unitserve rips vs core rips

David Hendon posted:

Keler Pierre posted:

Bart posted:

Keler Pierre posted:

And also we are just french guys, good enough to eat frogs....so our questions don't interest this english forum. ( je plaisante).

 

This must be terrible to go through life that way (and I don't mean as a Frenchman).

i was joking 

Joking is good. Many of us take ourselves too seriously much of the time.

(Personally I prefer escargot, hopefully with a well-chilled Loire wine to follow that lovely buttery, herby, snaily taste.......)

best

David

you have good tastes. escargots de bourgogne, good wine....nice program.  But i apologize for other who may found me agressive....

No worries Pierre, and here in my home we love to visit France and drink mostly Bourgogne.  Whites and reds.  And Sancerre, Pouilly Fume, Bordeaux of course, et les Champangnes.  And some Northern Rhones...and of course Southern Rhones...  You get the idea!

Stevesky posted:

Simon-in-Suffolk posted:

Steve - the findings I have provided are nothing to do with CD offsets as far as I am aware .. I assume you  are correcting posts that were talking about CD offsets etc. You can private mail me if you want more information - but my finding using debug tools have shown that WAV files produced independent of offsets are incorrect (in my opinion as information is lost) compared to other rippers- and do produce sample discontinuities in some cases, noticeable with continuous CDs  - again I assume you are looking into this - again if you want more information from by debugs please feel free to  contact me privately - and the differences I measured were greater than 48 stereo samples

 

Hi Simon, 

From the various checked that we did last week we have seen no evidence so far of substandard rips.  For example we did Core vs dbPoweramp vs EAC vs two different Unitiserves and apart from offset and a subtle WAV header / tagging difference the audio data part is identical across all of them.   We also tried different drives.  

As you're part of our beta test program feel free to open a thread up there and we can go through various samples.   Unless a disc is damaged the rip should be bit perfect.  

Best regards

Steve

 

I think we can all relax now, and leave it up to the experts !

....what a weekend, how many other home entertainment companies would be engaging with their customer base  like this !

Bart posted:

No worries Pierre, and here in my home we love to visit France and drink mostly Bourgogne.  Whites and reds.  And Sancerre, Pouilly Fume, Bordeaux of course, et les Champangnes.  And some Northern Rhones...and of course Southern Rhones...  You get the idea!

Not unlike that here! Right now. Having started the evening with a straightforward Chablis, I am currently well into a Pauillac, Ch.Lynch-Moussas 2006 which truthfully I think I should have left alone for a couple more years before pulling the cork.

best

David

David Hendon posted:

Bart posted:

No worries Pierre, and here in my home we love to visit France and drink mostly Bourgogne.  Whites and reds.  And Sancerre, Pouilly Fume, Bordeaux of course, et les Champangnes.  And some Northern Rhones...and of course Southern Rhones...  You get the idea!

Not unlike that here! Right now. Having started the evening with a straightforward Chablis, I am currently well into a Pauillac, Ch.Lynch-Moussas 2006 which truthfully I think I should have left alone for a couple more years before pulling the cork.

best

David

Oh we love Chablis.  I get sucked into buying some premier cru and grand cru chablis almost every year.

Reds - I'd rather drink them too young than too old.  But for New Years dinner had a 1982 Ducru Beaucaillou that was in PERFECT condition.  I bought 2 and have one left

That iconic yellow label. Not something I can run to mostly, but I do have a special weakness for Ch Leoville das Cases, which is also a Naim-like habit, price-wise.

best

David

Stevesky posted:

Dear All,

In Naim R&D we have been following this thread and felt it was time to respond, as there is some misinformation and wrong assumptions.   I have also been in contact with the author of the EAC library, Andrea Wiethoff and designer of AccurateRip to review this thread and sanity check a few other areas.

On CD red book ripping there is no absolute 'music audio samples of track 1 start here', unlike reading a file from a computer hard drive or CD data disc.  It varies on a per disc pressing master basis, but typically a large amount of silence is mastered in on start of track 1 and there are silence boundaries between tracks. 

With a gapless album the exact point of where track x ends and track x+1 starts is consistent (there are no lost samples), but based on where the chop between tracks is done, a few samples will end up on one side or the other of the transition point.

