Incense

by Bart Arondson

submit your photo


Hall of Fame
View past winners from this year

Please participate in Meta
and help us grow.

Take the 2-minute tour ×
Photography Stack Exchange is a question and answer site for professional, enthusiast and amateur photographers. It's 100% free, no registration required.

I've been a happy Canon user for a while. When I was doing my initial research on what brand to buy several years ago, I came to the conclusion that both Nikon and Canon are equally good (which I think is mostly true too).

However, I spent some time on DxOMark and noticed that the overall score (and also the portrait, landscape and low-light ISO scores individually) have been significantly higher for Nikon cameras than Canon ones after 2010. This plot illustrates the difference (I edited the screenshots so that Nikon points are red and Canon ones are blue):

Compare plot

Does it mean that as of now, it's better to buy Nikon SLRs than Canon? Or are these scores not trustworthy enough? Thoughts?

PS: I don't intend to start another Canon-Nikon fanboy war. Objectivity would be appreciated!

share|improve this question
3  
possible duplicate of How relevant are the DxOMark scores and tests? –  Philip Kendall Sep 4 at 9:55
    
That's a good reference, but there is also a technical explanation beyond DxO which I'm putting in an answer. :) –  James Snell Sep 4 at 9:56

4 Answers 4

The DxO Mark scores are misleading, but that doesn't mean the gap in performance isn't real!

Several Nikon bodies (D800, D600 many of the D3xxx and D5xxx series) are using Sony Exmor sensors which feature a cutting edge ADC/read noise reduction system to achieve massive gains in dynamic range compared to Canon sensors, which are designed and fabricated in house.

This is reflected in the DXO "landscape" score, but is better demonstrated by looking at the dynamic range graph:


(c) DxO Labs

As you can see the gap in dynamic range disappears after ISO800, this is where it moves from being read-noise limited (where the D810 is significantly better) to being photon-noise limited, where both cameras perform similarly as photon noise depends mainly on how many photons you capture, which is determined by sensor area.

The DxO low light score is massively skewed towards the Nikon camera as it considers colour accuracy as well as noise. Canon colour accuracy is slightly lower due to a design decision to optimise for sensitivity and performance under flourescent lights. The signal to noise ratio graphs tell a more accurate story:


(c) DxO Labs

The measured results are very close, which is exactly what you'd expect if noise is photon limited since two sensors of equal size will capture the same number of photons. Signal to noise ratio isn't the only concern and people will often qualitatively pick one over the other. But the fact remains low light noise performance is similar.

What is not similar nor open to interpretation is the gulf in shadow noise at low ISO. Canon are invested in sensor production so lack the ability to "shop around" for the best model, so it's unclear when they will be in a position to turn the situation around. Low ISO dynamic range is just one on many factors that make a good sensor (or camera) but if it is important to you then it may be worthwhile switching, as Canon have not moved on this front for quite a few years now, they are either unable to improve (due to manufacturing limitations or patent issues) or don't regard it as a priority (or both).

share|improve this answer

Photography is about the whole package, not just the imager performance but also the user interface, mechanical design, lenses and accessories etc. For example STM lenses (Canon) or integrated wireless TTL capabilities (Nikons always had them but they're quite new to Canon.) This answer will only offer some context in the process and production of Canon and Nikon that may be relevant. How relevant are the DxOMark scores and tests? provides a good critique of DxOMark's process and scoring.


This article (caution: technical content) from ChipWorks tells a significant part of the story from around the time when Nikon seem to have made a step change in their sensors.

The main technical difference appears to be that Nikon are using a much smaller fabrication process for their sensors than Canon is. The Sony sensors that Nikon use have come from their 0.25µm and 0.18µm foundries. Nikon also use Renesas/TSMC's 0.25/0.35µm for their own designs. Whereas Canon's in-house facilities are working to a 0.50µm process. Nikon's smaller processes allow for more detailed parts. This is a significant benefit from Nikon's decision to use outside foundries who produce at higher volumes and can afford a much bigger investment in infrastructure and equipment.

