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

[steve_1979]

This is worth reading.

+1

This is very good explanation of dithering which is written in 'relatively' simple to understand layman's terms and is short enough to be easily digested.

having read the article and surely heard the attached audio files, do you still believe that 16 bits offer full 96dB of dynamic range, Steve? unless you prefer to listen to distortion or hissssss.....

Those audio files are 4 bit not 16 bit. They are there to help explain how dithering works. :read:

 

[steve_1979]

And yes 16 bit has 96dB of dynamic range.

Clicky

 

[jcbrum]

The OP called for "the science, or the lack of" it, to be examined and discussed.

I think this thread is doing that admirably.

JC

 

[davedotco]

It should be pointed out that this is a public forum and members are entitled to post their view whether you agree with it or not.

Its not here for your personal use................

I suppose he had to find out some day ....

I know, I know........ :shame:

It just strikes me as a bit sad just going onto threads that you have no interest in just to try and put people down. They are perfectly entitled to do so, just as we are entitled to consider such a poster as rather a sad individual and a bit of an............. :silenced:

You mean like your two trolling posts on this thread below Dave? Neither of which added to the discussion. People with glass houses...

http://www.whathifi.com/forum/hi-fi/every-now-and-then?page=1

That was last year. My old and completely fruitless attempts to bring a little humour to the forum.

My resolution to be polite effectively bans humour, as no one on here has the slightest sense of it so it is easily miss-understood.

 

[jcbrum]

HeeHee

 

[davedotco]

Glad you have noticed, my new years resolution to be polite is holding so far...

Matter of opinion rather than fact, though, eh.

And I thought I was doing so well too......

I was contemplating a 'No more Mr Nice Guy' approach but thought being polite would be more annoying, to some members at least..... :twisted:

I don't think anyone would pick politeness as your most annoying characteristic

I wondered what had happened to him? It's been making me feel a bit nauseous TBH watching him pick his words so carefully!

Bring back the old Dave :twisted:

Nauseous eh...... :twisted:

See, my new super polite postings are still managing to upset people.

Looks to me that I just can't get it wrong......... :?

 

[the record spot]

Seeing as you posted the same thing twice in that thread Dave, presumably to "get a rise", that's trolling in my view. Just sayin' and all's good now of course.

 

[oldric_naubhoff]

This is worth reading.

+1

This is very good explanation of dithering which is written in 'relatively' simple to understand layman's terms and is short enough to be easily digested.

having read the article and surely heard the attached audio files, do you still believe that 16 bits offer full 96dB of dynamic range, Steve? unless you prefer to listen to distortion or hissssss.....

Those audio files are 4 bit not 16 bit. They are there to help explain how dithering works. :read:

those were 4 bit files, true. but it shows what would happen to SQ if you tried to use too much of available dynamic range from 16 bits format. if you recorded down to 4 bits you'd get exactly that kind of SQ for the quiet parts as you got in those examples on TNT website. if you still don't get it then I'm afraid even another 30 pages of this thread won't help you understd it either.

 

[davedotco]

Seeing as you posted the same thing twice in that thread Dave, presumably to "get a rise", that's trolling in my view. Just sayin' and all's good now of course.

Yep, that was the 'old' me, I've turned over a new leaf.

No trolling, no attempts at humour, no put downs, not that I did a great deal of that anyway.

 

[jcbrum]

. . . those were 4 bit files, true. but it shows what would happen to SQ if you tried to use too much of available dynamic range from 16 bits format. if you recorded down to 4 bits you'd get exactly that kind of SQ for the quiet parts as you got in those examples on TNT website. if you still don't get it then I'm afraid even another 30 pages of this thread won't help you understd it either.

Oldric, your premise is not valid.

This is a well known and accepted explanation which might help you understand. It's a few years old, but concise and accurate.

In particular this bit addresses the points you raise.

The concept of the perfect measurement or of recreating a waveform perfectly may seem like marketing hype. However, in this case it is not. It is in fact the fundamental tenet of the Nyquist-Shannon Sampling Theorem on which the very existence and invention of digital audio is based. From WIKI: “In essence the theorem shows that an analog signal that has been sampled can beperfectly[/b] reconstructed from the samples”. I know there will be some who will disagree with this idea, unfortunately, disagreement is NOT an option. This theorem hasn't been invented to explain how digital audio works, it's the other way around. Digital Audio was invented from the theorem, if you don't believe the theorem then you can't believe in digital audio either!!