 So to visualize it:

 With one offset:

|  Track 1  |  Track 2 |

|00abcdefghijklmnopqr00|

With a different offset:

|  Track 1  |  Track 2 |

|0abcdefghijklmnopqr000|

 Appreciating the TRUE physical start of a CD is extremely hard to resolve, based on how CD as a physical format is defined and it can be skewed +-75 sectors when the disc is mastered and pressed.   When the AccurateRip system was created, Andre experimented with a wide range of CD's and came up with a relative offset compensation system that seemed sensible.  Since doing that there has been evidence based on someone using professional CD mastering equipment that the calculation is actually 30 samples out, but in practice it does not matter.    Samples intended for end of one track can end up on the start of the next one.  This will be typically digital silence or tape hiss and its a tiny period of time (for example a 6 sample delta is 136usecs).   Offset compensation is only really needed if you want to do a comparison system of rip x from drive y, is the same as rip x from drive z. It doesn't mean the actual rip reflects the original master files on start/end transition points - just that it’s always the same for the same CD.  

 With the Naim servers we aim to ensure that the offset is close to theoretical CD pressing offset.   In Core we believe this resolves as:

+48 sample drive offset - (30 samples + 6 samples) = ~12 theoretical real world sample offset.    On Serve it is ~18 real world samples.  That is why there is a difference if you compare the two.   Please take into account the far more larger real world factors mentioned above though.

 Sample offset does not affect the sound as it's a case of the stereo samples for one period are just offset by (4 bytes x numberSamples) in memory.  In practice the whole track is not held in memory, it’s being paged in to circular buffers, where the current read vs write positions could be anywhere on each time it plays based on what has played before.

 Examples where 'a' is current read position, 'q' is current write position:

Circuilar Buffer ->  |opq      abdcdefghijklmn|

Circuilar Buffer ->  |     abdcdefghijklmnopq |

 So to summarize and key facts:

1. Your rips are good. The audio is correct and all you are seeing is logical offset differences where Core is slightly different to Serve.
2. There are no samples lost on gapless material. Just look at the next track to find them.
3. In CD extraction the offset of the rip is a logical value and typically is not precisely aligned to the physical disc offset.
4. A traditional CD player will chop off many samples when doing next/prev/goto track x, hence no music is placed on these track transition points for compatibility sake.
5. When differencing WAV files remember each sample is 16bits (2x 8 bytes) and they go in Left, Right pairs. So 4 bytes per stereo set of samples.

 

If you would like any extra clarifications our help desk can provide further information, but naturally feel free to chatter between yourselves here.

Best regards

Steve Harris

Software Group Director

Steve, could you please elaborate on the "disinformation" and "wrong assumptions" you have found in this thread?

To the best of my understanding, the comparisons done by different contributors have independently established that, for single tracks and consistently through 4 tracks, the differences between Jan's US and Core rips can be described pictorially as the differeces between

00abcdefghijklmnopqr0000

and

0000abcdefghijklmnopqr00

I believe that these observations are non debatable. But, as I mentioned, I am not sure which conclusions (if any) can be drawn from these observations. Thus, I would be grateful if you could answer the following questions:

1)  Are the above differences between US and Core rips to be expected?

2) If the answer to 1) is positive, do you expect these differences to be identical on all US and Core units or do you expect the differences to vary between units?

3) If the answer to 1) is positive, can you guarantee that these differences have no impact on the sound quality and on gapless playback capabilities?

Best, nbpf

Bart posted:

No worries Pierre, and here in my home we love to visit France and drink mostly Bourgogne.  Whites and reds.  And Sancerre, Pouilly Fume, Bordeaux of course, et les Champangnes.  And some Northern Rhones...and of course Southern Rhones...  You get the idea!

We do! Lived in Lyon for two years and came back with a Côte du Rhone habit. The high points were the St-Josephs.

We've drifted to wine, so the thread seems to have run its course.

Cheers all and thanks !

Jan

Stevesky posted:

Simon-in-Suffolk posted:

Steve - the findings I have provided are nothing to do with CD offsets as far as I am aware .. I assume you  are correcting posts that were talking about CD offsets etc. You can private mail me if you want more information - but my finding using debug tools have shown that WAV files produced independent of offsets are incorrect (in my opinion as information is lost) compared to other rippers- and do produce sample discontinuities in some cases, noticeable with continuous CDs  - again I assume you are looking into this - again if you want more information from by debugs please feel free to  contact me privately - and the differences I measured were greater than 48 stereo samples

 

Hi Simon, 

From the various checked that we did last week we have seen no evidence so far of substandard rips.  For example we did Core vs dbPoweramp vs EAC vs two different Unitiserves and apart from offset and a subtle WAV header / tagging difference the audio data part is identical across all of them.   We also tried different drives.  

