Press "Enter" to skip to content

What's the difference? (sample rate/bit depth test)

Test yourself. Create a new session in your DAW at 96kHz/24 bit. Import this file.


The bit depth and sample rate change with each sample. Can you hear the difference?

The source is a an original piece of music played by an orchestra (in a fairly bad room) captured with Neumann mics into Nagra 6 HD recorder at 96/24. Edit/mix/master by me.

4 versions were rendered from the mastering session with highest quality resampling mode and Ozone 4 MBIT+ dither.

turn your computer upside down for the answers

ʇıq ㄣᄅ zHʞ96 (ㄣ

ʇıq 9Ɩ zHʞ96 (Ɛ

ʇıq ㄣᄅ zHʞƖ˙ㄣㄣ (ᄅ

ʇıq 9Ɩ zHʞƖ˙ㄣㄣ (Ɩ


  1. Christopher w
    Christopher w May 24, 2011

    oh, now that’s strange. I thought the difference would be like night and day but I honestly can not tell the difference between the higher sample rates, I can however tell the difference between the bitrate.

    I thought the second one was my favourite with clarity and depth.

    Interesting, very interesting indeed.

  2. online mastering
    online mastering May 25, 2011

    I was just about to write the differences may not be as big as you might imagine. Numbers play a more important role when there is a lot of maths acting on the numbers IMO. Like ITB processing.

    Downloaded for a listen.


  3. Georg Nikodym
    Georg Nikodym May 26, 2011

    My listening environment is far from ideal. Like most, really. I have the usual noise and such. My audio interface may or may not be up to the task. My ears no longer hear the crazy frequencies that can drive teenagers crazy…

    So I am totally ok admitting that I don’t hear a difference. And frankly, I’m not really surprised.

    I offer up exhibit A:

    44.1kHz sample rate is enough to capture signals up to 22.05kHz (see Nyquist-Shannon)
    96kHz sample rate is enough to capture signals of up to 48kHz

    This might come as a surprise but chances are high that your speakers don’t make any vibrations above some 20kHz.

    Bit depth defines how many distinct values can be represented on the y-axis.
    16-bits yields 65536 values (-32767 through 32768).
    24-bits yields 16777216 values (-8388607 through 8388608)

    Ultimately, these digital values are converted to a +/- voltage in your speakers.

    I really doubt that the human ear is capable of detecting the difference (in single samples).


    we play with multitrack recording. Where all the tracks in a recording are added together. My intuition is that any “rounding errors” associated with lower sampling rates and bit depths will combine and might actually be detectable.

    Here’s an experiment I’d like to see:

    1. Record many tracks with max fidelity (96kHz/24-bit).
    2. Do mix but don’t render.
    3. Duplicate project.
    4. In one project, convert all of the tracks to 44.1kHz/16-bit
    5. Render both and compare.

    IOW, does the summing of lower quality material result in something that is *perceptibly* different from the higher resolution process.

    I think this goes to the question that many of us ask. Is it worth the huge hit in disk space and processing power required to work at very high resolutions.

  4. Jon
    Jon May 26, 2011

    The frequencies within the human hearing range are also captured with more accuracy when recorded at higher sample rates.

    From this test, and actually I compared 16/44.1 No Dither to original and while there was the most loss (as heard in null test) it still didn’t sound very much different. Maybe the difference between 320kbps mp3 and 256kbps mp3 (disregarding the problems inherent with that format). Not the difference between VHS and DVD as you might expect.

    And that leads me to think that its all in the capture that makes a difference.
    Something captured at 96kHz, and converted to 44.1kHz will still sound better than capturing at 44.1 to begin with. The problem is this is very hard to test.

    Thanks for commenting folks!

    • Xperienced
      Xperienced August 20, 2011

      “The frequencies within the human hearing range are also captured with more accuracy when recorded at higher sample rates.”


      “Something captured at 96kHz, and converted to 44.1kHz will still sound better than capturing at 44.1 to begin with.”

