Just in case you’re disappointed that this is about to get really nerdy and fast, here’s a picture of Weird Al on a Segway at his last concert in Buffalo, NY. This was from his song, “White and Nerdy.”
Sorry it’s so grainy. It was really dark back stage. I was the video switcher for that gig. I’ll post more about that show including more pics later. Right now I have to get nerdy. White and nerdy, I guess.
Sample Rate Great Debate
44.1 – 48 – 60 – 88.2 – 96 – 192 – and up, up, up… This is a topic that’s debated to no end and I’m not trying to start a fight. I simply want to share a mix of my own research with my own experience. Take what you want from it, and like everything else, your mileage may vary.
Google it if you want a long winded explanation with history and bio but what I will simply state about the Nyquist’s Theorem is that it’s a mathematical formula proving that any sound can be precisely recreated as long as it is limited in bandwidth and sampled at at least twice its own frequency. Since humans can only hear 20Hz to 20,000Hz, 44.1kHz is all that’s required.
Greater Fidelity Does Not Equal Better
So, if 44.1 converters can produce greater fidelity than vinyl and tape then why do most people like the sound of tape and vinyl, old analog, more? It’s the imperfections, the distortions, not like a guitar distortion, but harmonic content as a result of the, well, lack of fidelity that we find pleasing to our ears. Transients start to smooth, sound gets rounder and richer due to the inaccuracies.
Today’s Mac laptop has better fidelity than most converters from 10 or more years ago, not to mention tape decks, boom boxes, and turn tables.
There was a time that having an external clock to sync all the sources did make a positive difference in keeping everything working with each other correctly, or with the least amount of jitter. Still a lot of people swear that an external clock makes a difference for the better, but the reality is that most internal clocks of today’s converters are superior to any external clocks. Even Big Ben.
Early on converters lacked sufficient clocking and introduced jitters. Basically, these are time based anomalies and often heard as high frequency distortion. (Not pleasing.) Clocks mattered then.
I’m not going to say that there’s no impact that external clocks have today. I’m simply saying that the addition of a clock can likely create time based anomalies that the listener may decide are pleasing. (Or not.) That’s a completely subjective decision by the listener.
Converters from the previous decade lacked decent anti-aliasing filters which are used to filter out the super-sonic frequencies that we can’t hear.
Why do we need to do that? Nyquist hasn’t been disproved, first of all. There’s no added benefit to increasing the sample rate when recording. In fact, if the anti-aliasing filters are set too high into the super-sonic frequencies, we’ll likely create distortion in the audible frequency range. (I’m trying to keep this article concise and simple. Google the topic if you want to get deeper into anti-aliasing filters and super-sonic frequency intermodulation distortion.)
Oversampling – Why bother then?
First of all, let me say that the recording doesn’t have to be over 44.1 or 48kHz. The processing does. The short answer for this is that a lot of earlier plug-ins actually processed audio better at higher sample rates, 88.1 or 96kHz. (Not higher, though.) So upsampling your session when doing your audio processes would be helpful. Many non-linear EQ, fast Compressors, and other plugins perform better at higher sample rates.
Modern day plugins, however, already have upsampling written into their code and therefore don’t actually need your session to be upsampled. This is a good thing because higher sample rate sessions use a lot more disk space than smaller ones. However, disk storage has gotten really cheap now so do whatever your heart desires. (And ears tell you.)
I still work at 44.1 and 48 for broadcast. Brad Blackwood, a mastering engineer of the greats, and someone I respect quite a bit, doesn’t recommend anything over 44.1. He doesn’t see (or hear) the benefit. (From a Facebook conversation that was had between he and a few of us in an engineering group.)
I never. You can go ahead. If you think it sounds better… then it does. I won’t debate it. What I will say is that working at that sample rate is likely creating anomalies that to some will sound bad and to others will sound good. It’s not improving anything. It’s just different. Whether or not it improves anything is up to the listener. I have a friend that has a Master’s in Electrical Engineering. He builds studio gear. He swears by 192kHz. We obviously don’t agree.
Dan Lavry, a well respected ADDA Converter builder, someone that I respect and trust, is a huge opponent of 192kHz. In fact, he doesn’t see any reason to work with 88.1 or 96kHz. Dan has been quoted as saying that more isn’t better like pixels are to video. In fact, due to Nyquist’s Theorem, as soon as you go to 60kHz, you’ve already gone well over the level of human hearing. He does go on, however, to say that designing a converter to operate better at 88.1kHz is actually easier than 44.1, and you’re not hurting anything except maybe using processing power and disk space by working at that rate. It should be noted that his converters operating at 44.1kHz have continued to outperform other converters at higher sample rates. The fact is that a really well designed converter is the important factor. Not the sample rate.
That means that although a converter can outperform itself at a higher sample rate, it does not mean it will therefore beat another converter at a lower sample rate. A good design is the ticket.
So, if it’s cheaper and easier to make a converter work better at a higher sample rate, that means we should work at a higher rate then?
That’s for you to decide, in your own studio, but it requires that you compare multiple converters at different rates. You can’t get the goods from comparing the same converter to itself.
Furthermore, experiment with this. Anything you hear differently about a higher sample rate, will still be audible after you downsample to 44.1kHz. Try it for yourself and see what happens. What does that mean? It means whatever you’re hearing can still exist at 44.1kHz and has actually nothing to do with the sample rate but everything to do with the converter.
My Goods – My Approach
I use Lynx Aurora 16 converters and no clocks. I record and mix at 44.1kHz. I get broadcast work at 48kHz. I’m going to be working at 88.1 and 96kHz on some future projects and doing some comparisons to see if I can hear anything. I’ll perform some null tests, with plugs, and check results. (Modulation plugs won’t null, I know.)
I’m not going to preach that one sample rate is better than another. Do whatever you think sounds best and I’ll gladly mix it for you. I don’t care.
It doesn’t hurt to work at 88.2 or even 96kHz as far as fidelity is concerned. However, I do believe that when you get into the 192kHz area you will start to likely get super-sonic, intermodulation distortion. (Phantom Fundamentals)
That means 192kHz will sound different. You can call it “better” if you want. I just still hear “different.”
If you made it through this boring post, I commend you. I also appreciate you. If you made it this far and don’t understand most of what was written and/or don’t care but still made it this far, I am outright impressed.
Optimal is a matter of opinion. Sample rates any higher than necessary to capture what we can hear reduces audio accuracy. Period. (Until the theory is disproved, until gear works more accurately in super-sonic frequencies while producing better results in the audible spectrum, or until we can hear above and below 20Hz and 20,000Hz.)
You tell me if you think it sounds better.