Devendra Banhart - Live @ KCRW; Lossy Source? (Freq. screencap inside)

[rabble]

So, got this Devendra Banhart @ KCRW's "Morning Becomes Eclectic" in a postal trade the other day, and since it had quite a large ? in the lineage (Studio > FM Broadcast > ? > FLAC), I decided I'd try my first frequency analysis to see if it was lossy source (being that the webcast is the usual source for KCRW's without complete lineage) Soon discoverd that FM broadcasts and mp3 sources have a tendency to look similar in a freq.anal., so I thought I'd run it by the collective intelligence of TTD. So, is this lossy or not? The neat "haircut" right at 16K makes me tend to believe that it is, but I'm hoping against hope that's the FM compression.

Thanks!

[AAR.oner]

whats up rabble :wave:

i'd say the 16k line is more typical of FM than it is webcast...most of the webcast-captured stuff i've run across cuts out around 18k...

Five can tell ya more...you shuld check the new Spotting Lossy sources forum:
http://www.thetradersden.org/forums/forumdisplay.php?f=47

i'll get one of the mods to move this over there...you should get more replies that way

[rabble]

Awesome, thanks alot, g.

Look forward to that trade we've got coming up!

[DDSTree]

done

[Five]

hey rabble

from what I can see so far it could go either way

if you could post a SA of a full song and another zoomed on about 2 seconds that would be great. and FA too if possible.

good if you can keep the little time numbers at the bottom, too.

[rabble]

hey rabble

from what I can see so far it could go either way

if you could post a SA of a full song and another zoomed on about 2 seconds that would be great. and FA too if possible.

good if you can keep the little time numbers at the bottom, too.

Will do, I'll post as soon as I make it back home.. Thanks again, Five..

[rabble]

Ok, attached are a spectral analysis of the full song, spectral analysis zoomed in 2 sec, and a frequency analysis.

Thanks again for the help, ya'll!

[Five]

yeah, I'd call that mp3.

the FA isn't displaying correctly. CEP has a bug where you have to move the cursor to sort of wake up the Frequency Analysis display. try opening the file again, opening the FA window then clicking around the cursor on the wave display a few times. It would be cool if you could post a pic of that.

[rabble]

yeah, I'd call that mp3.

the FA isn't displaying correctly. CEP has a bug where you have to move the cursor to sort of wake up the Frequency Analysis display. try opening the file again, opening the FA window then clicking around the cursor on the wave display a few times. It would be cool if you could post a pic of that.

Argh.. That's not what I wanted to hear......

I couldn't get it to change the display by clicking around, but I did adjust HOW it was displaying, maybe this is what you were talking about. I've got both linear and non-linear views attached...

Gimme some good news, five!!

[range_hood]

Could you please do another sa shot of two seconds and save it in *.gif format? this would be more detailed.

Is this an accoustical track. Does the artist speak much? I don´t know Devendra Banhart.

[rabble]

Could you please do another sa shot of two seconds and save it in *.gif format? this would be more detailed.

Is this an accoustical track. Does the artist speak much? I don´t know Devendra Banhart.

It's acoustic, I can post an mp3 sample if it would help. He has a VERY strange voice, too, if it matters any. It'll take me a few minutes to get another SA worked up, I'll post in a few.

[rabble]

Actually, there's no need for me to post a sound sample, you can listen to a webcast of this show here:

http://www.kcrw.com/cgi-bin/db/kcrw.pl?show_code=mb&air_date=5/5/04&tmplt_type=show

The SA & FAs that I have posted are from track 3.

[Five]

I believe the show in question in mp3-sourced. when you get a chance perhaps you could post another FA, you have to click around to sort of wake it up. The one above doesn't look accurate to the SA.

[rabble]

Could you please do another sa shot of two seconds and save it in *.gif format? this would be more detailed.

Is this an accoustical track. Does the artist speak much? I don´t know Devendra Banhart.

Gif 2Sec SA Attached, this one of a different part of the song that had a bit more variance:

[rabble]

I believe the show in question in mp3-sourced. when you get a chance perhaps you could post another FA, you have to click around to sort of wake it up. The one above doesn't look accurate to the SA.

And for Five:

(I clicked around like a bastard on this thing, couldn't get it to change a bit. I did, however, uncheck "Linear View", I hope that was what you were looking for.)

Thanks again for the help, guys!

[Five]

rabble, just use linear please.

I meant clicking around on the wave to move the cursor postition. the fa will update itself each time you click on a new place. the bit numbers at the bottom that say 0:00.000 will change to different numbers each time you change the cursor positon. Likewise, the fa will change everytime you switch postiton. CEP has a little but that sometimes the fa gets frozen so I just want to make sure that we're getting the right readings.

here's some illustrations too:

[Five]

after you move the cursor, the little yellow triangle will be over at the right, and the time display and the fa display will be different.

[Five]

change the cursor positon and watch the fa changing. take a screenshot of an average-looking result.

