Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: What's the point of higher sampling rates in audio? (Read 105914 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

What's the point of higher sampling rates in audio?

Reply #75
Having digitized some vinyl I know for sure I won’t go through that again.
A very time consuming affair.
If I had to do it over again, I would go for as much overkill I can afford.
Better save then sorry

TheWellTemperedComputer.com

What's the point of higher sampling rates in audio?

Reply #76
But still, it leaves out all the sounds that hasn't been recorded through a mic. Synths, or a lined bass.
And more important (i'm guessing) it leaves out all the sounds together, in a mix.
It's a good job none of this really matters, because the measurable distortion in speakers is several orders of magnitude larger than the measurable distortion in air. If this mattered, these recordings with high levels of ultrasonics would be unlistenably bad.

Luckily, they're mostly OK. And any distortion in air, 10s of dBs lower than the distortion in the speakers, is irrelevant.

Unless, I suppose, you're running a PA rig at home - but then it's the distortion (and damage) in your ears that should concern you.

Cheers,
David.

What's the point of higher sampling rates in audio?

Reply #77

And are you 100% sure that the plugin is samplerate aware???

I've been part of this past year fixing most of the plugins of my program* so that they don't perform differently at different sample rates.


If nonlinear processing is done in the digital domain, certain high harmonics can reflect back down into the audible range. Any signal that is generated in the digital domain above the Nyquist frequency  aliases around the Nyquist frequency.  For example, an 8 KHz tone in a 44.1 KHz sampled sytem is distorted by a fourth order digitally-implemented nonlinearity and would be expected to produce a fourth harmonic at 32 KHz.  Since 32 KHz is 10 Khz higher than the 44.1 KHz Nyquist frequency of 22 KHz, the  4th harmonic is aliased down to 12 KHz where it is far more audible than it would be in a digital system with a much higher sample rate or an analog system. In a 96 KHz system, the same nonlinearity would produce a foruth harmonic at 32 KHz where it would not be audible.

Aliasing can also impact certain dynamic range compression or expansion algorithms that are based on nonlinarities implemented in the digital domain.

Most kinds of processing that are used in music synthesis and mixing are linear, so aliasing is not a common problem.


Aren't algorithmic reverbs based on non-linear transformations? Especially these days they offer many types of adjustments to the parameters for effects.

What's the point of higher sampling rates in audio?

Reply #78
Aren't algorithmic reverbs based on non-linear transformations? Especially these days they offer many types of adjustments to the parameters for effects.


How so? huge IIR filters are not nonlinear.
-----
J. D. (jj) Johnston

What's the point of higher sampling rates in audio?

Reply #79
How so? huge IIR filters are not nonlinear.

I was under the impression that standard reverb algorithms introduced time-variant "stuff" in the filter topology to reduce audible artifacts, making them non-LTI?

Nevertheless, I assume that the kind of nonlinearities "warned" about by Arnold in terms of aliasing was of a very different kind: Signal-dependent gain with non-smooth envelope.

-k

Re: What's the point of higher sampling rates in audio?

Reply #80
The highest frequency at 44.1 kHz 16 bit rate is 22.05 kHz according to Nyquist. Right?

In one second, the 44,100 samples would just pick up the crest and trough of the 22,050 waves. Do our ears just need the alternation between top and bottom of every wave to hear it? Or do we fill in the blanks?

Wouldn't higher sample rates like 192 kHz help when lots of instruments are playing at once? Take the 16 vs 24 bits out of the equation since it is only used for volume. Let's assume you only need the crest and the trough, which would expressed as a 1. Anything in between, and also no sound, would be expressed as a 0. If there are multiple instruments, then you would get a 1 for that sample, if there are any instruments at all it would be 1, but if there are no instruments hitting a frequency crest or trough, then you would get a 0.

Playback would be do the opposite. The player would look for rates of repeating ons and offs to produce a pitch. The problem I see here is that a 1 kHz note would perfectly overlay a 100 Hz note. And you would just hear the 1 kHz note. Of course real notes rarely line up like that.

