96Khz. Why?

[theophilus]

it's like taking a picture with a higher resolution... pixels are smaller and thus you can zoom in deeper while keeping lots of detail.

This is actually plain false and a lot of the confusion about this topic stems from exactly this misleading analogy. 96kHz does NOT lead to a finer resolution of audio than 44.1kHz or whatever...

no, he's exactly right... it _is_ a finer resolution picture. just normally, you don't need that, because it doesn't change the output when simply played back. which is why we need to differentiate playback and processing.

even with the finer resolution, the output waveform will be exactly the same (*) for a 96KHz signal as for 48Khz (keeping 2:1 for simplicity).

however, if you take a simple 4.8KHz sinewave, this will be described in 10 samples for the 48KHz, and 20 samples for the 20 KHz signal.

in many cases this doesn't matter, in some cases, having that extra detail can make the processing more exact; if you have feedback for instance, you can calculate feedback 2x as often, which because of the wild nonlinearities in feedback may make for a significantly different output waveform.

it's very analogous to taking pictures with e.g. 12 megapixels. a 5x7 picture at 300dpi resolution is only ~3.15 million pixels; anything over that is wasted. but there still may be reasons to capture at 12MPS; for starters, it may help autofocus. but the output image will be exactly the same.



(*) assuming band-limiting to 20KHz, avoiding changes in the sampling filters, etc. which may or may not affect the output, but aren't related to the bitrate directly.

[Vathek]

Fair enough, so let me put the emphasis on this: the analogy is misleading. Because it implies that you somehow get a more detailed digital representation of the recording. In a digital picture you actually do capture more detail when upping the pixel count and can print it out larger for example (you actually always lose information when digitally capturing a picture), but a sound is perfectly represented at any sample rate as long as it's twice as high (plus some margin for real world implications) as the pitch you're recording/playing back with no loss of information at all. The additional information does not add detail or somehow more accurately represent the audio, that's my point.

I agree (and mentioned it myself) on the difference between recording and playing back vs. creating or manipulating digital audio where higher sampling rates may or may not have their benefits (depending on the implementation of the coders of the synth / effect in question).

[jlgrimes]

There is a website that was done by FL Studio that talks about various Audio myths (it more emphasized the myths surrounding the sound quality of different DAWS). I think it is called DAW Wars. It goes into the topic and have a more detailed link about 96 khz audio and about why it doesn't matter. It even states that in some cases very high sample rates can be bad because most speakers aren't designed to play back very high frequencies and it can add distortions to your audible range.


The website had plenty of good information.

[Tarekith]

Here's what started it: https://xiph.org/~xiphmont/demo/neil-young.html

Here's the video replies:

https://www.youtube.com/watch?v=FG9jemV1T7I

https://www.youtube.com/watch?v=cIQ9IXSUzuM

[H20nly]

One beneficial point worth mentioning about higher sample rates is that they reduce the latency of the system.

what do you mean by "the system"?

[jestermgee]

For recording, 96Khz sometimes can make a huge difference if you like to torture your audio files.
The most obvious situation is when you want to pitch down a sound; it's like taking a picture with a higher resolution... pixels are smaller and thus you can zoom in deeper while keeping lots of detail.

This was the only point I was going to make. Having done a lot of location and film recording which is always either 24/48 or 24/96 I have used 96k recordings slowed down massively to create completely new sounds which sound much more realistic than when recorded at 44/48.

Other than this situation though I personally notice no difference between beyond 48k but the debate will continue for ever and a day.

