I watched Tony Northrup’s video on how to interpret the test scores on DxOMark (click here to watch this). Tony and Chelsea’s videos are always well researched and I recommend them as a source of objective information. Moreover – Tony takes a scientific approach to the research. What does this mean? Well, science is all about fact and evidence, and if you find evidence that contravenes your hypothesis, you need a new, or least modified hypothesis. Tony does this – if he finds evidence that what he thought before was wrong, he changes what he says. Not everybody can grasp this as can be seen in the comments below his videos 😛
In this video he observed from DxO’s test charts that especially for the sensors in Nikon bodies, there was almost a 1:1 trade off in dynamic range for every stop you gained in sensitivity – so in theory, you were not really gaining anything. E.g., if you shoot at ISO 100, and then same thing again at ISO 200, you gained a stop of exposure, but lost a stop of dynamic range. The extra dynamic range in the ISO 100 shot, allows you to bring it up a stop in post to match the exposure of the ISO 200 shot, with pretty much the same results.
I have a bunch of Nikon bodies so I thought I’d test this. I used my D810 for these test shots. I took 3 shots in manual exposure. One at 1/320th of a second, f/8 and ISO 3200. Then I took the same shot but at ISO 100, and added 5 stop of exposure to it in Lightroom. As a reference I then kept it at ISO 100 and dialled in 5 stops of exposure time to get a clean shot at 1/10th of a second, f/8 and ISO 100. Here they are:-
Now, you can’t tell too much from these small images (even if you click on them to see them at 1024 pixels wide). What you can see though is that there is a small colour difference on the middle shot. The ISO 3200 shot on the left looks pretty much like the reference shot on the right. This was lit by the bicycle light that came in that black box in the shot btw – and these lights are just brilliant (literally). The T6 daylight balanced LED gives a focused, but soft ball of light and lasts for hours connected to it’s lithium battery pack. 7dayshop.com sell these for about £21. Best of all, as they are meant to attach to bicycle frame tubing or handlebars, the little rubber straps also attach very easily to light stands. I have 2 on a boom arm I use to light video.
Anyway – see below for the 100% crops of these images where we can see the noise (ISO 3200, ISO 100 raised in post, ISO100 with 5 stops more of shutter time as clean reference) :-
There is a small difference in the noise on the second shot, that was raised in post, mainly in the darker areas which seem to be brighter. It’s small though, and is probably down to the blunt instrument of adding 5 stops of exposure (which means “mid-tones” in Lightroom). I would say this pretty much bears out the test charts on DxOMark, and confirms that ISO, is, well, 90% BS to be honest. I’ll still use it, if for no other reason than to get an accurate preview of my images on the camera monitor, however, if, like me you sometimes forget to set the ISO back to base, after using a higher value, and you shoot the rest of the day at ISO 3200, then you could adopt the strategy of just shooting everything at ISO 100 and raising up the odd shot where you needed more exposure later. Whatever you decide though, it’s nice to know you don’t need to worry about underexposure where you “shoulda raised the ISO, dammit!” – forget it and raise the exposure in post – it’s 90% the same result.
Edit: The control shot at 1/10th of second is less noisy, because it has 5 stops **more light** it would have 5 stops more light at any iso. You could shoot it at iso 1000 and pull it down by 3 stops – and the noise would be the same.