The Great Sample Rate Debate
© 2018 All Rights Reserved
It’s been a good week writing Musician’s Survival Guide to a Killer Record. At the moment I’m deep into the Mechanics chapter, but today, I’d like to share with you a section from The Basics chapter.
Sample Rate and Bit Depth
This is the shit that will make your eyes roll into your skull. It’s exceedingly technical in nature, and practically speaking, it’s not something to spend much time thinking about.
We all know that a CD (going 40 years strong!) reproduces at a sample rate of 44.1 kHz and 16 bits. But what does that mean exactly?
The sample rate describes the number of samples in a second. In the case of a CD that’s 44,100 samples every second. And yes, I realize CDs are becoming less common, (I don’t even have a CD player anymore), but they are still the format of choice for many independent Artists and bands in the US, and for a very good reason. They can be sold. Besides, the only delivery format that is of a higher sample rate than CDs is . . . I joke! There isn’t one. But you wouldn’t know that by reading engineering forums. If you read those, you’ll come away believing we all listen to records at 96 kHz or higher.
By definition, the top frequency response of your recordings is equal to half the sample rate. In the case of a CD that would be just above 22 kHz, which is 2 kHz outside the human range of hearing. There are some that feel this is insufficient and suggest that the frequencies that extend all the way up to 48 kHz improve the fidelity of the sound. Which is certainly possible. It’s also irrelevant when it comes to making a Killer Record.
Another argument for employing a higher sample rate like 96 kHz is that the converters introduce less distortion, which is a dubious argument given the levels of distortion we’re talking about here.
Then, of course, there’s the bit depth, which describes the resolution of the capture and which determines the dynamic range available. A CD is 16 bits, but for the past 20 years, we all record at 24 bits, mostly because that’s what the DAW comes set at, but also because it allows us to record at lower levels.
Why would we want to record at lower levels?
Well, for starters, the moment you go above 0dBFS (Decibels Relative to Full Scale) in digital you’re clipping, which, if you don’t know, is rather nasty distortion. Which means we have a ceiling. And I realize there are times that some of you Industrial Artists out there might like to use the sound of clipping as an effect, but there are plugins that will do that for you far more effectively and with considerably more control. Your converters are best used to record and reproduce audio.
While it’s very easy to record something too hot, it’s actually quite difficult to record something so low that it’s unusable. Oh, you can do it if you’re really super stoned, but so long as you keep the lower parts of your audio generally above the -18 mark as you record, you’re not going to have any problems. And if you have a super dynamic part, then record it at a lower level. I just mixed a track in which the breakdown drops in level to just below -24 dBFS. On the whole mix! If the Internet ever found this out it might explode.
Let’s just examine this for a moment. At 16 bits you have a maximum dynamic range from ceiling to noise floor of 96 decibels (dB). At 24 bits you have a maximum dynamic range of 144 dB. That might not mean a whole lot until you consider that every 6 dB of level is equal to about double the volume. So 144 dB is well beyond an enormous dynamic range. It’s certainly far more range than you’ll ever need. I mean, hasn’t loudness been the goal for the past two decades? We have to worry about dynamic range, why? Unless you’re recording classical music, and even if you are, your dynamic range is never, ever going to be an issue. I mean, not even close.
Here’s the thing. The reason everyone freaks out about recording too low is because you start to lose bits below -18. And somehow along the way, losing bits became worse than clipping. It’s not. It’s okay if your lowest parts are below -18 if that’s how you need to record it in order to avoid clipping. And you don’t need to always be up in the yellow of your meters either. The goal is not to get as close to 0 as possible. The goal is to record a magical take that isn’t completely fucked because you kept clipping. As a result, many engineers will put an analog compressor ahead of the input to the DAW to protect against overs. We’ll discuss compressors in depth later.
You’re not going to find many people arguing over optimum bit depth. Unless you’re on some sort of legacy system from the nineties, you’re operating at 24 bit these days. And if you’re operating at 16 bit that most certainly won’t negatively affect your results, even if most engineers will snort at the thought of it. Every hit record going back at least 30+ years has been released as a CD in 44.1/16 format. How many Killer Records does that make? And these days consumers listen to MP3s. You’re going to worry about sample rate and bit depth?
