the daily dub

Ever wonder what the difference between 8-bit, 16-bit, and 24-bit audio files is? Here’s a quick, english explanation.

Imagine a horizontal line:

That’s silence on an audio file, a straight line of nothing-ness. In the analog world (aka, your voice, sounds you hear from your computer speakers as they travel to your ears), a sound “looks” like a sine wave:

Figure 2. A single analog audio tone

Where the blue line reaches the first dotted horizontal line, call that the “maximum” volume. When converting this analogue sound wave to digital, it makes a transition from a wave to a sample. A sample can only have a certain range of values, where an analogue wave can have any amplitude (or volume) imaginable. This is why many people think analog sounds better - it’s definitely more forgiving, anyway.

Alright, back to the sample. Imagine after some amount of time, say 1 second, we take a measurement of how high the blue line gets. Then, 1 second later, we take another measurement. And so on. That process is what happens in an analog to digital (A/D) converter. How often that happens is determined by the frequency of the A/D converter (or sampling rate). These are usually 44.1 kHz, 48 kHz, 96 kHz.

1 Hz means 1 sample per second. 96 kHz means 96 thousand samples (or measurements of the wave) per second. It should be becoming obvious that more samples per second means a more accurate representation of the sound wave you’ll have after the conversion, right? Good.

Now, audio resolution. Every A/D converter has a resolution (sometimes known as “bit-depth”). Normal values for this resolution are 8-bits, 16-bits, and 24-bits. What do these mean?

Okay - remember that dotted line? That’s the absolute maximum our digital signal can take. This is where analog and digital differ - in the digital world, once you hit the limit, you cannot go any louder. If that happens, the wave “clips”, and your audio sounds like crap. Check out these two images:

Figure 3. Clipping (aka Distortion)

Notice how the clipped wave doesn’t look like the pure wave? That’s distortion. Sometimes it sounds good. Sometimes it makes loud clicks, and pops, and sounds terrible.

Moving on.

Back to your A/D converters. Lets take an A/D converter with an 8-bit resolution as our first example. In the digital audio world, everything is converted to binary. 8-bits means the sampled value from the original audio wave will be converted into a value from 0 to 255.

How do I get this number? Let’s review binary real quick (If you know binary, skip to the next paragraph): each bit is like each “place” in decimal, only instead of being based on 10s, it’s based on 2’s. So, 1001 in decimal as a 1 in the thousand’s place, a 0 in the hundred’s and ten’s place, and a 1 in the 1’s place. That same number in binary, sometimes expressed 0b1001 has a 1 in the 8’s place, a 0 in the 4’s and 2’s place, and a 1 in the 1’s place. Hopefully that makes sense.

As a shortcut, an A/D converter can create samples from an analog wave with values in this range: 0 to 2^(X) where X is the resolution of the A/D converter. So, a 16-bit A/D converter can create samples with values from 0 to 65536. Are you seeing where I’m going with this?

With only 8 bits of depth (or precision), the audio in your chain of processing, recording, and monitoring can only do so good a job at recreating the original audio signal. Without special processing in your A/D, the best you can do to recreate the original analogue signal out the other end of your effects, recording, and monitoring chain is this:

Figure 4. 8-bit digitally sampled audio signal

Is this starting to become clear? The more bits of precision, depth, or resolution (they all mean the same thing in this context), the better job your software and hardware can do recreating the original sound. The same is true of sampling frequency.

So, for recording, I generally stick with 24-bit audio at 96 kHz. For live audio (if I’m not recording), I drop that to 48 kHz. Anytime I plan to do further work on the audio (post-processing, mastering, or re-EQing after a show), I stick to higher sampling-frequencies.

Hope this helps illuminate anyone who is confused by all this non-sense.