JC

 

[CnoEvil]

Seeing as you posted the same thing twice in that thread Dave, presumably to "get a rise", that's trolling in my view. Just sayin' and all's good now of course.

Yep, that was the 'old' me, I've turned over a new leaf.

No trolling, no attempts at humour, no put downs, not that I did a great deal of that anyway.

You mean my spleen is safe for the rest of the year! :bounce:

 

[jcbrum]

p.s. Even if you recorded down to 4 bits, you would still have 72dB of utterly silent dynamic range available, which is much more than you need, anyway !

 

[davedotco]

This is worth reading.

+1

This is very good explanation of dithering which is written in 'relatively' simple to understand layman's terms and is short enough to be easily digested.

having read the article and surely heard the attached audio files, do you still believe that 16 bits offer full 96dB of dynamic range, Steve? unless you prefer to listen to distortion or hissssss.....

Those audio files are 4 bit not 16 bit. They are there to help explain how dithering works. :read:

those were 4 bit files, true. but it shows what would happen to SQ if you tried to use too much of available dynamic range from 16 bits format. if you recorded down to 4 bits you'd get exactly that kind of SQ for the quiet parts as you got in those examples on TNT website. if you still don't get it then I'm afraid even another 30 pages of this thread won't help you understd it either.

Thats fair comment, I can understand that.

What I still do not understand is why the SQ of the 'quiet parts' matter, they are, according to my calculations, some 72db below full output.

Given that even the best classical recordings typically have a dynamic range less than half that, I am struggling to see why that matters.

 

[davedotco]

Seeing as you posted the same thing twice in that thread Dave, presumably to "get a rise", that's trolling in my view. Just sayin' and all's good now of course.

Yep, that was the 'old' me, I've turned over a new leaf.

No trolling, no attempts at humour, no put downs, not that I did a great deal of that anyway.

You mean my spleen is safe for the rest of the year! :bounce:

Absolutely...... :cheers:

No jokes, not even 'Irish' ones!

 

[CnoEvil]

No jokes, not even 'Irish' ones!

We'll see! :shifty:

 

[jcbrum]

Most power amplifiers provide about 40dB of gain.

Few people listen with maximum volume set on their amplifier. Most use a setting no more than half or three quarters of maximum.

That means they are using a gain level of 20 or 30dB, or there abouts.

That means the music they are listening to is contained within the upper 30dB of the available dynamic range of a 16bit recording.

Anything quieter than that, on such a system, is therefore inaudible. It is at the same level as if your amplifier was switched off.

JC

 

[steve_1979]

those were 4 bit files, true. but it shows what would happen to SQ if you tried to use too much of available dynamic range from 16 bits format. if you recorded down to 4 bits you'd get exactly that kind of SQ for the quiet parts as you got in those examples on TNT website. if you still don't get it then I'm afraid even another 30 pages of this thread won't help you understd it either.

The audio files in that link are 4 bit which means that the noise floor is MUCH MUCH louder than it is with 16 bit audio. If you recorded music in 16 bit so that the noise floor was at the same volume relative to the music as those 4 bit files then the music would be so quiet that you couldn't hear it.

Think about it. Do you hear any distortion from CD playback even during the quietest of quiet parts of the music?

No? Well that's because the noise floor is too quiet to hear with 16 bit playback.

 

[steve_1979]

Here's a simple experiment that anyone can try at home with their CD player to test how loud the noise floor of 16 bit audio is.

1. Find a track on one of your CD's where the music gradually fades away to silence at the end of the song.

2. Turn the volume up loud as the music fades away.

3. Does the music sound distorted as it gets quieter? *

4. This shows that even during the sections of music that are recorded at a very low volume the noise floor of a CD is still too quiet to hear.

* Bare in mind that you will get some humming coming from your amplifier when the volume is turned up high, but that's coming from the amplifier and will still be there whether the music is playing or not. The hum from the amplifier is completely unrelated to the noise floor of the CD.

