What's So Good About RAW?

Digital photographers talk about something called "RAW".  Usually accompanying this is some strange talk about a so-called "digital negative".  What's all this about?   What is a RAW file, anyway?  Why would someone use it?  These are just a couple of the questions we'll deal with here. 

If you've been around here for a while, you may wonder "What's with all the digital photography articles, anyway?"  Don't worry, I'm not ditching film.  Actually, these articles about digital are going up here so new readers understand that all this "I love film" talk is not out of some willful ignorance.    In other words, my choice of film is not because I'm some kind of holdover from the Civil War or something;  instead I've chosen it because I know exactly what digital has to offer, yet I prefer film anyway.  I don't reject digital completely;  in fact it's a great tool for many things.  (You'll see in these articles, hopefully, just how much I do appreciate digital photography.)

Anyway, let's get on with the topic at hand here.


What's Wrong With JPG?

In most cases, nothing. 

We're going to explore the special-use case of landscape photography.  Well, it's not strictly landscapes, but any photo that has large areas of tone gradient. 

Just so you know, I greatly prefer film for landscapes.  I even prefer the look of a grainy 110 color negative to an ultra-crisp Canon 5D Mk III landscape shot, most of the time.  That said, my favorite "serious" landscape medium is either 120 or 4x5  Velvia film.  However, I do bring a digital camera with me, just because.  And so, this is where JPG vs. RAW shooting becomes an issue.

A color image file has three color channels:  red, green, and blue (R, G, and B).  In a JPG file, each color channel has 8 bits of data, which means the number of shades available in each channel is equal to 2 raised to the 8th power.  That's 256 shades in each channel.  Three channels times 8 bits per channel yields 24 bits of color.  24 bits is equal to 2 raised to the 24th power, which is about 16.7 million colors.  In practical terms, that's enough color to look realistic to the human eye.  By the way, most digital cameras don't yield images with the full 24-bit color depth.  Just to pick out two common ones:  according to the DxO bencharks, the effective JPG color depth on the Canon 1100D and 650D is only 21.9 bits.  That provides only about 3.9 million distinct colors.  Well, in theory it should provide the full 24 bits, but chroma noise sort of takes away from that.  I still don't see how a little chroma noise could turn 16.7 million colors into only 3.9 million, though, but I'm not going to spend this article criticizing someone's test methods.  Maybe some other time.   The point is that whatever the "effective" color depth of these cameras, the JPG images look just fine if you get the picture right.  I shoot a lot with the 1100D (Rebel T3) and I have no complaints at all about the color depth.

JPG works great for a lot of things, but the one place it's really lacking is when you try to do a lot of adjustment to an image file, especially when there are big areas with color gradients.   Because there are only at most 24 bits of color information to begin with, the computer has a hard time filling in values when you start stretching the color and tone gradients.  Actually, the values aren't there;  the histogram gets a bunch of spaces or gaps in it. 

The real-life outcome of this is that your JPG files get "tone banding" or posterization.  Large, graded areas of color, such as blue skies, look artificial and nasty.  The same thing can happen with any JPG file, even if it's from a scanned slide or negative.  Once a picture gets crammed into the 24-bit color space, it doesn't have much room to spare for image edits. 

Here's a digital photo, shot in JPG, that can illustrate the point.  It was underexposed because, well, I think it looks cool.  The only reason I'd actually want to lighten this image is to compensate for the darkening that happens when you print it on photo paper.  (Photos always look brighter by emitted-light than they do when printed.)



Mouse-over to see the limitations of JPG.

First version:  the JPG as-taken.

Second version:   Extreme tone banding!  This is a little bit exaggerated just to show you the effect, but even fairly tame curve adjustments can start to show color zoning. 


Poorly-written software makes matters worse.  On my old flatbed scanner, the bundled "film scanning" software was so bad that it caused tone-banding on every scan that had big expanses of sky.  Just using different software made a big difference.

Using better software has its limits, though.  Even the best-written software is still going to leave gaps in the histogram when you do curve adjustments, drastic saturation changes, etc.  Those gaps in the histogram indicate the boundaries of the tone bands.  There is no longer a smooth, natural-looking gradient of color and tone.