      Just, no. Not unless your recorder is defective or badly designed, no.
      If it is, get another recorder. If you do, send me that Nagra! I will help you dispose of it! Heck, I’ll even pay you the shipping in the name of friendship!

      • Jon
        Jon August 20, 2011

        Nice try.
        Do you have anything to back up your argument?

        • Xperienced
          Xperienced August 20, 2011

          My “argument backup” would be anything that correctly describes the sampling theorem. On an ideal sampler the response is exactly the same for everything below sr/2. In the real world of course this is not so, but the differences are only in very low fequencies (below 10Hz or something like that, read the ADC chip’s datasheet) due to the DC filter and in the upper end due to un-ideal lowpass filter, however upper end here is higher than most people thinks (it’s even higher than the “crazy frequencies that can drive teenagers crazy”, so if you can’t even hear those, yeah – think the last couple kHz, for 44.1 that’d be from 20kHz to 22.05kHz) and even then I don’t think it sounds as bad as people imagines, in my experience the effect is actually rather subtle (just play with filters at audible frequencies and hear for yourself – they work exactly the same way at high frequencies).

          Also bear in mind that nothing is actually being captured at 48kHz or 192kHz (I hate 44.1 for unrelated reasons – CDs are obsolete anyway as far as I’m concerned), the vast majority of ADC chips are sigma-delta (a.k.a. 1-bit) at a few MHz, and the 44.1~192 streams you hear are all derived from the same 1-bit stream. When you downsample from 192 to 48 you’re only repeating what the chip is already internally doing, you might as well let the chip handle it and save yourself a layer of processing which is always a good thing.

          You can prove yourself the sampling theorem with a DSO (digital storage oscilloscope – analogs won’t do it because your memory is not that good), just apply a DC voltage to your favorite ADC-DAC loop and see how the response is, besides being slower or faster, the same for any sampling rate. Any other glitch you see is either a design error (which means “stay away from either that ADC or that DAC”) or a problem with the way you’re applying the DC voltage (which means “science: ur doin it wrong”).

          Gee, that’s a whole lot of parentheses.

  5. Rupert Brown
    Rupert Brown May 29, 2011

    To be honest I cant hear much in it. I thought I could hear a difference but struggled to pick a “best”.
    Keep em coming Jon these are really great for training the old ear holes.

  6. menno keij
    menno keij July 22, 2011

    29 years old ears. Didn’t hear a difference neither on speakers (fire face 400, genelecs, bad treated room) nor on headphones (beyerdynamic dt 770 pro). A B compared the first and last on headphones with a high pass filter on a very high setting. Could also hear no difference. Is it beneficial to record at > 44.1KHz? I thought that higher sample rates are better for post-production, because stacking up fx in your daw will degrade your sound gradually (except for effects that use oversampling). Any comments?

    • Jon
      Jon July 22, 2011

      Comes down to personal preference in the end.
      In this test I demonstrated that when you start with the HD version, even when you convert it down to cd standard it sounds about the same. It is difficult to test that the HD version started with more audio data compared with a 44.1kHz capture. I think what you gain to start is more than you lose in the end. You still lose about the same amount on mixdown even at 44.1kHz.
      I feel that you have an advantage starting with the higher sample rate capture. Even if you work entirely ITB, some processors sound way better at higher rates, amp simulators are the most obvious.

    • Xperienced
      Xperienced August 20, 2011

      “because stacking up fx in your daw will degrade your sound gradually (except for effects that use oversampling)”

      “Any comments?”

      Yes, the reason stacking up FX in your DAW will degrade the sound quality is because you suck at mixing, and I regret to inform you that oversampling doesn’t really help.

      I kid, I kid. But no, that’s a myth. Stacking FX doesn’t technically degrade anything unless again someone (plugin developer) did something wrong. Oversampling doesn’t matter. There are only very few instances where oversampling actually makes a difference, and in those few cases the oversampling *should* be built-in, otherwise it’s bad design.

Leave a Reply