as much area as a modern FF sensor (a 1.6x crop is only 38% the area of FF). So I would estimate an, observed, improvement of 1/3 stop assuming all other things are equal (NR, etc.) Any more that an incremental Hi ISO improvement and you can be sure that Canon would, already, be advertising it aggressively.
Sep 08, 2009 at 03:19 PM
p.6 #97D noise vs 50D numbers
Daan B wrote:
Well, actually I do feel a bit shortchanged. That has also something to do with 5D2 prices dropping by 170 euros right after I bought mine 
But I can't believe that Canon didn't have all the new 7D technology (AF, metering, etc) available at the time of developing the 5D2.
I do believe that it was mostly marketing what drove them to shortchange the 5D2 on a few key areas (AF). I mean, there was no need to make it a better camera, because the only direct competitors were the Sony a900 (which isn't top notch when it comes...Show more →
Sorry you feel short changed, but it's easy to understand; Canon "boned" everyone for the second time with the 5d2. I'm sure that if the economy didn't tank, the 5d2 would be selling for $1995 (US) which is where it always belonged.IMO (or perhaps lower with the new Sony out)..
The 5 series is simply a FF sensor in an xxd body. The original 5d (a FF sensor in a 30d body) was $1,000 over priced when released and the 5d2 is at least $700 over priced today. In digital, the 5 series combines second tier parts/features stacked on top of a great sensor in the 5d and a pretty-good/OK sensor in the 5d2.
The 7d is probably 2007-2008 sensor technology on top of similar body technology (probably no 2009 technology in it). I'm pretty sure that the AF has been around for a while and could possibly have been used on the 5d2 (an educated guess) but since Canon screwed the pooch on the mark 3 series AF, I think they decided to hold back a while and see what happened.
I'm not a candidate for the 7d, but I sure hope it turns out to be a good camera.
Yohan Pamudji wrote
I can sympathize with any 5DII owners who might feel shortchanged (good luck getting very many to come out and say it though)But I can't believe that Canon didn't have all the new 7D technology (AF, metering, etc) available at the time of developing the 5D2.
I do believe that it was mostly marketing what drove them to shortchange the 5D2 on a few key areas (AF). I mean, there was no need to make it a better camera, because the only direct competitors were the Sony a900 (which isn't top notch when it comes...Show more →
Sorry you feel short changed, but it's easy to understand; Canon "boned" everyone for the second time with the 5d2. I'm sure that if the economy didn't tank, the 5d2 would be selling for $1995 (US) which is where it always belonged.IMO (or perhaps lower with the new Sony out)..
The 5 series is simply a FF sensor in an xxd body. The original 5d (a FF sensor in a 30d body) was $1,000 over priced when released and the 5d2 is at least $700 over priced today. In digital, the 5 series combines second tier parts/features stacked on top of a great sensor in the 5d and a pretty-good/OK sensor in the 5d2.
The 7d is probably 2007-2008 sensor technology on top of similar body technology (probably no 2009 technology in it). I'm pretty sure that the AF has been around for a while and could possibly have been used on the 5d2 (an educated guess) but since Canon screwed the pooch on the mark 3 series AF, I think they decided to hold back a while and see what happened.
I'm not a candidate for the 7d, but I sure hope it turns out to be a good camera.
Oct 09, 2009 at 04:30 PM
brainiac
Offline
[X]
Details
p.6 #107D noise vs 50D numbers
PhotogDave wrote:
So now, lets see a 5D2 update with the new AF/Metering etc. I really hope the 1D4 doesnt have more than 12-14MP's...in fear that packing them in will really hender the potential of extreme clean High ISO.
Please read this:
http://cyberphotographer.com/megapixelmyth
So now, lets see a 5D2 update with the new AF/Metering etc. I really hope the 1D4 doesnt have more than 12-14MP's...in fear that packing them in will really hender the potential of extreme clean High ISO.