[range_hood]

Some loss is in there. Maybe some noise reduction to remove the 19 kHz pilot tone?
I don´t think it´s mp3 (no square blocked out area).

[rabble]

after you move the cursor, the little yellow triangle will be over at the right, and the time display and the fa display will be different.


Ohhh, ok, I get what you're saying, cursor as in little yellow thing, not cursor as in mouse. Is this what you mean? (Linear View, chose an averagish looking one):

[rabble]

Some loss is in there. Maybe some noise reduction to remove the 19 kHz pilot tone?
I don´t think it´s mp3 (no square blocked out area).

What's the 19khz pilot tone? If this was the case, would it still be tradable (ala "lossless", even though it technically isn't?)

[Five]

this one is weird. I'm not so sure now. in the SA it cuts off ruler-straight at 16kHz which is a bad sign, but it doesn't really look blocky enough and the FA looks good. I thought the FA might be displaying wrong since there's no response above 16kHz in the SA how is it there's lots going on up there in the FA?

so now I'm thinking this might not be mp3 after all. but I can't figure out how everything about 16kHz is missing so perfectly in the SA :confused:

[rabble]

this one is weird. I'm not so sure now. in the SA it cuts off ruler-straight at 16kHz which is a bad sign, but it doesn't really look blocky enough and the FA looks good. I thought the FA might be displaying wrong since there's no response above 16kHz in the SA how is it there's lots going on up there in the FA?

so now I'm thinking this isn't mp3 after all. but I can't figure out how everything about 16kHz is missing so perfectly in the SA :confused:

If you'd like, I can upload a track from it to my webhost for you to download and look around at if it would help.. Would you mind, or do you think that it's safe to say this is lossless at this point?

[range_hood]

I thought the FA might be displaying wrong since there's no response above 16kHz in the SA how is it there's lots going on up there in the FA?
Do you have a TFT monitor?. I have one. I have to look from above of the monitor to recognize rather (I think it´s) dark blue than black (no response) above 16k.

What's the 19khz pilot tone?
Here´s some overview on Frequency Modulation (FM) (credits to : 4candles (http://www.thetradersden.org/forums/showthread.php?p=140599#post140599))

http://hyperphysics.phy-astr.gsu.edu/hbase/audio/radio.html#c2

[rabble]

Anyone else who wants to grab the track from this to take a look at, it will be available temporarily here (http://www.holotone.net/five/devban2004-05-05_03.flac). Once we get this mystery figured out, I'm taking it down, so if you crazy future people out there stumble upon this thread, the link probably won't work.

[range_hood]

I would say it´s some sort of noise reduced fm. Or a non-mp2 souced cableFM.
As I see on db.etree.org (http://db.etree.org/lookup_show.php?shows_key=242775&show_users=true) this show is not common in the lists.
I think it´s getting time to do an encyclopedia of known spectrals and for some sort of fill in form in the new-thread page of this sub forum to keep an overview. :)

[rabble]

I would say it´s some sort of noise reduced fm. Or a non-mp2 souced cableFM.
As I see on db.etree.org (http://db.etree.org/lookup_show.php?shows_key=242775&show_users=true) this show is not common in the lists.
I think it´s getting time to do an encyclopedia of known spectrals and for some sort of fill in form in the new-thread page of this sub forum to keep an overview. :)

Yeah, the entry you see for me there is actually for the RM Video Stream > VCD of this set I did for my own enjoyment.. Haven't even added my audio copy yet until I can verify this is lossless... So, knowing what you know, would you say that this is lossless "enough" to be tradeable?

[Five]

Okay I had a look and a listen. There's almost no activity above 16kHz, just a slight bit of noise. Using better FA shows quite a steep drop at 16kHz just like the SA is showing. The frequency response is sawed off right at 16kHz in a way that doesn't look like nr. I now believe this is a webcast/mp2 broadcast that's been thru a very clean analog stage which has added that tiny bit of noise in the highs. In terms of sound there's some slurring audible in the sibilants which is typical of FM and lossy but the sound is maybe a bit too nice for a typical RM webcast. I would make note of this information in the .txt file and trade it as a "best available source" most likely SBD>?>?mp2>?FM>unknown (most likely analog) transfer>FLAC

I know that's a pretty sketchy lineage to read but its the most accurate and honest I can come up with after puzzling over this for a couple days.

Here's some better FA analaseez taken from the sample I recieved. Take a look up around 16kHz at that drop, I couldn't trade this and tell ppl it is lossless with a straight face.

[Five]

I think it´s getting time to do an encyclopedia of known spectrals and for some sort of fill in form in the new-thread page of this sub forum to keep an overview. :)
This is exactly what I want to do. I can dig up a bunch of wav - mp3 comparisons done on the same file and post those in a tidy way in a sticky at the top of this forum as I've done with your MD SA. we could also use FA for that same MD sample, preferably Audacity and AnalFreq because they are the most accurate. The CEP SA is perfect. This stuff will eventually become part of a lossy reference archive including WMA, OGG, RM, AC3 and all that.