...

Steve, you are missing the main point of the thread, I believe.

There have been no reports of differences in the non-zero audio part of US and Core tracks. On the contrary, in Jan's 4 tracks, these parts have been confirmed to be identical by inspection.

At the same time, it has been found (also by inspection) that the US and Core tracks exhibit different amounts of leading and trailing zeroes in all 4 tracks.

What has been debated in this thread is whether the observed differences in leading and trailing zeroes are to be expected or not and whether they could have an impact on sound quality and on gapless playback capabilities or not.

I would be grateful if Naim could provide a precise answer to this specific question.

For heaven's sake will you all shut up about bloody wine! 

Yours sincerely, 

Tony (halfway through a dry January)

nbpf posted:

Stevesky posted:

Simon-in-Suffolk posted:

Steve - the findings I have provided are nothing to do with CD offsets as far as I am aware .. I assume you  are correcting posts that were talking about CD offsets etc. You can private mail me if you want more information - but my finding using debug tools have shown that WAV files produced independent of offsets are incorrect (in my opinion as information is lost) compared to other rippers- and do produce sample discontinuities in some cases, noticeable with continuous CDs  - again I assume you are looking into this - again if you want more information from by debugs please feel free to  contact me privately - and the differences I measured were greater than 48 stereo samples

 

Hi Simon, 

From the various checked that we did last week we have seen no evidence so far of substandard rips.  For example we did Core vs dbPoweramp vs EAC vs two different Unitiserves and apart from offset and a subtle WAV header / tagging difference the audio data part is identical across all of them.   We also tried different drives.  

...

Steve, you are missing the main point of the thread, I believe.

There have been no reports of differences in the non-zero audio part of US and Core tracks. On the contrary, in Jan's 4 tracks, these parts have been confirmed to be identical by inspection.

At the same time, it has been found (also by inspection) that the US and Core tracks exhibit different amounts of leading and trailing zeroes in all 4 tracks.

What has been debated in this thread is whether the observed differences in leading and trailing zeroes are to be expected or not and whether they could have an impact on sound quality and on gapless playback capabilities or not.

I would be grateful if Naim could provide a precise answer to this specific question.

How can the zeros impact on sound quality if all they are is silence? And surely if the music is gapless there is no gap to be affected. 

Maybe it's time to stop demanding PRECISE answers RIGHT NOW and wait till you audition a Core. Hearing will, surely, provide the answers. I strongly suspect, from what I've read of this interminable thread, that it's a lot of fuss about nothing. 

Guys the zeros is a red herring. They key thing are the track boundaries. These will vary from different CD masters, and potentially different ripper CD ROM types as we see here - and where the boundaries and the boundary detection varies the  rips WILL also slightly vary.

The advice would appear to be best not to mix and match extracted files from different rippers of the same CD - especially for continuous replay.  This way, by using the SAME ripper for the SAME album,  there should be no loss of samples between tracks unless the disc is faulty. I suspect most will never mix and match rips from different rippers of the same album anyway, so the impact is minimal to non existent in the real world...

I think given the above there is nothing further I can add here.

Hungryhalibut posted:

nbpf posted:

Stevesky posted:

Simon-in-Suffolk posted:

Steve - the findings I have provided are nothing to do with CD offsets as far as I am aware .. I assume you  are correcting posts that were talking about CD offsets etc. You can private mail me if you want more information - but my finding using debug tools have shown that WAV files produced independent of offsets are incorrect (in my opinion as information is lost) compared to other rippers- and do produce sample discontinuities in some cases, noticeable with continuous CDs  - again I assume you are looking into this - again if you want more information from by debugs please feel free to  contact me privately - and the differences I measured were greater than 48 stereo samples

 

Hi Simon, 

From the various checked that we did last week we have seen no evidence so far of substandard rips.  For example we did Core vs dbPoweramp vs EAC vs two different Unitiserves and apart from offset and a subtle WAV header / tagging difference the audio data part is identical across all of them.   We also tried different drives.  

...

Steve, you are missing the main point of the thread, I believe.

There have been no reports of differences in the non-zero audio part of US and Core tracks. On the contrary, in Jan's 4 tracks, these parts have been confirmed to be identical by inspection.

