As the first part of our testing we put together an ISO noise test against several other cameras. If you’d like to read our series of posts on the 1D X and the D4 from the beginning, the introductory post was:
So, in addition to putting the cameras through their paces in real-life shooting, we also wanted to do some more controlled testing, so we set up a test scene at the studio. There are more details on the test setup at the end of the post, but for now, let’s take a look at the test scene:
The resolution chart in the background isn’t really relevant to this test and the cameras were focused on the bar code on the DVD box. We then selected crops on mostly the same plane that would let us look at shadow noise (crop 1), detail retention (crops 2 and 4) and finally the appearance of noise on different color patches on the color calibration target (crop 3).
You will find the 100% crops from these locations on our noise comparison page. Just click on the buttons to choose between different cameras, ISOs and crops. Hint: you can also double-click a button to change your selection for both sides of the comparison at the same time.
Please note that this was not a scientific test and we mostly ended up testing our own patience, but let us know in the comments if you have any questions or ideas about our testing or the presentation of the results.
Our take on all the cameras
The poor 7D was at a double disadvantage in this test – it is the oldest camera here and also the only one with a crop sensor – so its light-gathering area is only about a third (38%) of the full frame sensors that the rest of the cameras in this test have. But since we had one available and it is a more affordable alternative to the Canon sports shooter, we wanted to see how its 18MP APS-C sensor (similar to 550D-650D, 60D) performs.
Martin: In a word, not good. Obviously it can’t keep up with newer full-frame cameras at higher ISOs, but it’s the lower ISOs that disappoint – the image does not seem clean even at base ISO – the rest of cameras don’t get this noisy until ISO 400/800. It’s particularly visible in the lightest blue square – no wonder people complain about noisy skies on the 7D. The image also seems blurrier and less detailed compared to 1D X, which has the same number of megapixels.
Tom: Canon 7D is a camera I have been shooting occasionally here and there over two years. And I started slowly like the camera but never liked the sensor inside this well built reliable body. There is noise and no nice grain in those images. I have been just too spoiled by image quality of those 12MP full frame sensros. I dont know why we have 7D in the test here, Martin?
Canon 5D Mark III
Martin: Against the 1D X, the 5D3 starts out with a small resolution advantage (22MP vs 18MP), but it begins to fall behind as ISOs increase, with the 1D X seeming more crisp and contrasty. It’s pretty close until ISO 6400, but after that the 1D X pulls ahead. At 25600, the 1D X looks distinctly cleaner. Comparing 5D3 to the Nikon D600, it’s the Nikon that has the slight detail advantage, which, to my eyes, remains constant throughout the ISO range.
Tom: Agreed on that. The camera that tries to please everyone and generally doing it (not my AF-ON thumb).
Canon 1D X
Martin: The high-ISO noise on the D4 seems a little more subtle and less saturated, but the 1D X puts its megapixels to good use, producing more detailed 100% crops than the 16MP D4. The advantage seems pretty consistent throughout the ISO range, although at ISO 204800 both seem equally unusable :) There is a certain crispness to 1D X images that even gives the higher megapixel 5D3 and D600 a run for their money and provides better image quality from ISO 6400 onwards.
Tom: Mixed feelings about this. Canon holds somehow better contrast and blacks. This test scene didn’t clear out the situation between 1D X and D4 for me. We have a more interesting test about pushing and pulling coming up, so I hope that will clear the picture more. It is not surprise that Canon is back in high ISO game and sure it did very good job rendering those incamera jpgs.
Martin: The entry level full-frame Nikon performs very well here, showing low and medium ISO detail second only to the 36MP D800. It is slightly better in pure image quality than the more expensive 5D Mark III and only starts to trail the much more expensive high-ISO specialists at five-figure ISOs.
Tom: I can add, that ISO 12800 is very competitive (can back up my claim with real world samples – I have also mandatory cat photograph) and after that game pretty much over.
Martin: The 36-megapixel gorilla in the room. Its images have detail that the other cameras are unable to see, such as the vertical lines on the bridge’s stones on the banknote. The D600 can almost resolve them, too, but the 5D3 does not cut it. The D800 may seem a bit noisier at the top of its ISO range than the D600 and 5D3, but not enough to completely wipe out its initial detail advantage. The takeaway is that high pixel density by itself is not bad for high ISO performance, as the 7D might have led us to believe. It’s also interesting to compare the D800 to the 1D X to get an idea of what a 2x difference in megapixels looks like in terms of detail and object size at 100% crop.
Tom: Im sorry to say the gorrilla was wounded (some green noise anomalies in images) and I managed to get its 36MP sensor replaced AFTER the test so I dont take the results here too deep into my heart. Seems gorilla doing much better after the surgery.
Tom: At some ISOs Nikon feels better (especially if I look for skin tones) and somehow smoother. Less detail than 1D X makes difference not only here but also in real-life samples. But difference is not what I see but what I feel because it too small.
Martin: Just looking at the image detail, the D4 consistently seems somewhat worse than the 1D X, but I agree, it also looks cleaner at high ISOs, most visible in areas of solid color and smooth gradients. It’s almost like the typical noise reduction trade-off – less noise, but also less detail. Not sure if this is a choice that Nikon or Adobe have made for us, but the Canon approach seems preferable – you can reduce noise in raw processing, but you can’t really add back detail. In actual low-light shooting, however, the cameras were more similar than they were different in terms of detail and noise.
Some real-life samples
Tom: Oh, its out of focus and a bit grainy. Yea, but those cons are activating my emotions and thoughts.
Tom: Crisp and detailed – the way we want RAW-files to work with high ISO.
The test was performed at identical exposure settings for all cameras and the ISO 100 base exposure was 1/2 seconds at f/8. The lenses used were the Micro Nikkor AF-S 105mm F2.8 G IF ED VR for Nikon bodies and the 100mm F2.8 USM Macro (non-L) for the Canons. White balance was set to 2800K on the Canons and 2780K on the Nikons. Picture Control/Style was set to “Standard” with default settings for all the parameters.
You’ll notice that we have a couple of different image versions for each Camera / ISO combination. There are RAW images with default Lightroom noise reduction settings (Color at 25 and Luminance at 0), RAW images with Lightroom noise reduction off (both Color and Luminance at 0) and also JPEGs straight from the camera. For JPEGs, in camera high-ISO NR was set to “standard” on all cameras; as you’ll see, that’s a pretty strong setting, resulting in a lot of smudginess and loss of detail at higher ISOs.
For Canon cameras, we also have two versions of RAWs processed using Canons own Digital Photo Professional software, one with default settings (sharpening and NR dependent on camera model and ISO) and one without any noise reduction or sharpening.
Tom: Thank you Martin for all the hard work cropping the test scene and putting it on the web (still not sure about tilting my ColorChecker Passport)!