The bewildering array of audio output formats on my audio production software and differing opinions left me curious about what real world use these higher sampling rates provided. To settle my curiosity I went on a Googling binge to see what I could glean. For those of a similar nature here are my findings and conclusions.
My curiosity was piqued by the crazy output options I have on my DAW. If I really want to I can export audio at a staggeringly ear-popping (or not actually) 192kHz and 64bit. My question was… do I or will I ever really need to?? Will this make my eyes bleed!? Will bats fall out of the sky!? Well, No, turns out to be the answer.
The fact is we can’t hear much above 16-18kHz (depending on age) and we are not likely to cause any ultrasonic injuries or disturb bats as no amplifier or speaker system can reproduce anything much beyond 22kHz
Furthermore, I am not aware of any widely available equipment that records at these sample rates or any sample libraries available at 96kHz or 192kHz
Having given this a lot of research there are many arguments for high sample rates being irrelevant and no conclusive test to show that anyone can reliably hear a difference.
I carried out several audio exports using Sonar and when I was unable to hear any difference between audio at 48kHz and 96kHz I decided to use my eyes to determine if the additional frequencies existed in the audio. I did so by using a spectrogram. I did try 192kHz file but my DAW would not open it.
You can see the screenshot below, note that the lower dotted line represents the limit of a 48kHz file and the upper 96kHz sample rates. (click for full size image)
I noted that the highest visible frequency was around 24000Hz. Anything above was silent or a minuscule blip that was totally inaudible e.g. 24200Hz. Given that 44.1 cuts frequencies at 22000Hz it seems logical to use 48kHz for masters. When analysing files at 96kHz virtually everything was black (aka silent) above 24000Hz.
In terms of bit rate there is clear evidence that 24bit gives more dynamic range than lower bit rates but going above this seems pointless given that the highest possible bit any speaker system with current A/D converters can reproduce is 24bit. Also, some research showed that going to higher rates such as 96kHz can cause ultrasonic anomalies that cause intermodulation distortion when audio is converted to a lower rate for transmission via speakers.
Conclusion
While this test was not conducted under scientific conditions it was an interesting exercise and for me personally I believe I could hear the difference between 44.1kHz and 48kHz but nothing above that. I also felt I could hear the difference between 16bit and 24bit but 32bit and beyond made no difference to my ears.
As 24bit is the highest bitrate of my soft-synths and the fact that with human ears it seems illogical for me to use anything higher.
Recently I had a client asking for 32 bit 96kHz files. Not only are they huge they clearly contain no more data than 48kz. Maybe he likes the added silence and enjoys buying hard-drives 🙂
I am sure there are many fervently held views and beliefs on this subject but personally I am sticking at 24bit 48khz and going no further until the following two factors are true: –
- a) Sound systems are made that can reproduce these frequencies and
- b) I can grow bat ears! 🙂
While I am no expert in this highly technical area I benefited from some great information on the web, the best of which was this article http://people.xiph.org/~xiphmont/demo/neil-young.html
What’s your opinion? Share your thoughts and experiences in the comments below.
There are also a few other aspects of the ‘hi-res’ audio debate to keep in mind when comparing 64 bit/192KHz analog to digital conversion conversion schemes.
1. The conventionally established human hearing range of 20Hz-20KHz was derived from testing with artificially generated sine waves (without harmonics). There have been recent studies showing that listeners can hear the difference between a live sound and a recorded (digitally converted) sound.
2. The importance of harmonics beyond 20KHz (ultrasonic frequencies) lies more in the impact on the integrated sound; that is, the interrelationships/interactions of harmonics, some of which become prominent in the audible range.
3. Because sound recording formats and recording/reproduction equipment designs are not generally capable of rendering a 64/192 audio conversion resolution, direct comparisons with playback and graphic editors generate erroneous results. Hopefully, these technical bottleneck issues will be resolved in the future and allow us to maintain a ‘hi-res’ audio conversion rate throughout the sound recording, editing and reproduction process.
Thank you for the insightful article!
Why is it that my audio interface produces lower latency at 96khz versus 48khz? For live performance I am seeing a 4 millisecond difference between the 2 and its definitely noticeable.
At 96khz the sample buffer gets filled quicker and can therefore be flushed quicker, meaning less latency. The CPUs on both sides doing A/D conversion are going to be more strained however.