[CryingHorn]

Hey everyone

So I decided to do a comparison between uncompressed audio files and the best way encoded mp3. I'm sorry if something like that was already posted or analysed here before, I just thought I'll share, since it was quite fun to experiment.

I did this experiment with several different types of uncompressed audio music files, but what you see here is “S. Rachmaninov’s - O come and worship” choir, one frame sample. Choir is usually quite sensitive to compressions.

I must admit I could not actually here any difference between flac and mp3-320kbps with my sennheiser hd595 on 0 eq even in a completely silent room. However I’ve noticed some minor difference between highest VBR mp3 and uncompressed source, so at least it says something. With better equipment (...good amp, hd 650 maybe or better) I believe you could here a difference even on 320kbps.

Anyway hope this helps for someone who is interested in this as I've noticed people discuss a lot about quality of FLAC or APE vs mp3.

 

[Anonanimal]

Very cool test. I played around a bit with some test tones on my HD650s and noticed I could still hear the tone between 16000-18000 Hz (albeit faintly), which may explain why some people claim to hear problems in the highs on their MP3s.

 

[CryingHorn]

Hey thanks. That's actually a great idea to test how much it is possible to hear. I found this site where you can test your hearing

High Frequency Response and Hearing Test

I'm not at home right now, so I can't really comment how good or bad this test is. Anyway, I thought since it is actually quite hard to hear a difference after ~16khz I'm going to do another test by reseting with eq everything until ~16khz and see if I can here something on both Flac and mp3 320

 

[CryingHorn]

OK, I did some testing as I planned. I took few uncompressed music samples and killed everything bellow 16khz. I could still here some highs, so I went higher to 17khz - still could here something, then 18khz - almost nothing but gentle noise(...and my head started to hurt) 19-22khz just some strange gentle noise which was getting weaker and weaker, at 22 it was complete silence to me

.

There is also important thing I've noticed that compression is actually not only about disposing all extreame highs, it is also how the whole signal is regenerated and optimized. I've noticed that "framerate" of peaking is much lower with each lower step of compression and THAT actually counts much more then highs which you can barelly spot even when they play alone. So in general it's not that simple I guess.

Anyway I'll stay with my current oppinion that 320 is perfectly enough to listen without any regrets compared to uncompressed even technically you can easilly spot the difference. You can all try this whole experiment for yourself so you could ignore forever all the comments that screams "320kbs sucks, choose FLAC instead..." which is just absurd imho...

Cheers

 

[Zai]

Imho 320 cbr is overkill for a lossy format, v0 is the way to go. Maybe you can add another graph using this preset.

Quote:

"320kbs sucks, choose FLAC instead..."

It sucks because going lossy just means ditching source quality for a factor that isn't music related at all, storage space in this case.

 

[CryingHorn]

Quote:

Originally Posted by Zai /img/forum/go_quote.gif It sucks because going lossy just means ditching source quality for a factor that isn't music related at all, storage space in this case.

Well FLAC in this case is even worse, as it takes even more space.

Quote:

Originally Posted by Zai /img/forum/go_quote.gif Imho 320 cbr is overkill for a lossy format, v0 is the way to go. Maybe you can add another graph using this preset.

Yeah it seems you are right. I tested VBRs that I made with AltoMP3 and AudioCatalyst and they were indeed worse as you can actually sometimes hear it looses to uncompressed sample. But it is strange that I missed to test a VBR from razorlame, which is actually looks almost the same as 320kbs and takes less space. Thanks for pointing that out.

Here are results:

I think I'm also changing my mind about razorlame, as it seems, it is actually a better encoder then Alto after all. With this particular sample I show, I just did the same experiment by lowering everything below 16khz on uncompressed sample. I couldn't hear anything. I figured out that it's not always that razorlame stops at 16khz (and it is true, cause I encoded even more various music to see that until I did). I think it actually has some sort of threshold where it kills the signal where it is actually impossible to hear it. I guess that makes razorlame more intelligent encoder. In any case this is just my guess I could be wrong. Seriously if someone here have studied and knows algorithms of encoding signals it would be great if you would comment on what I just wrote here.

Cheers

 

[paaj]

Quote:

Originally Posted by CryingHorn /img/forum/go_quote.gif Well FLAC in this case is even worse, as it takes even more space.

Not really, it does not sacrifice audio quality for a non-audio benefit and actually is much smaller than the original WAV file it should replace. And you can add info to the file so it really is more useful even with a smaller size.

 

[Zai]

Thanks for the graphs CryingHorn

You have to try the lame "-v0 --vbr-new" preset it you haven't already.

 

[Herbgarden]

I really DON'T KNOW WHAT MOST OF YOU ARE SMOKiNG HERE, but I can hear tones of up to 23,000hz ,
 
AND, EVERY SINGLE LAME/MP3 Codec that is Higher then Version "LAME 3.93.1" has a BUiLT-IN and FORCED Hard-Limiter cut off that will Cut your frequencies at 16-18kHz range, NO MATTER WHAT COMMANDS YOU USE FOR IT in the Consol or otherwise !!!
 
