grumpybadger
Alan Rickman
- Messages
- 4,638
- Name
- Paul Beastall
- Edit My Images
- Yes
Thanks to a discussion with Anton Roland, I was alerted to the DxO Mark site that contains measurements of camera performance.
Since I’ve been worried for some time that my clunky old 1Ds Mark II (a model of camera that is now over four years old – a lifetime in digital camera world) was not giving me the performance I could get from a newer model. Specifically, I was interested in the high ISO performance as I shoot many of my pictures in poor light. I have also been concerned though that the new cameras get most of the headlines by offering a slightly better performance but allowing the user to select much higher ISO than they used to be able to, at the expense of image quality.
I’ve always thought that my 1D Mark II offered better noise performance than the 1DsII, as it should due to bigger pixels. However, reading Brutus Ostling’s blog, he preferred to use the 1DsII as his low light camera. Some of the choice comes down to what you want to use the camera for. If you want to use the correct lens to get the correct size image rather than cropping, then the 1DsII will have more pixels available than the 1DII. However, if you want to print that image, or resize it for on-screen display, you will reduce the number of pixels. This reduces noise as the noise can be averaged across a number of pixels as part of the downsizing of the image.
The DxO Labs site allows a comparison of both “screen” resolution (i.e. what you would see looking at a 100% image) and “print” resolution (what you would get if you down sized the image to produce a 12x8” print at 300 dpi – a 6 Mpx image). What this shows is that both my original view of the 1DII being better at Ostling’s view of the 1DsII being better are correct, depending on whether you are pixel peeping or printing. The DxO tests shown here are for signal to noise ratio (SNR) which is a ratio of how much of the data is noise and how much is wanted information or “signal. In this case, a picture is taken of an 18% grey card and SNR is plotted against ISO.
1. Comparison of signal to noise ratio of 1Ds Mark II and 1D Mark II at screen resolution
It can be seen that, at full resolution, the 1DII has a lower noise than the 1DsII for the same ISO. This explains why 100% zoom views in Lightroom look better from the 1DII. However, if we resize the image for print, then the opposite is true.
2. Comparison of signal to noise ratio of 1Ds Mark II and 1D Mark II at print resolution
Since I normally either downsize my images for web presentation, digital projection or prints up to A4, I think this is the metric I should use. It also allows a fair, like for like comparison between cameras of all resolutions, assuming you have the lenses to frame your subjects accordingly.
So, being on a bit of a roll, I then decided to look at the 1D Mark III as well as users of that camera seem very impressed with its high ISO performance. The test shows that, although it is better than the 1DII, it is still outperformed by my 1DsII.
3. Comparison of signal to noise ratio of 1Ds Mark II, 1D Mark II and 1D Mark III at print resolution
So, the next thing to look at. Is the 5D Mark II – the new low noise leader from Canon - any improvement? The answer to this one is a clear yes, the 5DII offers better high ISO performance than the 1DsII
4. Comparison of signal to noise ratio of 1Ds Mark II, 1D Mark II and 5D Mark II at print resolution
At this point, I feel I should apologise to the Nikon fraternity for focusing on Canon, but that is the system I have invested in and the easiest way to consider my options at the moment. I will come to Nikon shortly. But next, the Canon 50D. A consumer camera that has also been promoted as a good high ISO performer with the option to go to very high ISO settings. Will it out perform the 4 year lump I normally use? This time, the answer is no. By quite a long way. Although the 50D is good, packing that many pixels in a small sensor is difficult. It is the best Canon APS-C sensor to-date, but it is still not able to match the performance of the larger sensors.
5. Comparison of signal to noise ratio of 1Ds Mark II, 1D Mark II and 50D at print resolution
And now, on to Nikon. I have no doubt that the D3 is a fantastic camera and is clearly the class leader in the high ISO stakes at the moment. However, I’ve always had a niggling doubt that it isn’t as impressive as some users claim. The key issue for me as that the D2X, the previous generation Nikon pro body, is actually very poor at high ISO. Therefore D2X users who upgraded are bound to be blown away by the D3. This isn’t to say that the D3 is bad. It isn’t, it is a superb camera and takes superb images but more to suggest that it isn’t as far ahead of the competition as some of its fans would claim. The graph confirms that it is significantly better than my 1DsII but that the D2X really is poor in this respect.
6. Comparison of signal to noise ratio of 1Ds Mark II, D2X and D3 at print resolution
So, the final comparison was to plot the 1DsII, 5DII (Canon’s top high ISO performer) and the D3 (Nikon’s top high ISO performer) on the same graph. This is very interesting as the 1DsII is “beaten” by both the newer models, but not as much as you might expect. There is less than a stop of difference for a signal to noise ratio of 30dB (DxO Labs proposed cut-off point) between best and worst. Yes a stop can make or break an image and people spend a lot on lenses to get an extra stop but the lack of difference surprised me. The other thing to note is that the D3 is still king of the hill when it comes to high ISO performance but the 5DII runs it pretty close.
