hi

i trade bootlegs and don't like receiving mp3 sourced bootlegs. Is there a sure fire way i can analyze the CD to tell if it is mp3 sourced. For instance, if i extracted one of the tracks to wav and did a frequency analysis in Cool Edit Pro 2 would that show the frequency clipping etc. Any help is appreciated.
stuart

This is a good question. I presume that quite a few black market CD-Rs might originate from MP3s or similar. I remember this being discussed a long time ago (at r3mix) and best suggestion back then was to look for lowpassing. MP3 encoders are usually very aggressively cutting above 16khz freqs, so that could give you hint about the source of the CD, but nothing definite of course.

I am also very interested if there would be any methods for this?

With a little practice, it's quite easy to see if the source was the original CD or an MP3. Look at various MP3 and WAV files in CEP (go to "View" and activate "Spectral view") and you'll see what i mean.

The frequency analysis is also an accurate way of telling if a CD is mp3 sourced: any original disc always reaches fully 22050Hz without cutting the 20+Khz area.

yep, spectral view is the way to go. it's the only way to tell mp3s from md recordings anyway. listening to the channel difference might help in some cases as well..

just some thoughts. i am considering to write up a page on this subject during the upcoming holidays..

This is true for MP3 indeed!
BTW, OGG files don't have this aggressive cutting :-)

All right, but what about audio which was recorded a few decades ago, stored in a poorly controlled environment, and transferred from the masters to the digital domain... but which, in order to provide an ear-friendly listening experience has undergone severe restoration? This isn't a non-issue; I do audio restoration (among many other things) and have on occasion used as many as thirty discreet passes to cut "noise" while preserving as much of the "natural" sound of a recording as possible. Not to mention the time involved in making sure all glitches were noted, and any stretched tape accounted for.

Since the source I have in mind had very little in the way of surviving high-frequencies anyway, after all the restoration there was a visible cutoff around 12 KHz. Let me emphasize, this was the only surviving copy of the recording in the first place, and it had been thrown in a box to be forgotten for the next thirty-some years by the guy who recorded it! Now, I spent a good twenty hours working on that audio... and when I had finished - and properly treed the show, with every step of the lineage documented! - one fellow irately wrote accusing me of treeing an MP3 source, "because I can't see any high frequencies in Cool Edit." (!)

What was I to do? I sent the @#!^ a copy of the original, unrestored show so he could compare... and he accused me of doing something to degrade the sound of that one, because he couldn't figure out how anything that sounded that bad could sound as good as what I sent him originally.

The point is, yes, the spectrogram will give you good results - sometimes. But it's better to know your source, and trust the person who mastered it.

- M.

EDIT: For anyone interested in trading such unreleased recordings - and being certain of their lineage! - there are several avenues. Some of them can't be mentioned here, but one which can is eTree. This group specializes in trading lossless audio of taper-friendly bands (Grateful Dead, Phish, Medeski, Martin & Wood... heck, for a better list, check their database of what's in circulation). eTree shows will generally be compressed as Shorten (SHN) lossless audio. Shorten is hardly the favorite on this board due to being older and not as effecient as modern lossless codecs, but it is quite serviceable. There is also a growing movement toward the use of FLAC. Trades are generally done via FTP, or by sending lossless files on CD-R through the mail.

You can sometimes tell if a recording is MP3 sourced, and it's not hard to take a look since you can use the analysis tools already built into Exact Audio Copy. For a discussion with both frequency analysis (db vs hz) and spectral analysis (hz vs. time), and how to use EAC for it, at this link:
http://mysite.verizon.net/vze1k03a/fa/fa.htm

It's not accurate to say that you can always tell, at least with anything other than a modern studio recording, since you don't have the original recording and don't know what the profile looked like before. Interestingly, the best encoder (LAME) is the easiest to identify because the frequency cutoff in the spectral analysis is so consistent, the haircut is obvious. Some of the others are actually harder to spot. See the link for illustrations.

It's not accurate to expect every recording to go up to 22khz, as there are a lot of factors that go into that - especially for live recordings, like for example what kind of miking was used?

And for older vintage stuff, it can be hard to do. Bootlegs from the 70s or 80s, or music taped off FM radio, may not have had high freqencies to start with. And as noted above, some music gets mixed in various ways that affect the high freqencies.

But it's still worth a look - easy enough to rip a WAV from a disc with EAC and look at it with EAC. You owe it to yourself and those you're sharing music with to look at the stuff you send out, and maybe the stuff you get too. I've gotten MP3 sourced discs from people who should definitely know better.

One last note, please don't make the mistake of thinking that the freqency cutoff profile suggests any particular encoder is better than another. It doesn't, it's what you can HEAR not what you can SEE that's important for encoding. These graphs tell you precious little about how a human hears sound.

HMM. an interesting topic. is there really no tool that can tell us 100% by looking for mp3 charakteristics in a wav/cdda file??

what do you recommend?? EAC??

EAC's Wave Editor will do fine for the spectral view if that's what you mean, but it can't make the decision for you whether the source is MP3 or not. I can't imagine how any tool could do this reliably.

bad luck

so we all have to guess from the spectral view.

but tnx

After spectral analysis you can encode one track in mp3 (e.g. 160/192kbps):
If you ear severe artifacts there's a better probability that CD was mp3 sourced.

People say there is such a beast as inverse decoder.

dreaming: If only that program was available somewhere... /dreaming

hmm. I think I figured everything out. it works quite fine with EAC. tnx for the great link ( http://mysite.verizon.net/vze1k03a/fa/sa.htm ).

sadly enough I'm still not sure about my nirvana bootleg. sniff.

