Hydrogenaudio Forums > HELP! Lossy or Lossless?

Adamn

Jan 18 2005, 11:34

I have there a concert and I think its mp3 sourced, but im not sure. The file format is flac. Here is a picture of a spectral view of the file:

Could anybody tell me if this is the lossless version or a lossy mp3 sourced file?
Thx in advance!

jokull

Jan 18 2005, 12:51

Well how large is the file? FLAC is considerably larger byte-wise.

guruboolez

Jan 18 2005, 13:14

Adamn is talking about the source, not about the current format (flac, apparently). A lossless encoding of a lossy source is still very big.

There's a clear 16 KHz lowpass: it looks like a lossy encoding.

PoisonDan

Jan 18 2005, 13:20

Did you try Tau Analyser yet? You will need to burn an audio CD of the FLAC first...

Adamn

Jan 18 2005, 13:56

QUOTE (PoisonDan @ Jan 18 2005, 04:20 AM)

Did you try Tau Analyser yet? You will need to burn an audio CD of the FLAC first...

Hi everybody!
Thx for all of your help.
No, i dont need to burn a audio CD first, there is another version of Tau Analyser, without GUI.
Tau Analyser says:
"This track looks like MPEG with probability 95%"

So does this mean its from a lossy source?

jth

Jan 18 2005, 15:50

QUOTE

So does this mean its from a lossy source?

Very likely, yes.

precisionist

Jan 18 2005, 17:25

I bet it has been mp3 (or some other lossy codec). The hard cutoff at 16kHz is typical.
Did you try auCDtect ? It can analyse wav files to see if they have been a lossy codec.

edit: I've just seen that Tau Analyser is based on the same algorithm like auCdtect, but isn't free.

Otto42

Jan 18 2005, 18:22

If it's a live bootleg sort of thing, it could have been recorded by some dude with a digital recorder going directly to MP3 format or something. Later he converted it to FLAC since that's what all the bootleg trading is done in. Stupid? Yep.

Adamn

Jan 19 2005, 02:28

QUOTE (Otto42 @ Jan 18 2005, 09:22 AM)

If it's a live bootleg sort of thing, it could have been recorded by some dude with a digital recorder going directly to MP3 format or something. Later he converted it to FLAC since that's what all the bootleg trading is done in. Stupid? Yep.

Is it possible, that a Minidisc recording of a show could look like mp3 sourced? Im using a Sony MD MZ-N10, LP2 mode ... this doesnt look like a lossy format. Never tried the LP4 mode.

indybrett

Jan 19 2005, 04:05

My Hauppage PVR 250 has a hard cutoff at around 16khz as well. Looks like MP3 in spectral view. Most likely though, your audio was mp3 sourced.

I have several "silver" cd's from Europe & Russia that are mp3 sourced. Some of the lossless stuff in trading circles was also mp3 sourced. Fortunately, it's a small amount.

dreamliner77

Jan 19 2005, 06:15

As has been said, it's definitly possible that it's lossy sourced...BUT... It could also have also been recorded live at 32kHz.

Cygnus X1

Jan 19 2005, 06:16

iRiver IFP series flash recorders have a hard-cutoff at 16kHz as well, even at 320kbps. I noticed the same spectral profile as posted above when I was "ripping" some LP's direct to MP3.

Cygnus X1

Jan 19 2005, 06:19

QUOTE (dreamliner77 @ Jan 19 2005, 12:15 AM)

As has been said, it's definitly possible that it's lossy sourced...BUT... It could also have also been recorded live at 32kHz.

Although it's hard to judge by eyeballing it, I'd probably say that it is not a 32kHz recording. It appears that there is content to 16.5kHz in the graph; when recording at 32kHz, most encoders will roll off a little bit before 16kHz. But, I could be wrong.

precisionist

Jan 19 2005, 13:40

QUOTE (Adamn @ Jan 19 2005, 02:28 AM)

Is it possible, that a Minidisc recording of a show could look like mp3 sourced? Im using a Sony MD MZ-N10, LP2 mode ... this doesnt look like a lossy format. Never tried the LP4 mode.

Ah, that's a question for me.
I bet: No! (SP)
I own two hifi-MD recorders; both don't cause a hard cutoff at ~16kHz. MD-ATRAC is the only lossy format I've encountered so far which has this aspect. It is detected by auCDtect as being "lossy", though.
Please don't use LP modes, the quality is much worse there. The bitrate is then only 292/2=146kbps (LP2), well below the general limit of audibility at roughly 200kbps I think.

QUOTE (indybrett @ Jan 19 2005, 04:05 AM)

My Hauppage PVR 250 has a hard cutoff at around 16khz as well. Looks like MP3 in spectral view. Most likely though, your audio was mp3 sourced.

