PDA

View Full Version : Unrealistic Flac compression


Punkishlyevil
2004-12-29, 06:06 PM
Someone in the TSP/Zwan hub seeded a show today. They ripped it from a CD to a wav file and compressed it to level 8 flacs. The flac bitrates are ranging from high 200s to high 300s. This seems way to low of a bitrate for a flac. This seems more like an mp3 bitrate. When I mentioned this to him he reripped a track and flaced it and it did the same thing. He seeded a different show also today and did the same thing, but when he flaced it it did normal 600-900 kb/s bitrates. So does anyone have any clue why these bitrates are so low? Do you want frequency analysis pictures from eac/analfreq?

U2Lynne
2004-12-29, 07:09 PM
Well, the pictures would certainly help to tell the story, so I would grab a 1.5 to 2 second picture from the middle of a track (I think that is what Five usually asks for).

Punkishlyevil
2004-12-29, 07:17 PM
What kind of picture? A spectral analysis or a frequency analysis? EAC or analfreq or Cool Edit Pro 2.0?

U2Lynne
2004-12-29, 07:21 PM
I believe analfreq is the tool most are using these days. And we like to see both a spectral analysis and a frequency analysis.

RainDawg
2004-12-30, 12:03 AM
That is quite low, but no need to assume mp3. If the source material is a mono signal, bitrates this low can indeed happen with FLAC. In fact, if you take audio and go FLAC > MP3 > FLAC, the chances are that the second file will have almost identical bitrate as the first one....compressing to lossy formats and then back don't change the compression much.

I'd still like to see a spectral anlysis and frequency sweep plot, but I'm going to venture a guess that it's mono lossless rather than mp3...

Five
2004-12-30, 12:44 AM
this is easy to solve. make md5 sums for the original WAV files, then decompress the FLACs you made in another folder and check if the decompressed WAVs (from your FLACs) are identical or not.

Punkishlyevil
2004-12-30, 01:06 AM
Frequency/spectral pictures.

Punkishlyevil
2004-12-30, 01:08 AM
I'm not really worried if its mp3, because mp3s wouldn't make a flac compress so much. The person who ripped the CD and I were both just very confused by why it is so compressed. The mono explanation would make sense though.

feralicious
2004-12-30, 02:10 AM
Frequency/spectral pictures.Also, I would do the spectral analysis in CoolEdit as we discovered in another thread here that EAC SAs aren't very reliable. :rolleyes:

Mods: maybe there should be a sticky about that? I don't think that problem is well known.

Five
2004-12-30, 02:19 AM
Also, I would do the spectral analysis in CoolEdit as we discovered in another thread here that EAC SAs aren't very reliable. :rolleyes:

Mods: maybe there should be a sticky about that? I don't think that problem is well known.
CEP/Audition is the best for sure... EAC has proven not very reliable. We need a freeware solution for people, too :hmm: I'm thinking up a solution for this.

New Homebrew
2004-12-30, 02:30 AM
I'm not really worried if its mp3, because mp3s wouldn't make a flac compress so much. The person who ripped the CD and I were both just very confused by why it is so compressed. The mono explanation would make sense though.

It looks like an mp3 source to me based on the spectral view, although i don't think that's the cause of the small torrent size.

feralicious
2004-12-30, 03:01 AM
Yeah and that Anal Freq pic doesn't look too good.

Punkishlyevil
2004-12-30, 03:09 AM
Yeah and that Anal Freq pic doesn't look too good.

Yeah when I saw the AnalFreq I was a bit worried.

Punkishlyevil
2004-12-30, 03:11 AM
Cool Edit.

NINJA
2004-12-30, 05:16 AM
Cool Edit.


looks pretty normal for a 60's or early 70's audience recording


can also indicate noise reduction was used :clap:

Five
2004-12-30, 05:51 AM
Cool Edit.
umm... I don't see much above 16kHz... can you zoom in on 1.5-3 seconds, again with CEP spectral?

Also, it would help to see the cep frequency analysis (under "analyze"). also, thought I'd mention that you can screencap just the current window using Alt+Print Screen

RainDawg
2004-12-30, 11:51 AM
CEP/Audition is the best for sure... EAC has proven not very reliable. We need a freeware solution for people, too :hmm: I'm thinking up a solution for this.
AnalFreq really is the best, and it's free, though it can be hard to find online anymore.

I wonder if we should host the executable here on our site.

RainDawg
2004-12-30, 11:51 AM
Might I ask, what exactly is the recording we're talking about, what's the advertised lineage, and when was it recorded?

Punkishlyevil
2004-12-30, 11:52 AM
First picture is Spectral Analysis zoomed in.
Second is Frequency Analysis on that same zoomed in part.
Thir is Frequency Analysis of the whole song.

Punkishlyevil
2004-12-30, 11:53 AM
Might I ask, what exactly is the recording we're talking about, what's the advertised lineage, and when was it recorded?

