If we could only hear up to 20Khz, why do some BluRay movies provide us with 96Khz to 192Khz audio sample rate?

[tomjal]

Hi,

 

I've tried to search and google this question but what I got mostly are the answers for audio editing process. I kind of understand when you use higher sample rate recordings to avoid artifacts when stretching audio files. But, why do some movies in Blu-Ray provide us with higher sample rate though? When once again, we could only hear upto 20Khz?

 

thank you.

[191x7]

Sample rate and hearing frequency ranges aren't the same thing.

[Eigenvektor]

To accurately store and reproduce sound that is stored digitally, the sample rate needs to be at least twice the highest frequency you want to capture. Search for "Nyquist theorem". Simply put, if you sample a wave at the same (or lower) frequency, this can lead to incorrect results.

 

As a simple example, think of a simple sine wave, that alternates between 1, 0 and -1. Depending on the timing, you might sample the wave any time it is 0, so your signal is 0 and you hear nothing, even though in reality there was something. To avoid that, your sampling rate must be at least twice the frequency of that sine wave.

[RONOTHAN##]

Sample rate and reproduced frequencies are not the same thing, though they are somewhat linked. 

 

 

Digital audio cannot store the exact waveform that was generated by the artist as that would require an infinite amount of storage. The workaround for that is to outline a maximum resolution for the audio, you can think of it like a grid that the audio waveform is layed on and the box that the line crosses through is the one that's filled in. The horizontal axis, or how often the audio signal is polled to see the the value it's at, is the sample rate, and a higher sample rate means a more accurate makeup of the audio waveform. That effective step function that's created is then thrown through some complicated math (Nyquist Shannon) to make it back into something smooth that can be played out of your headphones, and one of the main requirements is that in order to properly recreate the audio file, the sample rate needs to then be over twice the highest frequency present in the waveform. It's why the sample rate of 44.1kHz is the default, it's a bit over twice the 20kHz humans can hear with some wiggle room above to allow for imperfect filtering of high frequency that happens before the audio is recorded. 48kHz is the step up option, where the filter accuracy is less critical for not much extra complexity. 

 

Now as for the ridiculously high sample rate like 192kHz, those are still kinda overkill. Technically it will still lead to a better represented waveform, but it only really allows more allows high frequency components to make it in (something something Fourier transforms) that won't really be heard by anyone other than dogs. My guess would be audiophiles just wanting it for the LOLs and saying they can hear it when they very much can't. It can make some sense in the production industry, but not sure on Bluray. 

[porina]

Nice to see understanding of Nyquist here. Let me introduce the next level: Oversampling https://en.wikipedia.org/wiki/Oversampling

[OddOod]

I've always heard that the rule of thumb is your sample rate should be a power of two multiple of your target frequency range. Probably wrong, but I like learning and Cunningham's Law is useful

Edited June 17 by OddOod
Forgot the word "multiple"

[Eigenvektor]

4 hours ago, OddOod said:

I've always heard that the rule of thumb is your sample rate should be a power of two of your target frequency range. Probably wrong, but I like learning and Cunningham's Law is useful

Yeah, power of two sounds wrong

 

That would mean to sample 20 kHz you'd want (20²=) 400 kHz. Nyquist says you need at least two times the maximum frequency you want to record, so (20×2=) 40 kHz. In practice you'll typically want at least 2.3x to minimize aliasing. 44.1 kHz was some trade off between high enough sampling rate, data storage and digital to analog conversion.

 

~edit: Here's a website that allows to play around with signal rate and sample rate:

http://webapps.chem.uoa.gr/efs/applets/AppletNyquist/Appl_Nyquist2.html

 

The screenshot shows the kind of issues you can run into, if your sample rate is too low. The blue line is the actual signal. The red dots are the points in time you take a sample. The dotted green line is the signal you think you have, based on these sample points. This is called aliasing, and is an artifact of too few samples.

 

[OddOod]

19 minutes ago, Eigenvektor said:

Yeah, power of two sounds wrong

OPE! Wildly messed up that comment. Edited to be correct. 
I meant to say that I thought it was supposed to be a power of two *multiple*. Like x 1, 2, 4, etc. 
Appreciate the correction on real life vs theory ^.^

[GoldenSound]

On 6/17/2025 at 7:55 AM, tomjal said:

Hi,

 

I've tried to search and google this question but what I got mostly are the answers for audio editing process. I kind of understand when you use higher sample rate recordings to avoid artifacts when stretching audio files. But, why do some movies in Blu-Ray provide us with higher sample rate though? When once again, we could only hear upto 20Khz?

 

thank you.

Generally we can only hear up to 20khz, though some, particularly younger listeners may be able to hear slightly higher.

The primary benefit of high sample rate audio is you can achieve a more accurate signal reconstruction regardless of the filter you're using.
Nyquist theorem states that we can perfectly reconstruct a signal up to half the sampling rate, so 22.05khz for normal 44.1khz content. BUT, the bit people miss is that this is only true if you perfectly band-limit. ie: instantaneously and infinitely attenuate exactly at the Nyquist frequency of 22.05khz, completely removing all unwanted signal components above that and leaving everything below it untouched.

In practice, this requires infinite computing power, so we can't do that.

Because DACs have limited computing power, the filter options almost never achieve this. Take the Topping D90 as an example of a DAC that in most ways, noise, distortion etc, is very good. However none of the filter options actually adhere to Nyquist theorem as none of them attenuate fully by 22.05khz, meaning there will be unwanted content remaining and the end result is not fully accurate:

 

 

Some other DACs may have filters that do attenuate fully by 22.05khz, but attenuate some of the treble under 20khz which means they could then sound audibly 'darker'.

Some DACs like the Ferrum WANDLA GSE (disclaimer: This is a DAC that I collaborated on so take that as you will) have separate dedicated compute components, and do the reconstruction with a higher performance filter, that leaves everything under 20khz untouched but does fully attenuate by 22.05khz, meaning it is accurate to Nyquist theorem:

 

 

And then some products like the Chord MScaler are entirely dedicated to this task, using an exceptionally powerful FPGA to run a filter with over 1 million coefficients (instead of the typical 128-1024 in most DACs) and thereby achieving practically instant attenuation right at 22.05khz:

 

 

This is only difficult though because 44.1khz source material only gives us 2.05khz between the 20khz generally agreed upon audible band limit, and the 22.05khz nyquist frequency.
With something like 96khz source material, the Nyquist frequency is now 48khz, so instead of 2.05khz between that and 20khz you have 28khz, meaning you can have a much simpler filter and still achieve an accurate reconstruction.

 

There is also some debate about audibility of more precise time domain behaviour from high sample rate material, but whilst existing study does show that some listeners can discern high-sample-rate material from regular 44.1khz stuff to a statistically significant degree, there hasn't been enough to determine exactly why or what the specific reason for that is.

 

https://secure.aes.org/forum/pubs/journal/?ID=591

 

"meta-analysis that combined over 400 participants in more than 12,500 trials. Results showed a small but statistically significant ability of test subjects to discriminate high resolution content, and this effect increased dramatically when test subjects received extensive training. This result was verified by a sensitivity analysis exploring different choices for the chosen studies and different analysis approaches. Potential biases in studies, effect of test methodology, experimental design, and choice of stimuli were also investigated. The overall conclusion is that the perceived fidelity of an audio recording and playback chain can be affected by operating beyond conventional resolution."