Tonal compression

[Summerstar]

Can someone tell me the tonal range of a RAW file?

I'm gradually getting my head round to working in digital having spent most of my life shooting film. The old rule of thumb was that monochrome film, correctly exposed, could record 128 distinct tones between pure black and base white. Colour film had a much more compressed range of about 32 tones.

How does this transfer to a digital scenario? Is it possible to look at RAW files in this way? And if so, is the tonal range the same as for film, or is it greater or lesser?

 

[HoppyUK]

Never heard it put like that, you mean dynamic range, and you certainly can't generalise so broadly.

It varies, according to camera, sensor type and size, processing engine, processing parameters etc. ISO makes a big difference, higher settings cutting shadow depth considerably. Mostly somewhere in the 6-10 stops range.

Have a look here. This test says the Sony A99 has a bit less than 9 stops, presumably at base native ISO, with close rivals very similar http://www.dpreview.com/reviews/sony-alpha-slt-a99/14 Not sure how accurate those figures are, some comparisons don't seem to stack up, but these measures tend to be a bit of a moveable feast. Check out DxO, that's always fun

 

[alexkidd]

and tbh more of an issue is the ability of the display medium to show the tonal range

 

[HoppyUK]

and tbh more of an issue is the ability of the display medium to show the tonal range

True! Not to mention the viewing conditions. It's damn good print that can show more than four or five stops.

 

[Pookeyhead]

Nothing is clear cut with a digital image. If you are talking about distinct tones, then technically a 24bit colour signal can reproduce 256 distinct grey tones... think of the RGB colour sliders in Photoshop... 0-255 = 256 distinct tones in each colour channel. That's where the 16.7 million colours of 24 bit comes from:

8 bits per pixel.. 3 channels per pixel... 8x3 = 24 bits per pixel. As each bit can record 2 colours (binary 0 or 1), then a 8 bit image is 2 to the power of 8.. = 256.

However... most cameras can produce 14 bit RAWs. So 2 to the power of 14 = 16384 colours per pixel, or 16384 distinct levels of grey per pixel... and 4398046511104 distinct colours across all RGB channels.

Great... then we come to the display. Most cheap displays only use 6bit panels and colour dithering to simulate 24 bit colour. Chances are, if you paid less than £400 for your monitor, you're only seeing just over 260 thousand colours, but the dithering gives an apparent 16 million. So your camera is capable of recording 16384 levels of grey.. but your display may only be seeing capable of 64. Even a VERY top end display can only produce 4096 (12 bit panels).

All this is also at the mercy of the video card output. The vast majority of PCs (and I include Macs in this distinction) have a 8 bit RGB video output.... so that's 256 levels of grey, and 16.7 million colours. So viewing your camera's output on your computer... you are limited to 256 levels of grey. Those using Quadra cards will have a 10 bit display (1024 levels of grey) but only when using display port, and ONLY if you have a true 10 bit or higher panel (we're talking Eizo CG range, or NEC Reference here). Usually, even on a quadra driven 10 bit input, cheap monitor will be the weak link.

This is why your monitor is THE most important part of your digital darkroom. Basically, if you paid less than £400 approx for your monitor, you're only seeing a true 64 levels of grey per channel being dithered to 256. This is why I'm always baffled by people who will sink 2 grand into a computer for their image processing, then almost think of the monitor as an afterthought, and spend perhaps £400.. and even then think their pushing the boat out! I'd much rather, if push came to shove, spend £400 on the PC and £2k on the monitor.

Print? Very hard to define... but a high end inkjet like the 4880 will be hundreds of shades of grey, and potentially billions of colours, but the colour gamut can vary wildly with medium, so it's very, very hard to give exact figures.. but there will be a few hundred thousand discreet colours certainly.

Unfortunately, your display is almost certain to be incapable of displaying all the colours captured... and your printer definitely incapable.

 

[HoppyUK]

Bits and tones are not the same as dynamic range.

If you have a speedometer that goes from 0-100mph, if it's marked in increments of 10mph, that's ten bits; if it's at 20mph increments, that's five bits etc. But the car still goes as fast.

