Understanding Sample Rate

[beerandmusic]

going to dinner and watch tv with wife....i must admit, it has me "thinking"....not sure if my conclusions are different, but more confused now than when i started (grin).

[tmtomh]

59 minutes ago, beerandmusic said:

 

DSD samples higher than CD, right?

So everyone in this thread believes DSD is hogwash, right?

 

You started out with a genuine curiosity in this thread, but now you're getting defensive and seem to be lashing out wildly. DSD samples much higher - but as @crenca notes above, it samples at only 1 bit of bit-depth!  DSD's native dynamic range is only about 6dB - that's six, not 60. Aggressive noise-shaping shifts most of all that horrible noise up beyond the range of human hearing, to around 28kHz or 30kHz if memory serves. But the point is that DSD's higher sample rate does not allow it to capture frequencies in the human hearing range any better or more accurately than PCM sample rates.

 

More fundamentally, it seems the part you can't get your head around is this: You are convinced that because a single frequency test tone is simpler than two people singing in harmony, which is simpler than an entire symphony, which is simpler than 10 million Elvises singing along to 20 million guitars, that higher sample rates must better at accurately capturing music when the music has a lot of different voices and/or instruments or frequencies happening in it.

 

There's nothing anyone can say about this beyond what multiple people already have tried to explain to you in multiple ways: In essence, you're simply wrong in how you're visualizing or setting up your thought experiments in the first place; and in particular you are flat-oud, totally, dead wrong about what higher sample rates allow one to capture/encode digitally. It's a fine line between working hard to explain something to someone who genuinely wants to understand, and trying to find different ways to persuade someone who doesn't want to be convinced. Since I for one did not enter into this thread even wanting or intending to persuade you, but merely to try to educate, I'm done. Others can waste their time indulging you - because that's what it's become, you indulging yourself.

[Ralf11]

maybe the thread title should be changed to Mis-understanding sample rate

[crenca]

7 minutes ago, beerandmusic said:

 

You hear a continually changing composite, with the gaps between samples averaged based on probability of samples, to keep the waveform continuous, but not based on factual samples between times?

 

A waveform has no "gaps", it is a rate of change - not a point like "here, then here, then here".  A waveform is not a stepping on individual stones (say in a river you are crossing), it is like the water flowing in the river.  Your confusing sample rate with rate of change.  Because human hearing (and sound itself through all mediums) is band limited, its rate of change can only be so much.  Thus, you only have to sample so much - you are thinking that the sampling is sampling the sound itself which you imagine to be particle like movement - it is not, it is sampling a waveform that just happens to exist in the particle based medium, which is like the flowing river and thus you do not need to sample every molecule of river water to measure the rate of flow.

[beerandmusic]

1 minute ago, tmtomh said:

 

You started out with a genuine curiosity in this thread, but now you're getting defensive and seem to be lashing out wildly. DSD samples much higher - but as @crenca notes above, it samples at only 1 bit of bit-depth!  DSD's native dynamic range is only about 6dB - that's six, not 60. Aggressive noise-shaping shifts most of all that horrible noise up beyond the range of human hearing, to around 28kHz or 30kHz if memory serves. But the point is that DSD's higher sample rate does not allow it to capture frequencies in the human hearing range any better or more accurately than PCM sample rates.

 

I have only considered frequencies within normal hearing ranges...i believe higher resolution provides more accuracy within the hearing range.  (e.g. i am, and only have been, concerned with what is within the audible frequency range).

[crenca]

5 minutes ago, beerandmusic said:

going to dinner and watch tv with wife....i must admit, it has me "thinking"....not sure if my conclusions are different, but more confused now than when i started (grin).

 

My wife and kids are at birthday parties and shopping, so I am just sitting here with my beer and music engaging you in this somewhat tedious imagining exercise  

[beerandmusic]

1 minute ago, crenca said:

 

A waveform has no "gaps", it is a rate of change - not a point like "here, then here, then here".  A waveform is not a stepping on individual stones (say in a river you are crossing), it is like the water flowing in the river.  Your confusing sample rate with rate of change.  Because human hearing (and sound itself through all mediums) is band limited, its rate of change can only be so much.  Thus, you only have to sample so much - you are thinking that the sampling is sampling the sound itself which you imagine to be particle like movement - it is not, it is sampling a waveform that just happens to exist in the particle based medium, which is like the flowing river and thus you do not need to sample every molecule of river water to measure the rate of flow.

 

what i meant by gaps is that at t1 you have freq x and t2 you have freq y, and you must connect the dots, so the detail between the dots is the gaps that is estimated, calculated, averaged, or whatever terminology you use....and that is where the details and accuracy are lost....between the samples.

