Can you tell the difference between: FLAC + MP3 + WAV ? - Page 2

 

 I remember when someone in high-school bought a hundred or so 1GB USB flash drives I think from china and sold them for $50 or so each and made some profit Lol.

 

 

Reduced filesize, FLAC files should only take up 50-60% of the space of the uncompressed WAV or AIFF  files. And tag support is huge, once you develop a large enough library you will need tags in order to bring sanity to a database.

 

320kbps is overkill by 30% greater filesize for a few very rare instances of higher quality over LAME's VBR 0. Some DJ gear can't process VBR files so 320 may be a necessity when you're spinning at a club or something. Otherwise, VBR is the best choice and the most common recommendation on Hydrogen Audio for practical purposes.

 

I remember in the 90's an "MP3J" wasn't a real "DJ".
 

 

Here's a good rundown on the pros/cons of all the major lossless formats.  There's a nice comparison table at the top of the article.

 

http://wiki.hydrogenaudio.org/index.php?title=Lossless_comparison

I can tell if I'm looking at a spectrogram (-:  I have many flac some going as high as 24-192.  I can't tell much difference, but space isn't an issue so I'll take the highest resolution available.

On all but the worst recordings/masterings, I can easily tell the difference between MP3 and CD. The giveaways are in the treble detail, the nuances (like if you try listening for the exact timbre of an instrument for example, especially cymbals) and in the soundstage. MP3 compression limits soundstage to "left-middle-right" instead of the gradual change that it should be, and that's usually the biggest giveaway.

I still convert my music to 320kbps MP3 for portable use because I'm usually not paying enough attention to care, and having the extra space is nice. However, at home I never listen to mp3s - it's either FLAC or vinyl.

I trust you've done ABX or blind tests.

 

Soundstage, from my experience, is far more reliant on the recording than it is on the recording's bitrate.

If storage is not an issue then this is a good philosophy. The highest resolution available is 24/96, while the highest quality available is a tie between 16/44 and 24/96.

G

Yes, I've done plenty of blind tests with my girlfriend doing the switching for me. Once you know what to look for it becomes really easy.

One of the techniques mp3s use to reduce filesize is called Joint Stereo. What it does is that if there's a sound that's similar in both the left and right channels, even if they're at different volumes, it combines them into a single mono sound. Then when it's played back, that mono sound ends up played back either entirely on one side or in the center, depending on what the encoder's threshold was when compressed.

It's also apparent sometimes when a certain sound has a slight delay between channels, which is normally done when the mastering engineer wants to give you the feel that the sound is coming from everywhere (such as you'd get with an audience clapping). That tiny delay is ignored, and it becomes compressed to a mono sound. Sometimes the delay is large enough for the encoder not to pick it up, but most of the time it does happen.

This Joint Stereo effect is always there regardless of the mp3's bitrate. The only time it's undetectable is if you're using improperly set up speakers (not spread out enough, out of phase, room reflections, etc), or if the track doesn't have a detailed soundstage to begin with (as is the case for a lot of Electronic music).

Exactly
Edited by Manyak - 9/9/11 at 7:33am

I posted this in another thread but since this is also a relevant thread I thought somebody may be interested as well.

 

 

 

EDIT: Done with Mediamonkey > E7 + L7 > Asgard > AD900

Edited by tuahogary - 9/9/11 at 8:26am

 

 

I don't think you quite get how joint stereo works. Each channel retains its own signal intensity, the joint stereo used in mp3 encoding is a simple math trick used to minimize data redundancy. See http://wiki.hydrogenaudio.org/index.php?title=Joint_stereo

I'm not seeing where that link says anything different than I did - combines similar stereo signals into one and then uses phase information to determine if it's on the left, middle, or right.
Edited by Manyak - 9/9/11 at 4:10pm

First there was the assertion that the stereo signal gets mixed down to mono, instead the information is transferred to both a mid-channel and a side channel (and calculating the offsets renders a stereo signal with no loss). It's a more elegant way of representing the same information. Also, the tiny delays are not discarded in favor of the "mono" signal, though some signals do become masked by others because of psychoacoustic modelling - a process that is separate from joint stereo. A proper implementation does not use the Intensity Stereo (IS) process you describe in favor of the Mid-side Stereo (MS) and the results are markedly different.

Edited by anetode - 9/9/11 at 7:50pm