Audiophile Myths Part 1: MP3 VS FLAC, Cables, Sample Rates, Tube Amps, ETC.

[dheerajin]

If SACD rips shows some 5000odd kbps....and flac shows 1000odd kbps , then surely that song is having more information contained than 320kbps..and this difference is clearly audible for me ...i

 

[Thad E Ginathom]

Whilst the difference might be audible, I think you confuse the number of bytes with the amount of information.

 

[jagdish_p]

Not different actually.. Anything encoded above 44khz will not be audible to human ears. It is same as putting 1 pair of jeans in suitcase. You can only wear that pair of jeans however the whole package is too big, thus wasting space.

Agree that encoding above 44Khz may not have any perceptible difference in hearing. But the sampling rate is for reproducing an analog wave in digitized form which theoretically should be infinite. However this has nothing to do with human ears frequency response. They are entirely different parameters. Even if for persons like me who have lost the ability to hear higher frequencies if presented with a pure 10 K Hz sound wave sampled at 44 K Hz and another at 22K Hz the difference will be audible.

 

[Ambio]

?... But the sampling rate is for reproducing an analog wave in digitized form which theoretically should be infinite......

That would been true about 20 years back but today they discovered that sound doesn't travel as infinite analogue wave to our brain. It get broken to beats and send as electrical pulses which our brain then interpret them as sound. The analogue wave ends as mere beats and electrical pulses in the basilar membrane part of our ears. So the continue analogue wave is no longer relevant. IMHO.

 

[Abhinav_Rakesh]

If SACD rips shows some 5000odd kbps....and flac shows 1000odd kbps , then surely that song is having more information contained than 320kbps..and this difference is clearly audible for me ...i

Clearly audible to you because you have image in your mind that more bits or bytes means more information. More bits or bytes is certainly good for computer but not necessary for human.

 

[Abhinav_Rakesh]

In the end, it is irrelvant, because, as Digital Grand-daddy and co-inventor of MP3, JJ, says... hey, just use FLAC. Where there's a choice.

Isn't it because mp3 is a lossy and flac is a lossless. He may be just want us to preserve our collection in FLAC. So that we can convert our songs to any format from FLAC rather than converting from one lossy to another lossy...

Keeping FLAC files is like keeping a master key. You can make any number of keys with it of different material.

 

[dheerajin]

Clearly audible to you because you have image in your mind that more bits or bytes means more information. More bits or bytes is certainly good for computer but not necessary for human.

Actually for you it works..
For my non audiophile friends and my wife, difference is clearly audible in my system...they even dont have any idea about what rips i am playing first..their reactions are- this song has more punch and more meat than earlier or vice versa...whichever format I play...
Other than my system i did not check the same...so can say , what most of the people say- that i cant find difference between flac and mp3...

 

[Abhinav_Rakesh]

Actually for you it works..
For my non audiophile friends and my wife, difference is clearly audible in my system...they even dont have any idea about what rips i am playing first..their reactions are- this song has more punch and more meat than earlier or vice versa...whichever format I play...
Other than my system i did not check the same...so can say , what most of the people say- that i cant find difference between flac and mp3...

Yes could be.. I know the discussion for this can go forever.. For me I would like to go with scientific way.

 

[prem]

For a 44.1 KHz 16 bit recording, the bit rate is 1411 Kbps. For DSD it's four times that. For Flac, it depends on the amount of compression done. Normally compression is 50 to 60 %. So bit rate should be 50 to 60 % of 1411

 

[Thad E Ginathom]

Isn't it because mp3 is a lossy and flac is a lossless. He may be just want us to preserve our collection in FLAC. So that we can convert our songs to any format from FLAC rather than converting from one lossy to another lossy...

Please excuse me not finding the post. Gearslutz is a huge forum, and it is not easy to find someonething one just happened to be reading a few days ago.

My impression, which could indeed be wrong, is that he meant that, at least at high bit rates, there is no longer any point in lossy compression, because, just use lossless.

Keeping FLAC files is like keeping a master key. You can make any number of keys with it of different material.

Couldn't agree more. For all sorts of reasons, not necessarily scientific (), I would never keep a lossy file if I could have the lossless one. Except for limited-space portable players --- and then I just make a lossy copy of the lossless "master." Yes, duplicate keys: easy

There is a neat MP3 editor which does not decode/recode. It does basic stuff like fades and track splits and even works with cue lists (MP3DirectCut if anybody needs it. Runs in Wine on Linux too). It is still nothing like throwing the power of, eg Audacity at a lossless file.