[tmtomh]

2 minutes ago, beerandmusic said:

 

I have only considered frequencies within normal hearing ranges...i believe higher resolution provides more accuracy within the hearing range.  (e.g. i am, and only have been, concerned with what is within the audible frequency range).

 

You can believe it all you want - it's not true. See my edited/expanded prior comment: If you're going to repeat a baseless article of faith over and over again, you're just wasting others' time. It's not a discussion, a conversation, or an educational situation. You're just indulging yourself. You can GUTB should have a good time together - that's exactly what he does too.

[crenca]

2 minutes ago, beerandmusic said:

 

I have only considered frequencies within normal hearing ranges...i believe higher resolution provides more accuracy within the hearing range.  (e.g. i am, and only have been, concerned with what is within the audible frequency range).

 

As far as waveform accuracy they don't, except secondarily in that filtering distortion comes into play - higher sampling rates can be used with filters that do less damage during the reconstruction of the waveform.  In the human hearing range, to fully (with no "gaps") calculate a waveform (you think a waveform is "sampled", and put together back together again like humpty dumpty, or a puzzle, or a mosaic - it is not, it is fully calculated).

[GUTB]

Just now, psjug said:

A narrow pulse, say 10us, repeated every 500us, is a kind of waveform.  Is it your position that if we can hear this we are hearing ultrasonics?

 

Of course not. Ooshai showed that humans receive and process high frequency sounds, but that is 100% the domain of neuroscience. It could be why people love ribbon tweeters so much.

 

A frequency is an oscillation over a period of time. We can’t percieve tones above 20 kHz. A 20 kHz frequency oscillates at a rate of once per 50 microseconds. If looked at it in the time domain, we shouldn’t be able to hear pulses less than 50 microseconds long — however, we CAN. The study showed we are able to distinguish 10 microseconds, even against background noise.

[crenca]

3 minutes ago, beerandmusic said:

 

what i meant by gaps is that at t1 you have freq x and t2 you have freq y, and you must connect the dots, so the detail between the dots is the gaps that is estimated, calculated, averaged, or whatever terminology you use....and that is where the details and accuracy are lost....between the samples.

 

Nope, because a waveform is calculated, not "sampled".  It is continuous...

[GUTB]

2 minutes ago, crenca said:

 

Nope, because a waveform is calculated, not "sampled".  It is continuous...

 

You do realize that we don’t listen to waveforms, right?

 

What we listen to are compressed air waves that are three-dimensional physical phenomenon that have a width, height, velocity and intensity. The electrical impulses that come out of a microphone are only a loose, linear approximation of that phenomena.

[psjug]

13 minutes ago, GUTB said:

 

Of course not. Ooshai showed that humans receive and process high frequency sounds, but that is 100% the domain of neuroscience. It could be why people love ribbon tweeters so much.

 

A frequency is an oscillation over a period of time. We can’t percieve tones above 20 kHz. A 20 kHz frequency oscillates at a rate of once per 50 microseconds. If looked at it in the time domain, we shouldn’t be able to hear pulses less than 50 microseconds long — however, we CAN. The study showed we are able to distinguish 10 microseconds, even against background noise.

 

We can hear pulses shorter than 50us because a pulse is not sine wave, and so it has lower frequency content.

 

 OK, I am not going to take it any further than this.  The point is the two types of transients tested have different spectral content in the audible. You are taking a different conclusion away from this than the authors did.  The authors conclusion, stated in the abstract, is:

 

It appears that discrimination of slight changes in the energy spectrum of the two transient signals, especially in the high‐frequency region (8000 Hz and above), underlies the ear's sensitivity to a temporal discontinuity.

[tmtomh]

22 minutes ago, beerandmusic said:

 

what i meant by gaps is that at t1 you have freq x and t2 you have freq y, and you must connect the dots, so the detail between the dots is the gaps that is estimated, calculated, averaged, or whatever terminology you use....and that is where the details and accuracy are lost....between the samples.

 

Your view is understandable - it's the most basic and widely shared misconception about digital sampling. But as @crenca just noted, analogue waveforms are reconstructed via calculation. There are no gaps because nothing unexpected, no "extra detail" is happening "in the gaps."

 

Here's an example: Say you're sampling at the CD rate of 44.1kHz. You've got a sound that's 15kHz. A 44.1k sample-rate system can encode that sound digitally (by taking at least two samples of it), so that the digital to analogue converter on the other end knows exactly what frequency that is supposed to be.