Edit: when outputting, it's not a bad idea to assume the lowest common denominator as what you see on a high-end monitor is unlikely to be replicated by another viewer. Lightroom has a 'soft proof' output that mimics what you're likely to get in a printed reproduction.

 

[Pookeyhead]

Bits and tones are not the same as dynamic range.

The OP didn't mention dynamic range... he asked about TONAL range. 8 bits per pixel (24 bit image) gives 256 discreet levels.. that's your actual tonal range. It can be perceptibly higher with dithering and aliasing of course... but you have 256 levels of grey in a greyscale image rendered on a 8 bit per pixel system. Most computer systems are 8 bits per pixel, so that's what you end up with on screen. As I said though, with a cheap monitor, it can be less.

 

[HoppyUK]

The OP didn't mention dynamic range... he asked about TONAL range. 8 bits per pixel (24 bit image) gives 256 discreet levels.. that's your actual tonal range. It can be perceptibly higher with dithering and aliasing of course... but you have 256 levels of grey in a greyscale image rendered on a 8 bit per pixel system. Most computer systems are 8 bits per pixel, so that's what you end up with on screen. As I said though, with a cheap monitor, it can be less.

Yes, I know, which I why I questioned it. If you're talking film and analogue prints, then the range of tones is almost infinite. How many of those tiny discrete increments the human eye can discern is another question.

 

[Pookeyhead]

Yes, I know, which I why I questioned it. If you're talking film and analogue prints, then the range of tones is almost infinite. How many of those tiny discrete increments the human eye can discern is another question.

It's all down to the dithering of tones caused by aliasing etc. Technically, a black and white print contains 2 tones... black and white... just like a half tone newspaper print, but we see an infinite range because the silver halide crystals are so small, and randomly placed. Place 256 tones next to one another on a mathematically precise scale, and yes, you may see banding, but mix it all up in a real photograph, and you will not.

 

[Summerstar]

Thanks for all the answers guys. When I made the original post I thought it was probably a stupid question but in view of the responses, perhaps not.

I said I am 'gradually getting my head round to working in digital having spent most of my life shooting film'. Insofar as the answers have informed me, it just illustrates the huge differences in shooting film and shooting digital. If one thinks about exposure, in terms of the zone system for example, then obviously the techniques and their application are going to be very different.

And I've also got a lot to learn just to understand some of your replies. My knowledge of 'dithering' to date has been limited to what elderly drivers wearing flat caps do at traffic lights...

 

[HoppyUK]

Thanks for all the answers guys. When I made the original post I thought it was probably a stupid question but in view of the responses, perhaps not.

I said I am 'gradually getting my head round to working in digital having spent most of my life shooting film'. Insofar as the answers have informed me, it just illustrates the huge differences in shooting film and shooting digital. If one thinks about exposure, in terms of the zone system for example, then obviously the techniques and their application are going to be very different.

And I've also got a lot to learn just to understand some of your replies. My knowledge of 'dithering' to date has been limited to what elderly drivers wearing flat caps do at traffic lights...

If it helps, in terms of exposure, digital behaves somewhat like slide film in that it is relatively intolerant of over-exposure. Unlike neg film, that can often take large amounts of over-exposure, if you go too far with digital, highlights will blow and then there's no chance of recovery.

On the other hand, what you can do with digtial post processing far exceeds anything possible with film.

The principles of the Zone System still count, but the practical application is very different. If the Zone System is a method for getting the absolute best out of neg film, then the digital equivalent is Expose To The Right (of the histogram) or ETTR.

Some good tutorials on that if you google, like this one http://www.luminous-landscape.com/tutorials/expose-right.shtml It's based on the fact that with most subjects, a 'correct' meter setting leaves a lot of wasted potential to the right of mid-grey. You need a bit of skill and knowledge of where the limits lie with your camera and post processing regime, but the basic trick is to push the histogram as far to the right as you can, making a judgement on areas you're happy to let blow (enable blinkies - highlight alert in the menus).

With this technique, you can often give an extra stop or two of exposure, occasionally more with a benign subject, and that puts a load more detail into shadow areas, and much less noise. Of course the image is then too bright, but in post processing you can pull that back down again and still retain rich, detailed shadows. Shoot Raw to make the most of this method.

 

[Summerstar]

Now that I can understand! Thanks Hoppy!