The article mentions that Canon have a 0.18µm facility but their other reports on newer Canon sensors aren't public (I'd expect that if they had moved to it that there would be more industry news.)

TL;DR Canon and Nikon (with Sony's help) appear to be playing leapfrog, don't expect the past to be any indicator of the future.

share|improve this answer

DXOMark scores are pretty problematic - they measure several aspects of the sensor, in low light on a heavily down-sampled image, then derive all sort of numbers from those measurements and then calculate a single score weighting all those numbers differently.

The reason Nikon looks better than Canon is that the Sony sensors Nikon are using have less noise in low light (especially when you downsize the image to 5 megapixels) and the DXOMark is heavily biased to low light noise performance.

So, if what you want is the cleanest raw files when you shoot in low light and downsize to 5 megapixels the DXOMark score apply to you - if you on the other hand you want good looking pictures you shouldn't really give the DXOMark score much weight.

share|improve this answer
    
Not quite right, the reason Nikon cameras score higher is much lower read noise at base ISO (and better colour reproduction which skews the sport score). Low light noise is very similar between cameras. –  Matt Grum Sep 4 at 15:31
    
btw. the normalised "print" scores are 8 megapixels, not 5. That is however inconsequential, they could have chosen 12 or 16, the relative scores would be identical. This normalisation is actually less misleading, but if you want you can click the "screen" option to see un-normalised measurements. The key thing with DxO Mark is to ignore the "scores" and look at the actual measurements, there is a lot of data there to help you achieve the cleanest RAW file, which incidentally has nothing to do with shooting in low light or downsizing to 5 megapixels. –  Matt Grum Sep 4 at 15:35

Take a look at DSoMark's comparison of the low light performance of the D800 and then the 5D Mark iii. The D800 is clearly the better camera in low light according to this (by a mile)...

Then go look at this real world test that looks at the actual noise performance in low light for a real world situation. Hmm, that's funny, the 5D mark iii looks clearer than the D800. (Yes, it is an in-camera jpeg test, but it still demonstrates that with the basic processing that both cameras do, there is little to no noticeable difference.)

Lab tests are only good for so much and DXoMark's scores oversimplify and lose the truth. The fact is that cameras are complex systems with many relative strengths and weaknesses. Don't look at the overall scores on DXoMark, but rather use the more specific stats as a guideline and then look at real world tests to see how they handle situations similar to what you will do with it. Then consider the strengths and weaknesses of the lens system, form factor, interfaces and price and decide what is worth it to you.

The two brands are close enough that there isn't a right or wrong answer. There are certainly certain situations where one does clearly have an image quality advantage for a particular model (for example, the D800 is far better than the 5D mark iii for ultra high quality B/W portraits), but for general use, they are both right in line with each other overall. (Personally, I went with the 5D Mark iii over the D800 because of low light performance and familiarity with the interface as well as the ability to use Magic Lantern with it.)

share|improve this answer
2  
The linked website should have compared the raw output instead of the JPEGs produced by the cameras. This way the internal noise reduction (which is somehow independent of sensor quality) is factored in. –  ziggystar Sep 4 at 14:46
    
Then maybe one should try to process both raws with the same external noise reduction. Because maybe the difference we see on the current pictures is caused by a more aggressive reduction inside the Canon camera. If one applies enough external processing, the effect of the on chip reduction should be overplayed (unless they result in a truely better final result). –  ziggystar Sep 4 at 15:21
2  
The low light score is skewed by the minimum colour reproduction requirement, nothing to do with on chip NR. All you have to do with DxO mark is ignore the "scores" and look at the measurements, which tell a pretty accurate story when it comes to noise (though it does ignore PRNU). Comparing camera JPEGs is by far the least accurate way to compare sensors. –  Matt Grum Sep 4 at 15:26

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.