 

Now, there are "gaps" in the sampling - the digital sampling system does not sample the continuous, analogue waveform. Instead, it samples it "only" 44,100 times a second. 

 

But - and this is the key - by definition nothing weird, unexpected or "extra detailed" is happening in those gaps: it's just sound waves. The only way something could happen within the gaps that the sampling system wouldn't "know" about, is if you doubled (or quadrupled, or whatever) the frequency. In a 44.1kHz sample-rate system, 15kHz and 30kHz are encoded identically. The 44.1kHz system does not "know" that the alleged 15k wave actually is a 30k wave that oscillated twice instead of once during the "sample rate gap time." Similarly, a 60k signal also would look the same as a 30k and 15k signal to a 44.1k sample-rate system: 44.1k is not fast enough to sample each of those waves enough times to tell the difference between them. That's why the signal has to be bandwidth (frequency range) limited to a max frequency of 1/2 the sample rate. If it's not filtered, then in the above example the 30k signal would be "aliased," decoded by the DAC as a 15k signal, adding sound in the audible range that was not present in the original signal, aka distortion.

 

That is why I and others keep trying to tell you that sample rate correlates not to "extra detail" but rather to higher frequencies: the extra detail" can only come in the form of higher frequencies.

 

Finally, here are a bunch of links - they're pretty much random, the first half dozen or so that came up when I googled "is a higher sample rate better." They all say the same thing: Higher sample rates allow the capturing of higher frequencies, and they give more "cushion" for the necessary bandwidth limiting filters. Not a single one says that higher sample rates capture the 20-20k range more accurately or in more detail - and two or three of them explicitly state that this is not the case..

https://sonicscoop.com/2016/02/19/the-science-of-sample-rates-when-higher-is-better-and-when-it-isnt/

https://www.soundonsound.com/sound-advice/q-should-i-use-high-sample-rates

http://productionadvice.co.uk/high-sample-rates-make-your-music-sound-worse/

http://carriagehousemusic.com/analog-mixing/does-a-higher-sample-rate-audio-really-mean-better-quality/

https://www.sweetwater.com/insync/7-things-about-sample-rate/

[crenca]

24 minutes ago, GUTB said:

 

You do realize that we don’t listen to waveforms, right?

 

What we listen to are compressed air waves that are three-dimensional physical phenomenon that have a width, height, velocity and intensity. The electrical impulses that come out of a microphone are only a loose, linear approximation of that phenomena.

 

Yep, I have been simplifying, bending, and mangling for didactic purposes.

 

On the other hand, the "compressed air waves that are three-dimensional physical phenomenon that have a width, height, velocity and intensity" that come out of our dynamic/planar drivers based on this recorded waveform are strangely true to "the real thing".  Perhaps it's the linear, pistonic sensory apparatus we all have...

[tmtomh]

54 minutes ago, GUTB said:

 

You do realize that we don’t listen to waveforms, right?

 

What we listen to are compressed air waves that are three-dimensional physical phenomenon that have a width, height, velocity and intensity. The electrical impulses that come out of a microphone are only a loose, linear approximation of that phenomena.

 

Irrelevant to the topic of this thread.

[beerandmusic]

1 hour ago, crenca said:

 

Nope, because a waveform is calculated, not "sampled".  It is continuous...

 

in my thinking there is the real audio, the recorded audio and the playback audio all of which are different and only the real audio is accurate ....

 

either way, i will see if i can dedicate more time to it and see if i can get a better understanding, but my logic tells me that only what is real is accurate, everything that is constructed by man using his tools is very marginal,, and can always be improved on....i believe we have more information today than we had yesterday, and that music doesn't end at 44.1khz.

[Ralf11]

19 minutes ago, beerandmusic said:

 

in my thinking there is the real audio, the recorded audio and the playback audio all of which are different and only the real audio is accurate ....

 

either way, i will see if i can dedicate more time to it and see if i can get a better understanding, but my logic tells me that only what is real is accurate, everything that is constructed by man using his tools is very marginal,, and can always be improved on....i believe we have more information today than we had yesterday, and that music doesn't end at 44.1khz.

 

 

but you cannot hear the real audio

 

all you can do is see the shadows on the walls of the cave, through a glass darkly...

[beerandmusic]

1 hour ago, tmtomh said:

 

Your view is understandable - it's the most basic and widely shared misconception about digital sampling. But as @crenca just noted, analogue waveforms are reconstructed via calculation. There are no gaps because nothing unexpected, no "extra detail" is happening "in the gaps."

 

Here's an example: Say you're sampling at the CD rate of 44.1kHz. You've got a sound that's 15kHz. A 44.1k sample-rate system can encode that sound digitally (by taking at least two samples of it), so that the digital to analogue converter on the other end knows exactly what frequency that is supposed to be.

