Pixiq

Bit depth is the number of bits of color data for each pixel in a digital image. To start, I’d like to bury the terms “24-bit” and “48-bit,” since what they refer to are just the expanded 8- and 16-bits from the three RGB channels. (If you take a 16 bit-per-channel file, and multiply it by three for its three channels, then you have 48 bits. I guess that sounds more impressive for marketing purposes.)

I’m going to use this nice little program on the Mac to help show this. It is called Digital Color Meter, and you can find it in Applications>Utilities.

Here’s how it works: Place the cursor on any color on your screen. Here, I’ve placed it on the little red button on the upper left corner (image A). Depending on what “readout” you’ve selected, it gives you the actual RGB color numbers the monitor is using. The dropdown menu supplies quite a few options, but for this example I’m selecting “RGB As Actual Value, 8-bit” (image B).

When I place my cursor on a white patch, I get 255, 255, and 255. Every color in an 8-bit RGB image—every color represented in digital form, for that matter—is made up of three channels in 256 individual steps, from 0 to 255. (Remember: 0, 0, 0 is a color, too!) For example, 255, 255, 255 is a different color than 255, 255, 254. (There is no such thing as 255, 255, 254.5.)

Since every color is a simple combination of these three channels, that multiplies out to millions of colors when you work with 255 values for each channel. To be precise, 16,777,216 colors. Let’s look at this shade of blue (image C). Its values are reading as 110, 170, 231.

Now, pull down “RGB As Actual Value, 16-bit” (image D) and put the cursor on the white again. Look at that number: 65,535; 65,535; 65,535 (image E). We now have 65,535 values between black and white, and 65,535 shades of red, green, and blue that we can combine to create a color. This gives us billions and billions of possible colors. Actually, it gives us 281,474,976,710,656 possible colors. Trillions.

Let’s look back at our blue patch now (image F). We now have 28,270; 43,690; 59,367. That is going to be a different color than 28,270; 43,690; 59,366. 16-bit files give you more color options—more crayons in the box.

The general advice is to use 16-bit files for the best quality, simply because you have so many colors to work with. By the time you edit 8-bit files in a conventional way, you will probably have gaps in your histogram. Here’s an example (image G). In this example I purposely clipped the highlights. Because I am pushing the data to ft the new adjustment, I stretched the file to the point of creating gaps between my steps of color.

Here’s the same adjustment made to a 16-bit file (image H). There are no holes in the histogram because I am working with so many colors.

This is good advice for a standard, destructive editing workflow in Photoshop; essentially you’re giving yourself enough colors to be able to throw some out without seeing the gaps. However, there are a couple of problems. For one, unless you’re using a camera with a large sensor, you’re not really capturing in 16-bit, so you aren’t starting with all that color—you’re just trying to re-map 12- or 14-bit files out to 16-bit. Second, you’re making a huge file, especially if you’re working with multiple layers of Smart Objects. Every Smart Object adds the complete “weight” of the file. Even my little Canon G9 comes out to more than 30MB layers in 16-bit processing.

If you’re working with a constructive workflow by building up layers of RAW files rather than stretching existing files out, my advice is to stay in 8-bit. It’s faster and more accurate, and the 16-bit overhead is just not necessary.