High sample rates: Can you tell the difference? UPDATED!

In the spirit of my analogue vs. digital article, it is time for another Mythbusters style debunking session. Today we are going to explore the common misconception that higher sample rates equate to better sounding audio. There is no shortage of rhetoric by self-proclaimed internet experts that by simple matter of dropping a faster processor into a digital unit that you’ll get better sound. This, they claim, is due to having the horsepower to run at higher sample rates. These experts believe that higher sample rates will give better sound with no tweaking of the underlying algorithms. Is it true (and does it matter to the home recording studio owner)?

Testing audio sample rate

My test includes three different audio tracks processed with the same plugin at four different sample rates. The input for each track is an analogue signal recorded at each tested sample rate. The input tracks are then processed at each target sample rate with the Gearbox Plugin from Line 6. Processed tracks are then rendered to variable bit rate mp3 files at 16-bit/44.1k audio. My VBR settings are a minimum of 128kbps to a maximum of 320kbps. Why are the sound samples rendered to the common 44.1k sample rate? Because there is not a single major audio distribution channel available today (or in the forseeable future) that does not require you to supply audio at this resolution. From physical compact discs to iTunes, 16-bit/44.1k audio is the order of the day. The VBR mp3 files I provide as sound samples are even of better fidelity than what you get on iTunes. Also, it doesn’t matter how great things sound in your home studio. The only thing that matters is how great it sounds once your audience is able to listen. When that audience listens, it is going to be at 16-bit/44.1k. A large part of the high sample rate myth is the claim that you should record at high sample rates even though you will downsample to 44.1k at the end. Experts abroad claim this is because you will retain definition through the whole process that will lead to an audible improvement of fidelity after the downsampling.

The sound samples

Here are the sound samples. I have provided sound samples of vocals, rhythm guitar, and lead guitar. Each of the three test tracks is processed at each of the four target sample rates. The sound samples are labeled as a, b, c, d, but are not in any particular order (in other words, a is not necessarily the lowest sample rate and d is not necessarily the highest). All three samples use the same ordering of sample rates though (so if you think c is the 192k sample rate on vocals, then c is also the 192k sample rate on the two guitars).

source clip a clip b clip c clip d
Vocals mp3 wav mp3 wav mp3 wav mp3 wav
Rhythm guitar mp3 wav mp3 wav mp3 wav mp3 wav
Lead guitar mp3 wav mp3 wav mp3 wav mp3 wav

Your task is to listen to the sound clips and decide for yourself which clip goes with which sample rate or if you can even tell the difference. Decide for yourself. Post here or email me at [email protected] with your guesses. I will wait until I have 1000 responses before I post the results. I fully expect there will be no clear ability to determine which clips were processed at a high sample rate. The real winner will be you, armed with your own informed decision about how important a higher sample rate is to you for your own home recording studio.

Update

I have had 19% of the responses I am hoping to get on this. No one has yet been able to tell which used the “higher quality” algorithms. About half of you have missed the point entirely. Here are some clarifications and additional background on the argument.

  • Experts claim that no matter what your final media, recording/mixing at a higher sample rate will result in higher quality final product. Thus, according to the argument, using mp3 as a test is appropriate as it is nearly the most listened to format in our modern world, from iPod to internet downloads.
  • Faced with this article, a few of our experts have begun to backpedal. Now they claim mp3 is not a fair test. Fine. I have added full bandwidth, cd-quality wav files of all samples. If mp3 is the most listened to format, compact disc is certainly a close second. With this extra evidence, I challenge you to see if you can still tell which files used the higher sample rates.
  • Experts claim that the listener’s sound system will prevent them from being able to hear the subtle differences. If the listener can not hear the subtle differences in this simple test, what makes one think they will be able to hear the subtle differences when listening to your album on that same system?
  • Experts have claimed the need to future proof your work. I have mixed well over a hundred songs for over a dozen artists spanning fifteen years. I have not yet been asked to remix or “futurize” a single one of those songs. If your particular studio work has a high chance of being asked to futurize one of your mixes in ten years, then by all means, do what is necessary to record at higher sample rates.
  • Experts have claimed the mixing process will be hampered at cd-quality sample rates because you run the risk of introducing artifacts. This is true when mixing with low quality algorithms. If you are running plugins with extreme audio manipulation effects and low quality algorithms based on sample and hold dsp then this is true. You do run the risk of introducing resampling artifacts. If your plugins are of high quality then they are not running on sample and hold algorithms and have no more chance of introducing artifacts at lower sample rates than they do at higher sample rates.
  • Experts have claimed that high sample rates are not as beneficial to single sources as they are to an entire mix. This is not an article about mixing at high sample rates. This is an article about the effect of higher sample rates on dsp algorithms found in plugins. I would love to do an article comparing whole mixes, but I don’t think my home studio computer could handle a 24 track mix at 192kHz right now.
Line 6 GearBox Plug-In Gold Bundle
Line 6 Gearbox Plugin
This plugin package from Line 6 brings you their modeling technology in a VST/RTAS plugin format. This is the plugin I used for processing the samples on this page. It is blowing out at stupidly good deals right now.
Share
Share this Post[?]
        

