Bit depth

Snowleopard

Most Nikon, Canon or otherwise cropped or full frame DSLR's are capable of 12 bit raw images. Higher end Nikon's, Canon's, etc will let you shoot 14 bit raw. Mamiya, and Hasselblad allow 16 bit raw file recording. (I am not sure that the Pentax 645z will do 16 bit).

I shoot a D700 in 14bit raw, I was looking at a D810, or a D4s, but I am tired of 14bit raw. I am looking for 16, 24, or 32 bit raw files. We have memory cards on the market that will store 512gb of data.

To put this in perspective, I can capture a raw image from a flat bed scanner in a "perceived" (I don't know if it really is) 48bit color Tiff file. Infact, scanners have been able to do 48bit color for years.

What this means to me as a photographer, is that there are that many more shades of grey or color available to more closely match analog or some of that fidelity that film has over digital.

Ironheart

Apples and oranges bro. A scanner with 48 bit color is 16 bits per each color channel. TIFF files are 16bit format, meaning 16bits per color. Your scanner outputs a TIFF file no? A 14bit RAW file is 14 bits per channel or 42 bits in scanner terminology. A 16bit camera RAW file has the same number of colors as a 48bit scan file.

haroldp

Apples and oranges bro. A scanner with 48 bit color is 16 bits per each color channel. TIFF files are 16bit format, meaning 16bits per color. Your scanner outputs a TIFF file no? A 14bit RAW file is 14 bits per channel or 42 bits in scanner terminology. A 16bit camera RAW file has the same number of colors as a 48bit scan file.

yes.

... H

Snowleopard

Thanks for the clarification on what the similarities and differences are. As for the music example, I have taken a an MP3 ripped from a CD at 96 kbit/s, re-ripped the track at 320 kbit/s and then did a third rip of the track using 24bit flac at 2000 kbit/s and put them on an ipod with the cheap Apple headphones and people asked if it was a different song because they could tell the difference.

This is with an un-trained ear on non-audiophile equipment (I-pod) because there is that much difference.

I would expect the difference to be similar with a 12bit, 14bit, 16bit or say a 24bit raw file for a photo.

kenadams

uhm... CDs sample at 128 kbit... resampling at 320 would mean to add information that wasn't there in the first place. MP3s rely on 2 things for compression: cutting off frequencies most people can't hear (anymore) and huffman codes I think. So even at 128 / CD sampling rate, you'd be missing information.

What's the bit depth of the human eye? Cause I sure know my girlfriend sees colors I don't, like peach, lilac...

paulr

Kenadams, What's the bit depth of the human eye? Cause I sure know my girlfriend sees colors I don't, like peach, lilac...

My wife has the same problem, she sees shoes ,hats , Coats, and jewellery which I cannot see!

haroldp

uhm... CDs sample at 128 kbit... resampling at 320 would mean to add information that wasn't there in the first place. MP3s rely on 2 things for compression: cutting off frequencies most people can't hear (anymore) and huffman codes I think. So even at 128 / CD sampling rate, you'd be missing information.

What's the bit depth of the human eye? Cause I sure know my girlfriend sees colors I don't, like peach, lilac...

CD's sample at 16 bit depth at a rate of 44.1 Khz.
This is the 'red book' specification and must be followed to be compliant with the Sony/Phillips patent license.

128 kbit is one of many MP3 formats.

.... H

Sports

CD's sample at 16 bit depth at a rate of 44.1 Khz.

Times 2 (left and right channel).
That's 1378 kbit/s.

Sports

I would expect the difference to be similar with a 12bit, 14bit, 16bit or say a 24bit raw file for a photo.

At some point, you'd just be storing noise. If it's possible to get, say, 13 bits of actual signal, and the next bits are random because they are burried under noise, then it's waste of storage space and processing power to handle those additional bits.
I don't know if the camera internallyi does 14 bit or 16 bit A/D conversion, but I read somewhere that 14 bits are enough for current Nikon sensors (which is why the 14 bit format was chosen, I'm sure).

ggbutcher

From bear-of-little-brain: color depth seems to run into the limits of vision rather earlier than dynamic range (can't worry color if you can't discern the relative EVs in the highlights and shadows of interest), and resolution (looks fine in my laptop's web browser, why is it so grainy when I look at it on the highway billboard?). Perception of colors is bound by our eyes to that band between IR and UV, no matter how big we enlarge or how contrasty the light in the scene we recorded.