So, what's the answer to this problem?  In concept it's very simple:  include more colors in the image!


Billions and Trillions of Colors

A TIFF file can have 16 bits per channel, for a total of 48-bit color.  That's not twice the number of colors in a 24-bit JPG image;  it's actually way more.  A 24-bit JPG can render 2^24, or about 16.7 million, distinct colors.  A 48-bit image file can render 2^48, or more than 281 trillion distinct colors!   A trillion is a million times bigger than a million.

So, digital camera manufacturers have designed different "RAW" files to be able to export somewhere between 16 million and 281 trillion colors.  The upper end of the scale gets into some very large file sizes.  If you think your camera's RAW files are big, try dealing with a 48-bit TIFF scan of color film, scanned at the highest dpi setting on the scanner.  If it does't freeze your computer, it'll wear out your hard drive bearings with all the memory swapping.  For a while, I got accustomed to dealing with 300-megabyte TIFF images from my film scans, until (believe it or not) my hard drive bearings actually did wear out.   The reason I scanned files that big was in the hope that I could archive them until image-editing software got better.  Problem was, my scanner was not that good (here's a better way to scan film.)

So anyway, full 16-bits-per-channel files are a little bit much.  They're good for archiving pictures (even though CD's and such are not archival), but for image editing they're actually more than we need.  Cameras seldom even offer the full 16-bits per channel in their RAW files.  The most you'll usually see is 14 bits per channel.

Even though the RAW file is really just the data from each sensor pixel (which is either red, blue, or green), the output of that file has to combine the channels to produce a picture that humans can look at.  That's why we often talk about RAW files as if all channels are combined, as they would be in a JPG or TIFF file (or whichever file type you choose when exporting from RAW).

The point is that when the RAW information is converted into a real image file, that image file can have up to 281 trillion colors.  It probably has less, because few digital cameras are even capable of recording a full 16 bits per channel.  Nikon "NEF" files, for example, can have 12 to 14 bits per channel.  With three channels, 12 bits per channel could export an image having 36 bits of color.  14 bits per channel would yield 42 bits of color.  Remember that the total number of colors rendered equals 2 raised to the number of total bits.  Thus, 12-bit RAW files (being able to output up to 36 bits total) could yield a picture having up to 2^36 colors, or about 68.7 billion colors.  14-bit RAW files (having 42 bits total) would offer images having up to about 4.4 trillion colors.  The output from a 14-bit RAW file has up to 64 times more colors than the output of a 12-bit RAW file.  That's a lot of colors.

Anyway, a RAW file that has 12 bits per channel is still a lot better than a JPG file that has only 8 bits per channel.  An output file having 68.7 billion colors is obviously going to be much better than a JPG having only 16.7 million colors.

Hey, wait a second!  If we know the human eye can differentiate only 4 to 16 million colors, then why would we need RAW files anyway?

The reason, again, is that if there are more colors, you can do more substantial image edits without causing tone bands and other unsightly artifacts. 

By the way, I've said before that megapixels are overrated.  That's generally true.   However, one advantage of having more pixels crammed onto a sensor is that curve adjustments are less likely to yield tone banding.  In the extreme, picture a sensor that had only 1,000 pixels on it.  There aren't even enough pixels to represent the tone gradient of a sky.  If you adjust the curves, adjacent pixels are going to have more drastic differences between them.   Now consider a sensor with ten million pixels.  This will be more able to display a good tone gradient, even when you start adjusting the curves.  I haven't sat down and figured out how much this effect contributes, but once you get past (say) 10 megapixels, the color depth is probably more important than how many pixels.


Digital Negative?

Once you have the ability to make drastic tone & color adjustments on a photo without tone-banding, you can then "develop" underexposed pictures to your liking.   As you might remember from Camera vs. Log, digital does very poorly with highlights, but it happens to do well with shadow detail.  Shooting in RAW and underexposing is the best way to handle this.   Then, your "developing" can turn into a chore that's even more time-consuming than scanning film.  Welcome to the club!