 

Now, there are "gaps" in the sampling - the digital sampling system does not sample the continuous, analogue waveform. Instead, it samples it "only" 44,100 times a second. 

 

But - and this is the key - by definition nothing weird, unexpected or "extra detailed" is happening in those gaps: it's just sound waves. The only way something could happen within the gaps that the sampling system wouldn't "know" about, is if you doubled (or quadrupled, or whatever) the frequency. In a 44.1kHz sample-rate system, 15kHz and 30kHz are encoded identically. The 44.1kHz system does not "know" that the alleged 15k wave actually is a 30k wave that oscillated twice instead of once during the "sample rate gap time." Similarly, a 60k signal also would look the same as a 30k and 15k signal to a 44.1k sample-rate system: 44.1k is not fast enough to sample each of those waves enough times to tell the difference between them. That's why the signal has to be bandwidth (frequency range) limited to a max frequency of 1/2 the sample rate. If it's not filtered, then in the above example the 30k signal would be "aliased," decoded by the DAC as a 15k signal, adding sound in the audible range that was not present in the original signal, aka distortion.

 

That is why I and others keep trying to tell you that sample rate correlates not to "extra detail" but rather to higher frequencies: the extra detail" can only come in the form of higher frequencies.

 

Finally, here are a bunch of links - they're pretty much random, the first half dozen or so that came up when I googled "is a higher sample rate better." They all say the same thing: Higher sample rates allow the capturing of higher frequencies, and they give more "cushion" for the necessary bandwidth limiting filters. Not a single one says that higher sample rates capture the 20-20k range more accurately or in more detail - and two or three of them explicitly state that this is not the case..

https://sonicscoop.com/2016/02/19/the-science-of-sample-rates-when-higher-is-better-and-when-it-isnt/

https://www.soundonsound.com/sound-advice/q-should-i-use-high-sample-rates

http://productionadvice.co.uk/high-sample-rates-make-your-music-sound-worse/

http://carriagehousemusic.com/analog-mixing/does-a-higher-sample-rate-audio-really-mean-better-quality/

https://www.sweetwater.com/insync/7-things-about-sample-rate/

thanks for sharing this...i will read tomorrow....overlooking it quickly i see it does touch on some areas i need to understand more.

[Spacehound]

7 hours ago, beerandmusic said:

 

I don't think physics can prove God either (wink).

That's not how science works. It's  an UNbelief system.

[Spacehound]

6 hours ago, mansr said:

Not cool.

I said near the beginning he doesn't want the answers. He believes  the "serpent" that lives in his safe space.

[Spacehound]

6 hours ago, beerandmusic said:

what? that i respect you but not kumakuma?

sorry, but even I have my standards....

Do you suggest I have to respect idiot followers that do little but ridicule?  He's far from the worst, but i would put him in bottom 20%.

Those that prefer wilful ignorance to reality  deserve to be ridiculed.

[firedog]

8 minutes ago, Spacehound said:

Those that prefer wilful ignorance to reality  deserve to be ridiculed.

No, ridicule is just mean and diminishes the one ridiculing. At a certain point, they deserve to be ignored. People here have been incredibly patient with beerandmusic, actually. 

After all the time and personal attention that's been lavished on him by others - especially in this thread - he really should take a step back and try to understand the material without his false preconceptions.

If he isn't willing to do that, the best thing to do would be simply not to engage with him. No one should keep wasting their time on a person if he shows he isn't willing to listen to helpful, honest replies. 

[Spacehound]

5 hours ago, GUTB said:

 

Of course not. Ooshai showed that humans receive and process high frequency sounds, but that is 100% the domain of neuroscience. It could be why people love ribbon tweeters so much.

 

A frequency is an oscillation over a period of time. We can’t percieve tones above 20 kHz. A 20 kHz frequency oscillates at a rate of once per 50 microseconds. If looked at it in the time domain, we shouldn’t be able to hear pulses less than 50 microseconds long — however, we CAN. The study showed we are able to distinguish 10 microseconds, even against background noise.

You're like a sponge. Absorbs anything. But real sponges have got an organ that expels crap.

[esldude]

@beerandmusic Read up on superposition principle, and view the graphic illustration.

 

Then watch the Digital Show and Tell video linked early in the thread. Watch carefully.  Watch several times till you understand. 

 

If you find something you disagree with you'll be incorrect.  Study it some more till it makes sense. Or ask pointed questions in the framework of that video and maybe someone can help you see how it works clearly.