Audio Bit Depth for recording: What is it? How to choose? 16 bits or 24 bits?

So you wrote the lyrics, arranged the song; set up a home studio, downloaded a DAW, but then a new universe of information opened up when it was time to hit record. Bits, sample rates and buffer... What are those? Today we are going to explain audio bit depth. You must set this parameter when opening a new project in your DAW. The "bit" is the unity of measurement that the computer understands. So the depth of bits is the precision that any nuance can be properly represented in the digital signal. In technical terms, it is bit resolution.

I just want to know which audio bit depth to choose!

• Magroove Suggestion: Today, in professional or amateur recordings, go for 24 bits without a doubt. 
• First thing to have in mind is: when working with a larger Bit Depth, your files will get larger too.
• Second: To convert from 24 to 16 bits it's possible. From 16 to 24 you will be injecting noise in your recording to cover up the bit wholes that don't exist in the original. If you want 24, record it in 24 from the beginning!
• Third: The number of bits determines your dynamic range. The bigger your bit depth, more dynamic variation you'll be able to capture.

Let's explain it a little further!

Digital and analog: what is this?

Without wasting time with philosophical explanations, analog is everything that can be compared to nature. Analog audio behaves as it does in the real world: through waves that propagate in the air. In the earlier days of recording, we used to capture analogically in magnetic tapes. Today, it all goes inside the computer. Differently from analog recording, digital recording is all about "0s and 1s". That way, after being captured by microphones, the wave goes through a A/D Converter ("analog to digital") in the audio interface input. There lies the challenge of representing the analog wave with a series of numbers. Because of that, there is a lot of effort to ensure that the digital audio will be as rich as before it was captured. After that, this audio will be limited by the rate you chose to record with. You can't create quality virtually; you can only do that recording it again. The ideas of Bit Depth, Sample Rate and Bit Rate only have meaning here, in the digital world.

Sample rate

Very quickly: Sample Rate is the number of samples you have, in a second, of you analog signal. In an analog camera, it is how many "pictures" or "frames" you have in one second of film. If in a film of thirty seconds you only have three pictures, the quality of your film will be terrible, right? When we talk about Sample Rate, we are precisely considering that: the more pictures I got of a certain event, better represented and more fluent it will be, whatever it is that I am filming. In our case, recording. Usual numbers of Sample Rate are 44.100 Hz, 48.000 Hz, 88.200 Hz e 96.000 Hz. Hertz is the unity that represents cycles per second. In our case, samples collected (pictures taken) per second. But are those the only things that determines recording quality?

Audio Bit Depth: the resolution of each sample

When it comes to coins, the 1 cent coin is the smallest unity there is . We can achieve any amount of money by adding cent after cent. In digital audio, that unity is the Bit. I can get as far as $18 only with 1 cent coins. That's a lot of coins, but I can certainly do it. Now, getting up to $18 only with $10 bills is a little complicated, isn't it? How to do it? Should I round it down? Pretend 18 is 20? Using this analogy, if you only have $10 bills and you need to get to $18, it means your Bit Rate was too low. If we compare again with a video, the Bit Depth is the resolution of  the pictures you took.  The quality of your film. With 44.1 KHz of sample rate, we take 44.100 "photos" each second. But what if they are bad photos, of low quality? How will our video be? Well, bad. As important as to have many captures, to be able to have fidelity to the real thing, it is to have good captures. And Bit Depth is precisely all about the quality of what we capture.   If with cameras we measure resolution in MegaPixels, in audio we do it in Bits. The most common values of audio bit depth we find today are 16 bits and 24 bits.

Bit depths in physical media

Here are some Sample Rates and Bit Depths of known media formats:

• CD: reproduces audio in 44.1kHz (Sample Rate) and 16 bits (Bit Depth);
• DVD: reproduces audio and video in 48 kHz (Sample Rate) and 16 or 24 bits (Bit Depth);
• Bluray: reproduces from 96 to 192kHz (Sample Rate) and 24 bits (Bit Depth).

It does't mean though, that to reproduce a CD, you must capture the audio in 44Khz and 16 bits. When it comes to recording, always choose the best resolution of Bit Depth available.

16 or 24 bits: what to choose?

Conversion

• 24 bits to 16 bits: To convert from 24 to 16 bits it's common practice.
• 16 bits to 24 bits: If we record in 16 and later we want to convert to 24, we won't have the same quality of a recording done in 24 bits from the start. When we convert 16 bits to 24 bits, the DAW (digital audio workstation) will fill in the blanks with a series of artificial information. This processes is what we call upsampling or oversampling. That new information - called "white noise" - will not make your audio better. It will only mask the problems you need to fix and the audio blanks.

