Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Perfect ripping (Read 14829 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Perfect ripping

Reply #25 –
Quote
I just ripped 300+ discs with EAC and accuraterip enable. AccurateRip identified about 5 discs that did not rip accurately, even though EAC said there were no errors. Reripping got accuraterips. AccurateRip will also correctly set offset corrections and I think it is the only way to go for people that really want PERFECT rips.


Anecdotal. Dismissed, apart from the offset corrections, which can be arrived at by other means.

Quote
The philosophy behind AccurateRip is quite simple - each time an audio track is ripped (recorded by computer) it is compared with rips from other people, this allows a confidence report to be generated. The report might say that 4 other people had exactly the same results, this would guarantee your rip was without error, or the report could say that 3 other people disagree with your rip, the likely hood is that your CD has a scratch and should be cleaned.


Unreliable.

Quote
What is the Point?

It gives a confidence when storing your CD collection onto computer that the resulting audio files are free from error, without having to listen manually to each track. You may have heard of 'C2' or 'Secure' ripping, whilst these help obtain error free rips, they do not have 100% error detection.


Incomplete and misleading information, besides encouraging a lazy methodology.

Test your drive's C2 (actually, CU) error detection.
IF 99.9% THEN rip ELSE exit;

Also:
Quote
Discs in Database:     16288 


Useless.

Perfect ripping

Reply #26 –
You are almost trolling with your casual dismissals, but I will give you one more try (in layman terms) -

> 1. What do you mean by "correct ripping result".

Correct as in if you went to the record company and compared files they would be identical, that is the only true CD rip, the correct one without 1 byte of errors.

You have a popular CD which 20 people have submitted results for (by the way those 16K discs in the database are individual CDs, not submission duplicates, there have been submissions for about 30,000 discs - which is a fair number if you consider that AccurateRip does not come as standard with any program, it will do with my stuff as soon as it can work with the online database, so expect 300K discs shortly after). Now you rip that CD, it tells you that for a track that 20 other people agree with your rip, in mathamatical terms the probability of those 20 all being wrong would be in the 100's of billions, so you can guarantee that the rip is accurate. What makes accuraterip secure is its confidence system, it is not a simple CRC but CRC and confidence. Again the oppersite is true, if I was using EAC and c2 and it told me that the rip was 100%, but accuraterip said that 20 people disagreed with my rip, then I would believe AccurateRip (you would have to check that all the tracks did not agree as it might be a different CD pressing).

In laymans terms, that anyone should be able to comprehend.

Perfect ripping

Reply #27 –
Quote
What makes accuraterip secure is its confidence system, it is not a simple CRC but CRC and confidence.
[a href="index.php?act=findpost&pid=231375"][{POST_SNAPBACK}][/a]
Why don't you use a good cryptographic hash algorithm such as SHA-1? These days, processors are fast enough that the extra overhead is negligible.

Perfect ripping

Reply #28 –
Quote
Test your drive's C2 (actually, CU) error detection.
IF 99.9% THEN rip ELSE exit;

How could you do that? not with DAE Quality, not with any test CD I think. Engineers don't just play cards when developing hardware they use these CD's to test their drives & yet they can perform badly in real life or different test CD's, deep scratches, light scratches, consecutive scratches, damaged reflective layer there's just too much to test. Check out CD Freaks they do many of these kind of tests.

Perfect ripping

Reply #29 –
Quote
Now you rip that CD, it tells you that for a track that 20 other people agree with your rip, in mathamatical terms the probability of those 20 all being wrong would be in the 100's of billions, so you can guarantee that the rip is accurate.
...
In laymans terms, that anyone should be able to comprehend.


That's not "layman's terms", it is a unsubstantiated claim.

Can you substantiate it?

Quote
How could you do that? not with DAE Quality, not with any test CD I think.


I am assuming you have used DAEQuality. What is your informed reason for claiming that a drive which is found to detect ~100% of all incorrectable errors has not been tested properly, or that such a test would not reflect real-life ripping?

Perfect ripping

Reply #30 –
Quote
That's not "layman's terms", it is a unsubstantiated claim.

Can you substantiate it?


I can, but I'm at work, and I must disconnect now. Wait 8 hours and I'll answer.

Perfect ripping

Reply #31 –
Quote
I can, but I'm at work, and I must disconnect now. Wait 8 hours and I'll answer.


Thanks Pio, I look forward to it.

