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

[John Duncan]

If anyone is interested, I did the test I mentioned earlier.

I used the free test file from Linn (a 24/192 track) and opened it in Audacity.

I resampled to 16 bits and 44.1 and saved another track.

I opened both and inverted one of them. I then combined them to a new 192 Khz flac file.

I then analyzed what was left...

There is a bit of noise at -84dB and some very high frequency at around -70 dB.

I played the file through the speakers...absolute silence...

I fin this very interesting. Could you do the same with some lesser bit rates like 320/256/192/128?

 

[CJSF]

I find most of what is in this thread pointless

Careful, some on here will tell you not to bother posting and buzz off somewhere else.

Yep, but it dont change what I believe as some one who likes listening to music on a hifi system that one has worked up to sound the way one wants it to sound using only a well tuned pair of ears . . . I dont need a whole load of complication in facts, figures and theories to tell me what I can hear perfectly well . . . 8)

:? . . . Not very scientific, to tell someone to 'buzz off' . . . :cheers:

CJSF

 

[fr0g]

If anyone is interested, I did the test I mentioned earlier.

I used the free test file from Linn (a 24/192 track) and opened it in Audacity.

I resampled to 16 bits and 44.1 and saved another track.

I opened both and inverted one of them. I then combined them to a new 192 Khz flac file.

I then analyzed what was left...

There is a bit of noise at -84dB and some very high frequency at around -70 dB.

I played the file through the speakers...absolute silence...

I fin this very interesting. Could you do the same with some lesser bit rates like 320/256/192/128?

I'm sure I can, but I am guessing there will be sound in that case. One of the points of lossy formats is it drops sounds that will be inaudible due to other sounds that are occuring at the same time...in other words, the test in that case is flawed. You may very well find you hear something, but it would be sounds that are masked while the music is actually playing. So we are back to ABX for that I think...

 

[John Duncan]

I don't understand why it would be flawed? If there is something audible missing from a 320k file compared to eg 16/44 then there is something audible missing full stop, no? I mean I know I can't ABX 320 and full fat, but why throw anything away if you have the space for it?

 

[fr0g]

I don't understand why it would be flawed? If there is something audible missing from a 320k file compared to eg 16/44 then there is something audible missing full stop, no? I mean I know I can't ABX 320 and full fat, but why throw anything away if you have the space for it?

Like I said. One function of a lossy format is that some sounds whilst audible on their own, will not be if they are playing at the same time as certain other sounds.

Lossy codecs "do" remove information. But they attempt to only remove that which you can't hear. If you listen to what has been removed, then you may very well hear it.

A bit like taking the sound of a clock ticking in the kitchen whislt you listen to a Metallica chorus. Or the sound of a DVD spinning whilst you watch Jurassic Park...

Silence the room and you'll hear both the clock and the spinning DVD.

 

[fr0g]

Ok. Here is an MP3 in low bitrate of the Linn 24/192 test track (freely available from the LinnMusic web site)

http://www11.zippyshare.com/v/77734118/file.html

Now here is the file I made of the differences between that and a 16/44.1 version I created in Audacity with the "resample" function...

This is at 192 KHz sample rate and 16 bit depth.

http://www11.zippyshare.com/v/68836622/file.html

What do you hear?

 

[John Duncan]

I guess my point - and, if a may extrapolate, the point of your original post - is to get to the truth of audibility of loss in various bit depths and sampling rates. If people choose to then ignore that loss (or gain, depending on how you look at it) because it's inaudible (or audible) *to them* then fair enough?

 

[Crossie]

Thanks fr0g for this demo. :clap:

I await the responses of those who can hear differences. :?

 

[busb]

Unfortunately, you can't null out a lossy with a lossless original - all the information not removed in the lossy conversion will be present after nulling. The same applies to comparing one lossy file to another at differing rates. All that can be done is to listen, blind or otherwise.

