Qusetion about the Nikon D7000 in 12-bit and 14-bit mode

[Abitconfused]

The D7000 manual, on page 320, shows that the size of a 12-bit file will be "15.5 megabytes" while the size of a 14-bit image will be "19.4 megabytes" given lossless compression for both files. Shouldn't a 14-bit file should be much larger than a 12-bit file?

[Graystar]

16,200,000 pixels x 2 bits / 8 bits in a byte = 4,050,000 bytes

4,050,000 bytes / 1,048,576 bytes in a megabyte = 3.86 MB

19.4 MB - 15.5 MB = 3.9 MB

Sounds about right to me.

[Mike Buckley]

Wouldn't it be simple to place the camera on a tripod and capture the same image using the two modes? You could then download the images to your computer and verify the size of each file.

[revi]

Well, you might get a surprise there (take a white sheet as subject)

The difference sounds about right on average, but compression makes the file size also dependent on the subject.

And in the case of a white sheet, compression could reduce the file size to almost nothing: split by colour, then use one integer to specify the horizontal size of the white region, one integer for the vertical size, and one for the intensity information, repeat 4 times and add meta information. The meta information will occupy many more bytes than the image, so no visible difference between 12'bit and 14-bit mode. I'm exaggerating of course (the sheet would have to be horrible over-exposed for starters ) but you get the idea.

[GrahamS]

Never mind the file size. Stick to 14 bit if you want the smoothest tonal transitions.

[dje]

Ed the 14 bit image will give four times as many levels per pixel as the 12 bit image. But the file size is just determined by the number of bits per pixel as Graystar has indicated,assuming we are talking about the raw file rather than the jpeg.

I believe Nikon offer some form of compression on the raw file as an option so the scene content will have some bearing on file size in this case.

Dave

[tclune]

Never mind the file size. Stick to 14 bit if you want the smoothest tonal transitions.

The DXOMark site indictes that the actual data is 14bits at ISO 100, and loses a bit for each doubling of ISO. So it doesn't seem that there is a compelling reason to keep 14 bits if you are shooting at ISO 400 or higher. FWIW.

[Graystar]

The DXOMark site indictes that the actual data is 14bits at ISO 100, and loses a bit for each doubling of ISO. So it doesn't seem that there is a compelling reason to keep 14 bits if you are shooting at ISO 400 or higher. FWIW.

Actually, high ISO is where you want to use 14-bits. See here...

http://www.nx101.com/12vs14.html

With 14-bit cameras the extra bits are in the shadows. When you shoot at high ISO values, you're amplifying the shadow signals. So the extra bit resolution is most valuable at high ISO.

[Manfred M]

The DXOMark site indictes that the actual data is 14bits at ISO 100, and loses a bit for each doubling of ISO. So it doesn't seem that there is a compelling reason to keep 14 bits if you are shooting at ISO 400 or higher. FWIW.

I think you are misreading the DxO data; what is being lost as ISO increases is exposure latitude, not bit depth.

[Abitconfused]

A camera that records data at 14 bits...that is 14 bits total or per color channel?

[dje]

A camera that records data at 14 bits...that is 14 bits total or per color channel?

14 bits per pixel - and each pixel is only one colour.

[tclune]

I think you are misreading the DxO data; what is being lost as ISO increases is exposure latitude, not bit depth.

What is happening is that the noise floor is moving up. I have no idea what "exposure latitude" means, but I do know what the noise floor is. If there's only 12 bits of usable data left, there's no reason to hang onto 14 bits. Throw away the bottom two bits and keep what's usable. If the interface allowed for dynamic adjustment of the data depth, it would make sense to just keep throwing away the bits as the ISO went up. I'm actually kind of surprised that the raw data doesn't do that automatically -- perhaps hanging onto an extra bit or so just to be on the safe side, but it doesn't make much sense to hang onto bits that have no information content.