High resolution audio. The science, or lack of...?

[steve_1979]

...I accept the science and I will use that as my basis for "facts". If the science is proved wrong, then I will change my own stance. I hope I am not presuming too much to say I believe Steve would say the same thing...

+1

'tis the scientific way.

 

[Alec]

and here we go again, some really do need to get a life.

Certainly people who join in and post dead end posts in a thread that doesn't interest them do...

I am a tiny bit sorry, though.

 

[Anonymous]

UNPUBLISHED

 

[steve_1979]

Unpublished

 

[steve_1979]

To answer Mr Lemon. The tests you link are irrelevant.

And I disagree with Steve on this. It is entirely possible that a higher res file will sound different. If you play back a file with 40 KHz and higher frequecies contained in it and one that has been limited to 22 KHz, they "may" sound different. It depends on how your speakers (which are likely unable to reproduce those frequencies, reacts to them.

Same quote again.

This is an interesting point that you make about how different tweeters may react to frequencies played above 22kHz and how it may effect the sound. IME (with my speakers/headphones/ears) including sounds above 22kHz has no noticeable effect on the audiable frequency range but with other peoples equipment (speakers/headphones/ears) maybe it could.

I haven't seen much research done into this but I would guess that as no human can hear frequencies above 22kHz anyway then asking a tweeter to reproduce that information could only add distortion in the frequencies that we can hear.

 

[Alec]

Mr Spot, it is dangerous t refer to "accepted science" in general, as it can be wrong, an that can be known by laypeople who dog a bit, better than someone who has an uderbrad degree in a subject who merely repeats their learning by rote.

I'm just speaking generally, Ihave no idea about Michael Blomqvist theory and don't, for a nanosrcond, mean to imply it is wrong.

 

[johngw]

Forgive me if I'm offending anyone, but I'd wager no-one here has the expertise even to understand the maths behind Nyquist-Shannon

You're wrong. (But I'm not offended 8) )

Thank goodness I didn't bet any cash on it then.

I don't really know how difficult the maths are: Fourier, Poisson etc. Can you give us a sense of what level of mathematical knowledge is needed? I know this is entirely academic, but I like entirely academic stuff.

In my case it was late in 2nd of year of an engineering degree, as part of a module in signal theory and adding on to earlier blocks of multivariable calculus, linear algebra, statistics and transformation theory (incl Fourier as you mention).

Can't say I remember much of any of it though.

 

[matt49]

[... ] that can be known by laypeople who dog a bit [...]

"dog a bit"?

Maybe the laypeople should stick to laying ...

 

[Alec]

[... ] that can be known by laypeople who dog a bit [...]

"dog a bit"?

Maybe the laypeople should stick to laying ...

Oh dear. I'm on me tablet.

 

[johngw]

Mr Spot, it is dangerous t refer to "accepted science" in general, as it can be wrong, an that can be known by laypeople who dog a bit, better than someone who has an uderbrad degree in a subject who merely repeats their learning by rote.

I'm just speaking generally, Ihave no idea about Michael Blomqvist theory and don't, for a nanosrcond, mean to imply it is wrong.

Indeed, and also speaking generally (by no means a swipe at Steve), a little knowledge can be a dangerous thing, particularly when applied too broadly. There are other considerations to digital audio recording/reproduction than the Nyquist sampling theorem. Some of which well summarised by matt49 in his earlier post!

 

[manicm]

If you read the original post, you will see it is nothing to do with boredom, but rather a genuine request for links to anywhere where the theorum has been disproved, as Mr Dalethorn seemed to be claiming was the case.

If so, I would be delighted to see it. If it was verifiable then I would be happy to change my current belief. Science is exactly that. There is no preaching, no "faith". It needs to be empirically repeatable and at any time, if it is disproved, then it is no longer the accepted science.

So, it would be nice if you refrained from the insinuations and insults and maybe posted something positive...Is that at all possible?

I'm not insulting anybody, not at least in the way BenLaw would do *chuckle*, but I disagree. Isn't this the umpteenth time either you or Steve have brought this topic up?? And next week it will be audio cables again (HDMI it is this week if a glance serves my memory?).

This is getting sooooo boring.

 

[johngw]

[...]

Here's a decent reference by the way for those who genuinely want to understand the engineering background to all this.