 

[John Duncan]

I don't understand why it would be flawed? If there is something audible missing from a 320k file compared to eg 16/44 then there is something audible missing full stop, no? I mean I know I can't ABX 320 and full fat, but why throw anything away if you have the space for it?

Like I said. One function of a lossy format is that some sounds whilst audible on their own, will not be if they are playing at the same time as certain other sounds.

Lossy codecs "do" remove information. But they attempt to only remove that which you can't hear. If you listen to what has been removed, then you may very well hear it.

A bit like taking the sound of a clock ticking in the kitchen whislt you listen to a Metallica chorus. Or the sound of a DVD spinning whilst you watch Jurassic Park...

Silence the room and you'll hear both the clock and the spinning DVD.

Yes I get the theory, but it would be an interesting demo, right? I mean we've proved above (for one track, admittedly, which means not proved, but let's run with this) that the difference between 24/96 is *actual* silence, let's see what the difference is in other formats?

 

[John Duncan]

Duplicate

 

[fr0g]

I guess my point - and, if a may extrapolate, the point of your original post - is to get to the truth of audibility of loss in various bit depths and sampling rates. If people choose to then ignore that loss (or gain, depending on how you look at it) because it's inaudible (or audible) *to them* then fair enough?

Not really. This thread was to acertain the value of HD music. I personally find 320 Kbps MP3 enough, but that needs ABX to prove to oneself.

The validity of not needing more than CD quality is a scientific fact, backed up quite nicely I think by real-world examples that don't require ABX.

Ultimately I hate "the loudness wars" and can spot a CD that has been compressed to clipping a mile off. It's horrible and offensive. I simply want good quality audio, and I simply want it in a sensible format. I also do not want to pay through the nose for HD music when it offers no meaningful advantage for playback.

I want to see audiophile mastered CDs and FLAC/MP3 downloads. Is it asking too much?

 

[steve_1979]

I guess my point - and, if a may extrapolate, the point of your original post - is to get to the truth of audibility of loss in various bit depths and sampling rates. If people choose to then ignore that loss (or gain, depending on how you look at it) because it's inaudible (or audible) *to them* then fair enough?

I agree that where 'lossy' codecs are concerned such as MP3 or AAC your comment is correct.* But let's not forget that 16/44.1 is a 'lossless' codec so you're not losing anything here.

* However with a high enough bit rate (256 VBR or higher) then 'lossy' codecs will still sound the same even though some of the sounds are missing. These removed sounds would have been inaudible anyway because of louder sounds playing at the same time.

In a silent room I can hear the noise of my breathing but at a Iron Maiden concert with a 100kW sound system playing at full volume I can't hear my breathing but that noise is still there it's just not audible. If I held my breath the sound of my breathing would be removed but nobody would be able to hear any difference because the 100kW Iron Maiden sound system was covering it up anyway.

 

[fr0g]

I don't understand why it would be flawed? If there is something audible missing from a 320k file compared to eg 16/44 then there is something audible missing full stop, no? I mean I know I can't ABX 320 and full fat, but why throw anything away if you have the space for it?

Like I said. One function of a lossy format is that some sounds whilst audible on their own, will not be if they are playing at the same time as certain other sounds.

Lossy codecs "do" remove information. But they attempt to only remove that which you can't hear. If you listen to what has been removed, then you may very well hear it.

A bit like taking the sound of a clock ticking in the kitchen whislt you listen to a Metallica chorus. Or the sound of a DVD spinning whilst you watch Jurassic Park...

Silence the room and you'll hear both the clock and the spinning DVD.

Yes I get the theory, but it would be an interesting demo, right? I mean we've proved above (for one track, admittedly, which means not proved, but let's run with this) that the difference between 24/96 is *actual* silence, let's see what the difference is in other formats?

The problem is, the "difference" will be audible. But it proves nothing. The only way of demonstrating the lack of real-world difference in that case is ABX.