[range_hood]

Is this seedable here? Or on another tracker with external link?

we could also use FA for that same MD sample, preferably Audacity and AnalFreq because they are the most accurate.
I´m not so in FAs.
Do you mean one of this two or both? Is there a reason why you choose a different FFT size. The AnalFreq cap looks better for me (I like this 1kHz steps more than the 3KHz ones of audacity). Though Analfreq quits after choosing the channel with an error.
Are these shots of the whole files?

This stuff will eventually become part of a lossy reference archive including WMA, OGG, RM, AC3 and all that.
Would be cool to do such a thing.

[Five]

AnalFreq crashed on me too, but at the very end of the test. I ran the whole track, then when the error box came up at the end I moved it aside and took a screenshot of the final results. I'm not sure why AnalFreq likes to crash sometimes, its an old prog that is now available as $hareware (now called SpectraScope) I imagine the newer version has this issue fixed. Hopefully you can get AnalFreq working on your comp somehow it's a great tool.

The Audacity FA is something like 21 seconds, that's all it can handle at once. Highlight a section, choose view>plot spectrum and it will take the 1st 21 or so seconds of whatever you selected and analyze it. In the past when I was having trouble getting a FA to display something useful Audacity has come thru for me not quite sure why that is. My knowledge tends to be pragmatic in nature, the math and esoteric stuff is sometimes beyond my depth of knowledge. But I have a pretty good idea of what works for my purposes and trust the ppl with university degrees to step in and explain exactly why lol.

[range_hood]

AnalFreq crashed on me too, but at the very end of the test. I ran the whole track, then when the error box came up at the end I moved it aside and took a screenshot of the final results. I'm not sure why AnalFreq likes to crash sometimes, its an old prog that is now available as $hareware (now called SpectraScope) I imagine the newer version has this issue fixed. Hopefully you can get AnalFreq working on your comp somehow it's a great tool.

AnalFreq crashes every time before it puts out anything.
Why don´t you like the Cool Edit FA?

[Five]

the CEP FA is okay, its just that I've done FA using everything I can get my hands on and I seem to be able to read audacity and analfreq more easily than CEP. Especially when the dropoff goes down like stairsteps its hard to see that with CEP FA. I like audacity the best for FA, cep for SA at this point.

[ssamadhi97]

I'd say this one is mp3 / similar lossy webcast source. Probably even rm.

1) listen to the side channel / stereo difference. *warble*.

2) check out the impulse response whenever strings are touched. pre-echo like whoa. typical flaw of low-bitrate transform-based lossy codecs when it comes to transient signals.


And I'd say it's most likely not (or not only) mp2 - reason 2) again, most mp2 encoders have excellent transient handling unless you really starve them for bitrate (which would mean introducing a world of other audible flaws, too)

[rabble]

Well, craptacular. Looks like I won't be seeding this one then, I'll just have to keep my eyes peeled for a lossless source.

Thanks again for the help, ya'll!

[4candles]

Okay I had a look and a listen. There's almost no activity above 16kHz, just a slight bit of noise. Using better FA shows quite a steep drop at 16kHz just like the SA is showing. The frequency response is sawed off right at 16kHz in a way that doesn't look like nr. I now believe this is a webcast/mp2 broadcast that's been thru a very clean analog stage which has added that tiny bit of noise in the highs. In terms of sound there's some slurring audible in the sibilants which is typical of FM and lossy but the sound is maybe a bit too nice for a typical RM webcast. I would make note of this information in the .txt file and trade it as a "best available source" most likely SBD>?>?mp2>?FM>unknown (most likely analog) transfer>FLAC


Why doesn't anyone research these things? KCRW broadcasts a 128kbit/s live MP3 stream (which isn't archived). IMO, this is the most likely source.

If the "taper" made the common mistake of recording the webcast by simply pressing record in a sound editor (instead of capturing the original MP3 data), then he/she was recording the analogue sound inside the soundcard - which would add the "tiny bit of noise in the highs" you describe.

But what I'm interested to know is the effect of analogue stages on MP3 compression. i.e. if you take an MP3 file and pass it through an analogue generation what does the spectral analysis look like?

[Five]

d'oh! good job, a bit of research goes a long way.

But what I'm interested to know is the effect of analogue stages on MP3 compression. i.e. if you take an MP3 file and pass it through an analogue generation what does the spectral analysis look like?
probably exactly as seen here. not a lot of difference in the SA, just a light haze in the highs, FA would show a lift in the highs which is only caused by noise. It also depends on the quality of the sound card how much noise gets added. Also, the blockiness would appear somewhat "smeared" in SA. Some day I'll take the time to pass an mp3 out to analog and back again thru a couple different sound cards and post the results.

[range_hood]

Got some spare time. Did some recordings off KCRWs webcasts. The shots on the right side are actually black/white in cool edit (under options > setting > colors if someone cares).
So it seems the source in question contains real video format.

http://img147.echo.cx/img147/9191/kcrwcompare4in.th.png (http://img147.echo.cx/my.php?image=kcrwcompare4in.png)