Perfect ripping

Reply #32 –
Quote
Quote
What makes accuraterip secure is its confidence system, it is not a simple CRC but CRC and confidence.
[{POST_SNAPBACK}][/a]
Why don't you use a good cryptographic hash algorithm such as SHA-1? These days, processors are fast enough that the extra overhead is negligible.
[a href="index.php?act=findpost&pid=231380"][{POST_SNAPBACK}][/a]



I dont think using SHA1 vs CRC would have made a big difference. I did propose ADDING Tiger Tree Hashes to the database so that if a rip was not accurate you could identify where the error occured and only re-rip that part until you had a perfect rip.

I of course invite everyone to go here:
[a href="http://www.audiosense.org/forum/showthread.php?t=157]http://www.audiosense.org/forum/showthread.php?t=157[/url]
and add your support

Perfect ripping

Reply #33 –
Quote from: DrDoogie,Aug 2 2004, 10:47 PM
Also, my impression is that the whole checksum calculation deals with indices and offsets for the CD, not the accuracy of the actual data ripped.


AccurateRip doesn't perform anything of this kind. It's not a FreeDB kind of identification. The CRC come from the integrality of the ripped audio data. No track lenght or offset value comes into play for the CRC calculation.


Quote from: DrDoogie,Aug 2 2004, 09:21 PM
I don't trust CRC's. My impression is that they are orders of magnitude less accurate than C1+C2 error-detection and correction.


This is a false impression. The CRC being related to all the audio data, a signle error in these data changes the whole CRC. The CRC pattern is not correlated with the audio data pattern, thus the probability to get the same CRC for two different set of audio data by chance, since there are many more possible audio data than CRCs, is given by the number of possible CRCs. Accurate Rip uses CRC32, thus the failure rate, as Spoon recalls, is one out of 2^32 = 4,294,967,296. In your DAEquality test, you demand a C2 accuracy >99.9 %. The failure rate is then 1 out of 1000.
So CRC are about 4,000,000 times more reliable than C1/C2 in your setup.


Quote from: DrDoogie,Aug 2 2004, 09:21 PM
Furthermore, if by my methodology (and I haven't written much about the "detect offset"-procedure in the guide yet) the CD rips perfectly, why should I care?


Because of no-nonsense, anal-retentive pedantic perfectionism ?


Quote from: spoon,Aug 2 2004, 10:21 PM
The thing is, C1 + C2 (especially badly implemented ones) and EACs rip and re-try can let an error through (it was demonstrated recently when EAC on someones Rip said the Rip was without errors, AccurateRip said it was inaccurate, re-ripping generated a rip that AccurateRip agreed upon, so the 2nd rip was the correct one).


Yes, but DrDoogie dismisses EAC's retry. Since he demands 100% quality with C2 on, then EAC don't have to re-read anything. Only rips with zero C2 errors, thus zero re-read performed are accepted.


Quote from: DrDoogie,Aug 2 2004, 11:42 PM
2. How many different submissions of "results" do you think there can be to accuraterip before a 100% reliable consensus on the "correct" result is reached?


I don't know, maybe a lot, if people rips their CD several times and upload the result each time, and if many people rip copies of the same wrong CDR. This can produce a lot of matching wrong result.

But it doesn't matter. The point is that you should be certain, with a probability of error equal to 1 out of 4,000,000,000, that you can't get from a pressed CD that you yourself didn't already upload to AccurateRip's database, the same CRC if your rip is correct.

I say "should", because in fact, some known bugged CD ROM drives can return matching wrong results even under the above "ideal" conditions : an old Toshiba that produce a C2 error at the same absolute time on all ripped CDs (though I don't know if it actually returns wrong audio data), and some Samsungs that mute the last part of every track.


Quote from: DrDoogie,Aug 2 2004, 11:42 PM
3. What can using such a "result" contribute to the accuracy of the methodology that I have outlined?
[{POST_SNAPBACK}][/a]


It can report with confidence that your rip is correct even if you got less than 100 % quality. No need to rerip in this case.


Quote from: DrDoogie,Aug 2 2004, 11:51 PM
Quote from: www.accuraterip.com
Discs in Database:     16288 


Useless.
[a href="index.php?act=findpost&pid=231285"][{POST_SNAPBACK}][/a]


A good thing with Accurate Rip is that it doesn't interfere with your ripping method. You can enable AccurateRip and not use it, while nonetheless contribute to the database population.


Quote from: spoon,Aug 3 2004, 09:28 AM
Now you rip that CD, it tells you that for a track that 20 other people agree with your rip, in mathamatical terms the probability of those 20 all being wrong would be in the 100's of billions, so you can guarantee that the rip is accurate. [a href="index.php?act=findpost&pid=231375"][{POST_SNAPBACK}][/a]


Actually, combining the probabilities of CRC32 being right by chance gives 2^32^20, that is one chance out of 4.6 x 10e192

The probability that everyone including you used a bugged drive, that you all bought a pirate copy with errors, or that the record label made a bad master from a CDR with errors, is infinitely higher.