Please read this:
http://cyberphotographer.com/megapixelmyth
Oct 10, 2009 at 03:15 AM
brainiac
Offline
[X]
Details
p.6 #117D noise vs 50D numbers
PhotogDave wrote:
This would have been a complete homerun had Canon went with say a 12MP sensor >with this new sensor tech, then we'd be talkin real ISO performance.
Please read this:
http://cyberphotographer.com/megapixelmyth
This would have been a complete homerun had Canon went with say a 12MP sensor >with this new sensor tech, then we'd be talkin real ISO performance.
Please read this:
http://cyberphotographer.com/megapixelmyth
Oct 10, 2009 at 03:22 AM
brainiac
Offline
[X]
Details
p.6 #127D noise vs 50D numbers
keithreeder wrote:
Bear in mind too that - according to DxO - the 5D only weighs in at 11.1, and yet they reckon the Nikon D200 has 11.5, which is more than the 11.3 of the 40D, the camera decribed by DPR in the following terms:
Bear in mind that DPR and DxO compare images both as 100% crops AND as images correctly resized to the same magnification, so it is not surprising that they can't even agree with themselves. Their methods are broken and, and their results inconsistent. They create extra work for themselves by trying to fit together pieces from two different jigsaw puzzles. They both need to read and understand this: http://cyberphotographer.com/megapixelmyth
Bear in mind too that - according to DxO - the 5D only weighs in at 11.1, and yet they reckon the Nikon D200 has 11.5, which is more than the 11.3 of the 40D, the camera decribed by DPR in the following terms:
Earlier quote
"One surprise came from our testing of the cameras dynamic range, it delivered a consistently higher range than both the competition [they specifically tested the D200 against it] and its predecessor."Bear in mind that DPR and DxO compare images both as 100% crops AND as images correctly resized to the same magnification, so it is not surprising that they can't even agree with themselves. Their methods are broken and, and their results inconsistent. They create extra work for themselves by trying to fit together pieces from two different jigsaw puzzles. They both need to read and understand this: http://cyberphotographer.com/megapixelmyth
Oct 10, 2009 at 03:27 AM
brainiac
Offline
[X]
Details
p.6 #137D noise vs 50D numbers
thw2 wrote:
Looks like Canon is now falling behind Nikon's sensor technology. Not as bad as the pre-D3/D300 era when Nikon was trailing far behind Canon, but certainly VERY disappointing especially after Canon's claim that they aim to improve dynamic range and high ISO performance.
That's one way of looking at it. Another is that this is an excellent 18 Mpixel sensor with superb high iso and a great price, and arguably Nikon has no competitor for it. I agree that DR matters a lot, but the success of the point+shoot market shows that many camera buyers and users don't care about it, and are not asking camera shop assistants which camera has the highest DR.
Looks like Canon is now falling behind Nikon's sensor technology. Not as bad as the pre-D3/D300 era when Nikon was trailing far behind Canon, but certainly VERY disappointing especially after Canon's claim that they aim to improve dynamic range and high ISO performance.
That's one way of looking at it. Another is that this is an excellent 18 Mpixel sensor with superb high iso and a great price, and arguably Nikon has no competitor for it. I agree that DR matters a lot, but the success of the point+shoot market shows that many camera buyers and users don't care about it, and are not asking camera shop assistants which camera has the highest DR.
Oct 10, 2009 at 03:38 AM
brainiac
Offline
[X]
Details
p.6 #147D noise vs 50D numbers
Daan B wrote:
Agreed on that the Nikon/Sony relation might be more complex. OTOH Canon can't just what it wants either, because they have to take the risk involved in developing something new themselves. They also have to sell enough units to break even or make a profit. But at least they can do the development themselves. Even more confusing why they choose to cramp their sensors with more MP's again and again, while IQ remains the same or becomes even worse (pattern noise/banding issues). Which leads me to believe that Canon sensor strategy is more marketing driven than Nikon's... at least...Show more →
Please read this: http://cyberphotographer.com/megapixelmyth
Briefly, there is not a shred of evidence in the market over the last ten years that limiting the number of megapixels provides better quality images. That theory is in direct contradiction of the only evidence there is, namely that image quality has improved dramatically alongside increases in pixel density, from each and every manufacturer.