 

[tcpip]

Most people who claim to have heard the difference between FLAC and MP3 do not specify what the MP3 encoding parameters were. It is pretty certain that highly lossy MP3 will sound audibly different from the FLAC or WAV sources. I have seen such debates where the anti-MP3 gang were using 128 kbits/sec MP3 encoded using some nameless encoder and proudly saying how much worse MP3 is than WAV. Not all MP3 are made equal.

Those who are aware that not all MP3 are the same usually restrict their investigations to the bitrate of the MP3. It is not necessary that bitrate is a guaranteed indicator of quality. It is not necessary that you need to have 320 kbits/sec MP3 to get near-lossless quality. MP3 encoding requires a dozen parameters to be tweaked for quality. It will be good if those who are interested in MP3 quality check out the manpage of the Lame encoder, and see how many parameters can be tuned to control quality.

Manual page of "lame", arguably the most advanced MP3 encoder

I encode my MP3 by using the command "lame --preset extreme". The "--preset extreme" sets appropriate values for each of those dozen encoding parameters, saving me the trouble of figuring out the optimisations individually.

The resultant MP3 I get is VBR, where the bit rate varies from 32 kbits/sec at the silent bit at the start of a track, right up to 320 kbits/sec or higher in busy passages. I have not been able to identify any difference when playing back my FLAC and my MP3 when playing both back through a top quality DAC+headphone amp and a pair of Etymotic ER4PT IEMs. There is absolutely no chance that I can hear any difference if listening through speakers -- a good pair of IEM are much more revealing than any speakers I have encountered.

Therefore, I feel that it doesn't make much sense talking in generic terms about lossless versus lossy. If you talk about "good quality MP3", using parameters at least as good as those which I have used for encoding, then the comparison debate becomes more interesting. I can't hear any differences. I have never seen any account of any double-blind listening tests where similar high-quality MP3 could be distinguished from lossless originals. On the other hand, I have read a few accounts where double-blind listening tests have failed to distinguish differences.

 

[dheerajin]

Ok ...have you listen to stockfish record or laya project..just try to listen to them ..anyway i want you to visit my home to listen to the ssme...mp3,flac and then SACD rip

 

[Ambio]

...Therefore, I feel that it doesn't make much sense talking in generic terms about lossless versus lossy. If you talk about "good quality MP3", using parameters at least as good as those which I have used for encoding, then the comparison debate becomes more interesting. I can't hear any differences. I have never seen any account of any double-blind listening tests where similar high-quality MP3 could be distinguished from lossless originals. On the other hand, I have read a few accounts where double-blind listening tests have failed to distinguish differences.

+1 to that!. However, the best recording quality that I have now is in DSD. Having said that, I cant tell the difference between a very good quality WAV and DSD. I am not saying you can't but my system is not resolving enough to reveal them. I have to accept my limitations.


Here is a direct digital recording in 16 bit 44,1kHz WAV, It was converted to 320 Mp3. The original wav and mp3 then combined and saved as a single wav file.

First 90 seconds is A. And the second 90 seconds is B.

Which one do you think is mp3?

Test files

 

[Thad E Ginathom]

First 90 seconds is A. And the second 90 seconds is B.

Which one do you think is mp3?

They are not identical. I think I can hear that, but after an initial quick listen, I opened the file in Audacity: the wave forms are not the same.

I have to go out for the rest of the day. Please don't give the answer for a couple of days! hyeah:

 

[Ambio]

They are not identical. I think I can hear that, but after an initial quick listen, I opened the file in Audacity: the wave forms are not the same.

I have to go out for the rest of the day. Please don't give the answer for a couple of days! hyeah:

Thank you for being such a good sport. But why stretch ourselves to determine the difference? One is Mp3 and the other is Wav. Both sounds equally horrible or good according to personal taste. If you cant pick a better recording at the first go then the difference is not important for musical enjoyment.

Anyway, will wait for your feedback. I thought of demonstrating the masking effect by combining the 6 preceding notes and see how much Mp3 is going destroy them.

Once again, Thank you.