Quote from: westgroveg,Aug 3 2004, 10:18 AM
Quote
Test your drive's C2 (actually, CU) error detection.
IF 99.9% THEN rip ELSE exit;

How could you do that? not with DAE Quality, not with any test CD I think. Engineers don't just play cards when developing hardware they use these CD's to test their drives & yet they can perform badly in real life or different test CD's, deep scratches, light scratches, consecutive scratches, damaged reflective layer there's just too much to test. Check out CD Freaks they do many of these kind of tests.
[a href="index.php?act=findpost&pid=231381"][{POST_SNAPBACK}][/a]


This is very true. A simple example is given by my [a href="http://perso.numericable.fr/~laguill2/dae/c2accuracy/c2accuracy.htm]DEAquality analysis[/url] on a Sony drive :



C2 accuracy in %, averaged on 70,000 errors under 753 errors per seconds, and on 14 seconds beyond, over error rate in mono samples, of the Sony DDU1621 DVD ROM drive firmware S1.6 with a bad CDR



Close up

We can see that for a good part of the analysis, the C2 accuracy was 99.3 %, but that it is inferior to 75 % for another important part. It is perfectly possible for a test CD to behave in such a way that the overall C2 accuracy result on this drive is 99.3 %, while another test CD can return 70 % !
So DAEquality tests must not be blindly trusted. C2 accuracy is a much deeper problem.

For example, look at this very peculiar audio data, that matches a C2 failure : the drive didn't report any C2 error for this zone, while the data is different than the original one :

http://perso.numericable.fr/laguill2/pictu...tectederror.png

I won't go into technical details, but this kind of combed pattern shows that the error correction chipset of the drive, performing CIRC decoding, did a mistake somewhere.
I didn't take the time to analyze the combed pattern, from which it should be possible to guess where in CIRC the error occured (C1, C2, delay, interleaving, etc). It might be interesting for you, Spath, feel free to ask if you want more details or data (or open another thread in order to discuss the matter).

Perfect ripping

Reply #34 –
>Actually, combining the probabilities of CRC32 being right by chance gives 2^32^20, that is one chance out of 4.6 x 10e192

Laws of probability was not my strongest point , I thought they would multiply up, but wanted to be sure. If any of the matches have a confidence of 2 or more, you could trust your life to that. Like you said, scratches shouldn't give a match with someone else (unless you are super lucky).

Perfect ripping

Reply #35 –
Quote
> read EAC homepage
> especially the technology section.
>
> Also check this out:
> CDFreaks EAC guide

Well, feel free to read them too. EAC (or any other ripper) doesn't perform any error
correction, but retries and educated guesses based on statistics, pretty much as
DrDoogie explained. As explained before error correction is just not possible outside
of a drive.
[a href="index.php?act=findpost&pid=231172"][{POST_SNAPBACK}][/a]



Okay, i guess I was a bit simplistic (or at least quick to post)

1) You're right, EAC doesn't do error correction like a normal cd player does.  It will re-read a section where there it encounters an error

2) I should've been clearer.  CD players do interpolation.
"You can fight without ever winning, but never win without a fight."  Neil Peart  'Resist'

Perfect ripping

Reply #36 –
Quote
The CRC come from the integrality of the ripped audio data. No track lenght or offset value comes into play for the CRC calculation.


That's actually what I secretly feared...

Quote
The CRC being related to all the audio data, a signle error in these data changes the whole CRC.


At the risk of sounding stupid: How big does the checksum need to be to detect a single byte's difference between two files of ~700 Mib size? I just don't see that happening with CRC32.

Quote
Accurate Rip uses CRC32, thus the failure rate, as Spoon recalls, is one out of 2^32 = 4,294,967,296. In your DAEquality test, you demand a C2 accuracy >99.9 %. The failure rate is then 1 out of 1000.
So CRC are about 4,000,000 times more reliable than C1/C2 in your setup.


First, the 2^32 part applies to how large a set of data again?

Also, I thought the 99.9% result was due to a rounding error in the DAEQuality utility itself. Has anyone ever gotten 100%?

Quote
Because of no-nonsense, anal-retentive pedantic perfectionism ?


But see, what is the reliability of the submissions to AccurateRip, what does it add of value to the methodology I have outlined, and why should I trust it?