Steve Spencer wrote
They may not have to beg Sony, but planning a new sensor for Nikon is likely to open some complex negotiations for Nikon, whereas Canon can just make whatever they would like.Please read this: http://cyberphotographer.com/megapixelmyth
Briefly, there is not a shred of evidence in the market over the last ten years that limiting the number of megapixels provides better quality images. That theory is in direct contradiction of the only evidence there is, namely that image quality has improved dramatically alongside increases in pixel density, from each and every manufacturer.
Oct 10, 2009 at 03:59 AM
p.6 #157D noise vs 50D numbers
brainiac wrote:
They both need to read and understand this: http://cyberphotographer.com/megapixelmyth
Bookmarked for future reference and use, Richard - excellent, persuasive article.
They both need to read and understand this: http://cyberphotographer.com/megapixelmyth
Bookmarked for future reference and use, Richard - excellent, persuasive article.
Oct 10, 2009 at 04:04 AM
brainiac
Offline
[X]
Details
p.6 #167D noise vs 50D numbers
Daan B wrote:
What is worrying me is the pattern/banding noise (at lower ISO's) in cameras like the 5D2 and 50D.
This problem is dramatically attenuated in first the 500D, and now the 7D. Results from those cameras do not suffer from banding in the same way that the 5D2 and 50D do. And the 500D and 7D each has more pixels, not fewer, than its predecessor. Onward and upward.
What is worrying me is the pattern/banding noise (at lower ISO's) in cameras like the 5D2 and 50D.
This problem is dramatically attenuated in first the 500D, and now the 7D. Results from those cameras do not suffer from banding in the same way that the 5D2 and 50D do. And the 500D and 7D each has more pixels, not fewer, than its predecessor. Onward and upward.
Oct 10, 2009 at 04:04 AM
brainiac
Offline
[X]
Details
p.6 #177D noise vs 50D numbers
jorkata wrote:
Megapixels are such a polarizing topic.
The fact is, detail gives photos depth and 'dimensionality'.
So, on one hand, adding extra megapixels does improve image quality.
Noise suffers, of course, hence the anti-MP sentiment.
No - there is no evidence that noise suffers, and in fact there is a wealth of evidence to the contrary. The anti-MP sentiment arises simply because people do not understand how 100% crops produce different magnifications. Please read this: http://cyberphotographer.com/megapixelmyth
But here’s the thing – despite the anti-MP sentiment, the 5DII is a hot selling product and seems to outsell the cheaper and yet better spec’d Nikon D700. So, it would seem that people like megapixels after all.
The 5D2 is yet more evidence that pixel density and high iso noise are not related. It is Canon's best ever low-light camera and it is also Canon's joint highest resolving full-frame camera.
Megapixels are such a polarizing topic.
The fact is, detail gives photos depth and 'dimensionality'.
So, on one hand, adding extra megapixels does improve image quality.
Noise suffers, of course, hence the anti-MP sentiment.
No - there is no evidence that noise suffers, and in fact there is a wealth of evidence to the contrary. The anti-MP sentiment arises simply because people do not understand how 100% crops produce different magnifications. Please read this: http://cyberphotographer.com/megapixelmyth
But here’s the thing – despite the anti-MP sentiment, the 5DII is a hot selling product and seems to outsell the cheaper and yet better spec’d Nikon D700. So, it would seem that people like megapixels after all.
The 5D2 is yet more evidence that pixel density and high iso noise are not related. It is Canon's best ever low-light camera and it is also Canon's joint highest resolving full-frame camera.