This still leaves you with having to split up the 44.1 thousand samples amongst several instruments in a full band. Rhythm guitar has 6 notes with half a dozen harmonics each (36), plus the one bass note (6), plus the four at most notes of a drummer times maybe 6 harmonics (24), plus the one to three notes at once of the lead guitar times 6 harmonics (18) equals 84 pitches that need to be sampled. This is actually more like a 100 pitches estimated since notes move up and down a little within even a second.

If the computer can only do 41,100 samples per second total, then each harmonic only gets 411 samples per second. This would be fast enough for low notes, but you can clearly hear high notes on full band CDs.

I've read somewhere that the samples are constant, so maybe the computer is predicting sounds for playback based on whether the switches are on of off for every one of the 41 thousand samples per second. If the computer sees a crest or trough at sample 1, 101 and 201, etc, but also sees something at 1001, 2001, etc. It probably drops the bass note.

I've noticed that when I play a bass note perfectly in time and pitch, I can't hear it, but if the note is way off, it's very noticeable. Problem is you need notes to be heard. Which leads me to conclude that the higher rate of 192 kHz would be better, so the computer can hear instruments that are almost overlapping and catch more stuff as long as the lower notes are not perfectly perfect.

You might not run into overlapping frequencies at higher pitches because they are spread apart like 980 Hz then 1030 Hz etc, but very low notes around 20 Hz are so close together they would overlap with higher frequencies more easily. ie. If a low B is 25 Hz, Low C is 27 Hz, and a Low C# is 30 Hz, then one of those frequencies is bound to be an exact fraction of a higher note so that the computer just doesn't see it.

I have a solution. The computer would record in pitch groups. First would be a section between 4 Hz and 7 Hz. It would run 7  samples per second just for this, then test every second to see if there is movement in this range. Next, the computer would record 15 samples from 8 Hz to 15 Hz, then test for this band, and so on. Because this has a doubling pattern, the code would simply be 2 to the x through 2 to the x + 1 minus 1.

The test would be harder, and would only work with real musicians/vocalists. The computer would look for the naturally slight pitch shifts and know that if a note disappears, but then comes back, when there are higher overlapping notes, then the notes still exists and should be produced at let's say 30 Hz.

The computer would use three samples for this check. First, is there imperfect notes before, second, are there imperfect notes after, third, in the empty space is there possibly overlapping notes that are half, third etc a higher frequency. The lowest bands would check less often, and the higher notes would check a lot, but this would be fine if you have 192 thousand samples.

The only problem is a bass player often isn't playing for half a second when the rest of the band is playing. At 240 BPM, you get 4 beats per second, and 8 eighth notes per second in 4/4. A measure at 240 is 1 second. If the recorder is sampling right in line with the beginning of the measure then there would be 2.8 thousand samples at 44.1 between the 1st and 2nd eighth note, but if the bass lines up perfectly as a fraction of a higher pitch, then the computer needs to see the sudden drop outs in relation to the beat. The bass player would drop out at the end of his note of course, but the computer should watch for when he drops out, then comes back suddenly anytime before the 2.8 thousandth sample. If he comes in anytime after 2 thousand or so and continues, then it is part of a new note being played.

This solution would prevent bassists from going any faster than 8th notes at 240 BPM or 16th notes at 120 BPM. Personally, I don't play any faster than this because it just turns into mush even still in analogue.

To wrap it all up, a higher sample rate would catch even the smallest differences so that what overlaps at 44.1 kHz doesn't overlap at 192 kHz. Then you don't need some complex work around to try and figure out where the bass notes are.

Re: What's the point of higher sampling rates in audio?

Reply #81
Quote
This still leaves you with having to split up the 44.1 thousand samples amongst several instruments in a full band.
How is it possible to know absolutely nothing about sampling and still have enough self-confidence to write FIFTEEN paragraphs of nonsense?

Re: What's the point of higher sampling rates in audio?

Reply #82
@shusse82 : Please, don't take this as an attack to you, but your post is extremely crazy.