Is this a standalone hifi MD-recorder ? Hauppauge...

?

QUOTE (dreamliner77 @ Jan 19 2005, 06:15 AM)

As has been said, it's definitly possible that it's lossy sourced...BUT... It could also have also been recorded live at 32kHz.

Good point.
Could you, Adamn, offer a detailed screenshot ? 32kHz sampling can't have any content above 16kHz audio frequency...

edit: mmm..."audibility limit" seems to be no English word...What would you say instead ? audibleness ? transparency level ?

Adamn

Jan 20 2005, 13:17

QUOTE (precisionist @ Jan 19 2005, 04:40 AM)

Could you, Adamn, offer a detailed screenshot ? 32kHz sampling can't have any content above 16kHz audio frequency...

Here is a frequency analysis of that file, hope this helps:

And for comparison, this is a file, i recorded with my Sony MD MZ-N10 in LP2 mode:

To this file, auCDtect says: "This track looks like CDDA with probability 100%"

precisionist

Jan 20 2005, 15:44

QUOTE

To this file, auCDtect says: "This track looks like CDDA with probability 100%"

Uh, uh..
Which ATRAC-type and which year is your MD-recorder ? I can hardly believe that auCDtect fails so thoroughly...even worse, it's LP. How did you transfer the MD-record to hard disc ? In an analog way ? On the other hand, it's good for MD-ATRAC if auCDtect can't detect it.
What I meant by "detailed" is a zoom-in to the 16000-16500Hz area on your first frequency view screenshot.

Adamn

Jan 21 2005, 13:03

QUOTE (precisionist @ Jan 20 2005, 06:44 AM)

QUOTE

To this file, auCDtect says: "This track looks like CDDA with probability 100%"

Uh, uh..
Which ATRAC-type and which year is your MD-recorder ? I can hardly believe that auCDtect fails so thoroughly...even worse, it's LP. How did you transfer the MD-record to hard disc ? In an analog way ? On the other hand, it's good for MD-ATRAC if auCDtect can't detect it.
What I meant by "detailed" is a zoom-in to the 16000-16500Hz area on your first frequency view screenshot.

Year of the MD-Recorder: 9/2002
ATRAC-Type: ATRAC DSP Type-S
I transfered the MD-record to harddisk by a MD-deck, no, not analog, digital!

Screenshot of that other file will follow later.

thx again for all of your help!

esa372

Jan 29 2005, 16:55

I'm having a similar issue with auCDtect ; maybe you guys can help me suss it out.

I have MoFi UltraDisc and UltraDisc-II gold CDs of Dark Side Of The Moon. Out of sheer boredom, I ran auCDtect on both sets of EAC-ripped WAV files and the results are less than reassuring...

Result for MoFi UD CD:

CODE

auCDtect: CD records authenticity detector, version 0.8.2
Copyright (c) 2004 Oleg Berngardt. All rights reserved.
Copyright (c) 2004 Alexander Djourik. All rights reserved.
------------------------------------------------------------
Processing file: [01 Speak To Me.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like CDDA with probability 86%

------------------------------------------------------------
Processing file: [02 Breathe.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like MPEG with probability 89%

------------------------------------------------------------
Processing file: [03 On The Run.wav]
Data analysis: [100%]
------------------------------------------------------------
Could not qualify the source of this track.

------------------------------------------------------------
Processing file: [04 Time.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like CDDA with probability 42%

------------------------------------------------------------
Processing file: [05 The Great Gig In The Sky.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like CDDA with probability 100%

------------------------------------------------------------
Processing file: [06 Money.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like MPEG with probability 99%

------------------------------------------------------------
Processing file: [07 Us And Them.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like MPEG with probability 85%

------------------------------------------------------------
Processing file: [08 Any Colour You Like.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like CDDA with probability 49%

------------------------------------------------------------
Processing file: [09 Brain Damage.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like MPEG with probability 92%

------------------------------------------------------------
Processing file: [10 Eclipse.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like CDDA with probability 100%

Why are the results so varied?
Check out tracks 5 & 6: CDDA @ 100% probability and MPEG @ 99% probability.

And when I scan the entire album as a single WAV file, I get this:

CODE

Processing file: [UD.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like MPEG with probability 79%

The analysis for the UltraDisc II is no better.
Result for MoFi UD-II CD:

CODE

auCDtect: CD records authenticity detector, version 0.8.2
Copyright (c) 2004 Oleg Berngardt. All rights reserved.
Copyright (c) 2004 Alexander Djourik. All rights reserved.
------------------------------------------------------------
Processing file: [01 Speak To Me.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like CDDA with probability 94%

------------------------------------------------------------
Processing file: [02 Breathe.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like MPEG with probability 92%

------------------------------------------------------------
Processing file: [03 On The Run.wav]
Data analysis: [100%]
------------------------------------------------------------
Could not qualify the source of this track.