Heh, I came across a method quite by surprise... plug in some headphones, and unplug them slowly, until you hear a warble. Depending on the type of warble you hear, you can quite often discern between mp3's and normal uncompressed audio.

Ruairi

sounds funny. and... it works???

This is a very interesting idea, wouldn't be that hard, should add one more way to tip the scales to yea or nay. I'd tend to think you'd want at least 192 just to eliminate the possibility of getting artifacts due to low bit rates.

Lately this happened to me, too - I was happy that I had found a problem sample that easily but I noticed the real reason in time.

I found out that you can reproduce it if you mix stereo down to mono with one channel inversed (like some simple karaoke/vocal removal tools/plugins do). Especially where the vocals had been before you hear severe artifacts, no matter if the source is a wav decoded from mp3 or vorbis. (Haven't tried so far, but I expect this to happen with all lossy codecs.)

I played an MPC --xtreme song through a friend's boombox, and enabled the vocal removal feature. I could hear underwater artifacts. The effect is probably more pronounced on some recordings compared to others, depending on what processes have been used to fatten up the vocals and/or drums. But yeah, this feature seems to work even when detecting MPC's.

Question: does encoding in Stereo (rather than Joint Stereo) affect this "artifact"?

Probably not. What affects this "artifact" is mainly the ammount/energy of sound located in the centre (=same frequency, amplitude L/R; no time shift; no phase difference) that is able to mask the "artifacts" and is removed by vocal removal what causes the artifacts to become noticable.

One extreme where there'd be no such artifacts would be 2 totaly independant channels, another two identical channels (if recognized as "mono" by the encoder correctly). In all other cases I think there are artifacts like this - the bigger "centre energy"/"total energy" ratio, the more obvious.

Hello all. I've resurrected this thread for some help. I bought a live CD off ebay (the seller did not mention it was on CD-R, but I've reported him). I'm trying to determine whether it's made out of MP3s, in which case I'd be pissed (more so). I figured I could tell from listening, but it's hard to know what is the result of compression and what is a result of recording conditions. Supposedly, it was recorded in a radio station studio or something like that. I tested a lot of files and I'm posting a couple examples to prove my point (or confusion).

A WAV file (Little Steven - Born Again Savage) Notice how the bars reach the top, I assume color is intensity?


A nice LAME 3.90.2 file(Southside Johnny - Sirens Of The Night), looks very similar, but frequency cutoff is obvious (it's 20100)


A much worse example, LAME 3.93 (Joe Grushecky - Brand New Cadillac) (cutoff is around 14000)


And now the CD I just bought. I can't really decide whether there's a cutoff or just an interuption, either way it doesn't look like any of the other files I've looked at. Or is everything above ~16000 just noise from the recording perhaps?

Spectral view is about the only thing to look at.

Unfortunately, this gets a bit more complicated with shows that were recorded live by audience members. Especially shows recorded in the 1970's and earlier. Some of that recording equipment gets rather sketchy in about the same area that MP3 starts to lowpass, making it harder to tell the difference.

Shows recorded via FM transmission also are like this.

Looks like bad recording to me.

Look at that solid line at 19kHz. That's a constant tone, probably from monitors in the room the recording was mad in.

Kind of reminds me of some of Rage Against The Machine's recordings.

The 19 kHz line definitely smells like a pilot tone to me. Pilot tones are usually used for transmitting purposes....

Some background information

My father (an old ham radio nut) just sent me an explanation of pretty much the same thing, so I think it was indeed recorded off radio (his reasons: only noise after ~15KHz, 19KHz stereo pilot carrier). He also said most receivers would take the signal out though, so don't know why it's still there. I feel a little better that at least it wasn't made out of radio=>mp3=>cd. But I guess there is little left to lose going to mp3 anyway, since everything above 15KHz is crap anyways.

Well thank you all for the helpful replies, I certainly was not expecting them that quickly.

P.S. Is it true some people can hear that tone? Looks pretty high amplitude, would it hurt their ears if they could?

That's absolutely not true, I have a couple of original CDs from the early 80's that are lowpassed at 18KHz, for example (Dio - Holy Diver and The Last in Line)

Nevertheless, on most mp3 based files you might note a clear line at 16KHz, and not a lot of content above it.

This is not exactly true, though. I have original CD's that have no frequencies higher than 16 kHz…for instance, some songs of the "Bloodhound Gang", but also newer peaces!

A good realtime spectrascope is an easy way to tell (baudline is good on linux,
there are probably plenty of winamp vis plugins for this).

MP3 has a very "steppy" frequency response due to freq-domain quantization,
and it is very easy to tell visually when PCM has been through an mp3 encoder.

Cheers,
Martin

Another thing to consider: if a show was recorded with a DAT at 32khz, the frequencies will cut off at 16. So a cut at 16khz isn't always a 100% reliable clue.

Sorry to bounce an ancient thread, but...

Is just my imagination, or was the some time ago (perhaps a year or so?) a software introduced on this site which was used to analyze audio/cd content and "guess" if the audio has been lossly compressed at some point of production?

I have this feeling that I saw this kind of program some time ago, but I am now unable to find it. The search terms just don't cut it:" mp3, lossy, compression, etc.".

Anyone recall this one? I want to wipe clean my CD-R collection of lossy stuff.

Sure. It's auCDtect (command line, freeware)/Tau Analyzer (shareware, GUI-based).

- M.

auCDtect seems to be a great program, but (just so you know) I found three factory-pressed CDs that auCDtect reported as possibly being MPEG. So be careful; it's not 100% accurate.

Here's the link to my findings:
http://www.hydrogenaudio.org/forums/index....ndpost&p=269169