Me, I want dynamic range. That's what I seem to wrestle with in post-processing the most. Then, resolution, but even that is secondary if I don't garner ambitions beyond web galleries.

Ironheart

The issue isn't how many colors we can see, about 10 million on average, but how many colors we need in order to do reasonable post-processing and colorspace manipulations. You can quickly show that with 8-bit color per channel (24 bit color, or 16 million colors) you start to posterize as soon as you start any colorspace manipulations. Laura does a pretty good job here:
http://laurashoe.com/2011/08/09/8-versus-16-bit-what-does-it-really-mean/

ggbutcher

Laura is a really good explainer...

Really, however, don't you posterize at some level when you do any manipulation of any recorded set of exposure measurements? That's the consequence of digital recording in any medium: the quantization (or bucketization) of analog measurements into digital approximations. So, when you goof around with the exposure, you're just taking the existing buckets of measurements and moving some of the measurements to other buckets. The larger the color space, the smaller the buckets, so the posterization is not discernable at 12, 14 or 16-bit color (to most of us, I'd guess). I just looked up 'posterization' in Wikipedia; that definition would direct us to 'noticeable' bucketing. Part of my day job is communications engineering, and the whole quantization/bit error thing is in my face almost daily, thank you Claude Shannon. And to think I took up a hobby that let me take that part of work home...

So, just arbitrarily converting your RAWs to JPEG loses reams of color information that you can easily see the absence of in post manipulations. But what's the benefit of 14 vs 12-bit recording with respect to post? I mean, you can easily express the difference in powers of 2, but can you see it, and in what finished products does it make a discernible difference?

Ironheart

The short answer is yes, you are always throwing away information every time you touch a photo.

The real answer lies in the final presentation, and what it is capable of. The whole reason video folks are excited about 4K is so that they can edit in that space (including color, contrast, resolution, etc...) and then re-sample down to 1080i for broadcast and have it look great. In photo editing we want to work in 16 bit/color (48 bit) space so that when we dumb it down for 8-bit/color (24 bit) space it still looks great.

Remember those 12/14/16 bits are representing saturation, brightness, and hue so the larger that space is to start with, the better, especially if you are going to heavily edit it. Also remember that every additional power of two doubles the number of buckets for quantization. Look at the output from a 16-bit capable sensor and you will see exactly what the benefit is.

Or you wont, if you are only going to look at an 8-bit representation on a 2MP screen.

ggbutcher

@Ironheart, you make a good point about sample-down. I first noticed it in video; the old family tapes I massaged to digital in VirtualDub with selected filters looked a lot better than what we were shooting with the newer consumer MPEG camera. In fact, I've not been very impressed with almost any of the digital video cameras for home use. But I don't shoot much of that anymore...

To retain as much color information as long as possible, I'm opening my RAW files in GIMP with the ufraw plugin, 16-bit, Adobe 1997 color space, and do all my adjustment and save the work in the GIMP native format. I just switched from shooting JPEG to RAW, so I don't have many images yet to compare, just some test shots in the front yard. In my recent trip to shoot train pictures, I struggled the most with dynamic range, so I'm evaluating whether pulling RAW from the D50 will give me any significant improvement in that.

There's a part of me that wants a newer camera to get dynamic range; I'd probably try to score a D7100 after Nikon announces whatever follow-on camera they have in the wings. But I've got to weigh that expenditure against being able to afford more train field trips next year. I'd rather be shooting them with the old D50 than taking pictures of my front yard with a shiny new D7100...

Ironheart

So, we got off topic on that other thread...
The question is how/why use 12/14/16 bit color when monitors are only 8 bit?
I think my description of posterization above gives one example of why.

Golf007sd

I have a feeling my screen is no more than 8 bits And I believe prints from JPEGS are somewhat like that . In short we are worried about resolution we can't ( normally ) see ( without pixel peeping ), color variations which we did not know existed

Invent first and then create a need for it ......

That is not how it works.