 

[Thad E Ginathom]

It's not a problem, it's a bit of a challenge. And if I can detect the difference between WAV and 320-MP3, then I need to admit it, and stop quoting authorities about it. Even if one of them is the guy who not only invented MP3 but probably knows more about digital audio than... well, a whole heap of other people.

Spent the afternoon enjoying a great Veena concert, but the evening concert was cancelled, so I'm back home.

I guess I'm going to have to split the files and then try to ABX it. That means seeing if Foobar works in Wine (it does) and Wine audio works for me (it usually doesn't).

In a way, I realise that I have cheated already: I know which has the most information, because I see the waves. I'm supposing that the one with the most information must be the WAV. I've only looked a the waveforms, not the spectrogram, and so it might turn out to be a question of levels, rather than information. And you can probably tell already that I have no real idea of how to read those things for "forensic" investigation.

It's a tough sample too, with the long gaps, but I suppose you chose that for assessing the decay tails.

Anyway...

 

[asliarun]

For a 44.1 KHz 16 bit recording, the bit rate is 1411 Kbps. For DSD it's four times that. For Flac, it depends on the amount of compression done. Normally compression is 50 to 60 %. So bit rate should be 50 to 60 % of 1411

A flac file had to be deflated (like a zip file has to be unzipped) before it can be played back. This is done at runtime by the player. The output will have the same bitrate as the original file (most cases, CD bitstream, but you could also use flac to compress high def music). So it doesn't matter how flac compresses the file, the output is guaranteed to be exactly the same as the original file. Just as we don't expect an unzipped file to get corrupt. So flac will have the same bitrate as the source file.

Now it can be said that the player had to do extra work when deflating the flac at runtime which might cause a weak player to introduce errors. There are also players now that will first deflate the flac in its entirety, put the entire bitstream in memory, and then play audio directly from the in memory bitstream (memory playback).

This is a technically superior solution to even a high end CD transport that has to play CD spindle and extract the bitstream and then stream it back all at the same time.

A small point and sorry for being nitpicky.
CD encodes 16bits at 44.1 kHz. That makes it 44100 * 16 = 660kbps.
DSD only uses 1 bit which is why it samples so furiously. But CD bandwidth is really not that bad compared to DSD.

 

[jls001]

A small point and sorry for being nitpicky.
CD encodes 16bits at 44.1 kHz. That makes it 44100 * 16 = 660kbps.
DSD only uses 1 bit which is why it samples so furiously. But CD bandwidth is really not that bad compared to DSD.

I will be even nitpickier

CD bit rate is 1411 kbps. There seems to a slight error in in your calc.

DSD64 has 64x44100 kHz sampling rate, DSD128 is 128x and DSD256 is 256x.

Addendum: 44100 kHz x 16 bits x 2 channels = 1411.2 kbps

 

[prem]

Asliarun, thanks for the Flac calculation. I didn't know it needed the source bit rate.

Your bit rate calculation for cd is wrong. There are 2 channels in audio. Hence you need to multiply by 2. The exact bit rate is 1411.2 jobs for 44.1 kHz 16 bit

 

[Thad E Ginathom]

I have now gone mad and am writhing on the floor screaming No! No more! I can't stand it!

They are not identical. I think I can hear that, but after an initial quick listen, I opened the file in Audacity: the wave forms are not the same.

This was my first impression: part A sounds less "natural" than part B. I therefore would have said that part B is probably the WAV.

When I opened the file in Audacity, part B seemed to have more information, or maybe just a slightly higher level.

I downloaded Foobar and the ABX Comparator. Yes, my Wine Audio is working. Not normally used: I wouldn't like to say if it affects the sound quality, but anyway, it would be equal/equal. I split the file into two, A and B

I was working with the speakers. I could not even get as far as making a guess as to which was playing, nor could I tell any difference explicitly playing A and B.

I thought some secrets might be revealed with the headphones, but no... same story.

I ran out of patience and began writhing on the floor as mentioned above.



Big difference? No way. Sorry. Yes: my headphones are decent --- but no, my hearing is not, and that could very well make a difference, particularly if anything is there in the higher frequencies. Perhaps someone else would like to have a go with the ABX?

Regardless of anything else I might of said, I would have been chuffed if I could have found a difference. I didn't.

It might be there in the higher freqencies. It might be there in the length of the decay tails ...but who but the obsessively err, obsessive counts microbeats when listening to music? Some mridangists, maybe . I don't think anything felt chopped in either part.