------------------------------------------------------------
Processing file: [04 Time.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like CDDA with probability 49%

------------------------------------------------------------
Processing file: [05 The Great Gig In The Sky.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like CDDA with probability 62%

------------------------------------------------------------
Processing file: [06 Money.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like MPEG with probability 100%

------------------------------------------------------------
Processing file: [07 Us And Them.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like MPEG with probability 89%

------------------------------------------------------------
Processing file: [08 Any Colour You Like.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like CDDA with probability 92%

------------------------------------------------------------
Processing file: [09 Brain Damage.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like MPEG with probability 73%

------------------------------------------------------------
Processing file: [10 Eclipse.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like CDDA with probability 100%

Well, at least track 10 is consistent...

But check out track 6 (UD-II): MPEG @ 100% probability?! 100% ?!?

And scanning the whole UD-II album as a WAV gives the same result as the UD WAV:

CODE

Processing file: [UD-II.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like MPEG with probability 79%

So, I'm going to start testing a variety of different CDs, because these results definitely throw the accuracy of auCDtect into question.

But if any of you audio brahmins have any insight as to why this happens, please let me know.

Thanks in advance,

~esa

esa372

Jan 29 2005, 19:13

OK, I found another one: The MoFi UltraDisc-II gold CD of Steely Dan's Aja.

CODE

auCDtect: CD records authenticity detector, version 0.8.2
Copyright (c) 2004 Oleg Berngardt. All rights reserved.
Copyright (c) 2004 Alexander Djourik. All rights reserved.
------------------------------------------------------------
Processing file: [01 Black Cow.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like MPEG with probability 95%

------------------------------------------------------------
Processing file: [02 Aja.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like MPEG with probability 99%

------------------------------------------------------------
Processing file: [03 Deacon Blues.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like CDDA with probability 49%

------------------------------------------------------------
Processing file: [04 Peg.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like CDDA with probability 43%

------------------------------------------------------------
Processing file: [05 Home At Last.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like MPEG with probability 99%

------------------------------------------------------------
Processing file: [06 I Got The News.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like CDDA with probability 76%

------------------------------------------------------------
Processing file: [07 Josie.wav]
Data analysis: [100%]
------------------------------------------------------------
This track looks like MPEG with probability 89%

Could it have something to do with the years these CDs were produced?
The MoFi CDs of Dark Side were issued in 1988 and Aja was issued in 1990. I haven't found this problem on any of the later UltraDiscs I have (1990-1999)...

precisionist

Jan 31 2005, 13:37

Could it be that auCDtect expects a special (newish, 90s-) style of mastering (sensible, since lossy codecs didn't exist before)? Special frequency content, which its calculations are based on ?
Are the recordings actually from 1988/90, or were they rereleased/remixed/remastered at that time ?
Frequency content has slowly changed since the 70s...
I assume this CDs are all CD-DA and auCDtect fails.

esa372

Jan 31 2005, 15:07

Thanks for the reply, precisionist...

QUOTE (precisionist @ Jan 31 2005, 04:37 AM)

Are the recordings actually from 1988/90, or were they rereleased/remixed/remastered at that time ?

The recordings are from 1973 and 1977 and MoFi uses the original mater tapes for their gold CDs, but AFAIK they do not alter the recordings in any way (for example, no remastering or remixing - just an exact copy of the master tape.)

QUOTE (precisionist @ Jan 31 2005, 04:37 AM)

Could it be that auCDtect expects a special (newish, 90s-) style of mastering (sensible, since lossy codecs didn't exist before)? Special frequency content, which its calculations are based on ?
[...]
Frequency content has slowly changed since the 70s...

No doubt, but could that be the reason auCDtect has such trouble with these CDs?

QUOTE (precisionist @ Jan 31 2005, 04:37 AM)

I assume this CDs are all CD-DA and auCDtect fails.

Correct. Factory pressed; retail package.

precisionist

Jan 31 2005, 17:36

QUOTE (esa372 @ Jan 31 2005, 03:07 PM)

No doubt, but could that be the reason auCDtect has such trouble with these CDs?

I guess I'll analyze some old records with auCDtect...

QUOTE

QUOTE (precisionist @ Jan 31 2005, 04:37 AM)

I assume this CDs are all CD-DA and auCDtect fails.

Correct. Factory pressed; retail package.