Most consumer screens (and jpg) are 8 bit, that means

256 values (including 0) for each of R G and B color channels, which means 16,777,216 possible color values (256 X 256 X 256) for each nominal 'pixel'.

each 'bit' doubles the number of possibilities for each color channel, so going to 10 bit yields 1,024 per channel or 1,073,741,824 per pixel

12 bit yields 4096 per color channel or 68,719,476,736 values per 'pixel'.

In final display or printing, '8' bit may be enough, but having many more step values in processing , which is complex math, reduces cumulative rounding errors.

Rounding at the end of the process is usually OK, but recursive rounding thousands of times in processing is how we get flattened and artificial looking images. That is why jpg's are not good starting material for editing, and going from 12 to 14 to 16 bit in camera raw matters.

.. H

Golf007sd

At the risk of being warned by moderators for veering off the topic, I will continue with the "bit" .....

I have read in the past all those discussions about how it matters during processing between 12/14/16 bit and how much more playroom it gives to get the color depth/richness before converting it into the 8 bit JPEG for viewing/printing at the exact shade of color one desires... I get the reasoning behind it.

In short, it is said that the 20 billion possible colors give a vast advantage over ( the only possible ) few hundred million of the lesser bit one when viewing / printing the 10 miilion or so possible colors of the JPEG....

I just felt like saying Bullocks but I will be clever and not say that ! :-* . Instead I will just say that every human eye is different , sees a different tone than the other. So my friend who just spent half a day trying to get that exact tone / shade of turquoise in that Caribbean beach. I'm sorry but I will never be able to see what you intended for me to see. My eye has a different calibration.

And you my friend the professional photographer who worked 2 days perfecting that shot in the magazine, sorry but I looked at your image for 25 seconds before flipping over so I couldn't detect those few hundred different shades of green you put in there ! !What you did with the 16bit image was probably great but here is me with a different set of eyes and believe me I wouldn't have noticed if you had just did it on 12 bits

Just kidding of course .... What would you expect from someone dealing mostly JPEGs and spending between 45 seconds and 2 minutes editing an image :-*

Edit : not in response to you Harold ...Saw yours later .

Golf007sd

I don't understand. If our monitors are only 8 bit (16 million possible colors for each pixel) and we have a 12 bit RAW file (68 billion possible colors for each pixel) how can we edit our 12 bit photo in photoshop or lightroom when our monitor can only display 8 bits and not 12 bits? How can we make color selections among color values our monitors cannot even display? Wouldn't we need a 12 bit monitor to accurately edit the colors in a 12 bit RAW image?

Golf007sd

@donaldejose Exactly, unless you have a high bitrate monitor you cannot see all the colours that a 12 or 14bit RAW file can contain. Most of the colours we are talking about are minor shifts in colour, like different shades of red, green and blue. It's not like the colour is missing, just not appearing as it could. When you view a 12 or 14bit image on a 8bit monitor the colours are compressed, just like if you viewed a 8bit jpeg on a 12 or 14bit monitor. If you think that's bad, a printer has even fewer colours, so don't get too bent out of shape over it!

The bit depth you can see on your monitor depends on two things, the monitors output and your computers GPU. There are higher bitrate monitors, up to 16bit IIRC, but they are very expensive (over $1000 USD). Some, mostly older GPU's, max out at 8, or 12 bit output, although it's not common with modern cards, which would mean that even if you had a high bitrate monitor the GPU might not be able to output that many colours.

Golf007sd

I think the next time I buy a computer I need to find out if the monitor can display more than 8 bits. How do I do that? How do I know if the computers graphics card can display more than 8 bits? It seems this information would be helpful to all photographers who plan on editing RAW files.

Golf007sd

If you look at a monitors, and GPU's, specifications they should tell you the bitrate. Any basic TN panel will be 8bit. Generally speaking any monitor under $300 is very likely to be 8bit. Once you start getting into higher end IPS monitors, $500+ you'll start to see 12bit or even 16bit output on the highest end. All that depends on the size and quality of the monitor.

Golf007sd

Just doing some googling, I see 10 bit monitors being introduced late 2013 and 2014 which makes me wonder how many of us here working 12/14 bit files ( and asking for more ) actually had the 12/14 bit monitors to see those colors ? ( Come on guys, be honest ) I see Apple guys complaining in Aug 2013 forums saying " no sight of 10 bit monitor yet "

Apple 27" monitor specs says it is 24 bit and 16.7 million colors.... Wasn't 16.7 million for 8 bit ?