7. Comparison of signal to noise ratio of 1Ds Mark II, D3 and 5D Mark II at print resolution
The other key issue to look at to understand the performance of cameras and image quality is dynamic range, and this is a key area where cameras have improved over the last few years. Higher dynamic range means there is more information stored in the picture. Again, the D3 is top of the charts here but the 5DII runs a fairly close second.
8. Comparison of dynamic range of 1Ds Mark II, D3 and 5D Mark II at print resolution
Summary of Results
DxO Labs have defined their view of a suitable threshold for the upper ISO limit to produce suitable images. Every requirement will have a different definition, but there view is a signal to noise ratio of better than 30dB and a dynamic range of more than 9 stops. For some applications that may be too optimistic and for others too optimistic but it is a set of consistent comparison and I reckon pretty much what I have been following judging by how far I’ve been willing to push my 1D Mark II in the past.
Anyway, I’ve summarised a number of results below:
1. Canon EOS 1D Mark II – ISO 1003
2. Canon EOS 1D Mark III - ISO 1078
3. Canon EOS 1Ds Mark II - ISO 1480
4. Canon EOS 1Ds Mark III - ISO 1663
5. Canon EOS 5D - ISO 1368
6. Canon EOS 5D Mark II - ISO 1815
7. Canon EOS 40D - ISO 703
8. Canon EOS 50D - ISO 834
9. Nikon D2X – ISO 476 (!!!)
10. Nikon D3 – ISO 2290 (!!!)
11. Nikon D200 - ISO 583
12. Nikon D300 – ISO 679
Conclusions
So, where does this leave us?
Firstly, my old 1DsII can hold its head high. There is nothing available that can offer me sufficient advantage to pay to upgrade. The D3 as the leader is still less than 2/3 stop better!!!! Nothing Canon can offer me will give me an improvement of any more than 1/3 stop. So, there’s plenty of life in the old dog yet!
Secondly, the APS-C (1.5-1.6 crop) sensors are all worse than full-frame sensors. The APS-H (1.3 crop) sensor of the 1D series sits in the middle. When push comes to shove, physics is king.
Finally, although the D3 has moved the photographic world forwards and is a fantastic camera, it is only 1/3 stop better than the 5D Mark II. The biggest shock is that it is more than 2 1/2 stops better than the old D2X.
If you’ve stayed to the end, I hope you’ve found this post informative and interesting!
Paul
MODS: I believe this post complies with the “Fair Use” policy as described in section 1.2.2 of the conditions of use for the dxomark.com website. If there is any concern about copyright, I understand.
Since I’ve been worried for some time that my clunky old 1Ds Mark II (a model of camera that is now over four years old – a lifetime in digital camera world) was not giving me the performance I could get from a newer model. Specifically, I was interested in the high ISO performance as I shoot many of my pictures in poor light. I have also been concerned though that the new cameras get most of the headlines by offering a slightly better performance but allowing the user to select much higher ISO than they used to be able to, at the expense of image quality.
I’ve always thought that my 1D Mark II offered better noise performance than the 1DsII, as it should due to bigger pixels. However, reading Brutus Ostling’s blog, he preferred to use the 1DsII as his low light camera. Some of the choice comes down to what you want to use the camera for. If you want to use the correct lens to get the correct size image rather than cropping, then the 1DsII will have more pixels available than the 1DII. However, if you want to print that image, or resize it for on-screen display, you will reduce the number of pixels. This reduces noise as the noise can be averaged across a number of pixels as part of the downsizing of the image.
The DxO Labs site allows a comparison of both “screen” resolution (i.e. what you would see looking at a 100% image) and “print” resolution (what you would get if you down sized the image to produce a 12x8” print at 300 dpi – a 6 Mpx image). What this shows is that both my original view of the 1DII being better at Ostling’s view of the 1DsII being better are correct, depending on whether you are pixel peeping or printing. The DxO tests shown here are for signal to noise ratio (SNR) which is a ratio of how much of the data is noise and how much is wanted information or “signal. In this case, a picture is taken of an 18% grey card and SNR is plotted against ISO.
1. Comparison of signal to noise ratio of 1Ds Mark II and 1D Mark II at screen resolution
It can be seen that, at full resolution, the 1DII has a lower noise than the 1DsII for the same ISO. This explains why 100% zoom views in Lightroom look better from the 1DII. However, if we resize the image for print, then the opposite is true.
2. Comparison of signal to noise ratio of 1Ds Mark II and 1D Mark II at print resolution
Since I normally either downsize my images for web presentation, digital projection or prints up to A4, I think this is the metric I should use. It also allows a fair, like for like comparison between cameras of all resolutions, assuming you have the lenses to frame your subjects accordingly.
So, being on a bit of a roll, I then decided to look at the 1D Mark III as well as users of that camera seem very impressed with its high ISO performance. The test shows that, although it is better than the 1DII, it is still outperformed by my 1DsII.