The big problem I have with the source you're quoted comes in the first paragraph of chapter 3:

"The definition of proper sampling is quite simple. Suppose you sample a continuous signal in some manner. If you can exactly reconstruct the analog signal from the samples, you must have done the sampling properly"

A piano does not produce a continuos signal. It produces signals with transients, decay, that are not at a single steady frequency, but which has a fundamental frequency with harmonics overlaid on that frequency. It's a complex signal. Not the simple one discussed by your reference.

Quite right! But if you read on, they continue and expand from the trivial example, and answer this question.

"Three continuous waveforms are shown in the left-hand column in Fig. 3-5. The corresponding frequency spectra of these signals are displayed in the right-hand column. This should be a familiar concept from you knowledge of electronics; every waveform can be viewed as being composed of sinusoids of varying amplitude and frequency. Later chapters will discuss the frequency domain in detail. (You may want to revisit this discussion after becoming more familiar with frequency spectra)."

In other words. Sound is nothing but waveforms of different frequency, amplitude and phase overlaid on top of each other, adding up to the "complex signal" which is the sum total and what is recorded / what you hear. Which frequencies, and their amplitude, varies with time. But put simply, if you can represent the highest frequency signal (sample freq / 2) you can also represent any other lower frequencies in the waveform.

Incidentally one of the principles of both JPEG for images and MP3 (and AAC) for audio is to break up the signal in "blocks", then convert the signal from time domain (how a signal changes over time) to frequency domain (how much of the signal lies within each given frequency band over a range of frequencies). Storing coefficients for frequencies and their amplitudes takes less space than storing sample points in the time domain. So this is part of the compression - throwing away "inaudible" parts of the spectrum is another. If you compress a JPEG really hard you'll be able to see the block size visually.

Also, think about it the decibel range for a moment. In terms of wattage power it's logarithmic. 16 bit gives you 65,536 different possible values for representing the power at any particular sample moment. If you spread that out over 96dbs you're bound to get a coarsening of the sample at one end of the power range.

[...]

Now can anyone explain to me how our 16 bit sample, giving us 65,536 different possible values can capture our logarithmic power scale ranging from one quarter of a billionth to one without a lot of coarsening at one end of the scale?

I get that decibel is a logarithmic scale. But I don't get why THD would change at the lower or upper end of the amplitude scale. Can you explain this? Obviously if the entire recording was squeezed into the lower 8 bits (for example), and you therefore have a range of 256 possible values, you either get quantization errors or apply dithering in which case you raise the noise floor - is this what you're talking about here?

In fact I think dithering would always be applied in any case from the downsampling from the 32 or 24 bit master to the 16 bit end product..?

 

[steve_1979]

...There are other considerations to digital audio recording/reproduction than the Nyquist sampling theorem...

A 44.1kHz sample rate can accurately reproduce any wave upto 22.05kHz and 16bit audio can reproduce the full dynamic range that is audible to humans. What other considerations are necessary?

(A genuine question. Not just being awkward. )

 

[BenLaw]

If you read the original post, you will see it is nothing to do with boredom, but rather a genuine request for links to anywhere where the theorum has been disproved, as Mr Dalethorn seemed to be claiming was the case.

If so, I would be delighted to see it. If it was verifiable then I would be happy to change my current belief. Science is exactly that. There is no preaching, no "faith". It needs to be empirically repeatable and at any time, if it is disproved, then it is no longer the accepted science.

So, it would be nice if you refrained from the insinuations and insults and maybe posted something positive...Is that at all possible?

I'm not insulting anybody, not at least in the way BenLaw would do *chuckle*

Who said what in the where now?

 

[manicm]

A 44.1kHz sample rate can accurately reproduce any wave upto 22.05kHz and 16bit audio can reproduce the full dynamic range that is audible to humans. What other considerations are necessary?

(A genuine question. Not just being awkward. )

Reading from one of the links that CnoEvil provided the bit depth (16 as you refer to here) has got nothing to do with dynamic range, but the amount of data that can be stored at any interval of time, and which allows for smoother wave form over a given time - hence the potential to sound less sharp or 'digital'. And that is quite understandable to the layman.

 

[manicm]

If you read the original post, you will see it is nothing to do with boredom, but rather a genuine request for links to anywhere where the theorum has been disproved, as Mr Dalethorn seemed to be claiming was the case.