Golf007sd

@ Paperman you make some very good points and I agree with you a lot of what you have said
But if you working professionally, it is not you, who is the judge , it is the picture editor or webmaster and these guys can tell the difference

Man, I never worked or worried about 12/14 bit files but if I had, I would surely feel like a fool having learned ( like I did now ) that my 8 bit monitor lacked the capacity to show the color depths I thought I was editing ...

@sevencrossing

If what we talked about is correct, how can any editor can see the difference ?

Golf007sd

And let me add that I am still waiting for someone to come up and say we got it all wrong in our discussion about 12/14 bits & our monitors not showing those depths ... Can't be this simple, really ...It just seems impossible to me that such a large group of professionals could be led into working on something which is actually beyond their reach. Nor does it make any sense that the DSLRs can take 14-16 bit images when the capability to work on them does not exist .

Golf007sd

You can work on them, you just can't see them. The computer needs the extra precision to function internally and not suffer image degradation when doing math on the pixels. Photoshop, and Lightroom work internally in 16 bit (or 48 if you prefer to call out the three color channels). Actually the expeed processor does all the math in 16 bit, and then throws the least significant bits away to get back to 12 or 14 bits.
I learned the rules for "sig figs" in 8th grade science class:
http://en.m.wikipedia.org/wiki/Significant_figures
"Spurious digits introduced, for example, by calculations carried out to greater precision than that of the original data, or measurements reported to a greater precision than the equipment supports"

All modern monitors are 8bit (24 bits), however the gamut is much smaller than the 16million colors that the simple math would indicate. So again, read my links above, and ask the right question: How wide is the color gamut on my monitor vs my printer?
http://en.m.wikipedia.org/wiki/Gamut
http://www.cambridgeincolour.com/tutorials/soft-proofing.htm
The answer is.... It depends. On which monitor, which printer, and which image you want to compare.
The global answer is generally monitors and printers have the ability to reproduce a significantly smaller subset of the 16 million possible (theoretical) 8 bit (24 bit) color.

@ironheart
You can work on them, you just can't see them.
I guess we all know what that means ....

We do? I guess it isn't at all clear to me.
We work on things we can't see all the time, like electricity. And I can measure volts and amps, just like I can look at the numerical RGB values in a RAW file and know that the extra precision is working for me, even though I can't see more than 10million colors.
I'm trying to help you understand (I already get it), so snide comments aren't necessary.

The question is does it really matter? Why? If all you do is view images on a computer screen you'll never know the difference. If you print your images you won't notice either. Why? Printers cannot come close to 8bit files in terms of colour reproduction. Even high end printers use less than 100 colours, let alone millions.

Lets keep in mind that there is more to a 12 or 14bit RAW file than just colour.

Don't really get what was "snide" about my comment Ironheart? I thought we were on the same page ....The only thing I can't agree though is comparing working on/with colors without seeing them to working with electricity . You are not supposed to see electricity to work with it - it is not a visual act. The eye/visuals have nothing to do with electricity.

Golf007sd

I have a feeling my screen is no more than 8 bits And I believe prints from JPEGS are somewhat like that . In short we are worried about resolution we can't ( normally ) see ( without pixel peeping ), color variations which we did not know existed

Invent first and then create a need for it ......

That is not how it works.

Most consumer screens (and jpg) are 8 bit, that means

256 values (including 0) for each of R G and B color channels, which means 16,777,216 possible color values (256 X 256 X 256) for each nominal 'pixel'.

each 'bit' doubles the number of possibilities for each color channel, so going to 10 bit yields 1,024 per channel or 1,073,741,824 per pixel

12 bit yields 4096 per color channel or 68,719,476,736 values per 'pixel'.

In final display or printing, '8' bit may be enough, but having many more step values in processing , which is complex math, reduces cumulative rounding errors.

Rounding at the end of the process is usually OK, but recursive rounding thousands of times in processing is how we get flattened and artificial looking images. That is why jpg's are not good starting material for editing, and going from 12 to 14 to 16 bit in camera raw matters.

.. H