 

[shadders]

. . . those were 4 bit files, true. but it shows what would happen to SQ if you tried to use too much of available dynamic range from 16 bits format. if you recorded down to 4 bits you'd get exactly that kind of SQ for the quiet parts as you got in those examples on TNT website. if you still don't get it then I'm afraid even another 30 pages of this thread won't help you understd it either.

Oldric, your premise is not valid.

This is a well known and accepted explanation which might help you understand. It's a few years old, but concise and accurate.

In particular this bit addresses the points you raise.

The concept of the perfect measurement or of recreating a waveform perfectly may seem like marketing hype. However, in this case it is not. It is in fact the fundamental tenet of the Nyquist-Shannon Sampling Theorem on which the very existence and invention of digital audio is based. From WIKI: “In essence the theorem shows that an analog signal that has been sampled can beperfectly[/b] reconstructed from the samples”. I know there will be some who will disagree with this idea, unfortunately, disagreement is NOT an option. This theorem hasn't been invented to explain how digital audio works, it's the other way around. Digital Audio was invented from the theorem, if you don't believe the theorem then you can't believe in digital audio either!!

JC

Hi ,

Agreed, the theorem does state that the waveform can be perfectly reconstructed - but the filter required is not possible in the analogue domain.

In the digital domain it is possible with a very long delay - using a filter with a Sinc envelope for the impulse response - up to infinity.

This is not practical - so we approximate with analogue filters - hence an error will be introduced.

If we sample at a higher sampling rate - the interpolation of the analogue filter error is reduced.

If we use more bits - then yes - the dynamic range possible cannot be used - but the accuracy of the waveform is increased for the same full scale deflection as a 16bit quantised waveform.

If we can implement this correctly - 24bit 96kHz sampling or higher - then why not - have the best source possible.

Regards,

Shadders.

 

[Phileas]

Another good read:

http://www.soundonsound.com/sos/jun98/articles/digital2.html

 

[jcbrum]

. . . those were 4 bit files, true. but it shows what would happen to SQ if you tried to use too much of available dynamic range from 16 bits format. if you recorded down to 4 bits you'd get exactly that kind of SQ for the quiet parts as you got in those examples on TNT website. if you still don't get it then I'm afraid even another 30 pages of this thread won't help you understd it either.

Oldric, your premise is not valid.

This is a well known and accepted explanation which might help you understand. It's a few years old, but concise and accurate.

In particular this bit addresses the points you raise.

The concept of the perfect measurement or of recreating a waveform perfectly may seem like marketing hype. However, in this case it is not. It is in fact the fundamental tenet of the Nyquist-Shannon Sampling Theorem on which the very existence and invention of digital audio is based. From WIKI: “In essence the theorem shows that an analog signal that has been sampled can beperfectly[/b] reconstructed from the samples”. I know there will be some who will disagree with this idea, unfortunately, disagreement is NOT an option. This theorem hasn't been invented to explain how digital audio works, it's the other way around. Digital Audio was invented from the theorem, if you don't believe the theorem then you can't believe in digital audio either!!

JC

Hi ,

Agreed, the theorem does state that the waveform can be perfectly reconstructed - but the filter required is not possible in the analogue domain.

In the digital domain it is possible with a very long delay - using a filter with a Sinc envelope for the impulse response - up to infinity.

This is not practical - so we approximate with analogue filters - hence an error will be introduced.

If we sample at a higher sampling rate - the interpolation of the analogue filter error is reduced.

If we use more bits - then yes - the dynamic range possible cannot be used - but the accuracy of the waveform is increased for the same full scale deflection as a 16bit quantised waveform.

If we can implement this correctly - 24bit 96kHz sampling or higher - then why not - have the best source possible.

Regards,

Shadders.

I take your point, Shadders, but there are a couple of caveats to be considered.

Anti aliasing filters were in their infancy 30 or 40 years ago when digital systems were introduced to domestic HiFi equipment, and the quality of the filters is one of the things which identified the better equipment. Sony were generally in the vanguard of filter design then, but nowadays the practical problems are well understood and effective filters are readily available now.

