So, you’ve got a song in your head and you need to record it. You fire up your computer (humming the tune now), click on your DAW icon, wait for it to load up (humming the tune and tapping it out on your legs), try and think of a quick name for the session, almost ready to go before you forget it, and then…bit depth?..Sample…Wha?
What the hell is bit depth and sample rate? Is it important? Do I need to select the right ones? Will it affect my song? I’m confused.
OK, here it goes.
When you record your guitar into Pro Tools, Cubase, etc. this is what happens:
1. The instrument is played.
2. Sound waves hit the microphone.
3. The microphone diaphragm moves.
4. This movement is transmitted as voltage.
5. The voltage travels down the cable into the preamp stage of your interface.
6. The preamp spits the voltage into the A/D converter.
7. The A/D converter changes the voltage into numbers.
8. The A/D converter sends this numeric information to your computer.
9. The computer understands the numbers and this information is recorded by your DAW.
Now, when the converter reads the analogue signal (voltage), it does two things. The first thing it does is sample the signal; it replicates it as best it can. If your interface is running with a sample rate of 44.1kHz, it will take 44,100 samples (snap shots) per second of the sound your guitar is making. If you set the sample rate to 96kHz, it will take 96,000 samples per second. Imagine these snap shots as the dots in a dot-to-dot drawing of, oh I don’t know…a lion. The more dots you have, the more detailed the image. The lines are smoother, and it really looks like a lion. If you have only a few dots, there is no detail and it looks kind of like a lion…or some cat-like thing with a pointy face. Now, 44,100 dots is enough dots to make you think the lion is right in front of you. That’s why the drums, guitar, bass, voice, etc on a CD sound so good – you get 44,100 dots per second – that’s a lifelike image!
The second thing your converter does is quantize the signal. Think of this quantization as rounding up/down. A low bit-depth quantization takes your $24.76 and makes it $20.00. A high bit-depth quantization takes your $24.76 and makes it $24.75 – still with me? Every time you add a bit to the bit depth, it doubles the values available to it. So, it can round up/down to more precise values. This also translates as more dynamic range, as there are millions more values from min to max with higher bit-depths than with lower bit depths.
So, here it is, the million dollar question:
What’s the benefit of recording at 24bit/48kHz and so on?
If your music is going on a CD, which is 16-bit/44.1kHz – there is no point in recording at 48kHz. There is no significant quality gain involved in using a fractionally higher sample rate. The technical losses and time involved in sample-rate conversion (back down to 44.1kHz for CD burning) aren’t exactly constructive. But what about 96kHz, or even 192kHz? Again, if your target format is DVD, or you’re making music for a movie, or you’re a sound designer, fine. If it’s for CD, the files will need to be brought back down to 44.1kHz, and converting from such high sample rates only benefits the listener if they like ultra-high frequency distortion. It’s counterproductive.
There is, however, an advantage in recording your music at 24-bit, as this increases the dynamic range available to you. This translates into greater headroom and a reduced risk of overloads and clipping.
For the beginner to intermediate folks
If you are not familiar with dithering (I haven’t covered this yet), record at 16-bit/44.1kHz and carry on with your life.
If you are, then record your music at 24-bit/44.1kHz, and then apply a 16-bit dither when mixing down to burn to CD.
For the nerds (beginners not to read – it will damage your brain)
The audible human hearing range is 20Hz – 20kHz. The Nyquist frequency at 44.1kHz is 22050Hz – slightly outside our hearing range (this is why the CD sampling rate is 44.1). Some people, however, argue that frequencies above our audible hearing range can still be picked up on. So, recording at 88.2kHz is the best option as it makes the Nyquist frequency 44.1kHz (way beyond the human hearing range), and the down-sampling is easy because it only requires one step of filtering, since it’s exactly half the target sample rate of 44.1kHz. This results in a better quality file than down-sampling from other higher sample rates. A 96kHz file, for example, needs to be up-sampled to 14.112MHz before being down-sampled to 44.1kHz. This involves considerable filtering. It is also said that a file recorded at 88.2kHz sounds higher in quality and fidelity than a file recorded at 44.1kHz.
My experience is contrary to this argument. Modern sample-rate converters seem to handle down-sampling from any rate without noticeable quality loss. As for the 88.2kHz sample rate – stick a track originally recorded at 88.2kHz in the middle of a CD of tracks recorded at 44.1kHz and see if anyone can pick it. I don’t think anyone can really tell the difference, and some blind testing I did years ago (when my ears were young) proved that some people just like to ‘talk a big game’…as the Americans say. It is my opinion that higher sample rates on offer from certain interfaces and recording devices is simply a marketing ploy that suggests that the higher the number, the better the quality – this is certainly not the case.
I mean, recording at 192kHz may be useful…if you’re producing an easy listening album for bats.
If you are recording material that is to be professionally mastered – ask the mastering engineer what bit-depth he would like to receive. It’s usually 24-bit, meaning that you don’t have to truncate or dither at all.
If you are releasing to higher quality formats, then record with the target sample rate in mind.