 

[altruistic.lemon]

Ok. Here is an MP3 in low bitrate of the Linn 24/192 test track (freely available from the LinnMusic web site)

http://www11.zippyshare.com/v/77734118/file.html

Now here is the file I made of the differences between that and a 16/44.1 version I created in Audacity with the "resample" function...

This is at 192 KHz sample rate and 16 bit depth.

http://www11.zippyshare.com/v/68836622/file.html

What do you hear?

I don't hear the original files, for a start. What is it with you, you bang on about tests then give us a flawed one!

Doesn't prove anything by the way, we only know what you have claimed to have done, and, admit it, it is convenient neither to name the bitrate of the mp3 nor to provide the undoctored originals. 2/10, prof!

What do you do for a crust by the way?

 

[steve_1979]

Yes I get the theory, but it would be an interesting demo, right? I mean we've proved above (for one track, admittedly, which means not proved, but let's run with this) that the difference between 24/96 is *actual* silence, let's see what the difference is in other formats?

I agree that it's an interesting (if pointless) experiment to see exactly what's removed during a 'lossy' compression. Would you mind posting an MP3 comparison fr0g? I'd do it myself but I'm not at home and using a tablet at the moment.

 

[fr0g]

Ok. Here is an MP3 in low bitrate of the Linn 24/192 test track (freely available from the LinnMusic web site)

http://www11.zippyshare.com/v/77734118/file.html

Now here is the file I made of the differences between that and a 16/44.1 version I created in Audacity with the "resample" function...

This is at 192 KHz sample rate and 16 bit depth.

http://www11.zippyshare.com/v/68836622/file.html

What do you hear?

I don't hear the original files, for a start. What is it with you, you bang on about tests then give us a flawed one!

I used to like you Lemon. But these days you are too illogical and seem to have an agenda.

The original file is available from the Linn website.

If you really want, I can upload it and my 16/44.1 conversion, but it is irrelevant in this instance as you can easily do the same test yourself. The files I included were a small MP3 conversion to demonstrate the track used, and a file that contains only the difference of the HD version and my downsampled version.

 

[altruistic.lemon]

I don't understand why it would be flawed? If there is something audible missing from a 320k file compared to eg 16/44 then there is something audible missing full stop, no? I mean I know I can't ABX 320 and full fat, but why throw anything away if you have the space for it?

Like I said. One function of a lossy format is that some sounds whilst audible on their own, will not be if they are playing at the same time as certain other sounds.

Lossy codecs "do" remove information. But they attempt to only remove that which you can't hear. If you listen to what has been removed, then you may very well hear it.

A bit like taking the sound of a clock ticking in the kitchen whislt you listen to a Metallica chorus. Or the sound of a DVD spinning whilst you watch Jurassic Park...

Silence the room and you'll hear both the clock and the spinning DVD.

Yes I get the theory, but it would be an interesting demo, right? I mean we've proved above (for one track, admittedly, which means not proved, but let's run with this) that the difference between 24/96 is *actual* silence, let's see what the difference is in other formats?

The problem is, the "difference" will be audible. But it proves nothing. The only way of demonstrating the lack of real-world difference in that case is ABX.

Come on! proves nothing because you can hear the difference??

 

[fr0g]

I don't understand why it would be flawed? If there is something audible missing from a 320k file compared to eg 16/44 then there is something audible missing full stop, no? I mean I know I can't ABX 320 and full fat, but why throw anything away if you have the space for it?

Like I said. One function of a lossy format is that some sounds whilst audible on their own, will not be if they are playing at the same time as certain other sounds.

Lossy codecs "do" remove information. But they attempt to only remove that which you can't hear. If you listen to what has been removed, then you may very well hear it.

A bit like taking the sound of a clock ticking in the kitchen whislt you listen to a Metallica chorus. Or the sound of a DVD spinning whilst you watch Jurassic Park...

Silence the room and you'll hear both the clock and the spinning DVD.