Dynamic Range

Dynamic range is the amplitude variation that your recording can capture. In simple words, is the difference in volume from the loudest to the most quiet sound If you have on your instrument (or anything you may be recording) a bigger variation than your audio bit depth is able to capture, those extremes will not be recorded or will be distorted. Primarily, we need to understand that the bit (binary value) is the smallest unity of measurement in digital audio. Each bit corresponds to 6 db (decibels). That way, when we choose to work with samples of 16 bits, our dynamic range is of 96 db (6 x 16 = 96). If 24 bits, then we have 144 db (6x 24 =144) shown in the samples. The truth is that, today, no one records 16 bits. If you do that, you either have a very good reason for it, or you are probably inexperienced! Currently, 24 bits of bit depth and 44.100 of sample rate are pretty standard settings. Later on, if we need to convert to a inferior audio bit depth, there is no problem.

Quantizing - How does audio bit depth works

We already know that audio bit depth determines the depth of samples captured. The same way we know that the bigger the depth, the bigger the fidelity after conversion to the digital pattern. Imagine a ramp. The ramp is the analog wave. Now imagine that we should represent it by a ladder with steps. That is what the computer does when converting analog signal to digital signal. When recording with 16 bits of bit depth, there are more than 65 thousand possibilities to organize those binary digits. With 24 bits, it reaches more than 16 million possibilities. If that ramp was to be represented by a ladder of 16 million (24 bits) microscopic steps, it would still be a ramp to the naked eye, wouldn't it? The steps would be so small that they would be mistaken by a flat surface. However, the same ramp would be a lot less smooth if it was represented by 65 thousand mini steps. yes? This precess is called quantizing. The computer can only read and record "steps", but the human voice or any instrument (or any analog sound) is like a "ramp".

And Audio Bit Rate? Is it the same thing as Audio Bit Depth?

Aren't bit rate and bit depth the same thing? That is an excellent question! And the answer is simple: no! Imagine that I take 10 pictures (samples) per second to make my video. If each one of my pictures has 2 bits, it is 20 bits per second in my video. right? Bit Rare is nothing more than that: the amount of bits reproduced or recorded in one second. To calculate the bit rate, we use the unity of measurement bps (bits per second). First of all, consider how many samples we collected (Sample Rate); after you multiply it by the resolution of each sample (Bit Depth) -  finding the Bit Rate, that is the total of bits that your recording reproduced in one second. If it's a stereo file or recording, we also have to multiply that result by 2, since there are 2 channels.

Exemples of bit rate

It makes no difference if it is a recording or a reproduction. Bit rate is in both of them.

• To identify the bit rate of a pre existing recording, we must multiply the sample rate by the bit depth in which it was recorded.
• In the recording I am doing now, I find the bit rate when multiplying the sample rate by the bit depth I am using.
• To find the bit rate of a music file, you multiply its sample rate by its bit depth. The difference with the music file is that those files are normally stereo, which means, two channels - left and right - as if there were two recordings being played at the same time. Meaning that the result of the multiplication of sample rate by bit depth, still has to be multiplied by two to find the bit rate of each song.

Keep in mind that, in order to shorten the information, 1024 bits is 1 kb and 1024 kb correspond to 1 Mb. So, 1 Mbps means that we recorded 1024 kb in one second to reproduce the recording file.

Mp3 and Dynamic Range: the secret of compacting

When exporting a recording, besides WAV of AIFF, formats in high resolution, you'll probably make a copy in mp3. MP3s, besides having pre defined sample rate, normally don't have constant bit depth. That alone explains why those files are so tiny. The bits unused are eliminated, bringing only the essential. That causes each of your pictures to have different resolutions, because Mp3 files normally are already represented in bit rate, a unit normalized by second.

Conclusion

You will have to set the parameters of Bit Depth and Sample Rate in your recorder. They will determine the whole process. However, when transporting it to the DAW, you should have in mind everything we talked about. To make that possible, remember to start off with the best settings. If you recorded with 48 hHz and 24 bits of Bit Depth, keep it that way in the DAW. And never, never ever, do upscale unless the situation makes you do it so. Audio integrity should be your priority from opening the project to its closing. If you started with a lower Bit Depth than you wanted, keep it that way until you finish this project. Next time, when starting another session, remember to set it correctly.