[fishmonkey]

Fair enough, so let me put the emphasis on this: the analogy is misleading. Because it implies that you somehow get a more detailed digital representation of the recording. In a digital picture you actually do capture more detail when upping the pixel count and can print it out larger for example (you actually always lose information when digitally capturing a picture), but a sound is perfectly represented at any sample rate as long as it's twice as high (plus some margin for real world implications) as the pitch you're recording/playing back with no loss of information at all. The additional information does not add detail or somehow more accurately represent the audio, that's my point.

you are torturing the visual analogy to death here.

the audio equivalent of printing a picture at a larger size, or zooming in, is to slow down the playback of the sound. in this case, having more audio samples is analogous to having more pixels.

analogously, there is also an upper limit to how many pixels are required at at certain image size, viewing distance, and visual acuity.

[Da hand]

it's like taking a picture with a higher resolution... pixels are smaller and thus you can zoom in deeper while keeping lots of detail.

This is actually plain false and a lot of the confusion about this topic stems from exactly this misleading analogy. 96kHz does NOT lead to a finer resolution of audio than 44.1kHz or whatever, it only captures a larger range (half of which is outside of the range of the human ear).

no, he's exactly right... it _is_ a finer resolution picture. just normally, you don't need that, because it doesn't change the output when simply played back. which is why we need to differentiate playback and processing.

even with the finer resolution, the output waveform will be exactly the same (*) for a 96KHz signal as for 48Khz (keeping 2:1 for simplicity).

Actually, I see a bit of potential confusion there by both Vathek and theophilus, lol..... I agree with you both that for plain playback, there may not be a perceivable difference by humans - and I stress "humans" and "perceivable" - but there is an actual increase in the information and resolution carried by a signal with a higher sampe rate.

Let us start with the simple definition of waves - f = v/? .... or for sound... (frequency) = (speed of sound)/(wavelength)

Since the speed of sound in a given medium (non-dispersive) does not change, a higher frequency sound will be represented by a smaller waveform than a lower frequency sound.

The larger frequency range you mention Vathek - only includes more and more higher frequencies because you increase the sample rate. If you sample at 100hz, you will only have 100 points to work with and will only see (hear) clearly frequencies up to around 50hz... your 100 dots resolution will not enable you too see a 1000hz signal for example.

So, as you try to reproduce higher and high frequencies in the same time frame, more and more samples (or dots to fill in the picture of the waveform, if you will) will be needed to actually see (and reproduce) the smaller waveforms associated with those frequencies. Therefore, what you are in fact getting is a finer and finer resolution of the audio. You have more samples (or more dots) to fill in the audio waveform picture.

This is also why, theophilus, with a finer resolution, the output waveform will not be exactly the same. You will have more and more detail as you go up in sample rate.

Again, I agree that our ears (or maybe it's just our brains) may not have the resolution to pick up the difference, but when you look at the actual waveform, you will see more detail in the higher sample rate signal.

Back to Jur's statement... as with a picture, as you zoom in on the waveform, with a higher resolution waveform you will have more information to work with than with a lower resolution waveform. Whether all those pixels are useful or not in the context you use them in is one thing, but the extra information is available.

[Division Monarchy]

Would the pixel analogy make more sense if we compared it to bit depth in audio rather than frequency range?

[Vathek]

One beneficial point worth mentioning about higher sample rates is that they reduce the latency of the system.

what do you mean by "the system"?

The in/out latency of your computer goes down at higher sampling rates.

[Vathek]

Would the pixel analogy make more sense if we compared it to bit depth in audio rather than frequency range?

No, actually even less. I get the other guys' point regarding the sampling rate when drastic time stretching or such is applied, but when it comes to bit rate, no.

[Vathek]

it's like taking a picture with a higher resolution... pixels are smaller and thus you can zoom in deeper while keeping lots of detail.

This is actually plain false and a lot of the confusion about this topic stems from exactly this misleading analogy. 96kHz does NOT lead to a finer resolution of audio than 44.1kHz or whatever, it only captures a larger range (half of which is outside of the range of the human ear).

no, he's exactly right... it _is_ a finer resolution picture. just normally, you don't need that, because it doesn't change the output when simply played back. which is why we need to differentiate playback and processing.

even with the finer resolution, the output waveform will be exactly the same (*) for a 96KHz signal as for 48Khz (keeping 2:1 for simplicity).