Yes I get the theory, but it would be an interesting demo, right? I mean we've proved above (for one track, admittedly, which means not proved, but let's run with this) that the difference between 24/96 is *actual* silence, let's see what the difference is in other formats?

The problem is, the "difference" will be audible. But it proves nothing. The only way of demonstrating the lack of real-world difference in that case is ABX.

Come on! proves nothing because you can hear the difference??

Read the previous posts. And stop being so illogical. MP3 (and other lossy formats) throw away sounds you should not be able to hear at the same time.

If I use a drill at the same time as someone is talking in another room then you won't hear them talking.

An MP3 will simply remove the sound of the talking (at a very simplistic level).

However, if you remove the drill sound and leave only the talking, it will be quiet, but you will hear it. So it's quite probable that there will be sounds on a track that only shows the differences in the case of lossy formats...

As such it proves nothing to do that test with an MP3...Where ABX comes in. The differences ARE there. Nobody would argue that. But are they audible whilst listening to the music?

 

[John Duncan]

I agree that it's an interesting (if pointless) experiment

And there was me thinking you were the scientist around here.

 

[altruistic.lemon]

I don't understand why it would be flawed? If there is something audible missing from a 320k file compared to eg 16/44 then there is something audible missing full stop, no? I mean I know I can't ABX 320 and full fat, but why throw anything away if you have the space for it?

Like I said. One function of a lossy format is that some sounds whilst audible on their own, will not be if they are playing at the same time as certain other sounds.

Lossy codecs "do" remove information. But they attempt to only remove that which you can't hear. If you listen to what has been removed, then you may very well hear it.

A bit like taking the sound of a clock ticking in the kitchen whislt you listen to a Metallica chorus. Or the sound of a DVD spinning whilst you watch Jurassic Park...

Silence the room and you'll hear both the clock and the spinning DVD.

Yes I get the theory, but it would be an interesting demo, right? I mean we've proved above (for one track, admittedly, which means not proved, but let's run with this) that the difference between 24/96 is *actual* silence, let's see what the difference is in other formats?

The problem is, the "difference" will be audible. But it proves nothing. The only way of demonstrating the lack of real-world difference in that case is ABX.

Come on! proves nothing because you can hear the difference??

Read the previous posts. And stop being so illogical. MP3 (and other lossy formats) throw away sounds you should not be able to hear at the same time.

If I use a drill at the same time as someone is talking in another room then you won't hear them talking.

An MP3 will simply remove the sound of the talking (at a very simplistic level).

However, if you remove the drill sound and leave only the talking, it will be quiet, but you will hear it. So it's quite probable that there will be sounds on a track that only shows the differences in the case of lossy formats...

As such it proves nothing to do that test with an MP3...Where ABX comes in. The differences ARE there. Nobody would argue that. But are they audible whilst listening to the music?

And what bitrate did you use for your mp3 files?

Good to see you accepting using your ears instead of having blind faith in silence, but

 

[steve_1979]

I agree that it's an interesting (if pointless) experiment

And there was me thinking you were the scientist around here.

It is a pointless test. As fr0g says we know that sounds are removed during a 'lossy' MP3 or AAC compression. But as has been said (several times) the parts that are removed are the parts that you can't hear over the louder parts of the music so the end sound still sounds the same to our ears. On their own without the rest of the music playing over them you will probably be able to hear these removed parts.

But I do agree that it'd be interesting to see (I mean hear) exactly what gets removed.

 

[John Duncan]

It is a pointless test.

There is no such thing as a pointless test.

 

[fr0g]

I don't understand why it would be flawed? If there is something audible missing from a 320k file compared to eg 16/44 then there is something audible missing full stop, no? I mean I know I can't ABX 320 and full fat, but why throw anything away if you have the space for it?

Like I said. One function of a lossy format is that some sounds whilst audible on their own, will not be if they are playing at the same time as certain other sounds.