Quote
I don't know, maybe a lot, if people rips their CD several times and upload the result each time, and if many people rip copies of the same wrong CDR. This can produce a lot of matching wrong result.

But it doesn't matter. The point is that you should be certain, with a probability of error equal to 1 out of 4,000,000,000, that you can't get from a pressed CD that you yourself didn't already upload to AccurateRip's database, the same CRC if your rip is correct.


So what happens if all the teenage morans out there just jam their CDs into their players without any understanding of settings of EAC or whatnot, and submit their results? I don't know, but hey, if I know that -> I <- am doing it right, WHY SHOULD I CARE?

Quote
It can report with confidence that your rip is correct even if you got less than 100 % quality. No need to rerip in this case.


Okay. So if, in the only possible instance where I can get less than 100% quality as reported by EAC's log, by reading into Lead-Out due to offset correction, there are four possible submissions to AccurateRip:
1. Somebody hasn't checked the "replace with silence" in EAC
2. Somebody hasn't done offset correction
3. Somebody has done everything right, and the result agrees with mine
4. Somebody has done everything right, and the result does not agree with mine

How do I know which of these it is? And why should I care, it's the Lead-Out after all, and I replace it with silence anyway.

Quote
A good thing with Accurate Rip is that it doesn't interfere with your ripping method. You can enable AccurateRip and not use it, while nonetheless contribute to the database population.


And the lack of effort required is a good thing? No, I don't think so. FreeDB at least requires you to type things out, but even that database has ~30% errroneous and duplicate entries.

Quote
How could you do that? not with DAE Quality, not with any test CD I think.


Really? But you have the image on disk to test against... I don't get it.

Quote
We can see that for a good part of the analysis, the C2 accuracy was 99.3 %, but that it is inferior to 75 % for another important part. It is perfectly possible for a test CD to behave in such a way that the overall C2 accuracy result on this drive is 99.3 %, while another test CD can return 70 % !
So DAEquality tests must not be blindly trusted. C2 accuracy is a much deeper problem.


I'm actually interested in that too, if you could provide a link to more information regarding your claim that a reader can detect C2 (CU?) errors for one disk, but not another, that would be great, thanks.

Also thanks for taking the time to contribute your views!

Perfect ripping

Reply #37 –
Quote
am assuming you have used DAEQuality. What is your informed reason for claiming that a drive which is found to detect ~100% of all incorrectable errors has not been tested properly, or that such a test would not reflect real-life ripping?

Yes, I have and the results are posted on this forum. My informed reason is that pio has reported that depending the type of damage on the CD he would get diffrent results from his TEAC drive's C2 reporting same goes for the results at CD Freaks  & CD-R-Info depending on the test CD drives could perform better or worse.

At CD-R-Info the ASUS CRW 5232AS scored for the C2 tests they did,

Quote
ABEX TCD-721R 98.0%


Quote
ABEX TCD-726 64.9%


As you can see if they had only used the ABEX TCD-721R you would think the drive had 98% C2 accuracy when in fact it can be much lower in different conditions.

My drives failed to work with DAE Quality C2 test.

Have you used DAE Quality?

So far I have seen 2 web sites that have the CyberDrive CW099D with a score of 100% for C2 accuracy, may be the best drive you can get.

OT: EAC's ext m3u playlists are not working, time displayed is wrong, blast!.

Perfect ripping

Reply #38 –
Quote
due to offset correction, there are four possible submissions to AccurateRip:
1. Somebody hasn't checked the "replace with silence" in EAC
2. Somebody hasn't done offset correction
3. Somebody has done everything right, and the result agrees with mine
4. Somebody has done everything right, and the result does not agree with mine


That is why it took so long to implement it in EAC, every possible option in EAC had to be checked and any bad options that are switched on, then AccurateRip is switched off.

AccurateRip controls the ripping, it will not let you put in an offset, it will only submit tracks when it finds its own offset - so even a complete noob can submit valuble data to the database with no setting up of eac.

Quote
Okay. So if, in the only possible instance where I can get less than 100% quality as reported by EAC's log


No it is possible for EAC to report 100% and there still be an error.

Perfect ripping

Reply #39 –
I suggest that you perform a little seach on the Internet about what is a CRC. It seems that you don't understand the concept at all.


Quote
At the risk of sounding stupid: How big does the checksum need to be to detect a single byte's difference between two files of ~700 Mib size? I just don't see that happening with CRC32.


One bit would be enough, if you calculate the parity of the data. But since one bit can only have two values, there is one chance out of two that you get the initial value again after having changed the data. Using CRC32, this chance is reduced to 1 out of 2^32.


Quote
First, the 2^32 part applies to how large a set of data again?


