Sample rate, and the myth of accuracy


Hometracked has an interesting article on optimal sample rates at which to record.

Des McKinney writes:

The optimal sample rate at which to record is a matter of considerable debate. Proponents of recording at sample rates above 44.1 KHz typically claim that the higher frequencies yield greater detail. And while there’s a tradeoff – tracks recorded at 96 KHz need more than twice the storage space of those captured at 44 KHz – we’re assured that the increased detail means listeners hear more accurate recordings.

Don’t believe it. In recorded sound, accuracy is a myth.

Sample rateSamples per second defines the sampling rate. (© Graham Mitchell)

Des argues that mixing engineers strive to achieve transparency in mixes, in place of perfect accuracy. A mix should sound good on every system (studio monitors, stereo systems, iPod ear buds), not perfect on just one.

Read the article to see why Des thinks recording at sample rates above 44.1 KHz is only for few people.

On a similar note, check this article by Tweakheadz discussing 16 vs. 24-bit audio recording.

Cableguys TimeShaper
  • Looza

    I think this is total bull and works only if you do not edit/pitch/modify those recorded tracks at all.

    Two reasons :

    1. Resolution of high frequencies.

    Take white noise (the real white one, not pink or brown noise), use any Filter-VST (Crayon-Filter for example), set your DAW to 44.1 and slowly close the filter on the white noise. Now set your DAW to 96 or whatever is the highest you can do and close the filter again, compare. The result should be very obvious.

    The point of all this : if you are fiddling around with higher frequencies in any of your tracks the higher your frequency-resolution the better.

    2. Pitching

    sample something in 44.1, pitch it around a little bit and if you stay at 44.1 you might run into a problem because your DAW has to interpolate alot. It’s abit like scaling a 100px bitmap to 95px, it will look strange too.

    So I am sticking to my 96/32-processing in renoise for sure.

  • Looza

    Note to 1. : With crayon filter and the white noise, if you set it at 44.1 khz there should be a very visible “break” where your noise (or any other sample) suddenly gets much darker/filtered because with 44.1khz a proper smooth filtering can be achieved only from around 11khz downward.

  • Des

    Re: “I think this is total bull”

    You raise good points, Looza, but in fairness, my article deals more with the myth that higher sample rates “sound” more accurate. To quote myself, “some audiophiles claim that recordings lacking these very high frequencies are less accurate. In this context, accuracy is a myth.”

    Sure, 96KHz gives you more information to work with, which might be important for filtering/pitching/whatever. But that’s a separate issue from the old saw that high sample rates simply sound better because they capture higher frequencies.


  • Looza

    First of all, sorry for the rude tone, I had a hangover including a headache and was in a quite bad mood. Instead of commenting on blogs on the internet I should have watched a DVD to calm down or something. I apologize.

    In term of “sound more accurate” I think this might be true for those classic examples of hihats and cymbals. Those are instruments with a very complex frequency range and I have to say that some of the battery-included cymbals which are sampled at 96/24 sound abit brighter and more sizzling than the other 44/16 ones I have and are definately easier to control in mixing, the 44-ones become quite “muffled” when you start to filter them. So this might make sense for a few selected instruments.
    But again, as soon as you put your samples trough some heavy processing the more resolution you have the better, atleast in my opinion.

    (as a matter of fact, sample rates of 192khz seem abit over the edge for me. 88 or 96 is really enough).

    But in the end the music counts and not the technical details. (and the tb-909 has 6 bit and 12khz (I think))

  • The TR-909’s cymbals and hihats are said to have been sampled at 18KHz/6-bit.
    Thanks Des & Looza, you both bring up some good points!

  • The important part about recording at higher sample rates isn’t really because some people can sense >20kHz Frequencies. Even if they could, most microphones aren’t sensitive to these frequencies, so it becomes a mute point. When viewed on a spectrogram, most recordings at 96kHz are pretty empty between 20kHz and 48kHz. The most important reason for sampling at >44.1kHz, aside from pitch-shifting and other processing, is that a lowpass smoothing filter has to be placed at the Nyquist frequency in order to eliminate aliasing. As with all filters, the steeper the slope, the more it causes phase anomalies, especially near the cutoff frequency. Since there isn’t much extra room between 20kHz (the theoretical limit of human hearing) and 22.05kHz (the Nyquist frequency), the filter slope has to be especially steep, causing phase problems within the audible spectrum. If you record at a higher sample rate, say 96kHz, the filter is placed around 48kHz, well outside the range of human hearing, so the filter slope can be much more gradual. Because of this, the recording is, in fact, more accurate because the unwanted effect of the smoothing filter has been mitigated to a large degree.

    The idea that accuracy in recording is a myth is clearly ridiculous. While absolute accuracy is probably completely impossible, and very likely not even desirable, there is clearly a difference in accuracy between, for instance, an analog cassette recording and a modern digital recording on even the worst of professional equipment.


Driven Machine Drums 3 + MD Bundle