40D 14bit A/D converter means

Anyone can enlighten me whats does it mean by 40D support 14bit per channel?

I saw in photoshop there's only 8bit or 16 bit channel option. Does this mean it's not a generating true color(16bit)?

Is this also mean that these 14bit is still in development and not yet reach technology maturity stage as it's still a long way to go towards 32bits?

In a typical monitor, a true color is 24bit color display. Now they even have 32bit color!

I'm quite confused.

The 8 and 16 in PS is bit *per channel*.
On the monitor, 24 bit is 8 bit per channel (3*8 = 24. Gives millions of colors (255*255*255))

Canon DSLR cameras used to be 12 bit 'internal'. These 12 bits were converted to 8 bpc jpg files or, in a raw converter, to 8 or 16 bpc files.

Well a normal JPEG can store 8 bits of information per channel which is equal to 256 levels of discreet colour per channel. When you take all 3 channels into account an 8 bit image can store over 16 million different colours. A camera than can record 14 bits can store 16384 levels of colour per channel or (what I think is) just over 4 trillion colours all together. This all means that the 40D can record quite a lot more colour subtlety than a camera than can only record 8 bits.

Now that you have your file, you need to put it somewhere. There are normally three options. You can either put that data into an 8 bit file (but that'll mean losing a lot of it in order to make it fit), you can put it into a 16 bit file in which all the data will fit, and you have 2 bits left over or you can put it in a 32 bit file where all the bits will fit and you'll haev 18 bits left over. Obviously if you want to keep all the file information then the 16 bit file is the best option because it's the least wasteful in terms of file size.

As a (bad) analogy, assume that you have a loaf of bread (that is the raw picture data) and you want to put it in a box (the file format) to give to someone. You have three choices of box. One that is much too small, one that is slightly too large and one that is much too large. Now you can put the bread in the small box but it'll mean losing part of it. However, if you want to keep all the bread then the best option is the second box that is slightly too large.

As far as monitors go, they say that they're 24bit to confuse people. What it actually means is that they are 8 bit/channel with three channels (8 + 8 + 8 = 24). So most monitors can only display the range of colours provided for by an 8 bit image. Why would you want to keep a 14 bit image then? Because some expensive/commercial printers are capable of printing a higher range of colours than 8bit.

Thanks for your explaination. Then i guess even the older model which uses 12bit A/D is more than enough for use to us. BTW does Rebel and 10D use 12bit A/D or lesser?

Cyth0n wrote in post #4224182
Why would you want to keep a 14 bit image then? Because some expensive/commercial printers are capable of printing a higher range of colours than 8bit.

Not really that relevent, the human eye cannot see much more than an 8 bit image so having a device output more detail is irrelevent. To prove it try loading up your favorite photo ed package and creating a 2 pixel image which you zoom in to max. Set one pixel to the value 255,255,255 and set the other one to 254,254,254.

Assuming a decent monitor/colour space you will not be able to see the difference between the 2 pixels, it will just look like a uniform white rectangle.

However do the same with 255,255,255 and 253,253,253 and you may see a very slight difference down to 252 and you should definately see one, so we know that is the difference the eye can pick up.

The reason why you would want more so much more than 8 bit has nothing to do with trying to reproduce it accurately and everything to do with being able to process the data.

Say for instance you wanted to increase the brightness and contrast so you multiple every colour value by 4. If you do this to your original image then you will now have a step of 4 between 2 almost similar values which we already know is visible to the human eye (posterisation).

If however you already have values which only differ by a quarter of what the human eye can differentiate you can multiple by 4 without introducing any visible steps, the more "invisible" detail you have the more processing you can do.

THAT is the reason for wanting more bits, it has nothing to do with being a more accurate photograph of the object.


NOTE: I have left noise out of the situation for the sake of simplicity

And it's likely the reason the 40D has 2/3 stop more DR than the 30D had.

danielyamseng wrote in post #4224205
Thanks for your explaination. Then i guess even the older model which uses 12bit A/D is more than enough for use to us. BTW does Rebel and 10D use 12bit A/D or lesser?

As post 5 says, the extra 2 bits will help you out if you are doing extensive post processing. This has always been a bit of a non-issue for me though. I prefer to find ways of getting the image correct at the time of shooting, so only minimal post processing is required. I don't recall anyone ever complaining that their total bit depth was only 12 bits so you should be ok. As a photographer and not a computer scientist, bit depth is almost one of those things that you shouldn't worry about. Almost.

I'm pretty sure the Rebel is 12 bit, not sure about the 10D.

RedHot wrote in post #4225065
And it's likely the reason the 40D has 2/3 stop more DR than the 30D had.

Uh, no. The 30D has a little over 11 stops of raw DR at its optimal ISO for DR, ISO160:

http://www.pages.drexe​l.edu …rawhistogram/30​DTest.html

(to read, this, mouse over the figure; 12 stops is 72 dB, 11 stops is 66 dB). The 40D is not much different:

http://forums.dpreview​.com …rum=1019&messag​e=25253085

Note that both have raw DR less than 12 stops, which in a linear medium like raw means that 12 bits is sufficient to encode it. Moreover, the DR of the raw files is really no different between the two at the optimal ISO160. So the DR of the 40D has nothing to do with 14-bit encoding.

The "improved" DR of jpegs noted by the DPR review of the 40D is due to a different choice by Canon on where to clip the blacks when making a jpeg in-camera. The ~9 stops DR of the jpeg is smaller than the DR of the raw data because Canon throws away some of the DR when making a jpeg in-camera, and that is why you can recover more DR shooting raw. Typically the choice to clip the blacks is made because the last few stops of shadow DR is quite noisy; the image looks more pleasing (and contrasty) of you clip away those noisy shadows. The extra 2/3 stop of DR in 40D jpegs just means that 30D jpegs throw away more of the available DR when making a jpeg in-camera.

so is the 14bit A/D the reason why RAW files from a 40D are significantly bigger than those from a 400D despite both being 10MP cameras?