44.1kHz vs 192kHz (David and Goliath)
If you are short on time, just watch this. If you are interested in perspective, links to supporting data, and supplemental reference material, skip the video for now (it’s embedded further down on this page anyway).
Alright friends, a little further down the page is a link to some very compelling scientific data concerning audio sample and bit rates. In a nutshell, frequencies above 44.1Khz typically introduce distortion passed down to audible frequencies (intermodulation) from ultrasonic inaudible frequencies (the stuff beyond your hearing range) because audio equipment usually has no way of reproducing that content (and even if it did, you cannot under any circumstances hear it). “Ultrasonic content is never a benefit, and on plenty of systems it will audibly hurt fidelity.” I especially appreciate the distinction the article makes about why 24-bit audio is helpful during recording, mixing, and mastering but not so important in the final two-track master file. There are even some audio tests you can run to determine if intermodulation is coloring your subjective perspective.
I have elected to work at 44.1k because
- It allows me to run much bigger mixing and mastering sessions in my DAW,
- I’m convinced by scientific evidence that there is no benefit to working at higher rates,
- I’m exceedingly happy with the sound already,
and most importantly
- My final results don’t suffer from the artifacts that occur when down-converting from higher sample rates to reach the end target rate.
I have personally conducted A/B tests of my own and can promise that a track fully produced at 44.1KHz sounds noticeably better than one produced at a higher sample rate and down-converted to 44.1KHz for CD or MP3 distribution even using the best possible conversion algorithm.
The exception, of course, is if the end product is not a CD or digital music file. DVDs, for example, use a native bit rate of 48KHz. If that is your primary product (as in a film score that will be more popular in the context of a movie rather than a CD), then it may be prudent to compose, produce, mix and master at 48K. But if you are going to be selling more CDs and MP3s than DVDs, it’s probably better to work at 44.1KHz then up-convert to 48KHz for use on the DVD. It’s a judgment call. I suggest to you that people do more critical listening to music when they hear an album than when they watch a movie. In summation, if your primary goal is CDs, MP3s and other digital music files, 44.1 is your bet bet to start with.
I think CD-quality (16-bit/44.1kHz) digital has gotten a bad wrap because, up until recently, the audio plugins and processors in digital audio workstations (DAWs) haven’t been able to deliver the kind of tone and EQ curves reminiscent of the analog wonder years. That reality has changed dramatically in recent times, thanks in large part to plugins that faithfully emulate the most beloved hardware inventions of that era, and many legendary engineers are now admitting as much. That in turn trains us (engineers working in the digital domain) to be far more aware of what “sounds like analog”. Our mixes, consequently, are getting much better. OK, it’s time to drop some serious science:
I know the jury is still out for a lot of people, and some of you may strongly disagree with the author’s convictions, but I find that mixing and mastering has far more to do with the skill and creativity of the engineer than with sample and bit rates, period. I’ve read many reports in Mix and other leading magazines over the last decade featuring blind listening tests from the industry’s top professionals who, time and time again, can’t tell the difference between anything above 44,100 hertz. Eventually, I stopped buying the articles, because the debate always ended the same way. Basically, it’s the same article every year. Statistics prove opinions are completely random and subjective while passionate feelings and doubts rage on. Ultimately, I committed to 24/44.1 when I went digital for all the reasons stated herein and decided not to join any of the higher-frequency audiophile religions. A couple years ago, I also became aware that the damage caused by converting from a higher sample rate to 44.1k is worse than any perceived benefit gained in the first place. If you’re getting fed up with all this sacrilegious testimony and would like some substance, read this too.
Ultimately, we are splitting hairs here. Like I said, microphone placement, input chains, recording, mixing and mastering all play a substantially more important role than sample rates at or above 44,100 hertz. If it’s working for you, more power to you. In the end, if people are impressed by the quality of what you deliver, that’s all that matters. Me? I’m just glad I can run twice to four times the number of plugins in my DAW and still sleep well at night.
If you are still having a problem after swallowing the red pill, here’s a video that proves beyond any doubt that there is no advantage to higher sample rates when it comes to the resolution of audio information (if the video does not appear, use this link instead):
Here is another facet to this conversation that discusses how the filters on digital-to-analog converters within audio interfaces are not created equal and can skew a listener’s perception of what is happening inside the computer. It too contains extremely impressive links that you should also follow and read.
If using offline bounce, 44.1KHz is superior to higher rates (so long as you are using oversampling in your most critical plugins) because you won’t have down-conversion artifacts (which I can personally hear in A/B testing, even when using the worlds #1 advanced algorithm by iZotope), regardless of what you hear in your speakers (audio monitors) through your converters, since they have no bearing on the internal, non-realtime bounces. This underscores my previous claims that it is best to work at the sample rate of your target format (i.e. CD/MP3/etc., DVD, whatever). What I didn’t realize when pushing this assertion in the past is that the filters in the converters and the overall quality of the converters themselves are what is affecting subjective opinions. Both the converters and opinions on what is coming out of the speakers have no affect on the offline bounce at all. This, my friends, is a very big deal and puts the nail in the coffin of the debate regarding sample rates for 44.1 target song formats in 100% digital DAW setups. The only situation in which I can personally see a benefit to a higher sample rate is if your converter sounds better at a higher rates (an issue of interface build quality and not an inherent superiority of higher rates) and you have to rely on your converter to render your master in realtime (e.g. bouncing through outboard gear or to a real tape machine to generate your master). If your converter sounds equal or better at 44.1, then this advantage is nullified.
44.1 conveys (when properly converted using the best filters) a 100% accurate reproduction of sound with no loss of fidelity and is not inferior to higher sample rates. For those obsessed with the possible rolloff from 17k-20k in less optimal converters, you might want to pull up your favorite masters in a spectrometer. I’m willing to bet that you’ll see an intentional rolloff near 20k in the digital file (prior to D/A conversion) in any case. Trust me when I say that any spikes at 18k of significant amplitude will make you rip off your earphones or reach for the volume knob for your speakers if your equipment can reproduce those frequencies (most cannot). In fact, before I got rid of the mixer between my interface and speakers and replaced my cables with better ones (Mogami, if you must know), I couldn’t hear an 18k spike in my Events either. Now I can. But with my Ultrasone Pro 750s (8 Hz to 35 kHz), they become shrill terror and immediately apparent due to the extremely flat response curve from 20Hz to 20KHz.