15 Responses to “High sample rates: Can you tell the difference? UPDATED!”

Read below or add a comment...

  1. Ben Garney says:

    The comparison test is not sound because of two factors.

    First the MP3 algorithm is lossy – so even if you use VBR or a high fixed bit rate it will discard things that according to its psychoacoustic model people can’t hear. Second, playback is going to be limited by the sound card. Most people’s sound cards are pretty cruddy and aren’t going to be able to reproduce the full range of of a 44.1khz/16-bit sample, much less something better.

    Working with a high quality while mixing is desirable – especially a high bit depth – because it lets you do more manipulation of the audio before you run into artifacts. It also helps future proof you a bit against future technology shifts.

    With carefully designed algorithms, running at 44.1khz is going to work pretty well, especially running on simple scenarios. For broad spectrum sounds in complex mixing scenarios it may not be adequate.

    Ben and I already argued about this over lunch, though, so do your own tests. :)

  2. bvesco says:

    The test is not sound because it uses the #1 method of music distribution to demonstrate the effect it will have on your final mix?

    The test is not sound because it relies on a listener’s system when the point of the test is to demonstrate the effect high sample rates have on music that people will LISTEN to?

    Uh-oh, looks like we’re gonna have to have lunch again!

  3. Cenk says:

    Weird enough do i think that 44k sounds better then 96k in the tests i have done. Maybe B ´coz of my gear or ears ?

    Using Pod x3 live By the way

  4. bvesco says:

    Have you listened to my sample clips on your setup? If you have an opinion on which sample rate sounds better, maybe you can pick which of my samples sounds best.

    • Josh says:

      I personally don’t think that most human ears can hear the difference. And I’d be willing to bet that only the smallest fraction of the world’s population would be able to hear the difference.

      I have a pretty good set of monitors and a nice audio interface and I can’t hear the difference.

  5. Krontab says:

    It’s a shame that thw wav file links are only for the vocal track.

  6. bvesco says:

    They aren’t ;)

    p.s. Links fixed…

  7. Shapes says:

    There is an obvious difference when applied to video, so isn’t this comparable to audio?
    I would think that when the compression ratios for mp3s change in the future, the difference would become clearer.

  8. bvesco says:

    Well, this is an audio blog, not video, so I can’t really make any assertions about the differences in terms of video. Compression ratios of mp3 files are not really relevant unless you can hear the difference on the original, uncompressed audio files. Can you?

  9. Boz says:

    “The input for each track is an analogue signal recorded at each tested sample rate. ”

    An analogue signal doesn’t have a “sample rate”. Please clarify the original source recording. did you MULT the take into different systems, or do different takes, or what.

  10. bvesco says:

    Analogue signals don’t have a sample rate. Digitally recorded signals do. The input is a single analogue signal recorded digitally at each sample rate. Focusing on the input you are missing the point of the test. This is about processing at high sample rates, not recording at high sample rates.

  11. Attila says:

    Hello bvesco. Thank you for the awesome info. I got a short question for you.

    I’ve to decide between pod xt and behringer vamp 3.
    Pod got 44.1 and v-amp 3 got 33.1 sampling rate. 33.1 is lower than CD rate(44.1) so will it actually sound a bit inferrior compared to pod? Please tell me YOUR opinion and dont tell me to go try`em in a guitar shop, i simply cant :))

    Cheer buddy,
    Attila

  12. bvesco says:

    V-Amp is still a direct copy of the technical capabilities of the original POD (which was also at 33.1k). POD XT was a 2nd gen modeler (which is the same tech used in POD X3 which has the same overall sound as XT). V-Amp (1, 2, and 3 all based on the same tech) therefore is a generation behind XT and X3 in terms of capability and sound.

    Another point to keep in mind is that some people will tell you that a guitar and guitar speakers have pretty limited frequency range and 33.1 is quite able to produce that range.

    POD XT can be expanded with model packs to have an enormous number of amps, speakers, and effects. I think V-Amp can not be expanded in this way. If I were choosing between them I would go for XT.

  13. Attila says:

    Thanks so much. Now I’ll move on to Zoom G9.2TT its got double 12ax7’s and haven’t read a bad comment yet!

  14. Always a worthy comparison, I have generally never understood the us of 48kHz (unless for video purposes) for music production when the target rate is 44.1khz for CD anyway, you arguably lose what you gained in SRC.