Ken Rockwell calls film "the real RAW", and he's right.  Film has, in theory, practically an infinite range of color and tone values.  In practice the number is limited (temporarily) only by the method we use to digitize it.   281 trillion colors is within reach, if you scan as 48-bit TIFF (i.e., 16 bits per channel).   Film also has that beautiful, deep tonality to begin with.  Speaking of which, the tonality of film is more than simply a result of color depth.  I say that because, even in a scan that has only 16 million colors, the film scan JPG looks richer than a JPG made on a digital camera.  There are probably at least a couple of reasons for this.

Anyway, if you can scan negatives or slides in 48-bit TIFF, there's enough color and tone information that it can sort of act as a "digital negative".   In fact, that's exactly what many people call these files.  (Save the real slides and negatives, though..... "digital negatives" are not truly archival, and they're nowhere near as information-rich as film.)

Ken is right about something else, too.  Most of the time you can just shoot in JPG, if you know what you're doing.   I know a lot of people who do this.  I usually do this.  Actually just the other day (before I wrote this) I was talking to a portrait pro who said she never shoots RAW.  It just so happens that she's also been shooting film for many years.   Hmm, I see a pattern here.  

We've seen that JPG has some limitations, and we've also seen that many pros don't care.  Why choose RAW?  Well, it depends on your choice of subject and shooting conditions.  For pictures with a lot of sky in them, RAW is a much better idea, especially if you have to underexpose to avoid clipping the highlights (such as cloud detail).    Even if you do get the exposure right, I've found that saturation adjustments on a JPG can cause tone-banding anyway. 

Why is JPG sufficient for most everything else?   Experienced photographers work to get the exposure right the first time, and if they can't, the "keeper" is probably in one of their two bracket shots.     JPG is viable there.   Of course, since JPG takes up less space on the memory card and writes a lot faster, it's easier to make those bracket shots.   I like shooting RAW for sunsets and skies, but waiting for these massive files to write to a memory card is not exactly fun when you're trying to shoot action sports.  (On a side note, my old bridge camera at a wedding reception was probably slower than if I'd brought a Crown Graphic 4x5 with flash bulbs.  f/8 and be there!)


Formulating a Workflow

The "workflow" is the series of operations you do on pictures to make them look the way you want.  It could also include getting the pictures ready for a printing process of some kind.   Thus, you could have a RAW workflow, a JPG workflow, or one that incorporates both at different stages.  (Just to nitpick:  There's no actual "RAW" workflow, technically, because as soon as you export that sensor data to a human-viewable image, it's not RAW anymore.  Common file types are TIFF or PNG, if you want the 48-bit color depth). 

The workflow should be designed to preserve the maximum quality, while not being ridiculous.   There are some editing operations that don't actually need to be done on a 300-MB TIFF file.  (Glad of it, too... and so are my hard drive bearings...)

There's one more thing I'd point out here.  Just because you have a program that can edit TIFF or imported RAW files, that doesn't mean that program will do the best job on the image.  In fact, it doesn't even mean that program will do as well as some other program that only works on JPG files!   The reason I say that is because I was using a certain program, and I noticed it had its own noise-reduction algorithm.  So I tried that, and I found that not only was there too much detail loss, but there was also.... tone banding!!  Arghh!    I think I said it before, but:  Poorly-written software or algorithms can introduce digital artifacts into your picture. 

Learn which functions of your favorite programs are the least destructive to your photo quality.  The "tone curves" setting in your RAW editor might be pretty good, but some other function might be very sloppy and ruin the picture.  In the example I mentioned above, I found that I could adjust tone curves, saturation, and levels without too much problem, but most of the other functions are probably better off being done after exporting to JPG and opening the picture in some other program.  That's part of how you develop a workflow that works.

What's the bottom line here?  JPG is good enough for most stuff;  RAW is great for landscapes and such.  (For your main landscape camera, consider film!)

This has been a look at JPG vs. RAW.  I hope you've found it helpful! You can help me out by shopping for your stuff through the links or banners on here.


                 


Thanks for reading!




Support from readers like you helps keep this site on-line. 



Contact me:

3 p o.t o .1 2 0 s t u d i o.. c o m


This won't directly copy and paste.  Please manually type it into your mail program.
No spaces between letters.





Home Page

What's New

Article & Gallery Index





All photos on this site are Copyright 2010-2013.  Copying or distribution for any purpose is prohibited without express written permission
.