Oct 10, 2009 at 04:09 AM
p.6 #187D noise vs 50D numbers
brainiac wrote:
Agreed on that the Nikon/Sony relation might be more complex. OTOH Canon can't just what it wants either, because they have to take the risk involved in developing something new themselves. They also have to sell enough units to break even or make a profit. But at least they can do the development themselves. Even more confusing why they choose to cramp their sensors with more MP's again and again, while IQ remains the same or becomes even worse (pattern noise/banding issues). Which leads me to believe that Canon sensor strategy is more marketing driven than Nikon's... at least when it comes to MP count.Please read this: http://cyberphotographer.com/megapixelmyth
Briefly, there is not a shred of evidence in the market over the last ten years that limiting the number of megapixels provides better quality images. That theory is in direct contradiction of the only evidence there is, namely that image quality has improved dramatically alongside increases in pixel density, from each and every manufacturer.
1) I haven't seen the dramatically improved IQ alongside increases in MP's. If you compare output sizes the differences are minimal (if there are any at all).
2) In the meantime, we all have to upgrade our memory capacity and PC power to handle those large files.
3) And things like banding and pattern noise weren't totally eliminated (which affect IQ negatively). Well, maybe Canon has finally got it right with the 7D?
Daan B wrote
Steve Spencer wrote
They may not have to beg Sony, but planning a new sensor for Nikon is likely to open some complex negotiations for Nikon, whereas Canon can just make whatever they would like.Briefly, there is not a shred of evidence in the market over the last ten years that limiting the number of megapixels provides better quality images. That theory is in direct contradiction of the only evidence there is, namely that image quality has improved dramatically alongside increases in pixel density, from each and every manufacturer.
1) I haven't seen the dramatically improved IQ alongside increases in MP's. If you compare output sizes the differences are minimal (if there are any at all).
2) In the meantime, we all have to upgrade our memory capacity and PC power to handle those large files.
3) And things like banding and pattern noise weren't totally eliminated (which affect IQ negatively). Well, maybe Canon has finally got it right with the 7D?
Oct 10, 2009 at 04:11 AM
brainiac
Offline
[X]
Details
p.6 #197D noise vs 50D numbers
jamesf99 wrote:
The 5 series is simply a FF sensor in an xxd body. The original 5d (a FF sensor in a 30d body) was $1,000 over priced when released and the 5d2 is at least $700 over priced today.
People buy 'em.
In digital, the 5 series combines second tier parts/features stacked on top of a great sensor in the 5d and a pretty-good/OK sensor in the 5d2.
Having used 5D, 1Ds3 and 5D2 extensively for real work in a wide range of conditions, I am curious why you think the 5D sensor is 'great' while the 1Ds3/5D2 sensor is only 'pretty-good/OK'. My experience is that the 1Ds3/5D2 sensor is waaay better than the 5D sensor. Perhaps you have been comparing 100% crops, in which case, please read this: http://cyberphotographer.com/megapixelmyth
Edited on Oct 10, 2009 at 04:34 AM · View previous versions
The 5 series is simply a FF sensor in an xxd body. The original 5d (a FF sensor in a 30d body) was $1,000 over priced when released and the 5d2 is at least $700 over priced today.
People buy 'em.
In digital, the 5 series combines second tier parts/features stacked on top of a great sensor in the 5d and a pretty-good/OK sensor in the 5d2.
Having used 5D, 1Ds3 and 5D2 extensively for real work in a wide range of conditions, I am curious why you think the 5D sensor is 'great' while the 1Ds3/5D2 sensor is only 'pretty-good/OK'. My experience is that the 1Ds3/5D2 sensor is waaay better than the 5D sensor. Perhaps you have been comparing 100% crops, in which case, please read this: http://cyberphotographer.com/megapixelmyth
Edited on Oct 10, 2009 at 04:34 AM · View previous versions
Oct 10, 2009 at 04:15 AM
p.6 #207D noise vs 50D numbers
Brainiac, can you stop referring to your article time and time again... we will read it OK 
Oct 10, 2009 at 04:19 AM