3. Comparison of signal to noise ratio of 1Ds Mark II, 1D Mark II and 1D Mark III at print resolution
So, the next thing to look at. Is the 5D Mark II – the new low noise leader from Canon - any improvement? The answer to this one is a clear yes, the 5DII offers better high ISO performance than the 1DsII
4. Comparison of signal to noise ratio of 1Ds Mark II, 1D Mark II and 5D Mark II at print resolution
At this point, I feel I should apologise to the Nikon fraternity for focusing on Canon, but that is the system I have invested in and the easiest way to consider my options at the moment. I will come to Nikon shortly. But next, the Canon 50D. A consumer camera that has also been promoted as a good high ISO performer with the option to go to very high ISO settings. Will it out perform the 4 year lump I normally use? This time, the answer is no. By quite a long way. Although the 50D is good, packing that many pixels in a small sensor is difficult. It is the best Canon APS-C sensor to-date, but it is still not able to match the performance of the larger sensors.
5. Comparison of signal to noise ratio of 1Ds Mark II, 1D Mark II and 50D at print resolution
And now, on to Nikon. I have no doubt that the D3 is a fantastic camera and is clearly the class leader in the high ISO stakes at the moment. However, I’ve always had a niggling doubt that it isn’t as impressive as some users claim. The key issue for me as that the D2X, the previous generation Nikon pro body, is actually very poor at high ISO. Therefore D2X users who upgraded are bound to be blown away by the D3. This isn’t to say that the D3 is bad. It isn’t, it is a superb camera and takes superb images but more to suggest that it isn’t as far ahead of the competition as some of its fans would claim. The graph confirms that it is significantly better than my 1DsII but that the D2X really is poor in this respect.
6. Comparison of signal to noise ratio of 1Ds Mark II, D2X and D3 at print resolution
So, the final comparison was to plot the 1DsII, 5DII (Canon’s top high ISO performer) and the D3 (Nikon’s top high ISO performer) on the same graph. This is very interesting as the 1DsII is “beaten” by both the newer models, but not as much as you might expect. There is less than a stop of difference for a signal to noise ratio of 30dB (DxO Labs proposed cut-off point) between best and worst. Yes a stop can make or break an image and people spend a lot on lenses to get an extra stop but the lack of difference surprised me. The other thing to note is that the D3 is still king of the hill when it comes to high ISO performance but the 5DII runs it pretty close.
7. Comparison of signal to noise ratio of 1Ds Mark II, D3 and 5D Mark II at print resolution
The other key issue to look at to understand the performance of cameras and image quality is dynamic range, and this is a key area where cameras have improved over the last few years. Higher dynamic range means there is more information stored in the picture. Again, the D3 is top of the charts here but the 5DII runs a fairly close second.
8. Comparison of dynamic range of 1Ds Mark II, D3 and 5D Mark II at print resolution
Summary of Results
DxO Labs have defined their view of a suitable threshold for the upper ISO limit to produce suitable images. Every requirement will have a different definition, but there view is a signal to noise ratio of better than 30dB and a dynamic range of more than 9 stops. For some applications that may be too optimistic and for others too optimistic but it is a set of consistent comparison and I reckon pretty much what I have been following judging by how far I’ve been willing to push my 1D Mark II in the past.
Anyway, I’ve summarised a number of results below:
1. Canon EOS 1D Mark II – ISO 1003
2. Canon EOS 1D Mark III - ISO 1078
3. Canon EOS 1Ds Mark II - ISO 1480
4. Canon EOS 1Ds Mark III - ISO 1663
5. Canon EOS 5D - ISO 1368
6. Canon EOS 5D Mark II - ISO 1815
7. Canon EOS 40D - ISO 703
8. Canon EOS 50D - ISO 834
9. Nikon D2X – ISO 476 (!!!)
10. Nikon D3 – ISO 2290 (!!!)
11. Nikon D200 - ISO 583
12. Nikon D300 – ISO 679
Conclusions
So, where does this leave us?
Firstly, my old 1DsII can hold its head high. There is nothing available that can offer me sufficient advantage to pay to upgrade. The D3 as the leader is still less than 2/3 stop better!!!! Nothing Canon can offer me will give me an improvement of any more than 1/3 stop. So, there’s plenty of life in the old dog yet!
Secondly, the APS-C (1.5-1.6 crop) sensors are all worse than full-frame sensors. The APS-H (1.3 crop) sensor of the 1D series sits in the middle. When push comes to shove, physics is king.
Finally, although the D3 has moved the photographic world forwards and is a fantastic camera, it is only 1/3 stop better than the 5D Mark II. The biggest shock is that it is more than 2 1/2 stops better than the old D2X.
If you’ve stayed to the end, I hope you’ve found this post informative and interesting!
Paul
MODS: I believe this post complies with the “Fair Use” policy as described in section 1.2.2 of the conditions of use for the dxomark.com website. If there is any concern about copyright, I understand.
