I’ve been wondering about noise and image quality and the differences between the various RAW settings available on the D800E (and D800). The most obvious differences are those of file size. With full-size FX images these are roughly as follows:
FX 14-bit lossless compression 40-48Mb
FX 14-bit lossy compression 34-38Mb
FX 12-bit lossless compression 28-34Mb
FX 12-bit lossy compression 24-32Mb
Actual file sizes using either form of compression depend on the subject, and any particular image will give a smaller file as you go down the list, with typically a 12-bit lossy file being a little over half the size of a 14-bit lossless one.
[It is possible to save files uncompressed, but doing so simply makes larger files that are slower to write and usually slower to load into processing software – hard to understand why Nikon provide the option and I assume it is a marketing decision – a useless feature that some reviewers would complain about if it was absent.]
So is there anything to gain from working with the larger file sizes? Opinions on-line vary, though as usual the most dogmatic are the least well-informed. But there are some whose opinion I normally respect who advise on choosing 14 bit rather than 12 bit, and I’ve been doing so for a while. But I’m no longer sure there is any point.
What changed my mine was reading a very detailed and rather mathematical paper, Noise, Dynamic Range and Bit Depth in Digital SLRs by Emil Martinec. It was written in 2008, over 5 years ago, but I think most of his conclusions are likely to still be true.
I can’t pretend to understand all of the maths, but I think I got the gist of it from the non-mathematical summaries. There are some rather surprising points made which perhaps go against the commonly held views on the subject.
At the time he concluded that “Twelve bits are perfectly adequate to record the image data without any loss of image quality” for the currently available cameras with just the Nikon D3 coming close to warranting the use of 13 bits. Any discernible difference would be because the slower reading of the sensor in 14 bit mode increases the accuracy of reading the values.
Things have moved on a little with current cameras, but I suspect that there is still little if any practical advantage in using 14 bit files.
Lower down on the same page, Martinec discusses the Nikon “lossy” NEF compression, coming to the conclusion that because of the noise unavoidably present in light, the compression actually results in no loss of visual information. If I understand what he says correctly, the loss is of the essentially random contribution of noise to the image rather than the true information present.
So I’ve decided there is little or no point in working in anything other than 12-bit “lossy” mode on either the D700 or D800E. Hard disks and large cards may be cheap, but it’s still worth cutting down on the space required, particularly with backups in mind.
There are also some other points of interest in the article, with a discussion about the ‘expose to the right’ rule and also on using high ISO.
Expose to the right is still a good rule, and “exposing to the right at the lowest possible ISO provides the highest image quality” but the reason usually advanced is wrong, and it is simply because this improves the signal to noise ratio.
As for high ISO, above a certain ISO you will get greater dynamic range by underexposing your raw files and correcting this in software.
I’m not sure how you decide at what ISO to start doing this, but on the D800E when working at night I’ve found in practice I get better results by sticking to ISO 3200 and underexposing. Its almost always necessary to underexpose at night by a stop or more in any case to get the effect of night.
But when working at low ISO’s I think he concludes that you should give correct exposure following the expose the the right rule for optimum quality.
There is more to the article than this (including some useful advice for camera makers) with a discussion of pixel size with a conclusion that will surprise some (it did me.)
But it isn’t easy going. A simpler article 12Bit Vs 14Bit Raw And Compressed Vs Uncompressed… Does It Matter? by freelance photographer Francois Malan has some real life examples that I think show fairly clearly there is little if any advantage in 14 bit over 12 bit on the Nikon D7000 (though there is a rather curious colour shift in the heavily underexposed examples.) His article was inspired by the more theoretical work of Martinec, to which he links.
I’m hoping one day it will stop raining here and I will go out and take pictures again. But there are a lot of things to catch up with – and this was something I’d been meaning to find out more about for a while. Now to go and change the settings on my cameras.