If so, I would be delighted to see it. If it was verifiable then I would be happy to change my current belief. Science is exactly that. There is no preaching, no "faith". It needs to be empirically repeatable and at any time, if it is disproved, then it is no longer the accepted science.

So, it would be nice if you refrained from the insinuations and insults and maybe posted something positive...Is that at all possible?

I'm not insulting anybody, not at least in the way BenLaw would do *chuckle*

Who said what in the where now?

Not in this thread, but let bygones be bygones.

 

[matt49]

...There are other considerations to digital audio recording/reproduction than the Nyquist sampling theorem...

A 44.1kHz sample rate can accurately reproduce any wave upto 22.05kHz and 16bit audio can reproduce the full dynamic range that is audible to humans. What other considerations are necessary?

(A genuine question. Not just being awkward. )

Well, Steve, I've already suggested what other considerations might be in play, e.g. the effect of anti-aliasing filters. Your response to that was that in your experience the effects weren't audible. But we're talking about science, not about your experience.

We could also talk about smoothing filters at the DAC stage and phase distortion intorduced by anti-aliasing filters.

The point is: there absolutely are other considerations than Nyquist-Shannon. I have no idea whether they make an audible difference, but to say that Nyquist-Shannon is the only consideration in play is surely wrong.

:cheers:

Matt

BTW as you know, I'm not casting doubt on the truth of Nyquist-Shannon.

 

[lindsayt]

As far as the limits of human hearing are concerned (maximum dynamic range and maximum/minimum frequencies that we can hear) 16/44.1 resolution data can be used to perfectly reproduce any analogue wave by applying Nyquist-Shannon theory.

You would only need more data than 16/44.1 can supply if our ears could hear a greater dynamic range or a greater frequency range.

This is not true.

http://home.earthlink.net/~dnitzer/4HaasEaton/Decibel.html

0 dbs is the threshold of excellent youthful hearing.

With age the threshold will increase to 10 dbs or 20 dbs. With hearing damge the threshold may be higher.

80 dbs is the threshold of possible hearing damage, if you're exposed to this level for 8 hours per day.

130 dbs is the threshold of pain.

If we're a tiny bit sensible and limit the maximum volume to 120 dbs, that gives us a range limit of human hearing of 100 to 120 dbs for people with non-damaged hearing. That's more than 16/44.1 can provide full stop. And it's way more than 16/44.1 can provide without noticeable amounts of distortion.

 

[Anonymous]

Lighten up.....

 

[fr0g]

A 44.1kHz sample rate can accurately reproduce any wave upto 22.05kHz and 16bit audio can reproduce the full dynamic range that is audible to humans. What other considerations are necessary?

(A genuine question. Not just being awkward. )

Reading from one of the links that CnoEvil provided the bit depth (16 as you refer to here) has got nothing to do with dynamic range, but the amount of data that can be stored at any interval of time, and which allows for smoother wave form over a given time - hence the potential to sound less sharp or 'digital'. And that is quite understandable to the layman.

This shows a complete misunderstanding of the theorum. NOTHING can sound "digital". It is ALL analogue, and so long as the sample rate is more than half the maximum frequency required, then that analogue wave that comes out of your speakers is identical to what went in...

 

[steve_1979]

Lighten up.....

Love it! :rofl:

 

[fr0g]

As far as the limits of human hearing are concerned (maximum dynamic range and maximum/minimum frequencies that we can hear) 16/44.1 resolution data can be used to perfectly reproduce any analogue wave by applying Nyquist-Shannon theory.

You would only need more data than 16/44.1 can supply if our ears could hear a greater dynamic range or a greater frequency range.

This is not true.

http://home.earthlink.net/~dnitzer/4HaasEaton/Decibel.html

0 dbs is the threshold of excellent youthful hearing.

With age the threshold will increase to 10 dbs or 20 dbs. With hearing damge the threshold may be higher.

80 dbs is the threshold of possible hearing damage, if you're exposed to this level for 8 hours per day.

130 dbs is the threshold of pain.

If we're a tiny bit sensible and limit the maximum volume to 120 dbs, that gives us a range limit of human hearing of 100 to 120 dbs for people with non-damaged hearing. That's more than 16/44.1 can provide full stop. And it's way more than 16/44.1 can provide without noticeable amounts of distortion.

