Is this mp3?

[desert_rat]

Originally Posted by RainDawg

Holy shit Five....are those plots taken with EAC at the top of the page the same audio as the ones taken with analfreq at the bottom? The analfreq one is telling me, without a doubt, that this is mp3. The EAC one looks totally different.

I am stunned right now. I'll make sure to recommend people never use EAC for frequnecy analysis ever again if these two plots came from the same audio file! I always use and recommend analfreq anyway, but it looks like people will need to hold onto their old freeware versions.

Thanks again Five and everyone for taking a look at this. Fortunately I have a friend who has this show on 2nd gen cassette. I think I'll bug him for a copy and redo this transfer.

Looks like I'll stick to analfreq and Adobe Audition from here on out.

[feralicious]

Originally Posted by Five

Here's screenshots of the analysis using EAC... keep in mind I've got a 2-second sample, so the sa is "zoomed in" compared to desert_rat's sa. I also attached two screenshots of the fa, 1st one at 512 fft size, and the 2nd at 65536 fft size (maximum). You can see the dropoff a lot better with the fft size at maximum!

Sheesh! Just the discrepancy between the two EAC fas is astounding. Glad I have AnalFreq. And I guess I'll load up CEP or Audition after I reformat and use that as well. Crazy.

[Lou]

Everything but that very first EAC screenshot is mp3.

With EAC you really have to look at the frequency and spectral analysis in conjunction with each other. The one thing that EAC will trip you up with is if there was noise reduction.

[dorrcoq]

My EAC no longer performs FA or SA for some reason, so I have been using Nero Wave Editor for analysis. Anyone have any comments about it?

[Beleaguered]

Someone should mention these discrepencies over at the EAC forums. The EAC developer posts there and might be able to clear up these issues.

The spectral view generated by EAC in the first post says mp3 to me, as I posted above... the program did its job.

Remember that these things could have been sampled from different parts of a song or show.

[dorrcoq]

Originally Posted by New Homebrew

Remember that these things could have been sampled from different parts of a song or show.

I'm a little confused by that statement - wouldn't the MP3 compression be consistent throught the whole song or show?

Originally Posted by dorrcoq

I'm a little confused by that statement - wouldn't the MP3 compression be consistent throught the whole song or show?

Yes but differences in the relative levels in the frequencies below 16kHz can affect your perception or recognition of various patterns. Try sampling many different points in a song or show and see what you get.

You've reached the wrong conclusion here, now that you all have a lot more experience with these, you ought to revisit this issue. EAC is a very good tool for this, if used at all correctly.

Look at the images in this post, I added the bold:

Originally Posted by Five

Here's screenshots of the analysis using EAC... keep in mind I've got a 2-second sample, so the sa is "zoomed in" compared to desert_rat's sa. I also attached two screenshots of the fa, 1st one at 512 fft size, and the 2nd at 65536 fft size (maximum). You can see the dropoff a lot better with the fft size at maximum!

This post (about halfway through the thread) has analyses with a much shorter time frame than the first ones posted; and the FA presumably has only music in it, not all the other stuff that can be on a track. I'm betting the first FA and SA were whole track (clapping, voice, etc.), not just music. EAC looks to be working just fine to me, as a few others have noted here.

I think what you've proved in this thread is:

- don't look at whole-track analyses. When you look at music-only, shorter time frames, the results in any program are much clearer.
- in particular, the first FA posted here was probably for the whole track. That's not going to be effective all of the time, especially in unusual cases.

The comments also show the tendency to read too much into too little info, though now that everyone has more experience with this it probably wouldn't happen again. But people were making all kinds of conclusions based on an odd example and too little info. Wehn you get something that seems a bit off the wall, like the first post here, you really have to say, let's look closer.

This analysis is best done when you actually have the tracks. But if you're going to do it with screenshots, you really need about 3-4 views, over the right time frames, to analyze a track. (Though by the time someone gets done doing those, you'd hope they'd have a good idea themselves, except on the odd cases.) Some info here:

http://wiki.etree.org/index.php?page=SourceAnalysis

As far as the FFT size - that graph at 65k is unreadable to me. It's probably a matter of personal preference; and it may be worth trying a few options; but I find something around 1-2k works for me.

[Five]

EAC is one of my favorite programs on my computer. Its perfect in so many ways it makes us want it to be perfect in every way.

So far it looks to me like EAC is weak for source analysis. The clip being discussed earlier in this thread is a classic 128kbps CBR mp3 source, the most common lossy source. EAC has to be able to spot this and more to be of any use for our purposes. I don't think it was designed for this. I mean no disrespect to the author.

The next "interesting" source I get I'll post pics from EAC, CEP & Audacity side by side. I'd love to be wrong about this.

[ssamadhi97]

EAC is uninteresting for spectral analysis thanks to artifact orgy from hell in treble range whenever there's something of considerable intensity is going on in the lower ranges.

Even if there's actually no treble content in the signal, EAC output still looks like a weaker version of the spectrum has been flipped and superimposed on the actual spectrum. (some kind of weird aliasing effect, I guess)

It's just too misleading (see: this thread)

[Five]

Originally Posted by buzzy

http://wiki.etree.org/index.php?page=SourceAnalysis

This Page Last Changed: Dec 9, 2004 13:46:01