As I say, the bit depth is hardly a point of contention. It’s the sample rates that produce the heated debates online, and I can’t for the life of me figure out why. The consumer still listens to MP3s, which are atrocious when compared to CD. They hear MP3s on Spotify. They hear them on YouTube. They hear them on Pandora. They hear them from their iTunes collection. So, why on earth would anyone be worried about the sample rate? Because you want to be able to deliver a higher fidelity master in the future? Your record is a snapshot in time, and if somehow it’s a hit record, it will be recorded again hundreds if not thousands of times. It’s the song that will stand the test of time not the recording. Stop worrying about your sample rate. Get a hit first. Then you really won’t care about sample rate, because your first big hit was at 48 kHz, so why the fuck should you switch now?
You see what I’m saying?
Now, there are a great many engineers, including many personal friends of mine, who swear that a sample rate of 96 kHz makes their life easier because they feel the plugins are optimized for it, and that’s probably true. And some engineers are convinced that if you record at a higher rate, it will dumb down to MP3 more betterer. But if your computer is long in the tooth, or if you’re light on RAM, or if you have limited storage, or even slow Internet speeds, then I can promise you, working at 96 kHz is not going to make your life easier. You may not even be able to operate efficiently at that sample rate, and if you can, your track counts could be limited.
Yet another argument for higher sample rates is that time stretching and tuning are more forgiving. I’m sorry but if you need to alter your timings so much that you’re getting obvious artifacts at a sample rate of 48 kHz, then you should probably keep recording the part until it’s close enough for stretching tolerances. Some call this practice. This is a Musician’s Survival Guide, which means you need to think like a musician, and as a musician, you should record a part until it’s great because it actually takes less time, it’s more fun, and it’s what makes you a better musician.
Time stretching and tuning are amazing tools. But there is nothing particularly musical about perfect tuning or timing. With MIDI you have some control over feel because you can adjust velocity and note duration along with timing. It’ll take forever, but it can be done. Not so with a recording. And if you’re a musician and you plan to perform your record in front of an audience, then it seems worthwhile to invest the time working out the part until you can play it with some feel.
As far as I’m concerned, you should use time stretching and tuning with impunity if that’s what’s required for the production. But it doesn’t make much sense to learn how to play your record after you’ve already recorded it. And it makes even less sense to record at 96 kHz purely as some kind of permission to play poorly. I would rather record a part for an hour until I can perform it, than to spend an hour moving bits of audio. Wouldn’t you?
Just to be clear, although it may appear otherwise, this is not an argument for working at lower sample rates. There are legitimate reasons to choose to operate at 96 kHz, including the ones I’ve already mentioned along with reduced latency. The latency alone could make it worthwhile to record at a higher rate. My point is if you need to record at a lower sample rate, there’s no reason to feel as though you’ve somehow compromised your recording. You really haven’t.
Here’s the bottom line on sample rates, and this is really the crux of my argument right here: No matter what you read online, I can promise you that your sample rate will have absolutely nothing to do with your record’s success or lack thereof.
No one will ever listen to your record and wonder what sample rate you used. You will never, ever, hear a fan, or even a detractor mention anything about the sound, or the bit rate, or the noise floor, or the sample rate of your record. Those decisions will have no bearing on whether people share your song. They will have no bearing on whether people react to your song, or like it, stream it, or buy it for that matter. The only people who will care about the sample rate and the bit depth are engineers, and they will argue ad nauseamon the Internet as if it matters. It may matter to them. We don’t make records for them.
The funny thing is, audiophiles–who only listen to sound–go gaga over my Ben Harper mixes. If they only knew that I mixed those records exclusively on the bookshelf monitors of the day. With some of the systems that those numbskulls use, they would have to be able to hear all sorts of shit that I never heard while mixing those records. There is such a thing as too much fidelity, and too much clarity, and I never give a fuck about anyone who puts sound above the music.
I produce my records at 48 kHz. Not only is it the standard for video broadcast, it doesn’t overtax my system or storage. Most importantly, it will never, ever, ever prevent me from making a Killer Record.