Secondly, the requirements of the recording process are different to the playback process, because of the headroom required for the manipulation of the recording during the mastering and production processes. A 32bit system is a practical minimum requirement in that respect, for example. Certainly the use of 96kHz sampling rate, also provides a similar safety margin for recording systems, but theses considerations are not required for playback only, since no manipulations are required. The file is not modified, you merely listen to it playing.

We are therefore left with the situation that 16/44 during playback is adequate for the requirements of human hearing.

JC

 

[andyjm]

. . . those were 4 bit files, true. but it shows what would happen to SQ if you tried to use too much of available dynamic range from 16 bits format. if you recorded down to 4 bits you'd get exactly that kind of SQ for the quiet parts as you got in those examples on TNT website. if you still don't get it then I'm afraid even another 30 pages of this thread won't help you understd it either.

Oldric, your premise is not valid.

This is a well known and accepted explanation which might help you understand. It's a few years old, but concise and accurate.

In particular this bit addresses the points you raise.

The concept of the perfect measurement or of recreating a waveform perfectly may seem like marketing hype. However, in this case it is not. It is in fact the fundamental tenet of the Nyquist-Shannon Sampling Theorem on which the very existence and invention of digital audio is based. From WIKI: “In essence the theorem shows that an analog signal that has been sampled can beperfectly[/b] reconstructed from the samples”. I know there will be some who will disagree with this idea, unfortunately, disagreement is NOT an option. This theorem hasn't been invented to explain how digital audio works, it's the other way around. Digital Audio was invented from the theorem, if you don't believe the theorem then you can't believe in digital audio either!!

JC

Hi ,

Agreed, the theorem does state that the waveform can be perfectly reconstructed - but the filter required is not possible in the analogue domain.

In the digital domain it is possible with a very long delay - using a filter with a Sinc envelope for the impulse response - up to infinity.

This is not practical - so we approximate with analogue filters - hence an error will be introduced.

If we sample at a higher sampling rate - the interpolation of the analogue filter error is reduced.

If we use more bits - then yes - the dynamic range possible cannot be used - but the accuracy of the waveform is increased for the same full scale deflection as a 16bit quantised waveform.

If we can implement this correctly - 24bit 96kHz sampling or higher - then why not - have the best source possible.

Regards,

Shadders.

Shadders,

We had this discussion some pages back. The challenge as an engineer is often where to draw the line - when is enough, enough. 'Lets sample at higher rate, lets have a few more bits' is all well and good, but why? and where do you stop?

Link below is to a 1.5 Giga samples per second, 16 bit DAC. Want to use that? Of course not. But according to you more is better?

http://www.idt.com/products/data-converters/digital-analog-converters-dac/high-speed-jesd204b-serial-interface-dac/dac1658q-high-performance-16-bit-quad-dac-high-common-mode

If you could explain what signal to noise ratio you think is necessary and what maximum frequency you want in your baseband, then we could have a sensible discussion about changing sample rate and bit depth, but until you can explain why 96dB and 20KHz doesn't do it for you, I am afraid it is all just pie in the sky.

 

[altruistic.lemon]

Here's a link to some high res files http://www.audiocheck.net/testtones_highdefinitionaudio.php There's also a section where you can blind test files of different resolutions.

HDDI members, I know you want to educate the clueless, but can you give it a rest for a bit? We are able to make up OUR OWN minds. Cheers, guys.

P.S We also know how to google as well as you do to get our info

 

[Anonymous]

All very well, it's just a suggestion. I would fight for your right to post your pointless drivel on this thread... But unless you are simply trying to troll, then wouldn't it be better to go and enjoy a thread that suits you?

You are the equivalent of a spawn-camper in a FPS (That's a synonym by the way)

I'm not the one trolling pal. I'll post where I like, as I said before. Your veiled and frankly pathetic insults are tiresome now.

 

[steve_1979]

Here's a link to some high res files http://www.audiocheck.net/testtones_highdefinitionaudio.php There's also a section where you can blind test files of different resolutions.

As I've said before. For it to be fair and scientifically valid test you need to convert the high res files to 16/44.1 yourself before doing blind comparisons otherwise there could be other factors such as the mastering which could cause them to sound different.