At the same time, it has been found (also by inspection) that the US and Core tracks exhibit different amounts of leading and trailing zeroes in all 4 tracks.

What has been debated in this thread is whether the observed differences in leading and trailing zeroes are to be expected or not and whether they could have an impact on sound quality and on gapless playback capabilities or not.

I would be grateful if Naim could provide a precise answer to this specific question.

How can the zeros impact on sound quality if all they are is silence? And surely if the music is gapless there is no gap to be affected. 

Maybe it's time to stop demanding PRECISE answers RIGHT NOW and wait till you audition a Core. Hearing will, surely, provide the answers. I strongly suspect, from what I've read of this interminable thread, that it's a lot of fuss about nothing. 

There is no notion of gapless music, just of gapless playback.

It is the fact that the files obtained by ripping with Jan's US and with a Core have been found to be different (and, to some of us, to sound differently) that has naturally raised the question of whether differences in the amount of leading and trailing zeroes can possibly have an impact on the sound quality and on gapless playback capabilities.

I think that this is a very interesting question, independently of the original observations that have lead to it. I also believe that Naim experts on ripping software are better qualified for answering this question than most of us, certainly they are better qualified than me.

I am also sure that the question cannot be answered by auditions and I am not myself interested in auditioning a Core. Finally, I agree that the thread is possibly a lot of fuss anout nothing. But this is more or less true of every thread, not only in this forum.

nbpf posted:

Hungryhalibut posted:

nbpf posted:

What has been debated in this thread is whether the observed differences in leading and trailing zeroes are to be expected or not and whether they could have an impact on sound quality and on gapless playback capabilities or not.

I would be grateful if Naim could provide a precise answer to this specific question.

How can the zeros impact on sound quality if all they are is silence? And surely if the music is gapless there is no gap to be affected. 

Maybe it's time to stop demanding PRECISE answers RIGHT NOW and wait till you audition a Core. Hearing will, surely, provide the answers. I strongly suspect, from what I've read of this interminable thread, that it's a lot of fuss about nothing. 

There is no notion of gapless music, just of gapless playback.

It is the fact that the files obtained by ripping with Jan's US and with a Core have been found to be different (and, to some of us, to sound differently) that has naturally raised the question of whether differences in the amount of leading and trailing zeroes can possibly have an impact on the sound quality and on gapless playback capabilities.

I think that this is a very interesting question, independently of the original observations that have lead to it. I also believe that Naim experts on ripping software are better qualified for answering this question than most of us, certainly they are better qualified than me.

I am also sure that the question cannot be answered by auditions and I am not myself interested in auditioning a Core. Finally, I agree that the thread is possibly a lot of fuss anout nothing. But this is more or less true of every thread, not only in this forum.

I think HH was expressing his view about demanding precise answers, imediately.

I personally don't have a problem with this being an interminable thread, I find it facinating that nobody seems capable of providing a clear, concise description of how music is stored on a CD, what is ripped and how it might affect subsequent  playback. In other words, many of us don't have a clue what we're talking about -yet !

status - Normal !

Don Atkinson posted:

nbpf posted:

Hungryhalibut posted:

nbpf posted:

What has been debated in this thread is whether the observed differences in leading and trailing zeroes are to be expected or not and whether they could have an impact on sound quality and on gapless playback capabilities or not.

I would be grateful if Naim could provide a precise answer to this specific question.

How can the zeros impact on sound quality if all they are is silence? And surely if the music is gapless there is no gap to be affected. 

Maybe it's time to stop demanding PRECISE answers RIGHT NOW and wait till you audition a Core. Hearing will, surely, provide the answers. I strongly suspect, from what I've read of this interminable thread, that it's a lot of fuss about nothing. 

There is no notion of gapless music, just of gapless playback.

It is the fact that the files obtained by ripping with Jan's US and with a Core have been found to be different (and, to some of us, to sound differently) that has naturally raised the question of whether differences in the amount of leading and trailing zeroes can possibly have an impact on the sound quality and on gapless playback capabilities.

I think that this is a very interesting question, independently of the original observations that have lead to it. I also believe that Naim experts on ripping software are better qualified for answering this question than most of us, certainly they are better qualified than me.

I am also sure that the question cannot be answered by auditions and I am not myself interested in auditioning a Core. Finally, I agree that the thread is possibly a lot of fuss anout nothing. But this is more or less true of every thread, not only in this forum.