Lossy codecs "do" remove information. But they attempt to only remove that which you can't hear. If you listen to what has been removed, then you may very well hear it.

A bit like taking the sound of a clock ticking in the kitchen whislt you listen to a Metallica chorus. Or the sound of a DVD spinning whilst you watch Jurassic Park...

Silence the room and you'll hear both the clock and the spinning DVD.

Yes I get the theory, but it would be an interesting demo, right? I mean we've proved above (for one track, admittedly, which means not proved, but let's run with this) that the difference between 24/96 is *actual* silence, let's see what the difference is in other formats?

The problem is, the "difference" will be audible. But it proves nothing. The only way of demonstrating the lack of real-world difference in that case is ABX.

Come on! proves nothing because you can hear the difference??

Read the previous posts. And stop being so illogical. MP3 (and other lossy formats) throw away sounds you should not be able to hear at the same time.

If I use a drill at the same time as someone is talking in another room then you won't hear them talking.

An MP3 will simply remove the sound of the talking (at a very simplistic level).

However, if you remove the drill sound and leave only the talking, it will be quiet, but you will hear it. So it's quite probable that there will be sounds on a track that only shows the differences in the case of lossy formats...

As such it proves nothing to do that test with an MP3...Where ABX comes in. The differences ARE there. Nobody would argue that. But are they audible whilst listening to the music?

And what bitrate did you use for your mp3 files?

Good to see you accepting using your ears instead of having blind faith in silence, but

Which MP3? You are missing the point entirely Mr Lemon.

The MP3 I linked was only there as an indicator to what music I was doing. I haven't done the tests with MP3, nor will be doing as , as has been mentioned, they are pointless.

THIS thread is nothing to do with arguments over the audibility or not of differences between lossy codecs. The test I did simply showed that the difference between a 192 Kbs encoded "studio master", and the same file downsampled to CD quality, was,.to all intents and purposes...nothing...unless you count some inaudible noise from 15 KHz to 20 KHz at 84 dB below the peak volume and some noise from 80 KHz at -70 -70dB.

 

[BenLaw]

I don't understand why it would be flawed? If there is something audible missing from a 320k file compared to eg 16/44 then there is something audible missing full stop, no? I mean I know I can't ABX 320 and full fat, but why throw anything away if you have the space for it?

Like I said. One function of a lossy format is that some sounds whilst audible on their own, will not be if they are playing at the same time as certain other sounds.

Lossy codecs "do" remove information. But they attempt to only remove that which you can't hear. If you listen to what has been removed, then you may very well hear it.

A bit like taking the sound of a clock ticking in the kitchen whislt you listen to a Metallica chorus. Or the sound of a DVD spinning whilst you watch Jurassic Park...

Silence the room and you'll hear both the clock and the spinning DVD.

Yes I get the theory, but it would be an interesting demo, right? I mean we've proved above (for one track, admittedly, which means not proved, but let's run with this) that the difference between 24/96 is *actual* silence, let's see what the difference is in other formats?

The problem is, the "difference" will be audible. But it proves nothing. The only way of demonstrating the lack of real-world difference in that case is ABX.

Come on! proves nothing because you can hear the difference??

Read the previous posts. And stop being so illogical. MP3 (and other lossy formats) throw away sounds you should not be able to hear at the same time.

If I use a drill at the same time as someone is talking in another room then you won't hear them talking.

An MP3 will simply remove the sound of the talking (at a very simplistic level).

However, if you remove the drill sound and leave only the talking, it will be quiet, but you will hear it. So it's quite probable that there will be sounds on a track that only shows the differences in the case of lossy formats...

As such it proves nothing to do that test with an MP3...Where ABX comes in. The differences ARE there. Nobody would argue that. But are they audible whilst listening to the music?

And what bitrate did you use for your mp3 files?

Good to see you accepting using your ears instead of having blind faith in silence, but

Wouldn't that be deaf faith in silence?