Nyquist says that you need 2*N samples to record frequencies up to and not including N. This means every frequency, even non-integer frequencies.
The bit depth can, in fact, help adding precision to non-integer frequencies, so bit depth is not only a volume issue (volume understood as max value where noise is under the threshold of hearing. Obviously volume is not limited itself by bit depth)

It needs 2*N since it samples a wave, and a wave oscillates between a peak value and a valley value. This representation is what becomes the frequency in itself when an amplifier and a speaker reproduces it. To understand how the ear interprets this variation in pressure that the oscillating wave produces is relatively easy, but you need to study about the cochlea and the inner ear ( see https://en.wikipedia.org/wiki/Ear and specifically the inner ear link). Suffice to say it is not about detecting a peak.


From here on, your post is where it starts to be crazy. I can't even try to guess where do you want to go. What I first think is that you imply that 44Khz is not enough, that you could use 192Khz, and then you seem to try to describe a way to encode audio differently, very limited to what it can store (because you focus only on instruments, so I may ask... will it be able to store vocals?) and even you describe difficulties in how you could store what you try to store.

That's why I won't reply about that other part, more so when you end saying that maybe with 192Khz you would have enough.

Re: What's the point of higher sampling rates in audio?

Reply #83
Reading that reasoning above on the same page with J.D. posting must be some kind of comedy  :)
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

Re: What's the point of higher sampling rates in audio?

Reply #84
How is it possible to know absolutely nothing...and still have enough self-confidence...(to invent your own "facts".)
Welcome to the new reality. Why bother to understand how things really are when it's so much easier to just make it up?

Re: What's the point of higher sampling rates in audio?

Reply #85
Funny how Meridian could contrive a test that could only manage to get a dataset that showed a 56% success rate.  With this latest wrinkle it's a wonder how digital audio could even work at all!

Re: What's the point of higher sampling rates in audio?

Reply #86
I wonder why those topics are such often repeated, even if I in the past also contributed to their length.

I think it is clear given todays knowledge.

48 kHz is the optimal sample rate for human audio and rates above that we can use, but it does not bring any practical (audible) nor theoretical (impacting spectrum of human hearing) advantages. We can do it but only for technical/futureproof reasons. 44.1 kHz is the CD standard and since majority of recordings is available in that format it is very important for DACs and other audio equipment to support it well.

16 bit audio is completely recommended for practice - audibly transparent playback. When we record in 24 bit we can keep it in that format (if space is not a concern) even for playback, then we do not have to dither/noise shape and we have theoretically better source for subsequent conversions, if required.

Until somebody proves - by ABX or other reproducible method of testing - something else, then these discussions do not bring new knowledge.


Re: What's the point of higher sampling rates in audio?

Reply #88
Doing a very simple Bayesian analysis of  90 successes out of 160 attempts in the style of my "Bayes Factors for ABX tests" post, the result is actually evidence for the null hypothesis.

Even if we start with a completely dishonest prior that distributes the 95% most credible values (of the parameter that describes the performance of the listener) above 50% with the peak at exactly 90/160%, the result is still only very weak evidence for the alternative hypothesis.


@greynol: Simply remove a few numbers from the quote timestamp and you'll end up with such dates.
"I hear it when I see it."

Re: What's the point of higher sampling rates in audio?

Reply #89
the quoter puts the time in ms in the tag but it's clipping the 3 least significant digits or something.

edit
ha, ninja'd

Re: What's the point of higher sampling rates in audio?

Reply #90
I wonder why those topics are such often repeated, even if I in the past also contributed to their length.
...and there you were thinking that being mocked and ridiculed with such vehemence was personal.  Sorry bro, neither you nor your placebophile tendencies/misunderstandings were(?) all that unique/special.

48 kHz is the optimal sample rate for human audio
Declaring what is "optimal" will likely get you into trouble; especially when the ability to hear pure tones does not neatly correlate with the ability to hear the same frequencies in actual consumer content, pesky psychoacoustics!

technical/futureproof reasons.
theoretically better source for subsequent conversions
The needle on my speculative//hand-waving bullshit meter just moved.  How fun!