The Smashing Pumpkins
1999/04/19 - Ogden Theater; Denver, CO, US

Source: MiniDisc Master ( SM58>MZR30 )> CDR
Conversion: CDR>Xing AudioCatalyst>WAV>FLAC

fatoldpig
2004-12-30, 12:06 PM
Source: MiniDisc Master ( SM58>MZR30 that's explains it. minidisk will have freq drop after 16-18K

RainDawg
2004-12-30, 12:08 PM
Yeah, it looks like a true MD, not an mp3. I am wondering about that compression rate, because it's really only possible if the source is mono. Would someone record this into mono? Maybe a single mic? Can you confirm that it's indeed a mono source?

Punkishlyevil
2004-12-30, 12:27 PM
Looks mono to me. Heres a picture of the waveform.

Punkishlyevil
2004-12-30, 02:09 PM
So is this recording shareable?

robkismet
2004-12-30, 05:31 PM
If it's an SM58 then it'll be mono - it's Shure's classic vocal mic. The frequency analysis looks harsher than typical MD, but I guess an R30 will be using the oldest ATRAC codec which may explain it. The following will also have a big influence I think:
http://www.jbsmusic.co.uk/images/detail_media/shure_sm58_responce.gif
The mic's frequency response is only 50-15,000 Hz.
http://www.jbsmusic.co.uk/validate07.php?action=show&cat=36&orig_rowid=528

Should be fine for seeding?

Five
2004-12-30, 06:11 PM
If you want to check if it's mono or not, copy the left channel to a new CEP window, then copy the right channel and mix-paste it out of phase (use overlap(mix) and check the "invert" box). If the left is identical to the right, you will get a perfectly flat line, and the recording is therefore mono.

So is this recording shareable?
MD is an acceptable source... the only thing that's touchy is the Xing software used for extraction. This was a few years back, who knows if that one CDR copied off the DAT still works or even exists. Or, for that matter, if the DAT has been taped over by now. I'd say this is okay to share, but RainDawg gets the final word on this.

h_vargas
2004-12-30, 06:23 PM
also, it maybe noteworthy that for a MD recorder, you can record in what's called (i believe) LP2 mode - which will produce a mono recording (and give twice the recording length on a disc, if i remember correctly), regardless of whether or not stereo mic(s) were used. plus, since it's a MD source, chances are that it underwent an analog transfer from MD > CDR... the source info given doesn't specify equipment used, but it's quite possible a cheapo soundcard going Mic or Line In was used.

Punkishlyevil
2004-12-30, 06:30 PM
I don't really plan on seeding it here becasue the quality is so bad compared to other shows that have almost exact setlists from the same tour. I do plan on sharing it in the lossless TSP/Zwan hub if anyone really wants it though. But the hub doesn't want things that are sourced from mp3s or anything like that, so I figured this would be a great place to get some info because there seems to be some experts here! As the thread title says, we were most confused about the size of the files (an 8 and a half minute song that is only 23.7 MB seems like a good reason to be confused if you don't know too much about this sort of thing). Thanks a lot for the help. I really appreciate it, and I'm sure the rest of the people in the hub will too.

ssamadhi97
2004-12-30, 08:53 PM
looks like the worst case to me (mono AND mp3). As opposed to noise reduction, mp3 cuts a lot of high frequency content when masked by lower frequencies.

Sure, same goes for md to a lesser extent, but this recording simply doesn't really look like md to me. Not even mdlp.


And I gotta agree with Five: the Xing AudioCatalyst software in the lineage is definitely suspicious as well - heh, I had the same kind of fun with some guy who ripped CDRs for me using some Real software. :wtf: I educated him about EAC and got some perfectly fine copies as replacement.

robkismet
2004-12-31, 07:04 AM
also, it maybe noteworthy that for a MD recorder, you can record in what's called (i believe) LP2 mode - which will produce a mono recording (and give twice the recording length on a disc, if i remember correctly), regardless of whether or not stereo mic(s) were used.Mono and LP2 are completely different. Mono continues to use 292kbps, but doubles capacity by only recording one channel. LP2 doubles capacity by recording in stereo at 132kbps (not 146kbps - some dummy data is needed to stay compatible with SP players). The MZ-R30 used here is very old and is not capable of LP2, though can do mono.

http://www.minidisc.org :thumbsup

h_vargas
2004-12-31, 04:51 PM
Mono and LP2 are completely different. Mono continues to use 292kbps, but doubles capacity by only recording one channel. LP2 doubles capacity by recording in stereo at 132kbps (not 146kbps - some dummy data is needed to stay compatible with SP players). The MZ-R30 used here is very old and is not capable of LP2, though can do mono.

http://www.minidisc.org :thumbsup

thank you for that clarification, i appreciate it. :)

buzzy
2005-01-08, 01:14 PM
Also, I would do the spectral analysis in CoolEdit as we discovered in another thread here that EAC SAs aren't very reliable. :rolleyes:

Mods: maybe there should be a sticky about that? I don't think that problem is well known.Pardon my french, but that's bullshit.

robkismet
2005-01-08, 02:21 PM
Pardon my french, but that's bullshit.http://www.thetradersden.org/forums/showthread.php?t=1331