Among several photographer friends of mine who shoot Canon 6D, which is marked as excellent at handling high ISO in terms of digital noise, there are frequent discussions on whether the degree of noise we see in our images is greater than expected. Everyone expects digital noise in the shadows. However, in our discussion someone brought up the fact that there are photographers out there who believe that every camera and/or brand have their “native ISOs” that presumably give you lower degree of digital noise then the numbers in-between (for example, if iso of 640 is “native” to your camera, and iso of 500 is not, you get LESS noise at 640 than you would at 500). The argument is that any ISO that is not a native iso is a digital pull-up or pull-down to a native ISO, thus generating more digital noise. This sounded quite interesting to me, so i decided to investigate further.
If you google “native ISO,” you will come upon hundreds of threads discussing this. there are debates on whether canons and nikons have same “native ISOs” or is it different, and what those numbers actually are. Some believe that native ISOs are in steps of 160 (i.e. 160, 320, 640), while others believe that they are in steps of 100s (i.e. 100, 200, 400, 500). There are many tests out there, that display magnified pixels of blackness, counting noise. There are tests that utilize computers at counting noise. There are even firmware hacks out there that would force your camera to only shoot at native ISO (even claiming that ISO of 75 would be native to canon, and starting ISO at 75, and going up from there).
I personally don’t believe that counting pixels translates into a better image, so i decided to shoot a real-life object, and look at the shadows with my eye rather than have a computer that counts pixels (of course, this introduces subjectivity into the process – but i hope you will too look at these images and evaluate whether there is an unexpected decrease, rather than increase, in the digital noise, at certain ISOs.
All images were shot in RAW and are entirely unprocessed – no exposure adjustments, no contrast, no color noise reduction, no sharpening, no luminance added (i actually un-did the usual LR import preset for this). I cropped in on the shadows, because that’s where the noise lives. I chose 1:1 crops, rather than anything more, because in real life that would be the closest crop actually useful. For all of these, i kept the exposure the same: at first, i kept the aperture the same, as i increased the iso as i increased the shutter speed by one stop each. When I maxed out my shutter speed (4000 on 6D), I increased aperture as I increased the ISO.
iso 12800iso 16000
To me, the noise progresses gradually, without any unexpected breaks in the trend. My conclusion is that, for at least my camera and maybe for all canon 6Ds, there does not appear to be any relationship between noise level and ISO levels, outside the usual predictable increase in the noise at higher ISOs. My test results are different than many others out there. However, in my opinion, just because you can prove a difference at 3:1 or 5:1 magnification, it does not mean that this will ever have any impact in the way the images actually appear in their usable form. In addition, we all process our images, and apply noise reduction to those images with noise. Lightroom is incredibly efficient at reducing color noise while preserving the image quality (i posted some examples below).
Can these results be applied to other canon models, or other camera brands? i am not sure. I suspect the answer is “yes” – but someone would need to do this with their camera and let us all know.
Thanks for reading my lengthy tutorial (better qualified as a test maybe).
Bonus: examples of how efficient LR5 really is at noise reduction.
The only processing in these images are: Sharpening +20, Luminance +20, and Color (noise reduction) +50
ISO 5000, unprocessed
ISO 5000, processed
ISO 12800, unprocessed
ISO 12800, processed
ISO 25600, unprocessed
ISO 25600, processed