LAME 3.93.1 and Earlier version leaves the WHOLE SPECTRUM of up to 22.05kHz per Channel completely intact !
 
This is where MOST Of the Controversy stems from in this FLAC vs MP3 comparison business,
 
Other then that, you can hear the MOST Difference between the two in recordings of Vocals and Instrumental music, NOT low-quality electronic music in which higher bit-rate cycles per second does not mean much !  
 
What you are looking for is very High-Quality productions which have very smooth and complete soundwaves

 

[Baxide]

  Quote:

Originally Posted by CryingHorn /img/forum/go_quote.gif Well FLAC in this case is even worse, as it takes even more space.

Not really, it does not sacrifice audio quality for a non-audio benefit and actually is much smaller than the original WAV file it should replace. And you can add info to the file so it really is more useful even with a smaller size.

And that is a conundrum. High resolution files (i.e. 192kHz etc) have more data, wich means a larger file size. But FLAC is supposed to be lossless and  just as good as WAV even though it is far smaller in size, which means less data. How can the two scenarios be mutually exclusive but equal in outcome? .  

 

[paaj]

I don't think I understand the point you are trying to make: a 192kHz WAV file is huge, the FLAC conversion has the same quality but is smaller. Its's like ZIP/RAR for audio.

 

[tomb]

 

  Quote:

Originally Posted by CryingHorn /img/forum/go_quote.gif Well FLAC in this case is even worse, as it takes even more space.

Not really, it does not sacrifice audio quality for a non-audio benefit and actually is much smaller than the original WAV file it should replace. And you can add info to the file so it really is more useful even with a smaller size.

And that is a conundrum. High resolution files (i.e. 192kHz etc) have more data, wich means a larger file size. But FLAC is supposed to be lossless and  just as good as WAV even though it is far smaller in size, which means less data. How can the two scenarios be mutually exclusive but equal in outcome? .  

 
 

I don't think I understand the point you are trying to make: a 192kHz WAV file is huge, the FLAC conversion has the same quality but is smaller. Its's like ZIP/RAR for audio.

This ^
 
Paaj is correct.  A 192KHz FLAC will be larger than a 48KHz file, but it will still be smaller than the corresponding 192KHz WAV file.
 
FLAC compresses white space, not anything with data.  Print a page out on a printer and look how much white space is there compared to the text - even a full page of text.  The analogy is valid. 

 

[ShadowVlican]

  I really DON'T KNOW WHAT MOST OF YOU ARE SMOKiNG HERE, but I can hear tones of up to 23,000hz ,
 
AND, EVERY SINGLE LAME/MP3 Codec that is Higher then Version "LAME 3.93.1" has a BUiLT-IN and FORCED Hard-Limiter cut off that will Cut your frequencies at 16-18kHz range, NO MATTER WHAT COMMANDS YOU USE FOR IT in the Consol or otherwise !!!
 
LAME 3.93.1 and Earlier version leaves the WHOLE SPECTRUM of up to 22.05kHz per Channel completely intact !
 
This is where MOST Of the Controversy stems from in this FLAC vs MP3 comparison business,
 
Other then that, you can hear the MOST Difference between the two in recordings of Vocals and Instrumental music, NOT low-quality electronic music in which higher bit-rate cycles per second does not mean much !  
 
What you are looking for is very High-Quality productions which have very smooth and complete soundwaves

not according to this:
http://wiki.hydrogenaudio.org/index.php?title=LAME#Technical_information
 
the recommended switch "-V 0" does not apply a low pass filter
 
it's engineered by the developers to be virtually transparent... personally, i can't hear a difference, though i don't doubt there are a select few who can

 

[paaj]

I'm not willing to go looking and/or doubt. Storage is cheap (at home where it matters) so why not keep it lossless.

...
And you don't have to worry. Now I've got my lossless library stored in ALAC, for portable use my iTunes converts automatically to 256kbs AAC and I'm even considering having my NAS convert the stuff to WAV on the fly for home use, bypassing the whole WAV vs FLAC thing. To have the best of three worlds

 

[Baxide]

   
FLAC compresses white space, not anything with data.  Print a page out on a printer and look how much white space is there compared to the text - even a full page of text.  The analogy is valid. 

There is at least one major hole in your analogy. There are no white spaces in a music file.. It is all 0 or 1.
To get an idea of where errors are bound to crop up in FLAC: a 16 bit WAV file has quite a few compromises in it after conversion from 24 bit (as used at the studio end for editing etc.) Indeterminate boundary bits consisting of alternative 0s and 1s in a WAV file are easily interpreted as a string of just 1s or 0s after conversion to FLAC. Much is made of the FLAC checksum before and after compression. But in reality, it should be the WAV file checksum that should be compared against the compressed FLAC checksum. If those two match, you can then have a better certainty that no bits were lost during the making of that FLAC file.