[Speederlander]

So how do people know a flac/wav was not converted to that format from a lossy format? I see people being able to make that distinction. What do they use to be able to tell this? Is there a program or tool that makes this determination easy? What should you look for? Thanks for any help.

 

[Rip N' Burn]

I use a program called Tau Analyzer. Basically, if the frequency drastically rolls off before 22kHz, it has come from a lossy source. However, it's not true with certain lossy encoders like Nero's AAC encoder which has no roll off. With MP3's there always will be roll off at higher frequencies.

 

[Speederlander]

This guy?

True Audio Codec Software - Tau Analyzer - Project Description

 

[AtomikPi]

Just look for high frequency rolloff. Around 16khz is 128 mp3, around 19 is 192, around 19/20 is v0, and 20 or so is 320. If you see it going to 21-22khz, it's probably not a transcode (so it is a legitimate rip).

You can use adobe audition, or for a lightweight and simple solution try spectro:
Spectro - Freeware Audio File Analyzer

 

[Speederlander]

So what's going on in this example? I see a faint line across the top that would lead me to believe there was roll off at 16K but it's not really cut off as there is valid information above that point.

 

[Speederlander]

Much more significant example....

 

[Speederlander]

Looking at the foobar spectrum it seems there's just a strong continuous component at that frequency. Weird.... And spectro shows the roll-off at 21.4 so that must be it.

 

[Berlioz]

So basically you're just looking for that horizontal line?

 

[Speederlander]

Quote:

Originally Posted by Berlioz /img/forum/go_quote.gif So basically you're just looking for that horizontal line?

I thought that was what mattered but it seems not. I have no experience looking at these frequency plots. But after thinking about it, it seems that solid line is just strength at that frequency, especially given there is additional information above it.

Note Spectro gave 21.4 when the line was well below that level.

However I notice Spectro will give a roll-off at a lower level when there is clearly information above it. I don't understand that, unless the algorithm that picks the roll-off point is just being fooled.

 

[AtomikPi]

Well I believe that sometimes horizontal lines result from interference that's not removed in mastering. It's especially common in live music. The horizontal line is NOT a cutoff. For reference, here's an example of a 128 kb/s mp3 cutoff (transcoded from flac right now):

I think your file is a legitimate rip because if it were an mp3, there would be a CLEAN BREAK; that fuzziness or whatever is probably just a result of a not great recording.

 

[xnor]

There's a very simple application called Audiochecker (auCDtect.exe). It analyzes the spectrum and gives you a probability value if the file was ripped from a CD or transcoded e.g. from MP3.

But this is just an estimation, so there are exceptions. (bad recording ..)