Actually, I see a bit of potential confusion there by both Vathek and theophilus, lol..... I agree with you both that for plain playback, there may not be a perceivable difference by humans - and I stress "humans" and "perceivable" - but there is an actual increase in the information and resolution carried by a signal with a higher sampe rate.

Let us start with the simple definition of waves - f = v/? .... or for sound... (frequency) = (speed of sound)/(wavelength)

Since the speed of sound in a given medium (non-dispersive) does not change, a higher frequency sound will be represented by a smaller waveform than a lower frequency sound.

The larger frequency range you mention Vathek - only includes more and more higher frequencies because you increase the sample rate. If you sample at 100hz, you will only have 100 points to work with and will only see (hear) clearly frequencies up to around 50hz... your 100 dots resolution will not enable you too see a 1000hz signal for example.

So, as you try to reproduce higher and high frequencies in the same time frame, more and more samples (or dots to fill in the picture of the waveform, if you will) will be needed to actually see (and reproduce) the smaller waveforms associated with those frequencies. Therefore, what you are in fact getting is a finer and finer resolution of the audio. You have more samples (or more dots) to fill in the audio waveform picture.

This is also why, theophilus, with a finer resolution, the output waveform will not be exactly the same. You will have more and more detail as you go up in sample rate.

Again, I agree that our ears (or maybe it's just our brains) may not have the resolution to pick up the difference, but when you look at the actual waveform, you will see more detail in the higher sample rate signal.

Back to Jur's statement... as with a picture, as you zoom in on the waveform, with a higher resolution waveform you will have more information to work with than with a lower resolution waveform. Whether all those pixels are useful or not in the context you use them in is one thing, but the extra information is available.

I get your point. And the extreme time stretching example makes sense. But where the pixel analogy fails is that a picture is always a snapshot of the 'whole thing' for lack of a better word, just more or less detailed depending on the pixels used to capture it. But in audio, a 50Hz wave can be perfectly captured and reproduced with a sampling frequency of a bit over 100Hz with no loss at all, it's a perfect representation and not just a approximation where the question is whether or not the amount of detail is sufficient to fool the human ear (or eye in the case of the picture). It's perfect.

[fishmonkey]

Would the pixel analogy make more sense if we compared it to bit depth in audio rather than frequency range?

nope. the audio bit depth corresponds to the colour bit depth used in an image. the most common currently used is 24 bit colour, which uses 8 bits for red, 8 bits for blue, and 8 bits for green.

the number of pixels sampled in a given distance, e.g. 300 pixels per inch, determines the visual resolution. of course how this looks also depends on how good your eyesight is, and how far away from the image you are. the auditory equivalent is the number of samples per second. here we are considering an image to be spatial, and sound to be temporal. of course sound can also have a spatial element, and vision can have temporal elements as well.

[fishmonkey]

I get your point. And the extreme time stretching example makes sense. But where the pixel analogy fails is that a picture is always a snapshot of the 'whole thing' for lack of a better word, just more or less detailed depending on the pixels used to capture it. But in audio, a 50Hz wave can be perfectly captured and reproduced with a sampling frequency of a bit over 100Hz with no loss at all, it's a perfect representation and not just a approximation where the question is whether or not the amount of detail is sufficient to fool the human ear (or eye in the case of the picture). It's perfect.

i think you are still torturing the analogy.

a 50 Hz sine wave (not just any 'wave') can be perfectly captured with a sampling frequency of just over 100 Hz because it is a slow and simple wave to reconstruct mathematically.

analogously, a very long straight line can be perfectly sampled with just a few data points.

[XSIMan]

If Live is set to 48khz and I export the render at 96khz does live run all the plugins at 96khz while it renders?.

If my sound card in Asio is 96khz can Live still manage to run all my plugins at 192khz during the render if thats set to export at 192khz.