I think HH was expressing his view about demanding precise answers, imediately.

I personally don't have a problem with this being an interminable thread, I find it facinating that nobody seems capable of providing a clear, concise description of how music is stored on a CD, what is ripped and how it might affect subsequent  playback. In other words, many of us don't have a clue what we're talking about -yet !

status - Normal !

It is fascinating but, in a sense, not completely surprising: motivated by somehow unexpected results (Jan's ripping outcomes), we have tried to figure out minimal requirements for bit perfect ripping.

What we had at our disposal were limited empirical observation and the experience from past rips comparisons reported by Simon.

In particular, we hadn't any pointer to documents providing precise specifications of what it means for a set of files obtained by ripping a CD to be bit perfect. We also were not given any reference example of bit perfect rips.

Thus, it is not very surprising that we have not been able to achieve definitive conclusions. In formulating the questions stated above, I have tried to summarize the understanding that we have been able to achieve with our limited means.

I was hoping that Naim experts could provide authoritative answers to these questions. They are simple, obvious and, it seems to me, relevant to the topic. I might be wrong, of course.

nbpf posted:

It is the fact that the files obtained by ripping with Jan's US and with a Core have been found to be different (and, to some of us, to sound differently) that has naturally raised the question of whether differences in the amount of leading and trailing zeroes can possibly have an impact on the sound quality and on gapless playback capabilities.

The "fact" that they "sound differently" to "some of [you]" is a red herring.  There is nothing in hi fi that doesn't sound differently to some segment of the listeners.  The green pen sounded differently.  The "fact" that to n participants they "sound differently" proves nothing, and is not a basis for launching into "Naim must prove to us that the zeros are irrelevant to sound."  Do you want them to prove that the green pen doesn't sound differently?

I believe (and so much of this is grounded in belief, and not in fact) that some technical matters may be dismissed as irrelevant to "sound."  And we may move on. Otherwise this never ends . . . the green pen comes back . . . "etc etc etc" (said in the voice of Yule Brenner in The King and I).

Bart posted:

nbpf posted:

It is the fact that the files obtained by ripping with Jan's US and with a Core have been found to be different (and, to some of us, to sound differently) that has naturally raised the question of whether differences in the amount of leading and trailing zeroes can possibly have an impact on the sound quality and on gapless playback capabilities.

The "fact" that they "sound differently" to "some of [you]" is a red herring.  There is nothing in hi fi that doesn't sound differently to some segment of the listeners.  The green pen sounded differently.  The "fact" that to n participants they "sound differently" proves nothing, and is not a basis for launching into "Naim must prove to us that the zeros are irrelevant to sound."  Do you want them to prove that the green pen doesn't sound differently?

I believe (and so much of this is grounded in belief, and not in fact) that some technical matters may be dismissed as irrelevant to "sound."  And we may move on. Otherwise this never ends . . . the green pen comes back . . . "etc etc etc" (said in the voice of Yule Brenner in The King and I).

Bart, in contrast to what you seem to subsume, I have not been claiming that the fact that some participants have found the files to sound differently proves anything. I have simply reported Jan's observations. I personally could not hear any difference between the US and the Core rips, as I have many times reported in this thread. But this fact, too, is irrelevant to the matter discussed here. I am also not arguing that the differences between the US and the Core rips (that the files are different is not debatable) must have an impact on sound quality or gapless playback. I have simply asked Naim whether these differences are to be expected or not and whether they could potentially impact sound quality (and gapless playback capabilities) or not. You may find such questions irrelevant but, to me, they are interesting. I am not saying that they are important to me in practice because I do not own a US and I am not interested in buying a Core.

Has the past week brought any new understanding of the reasons for the differences between Core rips and US rips? Do we meanwhile know why Core rips are different from US rips? Do we know whether they should be expected to be different or not? Are both Core rips and US rips "bit perfect" in spite of being different? Can these differences lead to differences in sound quality?

Yes yes yes yes no

jon honeyball posted:

Yes yes yes yes no

Succinctly put.

best

David

David Hendon posted:

jon honeyball posted:

Yes yes yes yes no

Succinctly put.

best

David

I'm not sure. Answers that do not contribute to a better understanding are never short enough, I would say.

Moderated Post: NBPF, as I see you have begun a new thread looking at the wider question of accurate ripping, I shall now close this one.