IPB

Welcome Guest ( Log In | Register )

2 Pages V  < 1 2  
Reply to this topicStart new topic
EAC T&C without C2. Overkill?
Pio2001
post Nov 27 2002, 22:28
Post #26


Moderator


Group: Super Moderator
Posts: 3936
Joined: 29-September 01
Member No.: 73



QUOTE (user @ Nov 27 2002 - 07:07 PM)
hmm, Halcyons test, was carried out with a special test-CD, with black stripes on it ?

Is this very realistic, or was that Sony drive so bad ?


I think it is realistic. A deep big scratch parallel to the groove should have quite the same effect.

QUOTE (user @ Nov 27 2002 - 07:07 PM)
you are just reading 2 times again by t&c, but eac has had the chance to read it 82 times in the first (or only) run..)


Yes, but, as discussed in http://www.digital-inn.de/showthread.php?t...?threadid=14600 , you might have caching issues reading 82 times. Some drives are reported to cache by Feurio and not by EAC, some by EAC and not by Feurio.
With test and copy, you eliminate any caching issue.

QUOTE (user @ Nov 27 2002 - 07:07 PM)
How big is the probability, that eac reads twice (no c2, secure) and thinks: no error, all is well, but it has read twice the same wrong value for that damaged sector.


Quoting the same link :

QUOTE
First, let's assume that the chipset can't correct E32 errors, and that the errors come from unreadable parts of the CD.
Therefore errors will be in frames with at least 3 wrong bytes. If they are random, the probability that the detection will fail is
1/256^3=1/16,777,216
It is the probability that the two readings are wrong (three wrong bytes) and identical.
Being 7350 frames per second, if all frames are wrong, there would be about two undetected errors per CD.

But this can vary very much.
1 - recent chipsets must have five wrong bytes per frame in order to generate wrong data : that would lead to one undetected error once every 32,000 CDs.
2 - errors should be random if the CD is completely unreadable, but for the other bytes to be OK, the damage must not be total. So there should be some parts slightly damaged, and these are likely to cause some problems since there can be just jitter errors in some EFM data.
I don't have the EFM table in text format, so I can't check if the possibility to misplace one transition can lead to an error as small as one modified bit in the decoded 8-bits symbol. If I had setup the EFM table myself, I would have tried to do so, in order to get LSB errors if possible in the case of jitter errors.
If it is indeed the case, we can imagine three unstable bytes leading to 1 wrong bit in each.
In this case, these bits being random, there would be one chance out of 8 to get twice the same wrong data.

So the theoretical efficiency is between 1/8 (jitter errors - E32 chipset) and 1/1,099,511,627,776 (damaged CD, E52 chipset).
That's quite an uncertainty isn't it  ?


QUOTE (user @ Nov 27 2002 - 07:07 PM)
which leaves more room for minor differences between the reads which aren't audible, but still exist


Why wouldn't they be audible ?
Isolated errors are often not audible, because the drive interpolates the previous and next sample, which performs a very good click removal, but Halcyon example showed that this was not always the case (it must have been a multisample error, or a drive that doesn't interpolate).

QUOTE (user @ Nov 27 2002 - 07:07 PM)
Unfortunately, if the surface is damaged, the original data is changed.


Yes and no.
It is quite impossible to change valid data into another valid ones by damaging the surface, because of all the CIRC encoding. However, if the error is perfectly localized on some given samples (it is a very uncommon phenomenon), as the drive interpolates, it will return twice the same interpolation because the valid samples have not changed.
The fact that errors can't affect the same samples every time is the key of error detection by reading twice. Any error will always return the same data (provided it doesn't affect contigous samples, which is ensured by the CIRC for light to medium errors), but errors will nearly never affect the same samples twice.

QUOTE (user @ Nov 27 2002 - 07:07 PM)
I'm still a little worried about the results of the NEC with Test & Copy


Try enable the cache option and/or non accurate stream. It already fixed some CRC issues on some drive that are detected as accurate and not caching.
(Memorex DVDMaxx1648, and Teac E540, or maybe Sony DDU1621, I don't remember. It was very hard to get mismatching CRC on "copies OK" anyway).

This post has been edited by Pio2001: Nov 27 2002, 22:30
Go to the top of the page
+Quote Post
Volcano
post Nov 28 2002, 16:23
Post #27





Group: Members (Donating)
Posts: 916
Joined: 30-September 01
From: Berlin, Germany
Member No.: 112



Pio2001:

QUOTE
QUOTE
which leaves more room for minor differences between the reads which aren't audible, but still exist

Why wouldn't they be audible ?


I was just guessing that if EAC's error correction kicks in, two reads may perhaps result in a difference of only a few samples (because an error isn't corrected in exactly the same way each time) - which wouldn't have to be audible at all, but still, a CRC check will pick up the difference.

My question is, is that theory right, or am I way off? smile.gif

I'll try your suggestions though, and report back.
Go to the top of the page
+Quote Post
liekloo
post Nov 28 2002, 17:18
Post #28





Group: Members
Posts: 456
Joined: 22-March 02
From: Belgium
Member No.: 1596



[Q]QUOTE VOLCANO: My question is, is that theory right, or am I way off? [/Q]

Because the difference is generally so tiny, CRC mismatches are very likely to be inaudible errors (not always, as Pio showed).