While your theory is ok, in practice, in a normal listening room, it's irrelevant. I know for a fact that in my house (in a very quiet area) that the noise floor is at 30 to 40 dB when the room is "silent". CD quality is more than enough, and way more than the DR of vinyl.

However I would be quite happy if they upped the standard to 24 bit. It wouldn't hurt and would help with digital volume control issues.

As for Nyquist-Shannon....it has nothing with that to do...44,1 covers any human need...

 

[manicm]

A 44.1kHz sample rate can accurately reproduce any wave upto 22.05kHz and 16bit audio can reproduce the full dynamic range that is audible to humans. What other considerations are necessary?

(A genuine question. Not just being awkward. )

Reading from one of the links that CnoEvil provided the bit depth (16 as you refer to here) has got nothing to do with dynamic range, but the amount of data that can be stored at any interval of time, and which allows for smoother wave form over a given time - hence the potential to sound less sharp or 'digital'. And that is quite understandable to the layman.

This shows a complete misunderstanding of the theorum. NOTHING can sound "digital". It is ALL analogue, and so long as the sample rate is more than half the maximum frequency required, then that analogue wave that comes out of your speakers is identical to what went in...

I wasn't talking about the theorem, I was talking about the bit depth - which you conveniently ignore in your response. In which case it's not a 'complete misunderstanding'.

 

[andyjm]

A 44.1kHz sample rate can accurately reproduce any wave upto 22.05kHz and 16bit audio can reproduce the full dynamic range that is audible to humans. What other considerations are necessary?

(A genuine question. Not just being awkward. )

Reading from one of the links that CnoEvil provided the bit depth (16 as you refer to here) has got nothing to do with dynamic range, but the amount of data that can be stored at any interval of time, and which allows for smoother wave form over a given time - hence the potential to sound less sharp or 'digital'. And that is quite understandable to the layman.

This shows a complete misunderstanding of the theorum. NOTHING can sound "digital". It is ALL analogue, and so long as the sample rate is more than half the maximum frequency required, then that analogue wave that comes out of your speakers is identical to what went in...

I wasn't talking about the theorem, I was talking about the bit depth - which you conveniently ignore in your response. In which case it's not a 'complete misunderstanding'.

Manic, I would recommend a bit of googling on the subject as your understanding is definitely adrift here.

Bit depth drives dynamic range.

Sample rate drives maximum frequency that can be correctly sampled.

 

[andyjm]

Could I respectfully request that anyone reading/commenting about 16/44.1 CD quality audio reads about and understands Nyquist-Shannon theory first?

Nyquist-Shannon theory is used to to perfectly reproduce any analogue wave within the limits of human hearing from 16/44.1 digital information.

in general you've got two processes occuring simultaneously when you digitize an analog signal; sampling and quantizing. sampling is based on N-S theorem and it is true that sampling is error-less up to its limit, ie. sampling half band frequency. sampling in analog to digital conversion is responsible for digitazing the frequency spectrum of the signal. however, quantizing is used to capture level of the signal being digitized and quantizing, by definition, will never be error-less, simply because you've only got a limited amount of volume levels which you can apply to analog signal during quantization. as Lindsayt rightly points out; 16 bit resolution gives you 65 536 (or 2^16) discrete levels, 24 bit resolution givvevs you 16 777 216 discrete volume levels, 32 bit resolution nearly 4.3 bilion discrete volume levels. aliasing is a by-product of quantizing process and the reason why it occurs is exactly the fact that you can't fully quantize analog signal.

but on the other hand one might argue that even for 16 bit resolution the number of discrete volume levels is high enough, provided you only use a limited DR within 16bit full DR range, that you could capture the analog signal very faithfully - in effect indistinguishably from the analog input signal. still, the reality is that more bits of resolution are better for the purpose of faithfully capturing of the analog signal.

Quantisation has nothing to do with aliasing,

Aliasing occurs when an input signal is sampled at less than twice its maximum frequency. Frequencies above (sample rate)/2 appear in the reconstructed signal as lower frequencies or aliases.

The higher frequencies act as if they have been reflected around (sample rate/2). So a 22KHz signal sampled at 40KHz shows up as 18KHz in the reconstructed signal, 30KHz sampled at 40KHz shows up as a10KHz alias. Strange effect, but there it is.

This is a avoided by having an anti aliasing filter before the sampling process to ensure no signal above (sample rate/2) is sampled.