Actually I meant that at first the source of a CD's audio is not yet determined, whatever way one has got the CD. Lossy artefacts on factory pressed CDs bought in a 'normal' way (hifi-store...) are very unusual, thus the CDs should be real CD-DA; mastering engineers don't deal with data compression. Therefore auCDtect obviously fails.

esa372

Jan 31 2005, 18:51

QUOTE (precisionist @ Jan 31 2005, 08:36 AM)

I guess I'll analyze some old records with auCDtect...

Please let me know what you find out...

QUOTE (precisionist @ Jan 31 2005, 04:37 AM)

Actually I meant that at first the source of a CD's audio is not yet determined, whatever way one has got the CD. Lossy artefacts on factory pressed CDs bought in a 'normal' way (hifi-store...) are very unusual, thus the CDs should be real CD-DA; mastering engineers don't deal with data compression. Therefore auCDtect obviously fails.

Agreed.

And earlier, I said:

QUOTE (esa372 @ Jan 29 2005, 10:13 AM)

The MoFi CDs of Dark Side were issued in 1988...

...but I have a correction. I just found out that, while the UltraDisc of Dark Side was in fact issued in 1988, the UltraDisc-II was issued in 1993.

I was checking my discs against the description found here to verify authenticity, and they list the UD-II as being issued in 1993.

Thanks for your help, precisionist.
I hope we can get to the bottom of this; I don't like the idea of auCDtect giving false results all this time...

nyarlathotep

Jan 31 2005, 19:17

Something even stranger IMO.

So I have that mp3 file:

I decoded it with foobar2000 to WAV (PCM, fixed-point), once with Dither and once without Dither. And here's what auCDtect reveals:

Can someone explain?

cabbagerat

Jan 31 2005, 19:57

QUOTE (nyarlathotep @ Jan 31 2005, 10:17 AM)

So I have that mp3 file:
...
Can someone explain?

Certainly. Dither changes the frequency of the quantization noise - this change in the frequency distribution of the file probably fools auCDetect's MPEG detection algorithm. The graphs Here (from the HA FAQ) should clear things up.

nyarlathotep

Jan 31 2005, 20:10

Thanks cabbagerat.

esa372

Feb 1 2005, 02:10

QUOTE (cabbagerat @ Jan 31 2005, 10:57 AM)

Dither changes the frequency of the quantization noise - this change in the frequency distribution of the file probably fools auCDetect's MPEG detection algorithm. The graphs Here (from the HA FAQ) should clear things up.

Great link, cabbagerat. Thanks for posting it.

I don't suppose you have any information on what might be causing the problem I've experienced...(?)

precisionist

Feb 2 2005, 13:44

Without having read the link, a short explanation why auCDtect may fail:
Dither adds a small amount of white noise. White noise contains frequencies from all over the spectrum at the same loudness, 0-22050Hz in this case. Thus there's some frequency content above 16kHz again.

itchyfish

Mar 6 2005, 16:04

nyarlathotep - i'm trying to get auCDtect to read my cd's and it won't .. I saw the command you used below on your screen shot - but it's not working for me. Can you give me a hand on what it is i'm doing wrong?

at the command prompt i've tried aucdtect *.wav - i've tried pathing to the cd drive and doing the same thing but the software just displays the different switches you can use with the software.

Any help would be appreciated!

Thanks!

precisionist

Mar 31 2005, 19:09

QUOTE (itchyfish @ Mar 6 2005, 04:04 PM)

nyarlathotep - i'm trying to get auCDtect to read my cd's and it won't .. I saw the command you used below on your screen shot - but it's not working for me. Can you give me a hand on what it is i'm doing wrong?

at the command prompt i've tried aucdtect *.wav - i've tried pathing to the cd drive and doing the same thing but the software just displays the different switches you can use with the software.

Any help would be appreciated!

Thanks!

Try copying the wav files into the directory with auCDtect.exe

Jojo

Mar 31 2005, 19:11

QUOTE (precisionist @ Mar 31 2005, 10:09 AM)

QUOTE (itchyfish @ Mar 6 2005, 04:04 PM)

nyarlathotep - i'm trying to get auCDtect to read my cd's and it won't .. I saw the command you used below on your screen shot - but it's not working for me. Can you give me a hand on what it is i'm doing wrong?

at the command prompt i've tried aucdtect *.wav - i've tried pathing to the cd drive and doing the same thing but the software just displays the different switches you can use with the software.

Any help would be appreciated!

Thanks!

Try copying the wav files into the directory with auCDtect.exe

I wouldn't rely on aucdtect...see http://www.hydrogenaudio.org/forums/index....ndpost&p=286779 and http://www.hydrogenaudio.org/forums/index....ndpost&p=287263