Any size from one bit to infinite (search for some docs).


Quote
Also, I thought the 99.9% result was due to a rounding error in the DAEQuality utility itself. Has anyone ever gotten 100%?



Sorry, I didn't know that there was a rounding error in EAC.

Quote
But see, what is the reliability of the submissions to AccurateRip, what does it add of value to the methodology I have outlined, and why should I trust it?


If something is unclear in my previous messages, please ask more specific questions.


Quote
if I know that -> I <- am doing it right, WHY SHOULD I CARE?



No problem, you can do it the way you whish. I personally don't mind an extra check of higher confidence, especially if it requires no effort at all.


Quote
Okay. So if, in the only possible instance where I can get less than 100% quality as reported by EAC's log, by reading into Lead-Out due to offset correction,


Any detected error will lead to less than 100 %. I was saying that instead of rejecting these results, which prevent you to rip a lot of CDs, you can have them checked by Accuraterip.

Quote
there are four possible submissions to AccurateRip:
1. Somebody hasn't checked the "replace with silence" in EAC
2. Somebody hasn't done offset correction
3. Somebody has done everything right, and the result agrees with mine
4. Somebody has done everything right, and the result does not agree with mine

How do I know which of these it is?


It's the third one. It's impossible that your drive and someone else drive have missed the exact same bytes on the CD. It is already very highly unlikely reading twice the same CD in the same drive, so reading another CD with another drive, it is completely impossible.


Quote
And why should I care, it's the Lead-Out after all, and I replace it with silence anyway.


It would not be the lead out, it would be what your drive believes to be the lead out, but it is actually the end of the last track. And it's not necesarily silent. But we're not talking about the lead-out, but about errors that EAC thinks it has corrected (no errors occured, quality <100%, red bars in the middle of the extraction )


Quote
I'm actually interested in that too, if you could provide a link to more information regarding your claim that a reader can detect C2 (CU?) errors for one disk, but not another, that would be great, thanks.


I'm not talking about this (of which you can incidentally find an example in the "Cactus Datashield experience" thread), but about failing to detect an error on one disc, while succeeding in another. Westgroveg just gave an example above :
Disc 1 : ABEX TCD-721R , 2% of errors missed
Disc 2 : ABEX TCD-726 , 35.1 % of errors missed.

Perfect ripping

Reply #41 –
Quote
Discussion about C2 accuracy splitted here : http://www.hydrogenaudio.org/forums/index....showtopic=25291
[a href="index.php?act=findpost&pid=232295"][{POST_SNAPBACK}][/a]


Well, thanks for your input all, and Pio2001 in particular. I considered adding AccurateRip, but EAC doesn't use the CD-image with AccurateRip, only individual tracks, so.

Also, when it comes to the claim that a drive may not be reporting all CU-errors for all disks, I feel the best I can possibly do is test the accuracy of a drive at different speeds, and stick with extracting at the speed for which the drive reports errors most accurately, and abort the extraction if any errors are detected. As well as recommending drives with 99.x% or better error-detection, of course.

I believe anything else would be mostly an academic discussion, really, since if there is nothing that can be done about a possible problem, it is no longer a problem but a fact of life.

Perfect ripping

Reply #42 –
Quote
Also, when it comes to the claim that a drive may not be reporting all CU-errors for all disks, I feel the best I can possibly do is test the accuracy of a drive at different speeds, and stick with extracting at the speed for which the drive reports errors most accurately, and abort the extraction if any errors are detected. As well as recommending drives with 99.x% or better error-detection, of course.

I believe anything else would be mostly an academic discussion, really, since if there is nothing that can be done about a possible problem, it is no longer a problem but a fact of life.


To solve that case for me, I use manual test & copy with C2 enabled on image extraction.
The name was Plex The Ripper, not Jack The Ripper

Perfect ripping

Reply #43 –
In that instance I think if you were to rip with two totally different drives, if the results matched then you could be reasonably sure that everything is as near 100% as you can get.

Perfect ripping

Reply #44 –
Testing at different speeds maybe overkill. Anyway, it should be impossible unless you find a software that sets the speed of the drive, and the c2extract exe doesn't reset it, because c2extract, at least the first version, don't let you set the drive speed.

Your method (100% quality) is more secure than EAC's basic secure modes. It might be more secure than checking with two drives, because it can detect a permanent isolated error on the CD, while two drives should not be able to, nor EAC (since all reads will be the same).
Only AccurateRip would be more secure (since it compares with another CD), but the problem is that it can't work for all CDs, only the ones in the database, and only the same pressings, for the time being.