Now, why are they so tiny in general?

The reason for this is statistics (not, as you think, varying error correction).
It is rather by accident that 2 reads over a scratch can give identical results. Normally errors on the CD will give different read results every time they are read.
The bigger the error, the less likely it is that the 2 reads will give identical results by accident...
A CRC mismatch means that a third read reveals the previous 2 reads were identical, and therefore accepted by EAC... but by accepting them EAC failed, as the CRC mismatch proves!


--------------------
"E S S E N T I A L" Guide for E A C :

http://users.fulladsl.be/spb2267/
Go to the top of the page
+Quote Post
liekloo
post Nov 30 2002, 12:18
Post #29





Group: Members
Posts: 456
Joined: 22-March 02
From: Belgium
Member No.: 1596



In fact Pio2001 already answered the question:

Interpolation will change isolated errors in most cases, so that they become unhearable anyway.


--------------------
"E S S E N T I A L" Guide for E A C :

http://users.fulladsl.be/spb2267/
Go to the top of the page
+Quote Post
spoon
post Nov 30 2002, 22:47
Post #30


dBpowerAMP developer


Group: Developer (Donating)
Posts: 2723
Joined: 24-March 02
Member No.: 1615



Here is what I have found on ripping audio CDs -

Discs with scratches can return constant errors no matter what speed is used,
Ripping the same disc with a different make of drive will highlight the error,
Applying a CD restore liquid helps alot.


--------------------
Spoon http://www.dbpoweramp.com
Go to the top of the page
+Quote Post
liekloo
post Dec 4 2002, 09:33
Post #31





Group: Members
Posts: 456
Joined: 22-March 02
From: Belgium
Member No.: 1596



T&C should also help to verify whether your drive really supports accurate stream right? (although a modern drive is very likely to)



Then a few more difficult issues...
***About caching, let's say we want to be sure our drive does not cache, so we turn it off and start ripping (T&C). How many CDs should be ripped that way to be sure...? (I know this is a tricky question wink.gif )

***Secure reading with T&C is not likely to find CRC mismatches on disks in perfect condition, but if the disk is (slightly) damaged the chance on CRC mismatches increases significantly. As pointed out before, this must be a theoretical chance somewhere between 1/8 - 1/100000000000... (per disk), which means: we cannot draw a conclusion, it's just worth further research.

Hence my qestion: Personal Experiences??? Quite some folks will use T&C if the disk has scratches, just 'to be sure'. So, what's your experience?
(note: one can only draw conclusions in case of a considerable amount of those rips)

This post has been edited by liekloo: Dec 4 2002, 09:33


--------------------
"E S S E N T I A L" Guide for E A C :

http://users.fulladsl.be/spb2267/
Go to the top of the page
+Quote Post
Pio2001
post Dec 4 2002, 12:39
Post #32


Moderator


Group: Super Moderator
Posts: 3936
Joined: 29-September 01
Member No.: 73



QUOTE (liekloo @ Dec 4 2002 - 11:33 AM)
***About caching, let's say we want to be sure our drive does not cache, so we turn it off and start ripping (T&C). How many CDs should be ripped that way to be sure...? (I know this is a tricky question  ;) )

One CD with many correctable errors is enough. Note that a CD with only correctable errors is something difficult to find.

QUOTE (liekloo @ Dec 4 2002 - 11:33 AM)
between 1/8 - 1/100000000000... (per disk)


I was wrong, the parity bytes can be corrupted too, leaving only one wrong audio bit beside the two to four wrong parity bytes.
So it's not 1/8 but 1/2.

QUOTE (liekloo @ Dec 4 2002 - 11:33 AM)
Hence my qestion: Personal Experiences??? Quite some folks will use T&C if the disk has scratches, just 'to be sure'. So, what's your experience?
(note: one can only draw conclusions in case of a considerable amount of those rips)


I usually rip CDRs with problems. All CDRs I burned two years ago and more are turning unreadable.
With such CDs, C2 on, no errors occured and CRC mismatch is common. In fact, if a track is short enough to fit entierely in the zone where errors are still correctable, I get no errors and wrong CRC all the time.
Go to the top of the page
+Quote Post
liekloo
post Dec 5 2002, 10:12
Post #33





Group: Members
Posts: 456
Joined: 22-March 02
From: Belgium
Member No.: 1596



QUOTE (Pio2001 @ Dec4 12:39)
So it's not 1/8 but 1/2.


Then it's even better! laugh.gif

(BTW My personal experiences aren't really useful here - most of my disks are in a too good condition, to get useful results.)

Of course scratching them would give me the opportunity to do a lot of testing...
but I don't think I'll do that wink.gif


--------------------
"E S S E N T I A L" Guide for E A C :

http://users.fulladsl.be/spb2267/
Go to the top of the page
+Quote Post
liekloo
post Apr 23 2003, 20:49
Post #34





Group: Members
Posts: 456
Joined: 22-March 02
From: Belgium
Member No.: 1596



Tests have been done meanwhile, and a conclusion is posted here. smile.gif


--------------------
"E S S E N T I A L" Guide for E A C :

http://users.fulladsl.be/spb2267/
Go to the top of the page
+Quote Post

2 Pages V  < 1 2
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 17th April 2014 - 14:17