Re: What's the point of higher sampling rates in audio?

Reply #91
The highest frequency at 44.1 kHz 16 bit rate is 22.05 kHz according to Nyquist. Right?

...
...
...

To wrap it all up, a higher sample rate would catch even the smallest differences so that what overlaps at 44.1 kHz doesn't overlap at 192 kHz. Then you don't need some complex work around to try and figure out where the bass notes are.

Amazing. I've seen heaps of people, first and foremost myself, completely misunderstand digital audio, but to follow a fictional understanding  with a fictional alternative is quite novel. Well, you have an idea. Go for it! See if it works. If you can.

In my experience of systems management and support, people who say things like, "the computer would..." actually don't have a clue how, or whether their request is even feasible. The have no clue how "the computer would..." Sorry, making it up as you go along does not write working software or create usable devices.

I am very much a novice at this stuff. But, even if I have not, and probably never will, master the maths of what happens in the gaps between those samples, at least I now understand that I am not, actually, as some of the audiophiles would have it, listening to something like speeded-up stuttering. I had to read a lot, watch some videos, listen to some actually-well-informed people etc etc to get this far. Now its your turn. If you choose to do so, then welcome to the club.  :)
The most important audio cables are the ones in the brain



Re: What's the point of higher sampling rates in audio?

Reply #94


technical/futureproof reasons.
theoretically better source for subsequent conversions
The needle on my speculative//hand-waving bullshit meter just moved.  How fun!

I know that meter (would not call it bullshit however) is very sensitive :), and it is good to have it working ...

I meant by those reasons digitizing very unique analog content when it is necessary/desired to digitally store  information e.g. up to 30 kHz. Then we can theoretically use 96 kHz sample rate, but no more and even then it is only for archival purposes, we still cannot hear that. For normal/standard audio recordings (including common vinyl records), there is nothing what calls for or forces us to use higher SR that 48 kHz.

I agree that word "optimal" is risky to use, especially in discussions here.

Similar to many audio processing standards like Mastered for ITunes or workflow of good engineers like Bob Ludwig, I prefer to use 24bit source for resampling/editing/creating MP3s if that processing is required and 24 bit source is available. This, however, does not claim that using 16 bit source for that purpose would be practically (audibly) different.







Re: What's the point of higher sampling rates in audio?

Reply #95
This, however, does not claim that using 16 bit source for that purpose would be practically (audibly) different.
Which is all that matters.

Re:Apple/ Bob Ludwig
I see you're reverting to hollow pleas to authority instead of providing evidence once again, sigh.
https://hydrogenaud.io/index.php/topic,111271.msg918333.html#msg918333
https://hydrogenaud.io/index.php/topic,111271.msg918387.html#msg918387

Re: What's the point of higher sampling rates in audio?

Reply #96
Just got done recording a bat conversation @ 576 KHz.  The problem is I couldn't hear any of it.  :P

Re: What's the point of higher sampling rates in audio?

Reply #97
Just got done recording a bat conversation @ 576 KHz.  The problem is I couldn't hear any of it.  :P

What sort of bandwidth does a bat chirp cover? As in, what are the typical lowest and highest frequencies present in a chirp?
Regards,
   Don Hills
"People hear what they see." - Doris Day

Re: What's the point of higher sampling rates in audio?

Reply #98
What sort of bandwidth does a bat chirp cover? As in, what are the typical lowest and highest frequencies present in a chirp?

A quick google says 10-100kHz, with some sources claiming >150kHz. That would leave some merit to sampling at 192kHz or even higher if you are into bat music or sound effects, or possibly dolphin and whale music as they can hear similar frequencies according to my quick google  :) Since sound is such an important sensory input for bats, like our vision, would bat music make them disoriented or "drunk"? Maybe a bat could ABX 16 bit/48kHz from 32 bit/192kHz?

But should we care?

Re: What's the point of higher sampling rates in audio?

Reply #99
I did a quick search on hearing range and read this wikipedia article before posting my funny post making fun of higher sample rates to at least be sure.