[Russ Arcuri]

Quote:

Originally posted by Joe Bloggs I haven't read much of this thread, to tell the truth, and someone might have addressed this before, but in that case I'd just say it again--

Joe, we've been through all that already, and you might be interested to learn that most D-D copies of CDs using computers are bit-for-bit exactly like the originals. At least, in my experience they are.

If you want to participate in this thread in a meaningful manner, you'll probably have to go back and read it all. The question we're on now (at least, I think it's the question we're on) is whether either random errors in a D-D copy OR jitter introduced in a D-D copy, OR a combination of those two phenomena can result in copies that sound consistently brighter than the source CD. I'm not even going to try to summarize the discussion thus far -- if you have the patience, read it all and see if you can offer anything not already mentioned.

 

[Joe Bloggs]

Russ,
Does your diagnostic have C1 and C2 error statistics? If you go to www.tomshardware.com and look at some (old) reviews of CDRW drives, you'd see that *every* drive burns CDRs with C1 errors on them. The reviewer doesn't even need to burn more than 1 CD from each drive to get a decent sampling size to assess error rate--there can be up to hundreds of C1 errors in *one* newly-burned CDR. A C1 error means that an error was detected and corrected by the C1 level of error-correcting code. Since C1 error correction is built into all CD drives, if C1 errors are all you get, you wouldn't even notice anything going wrong. Different CDRW drives have different error rates, but *no* drives has a rate of 0!

I have no doubt that bad digital copies of music CDs do not usually just sound brighter. However, I would say that it is much easier to make bad copies of music CDs than you think. Do you use EAC or some other rippers with secure modes? If not, then half of your copies may not be bit-exact

As for your argument with morphsci...

Hey, there's always that one in a billion trillion zillion chance that the errors will mutate the CDRs to produce a brighter recording and nothing else!

 

[dhwilkin]

The reason, Russ, that you can't think of any way a burned CD could be brighter is because of one of your basic assumptions. Namely, any jitter introduced is going to be random. What if the jitter is not introduced in a random fashion? What if the device you burn on has a clock (or is tied to a clock) that is consistently just a little faster or slower than it should be? This might result in the laser burning slightly longer or shorter pits, and having longer or shorter gaps between the pits. So, even though the data is bit-perfect, the timing is consistently slightly off. This is a possible explanation for a CD-R sounding slightly brighter. Now, all this is conjecture on my part, and I don't know which timing error (faster or slower) would cause brightness, or even if it would. But it's food for thought.

 

[aos]

>I don't see anything here that I necessarily disagree with. However, I do disagree with your implicit assumption that digital data representing text and digital data representing music are somehow fundamentally different. They're not. It's only after D/A conversion that the data becomes "a change of air pressure," and even then, the effects of jitter are ill-defined.

Maybe I didn't express myself clearly. But I was not saying that the digital data representing music and data are different in nature. In fact that is what's probably causing all the confusion in many people. Storing and retrieving is all the same. Error correction on audio CDs is different but that's beside the point. What I wanted to say is that the process of conversion is completely different, as the nature of information was. Actually, in case of storing something that can be fully described with symbols - an email or word document with all the fonts and borders, or a computer program - there is no fundamental difference between source and storage. Coversion to and from is just conversion of number base, as I was trying to say. But people think that the same would apply to any other information stored, such as audio or video. It does not, as the information stored is just a list of coefficient from the reconstruction formula, for example. The whole formula is depends on several parameters and the information stored is just one of them (although most important). The reason conversion is not simple is because what we are storing is not compatible with the storage. There is in fact information loss. That's what I was trying to say - not to expect true to the original and reproducable audio effect from two different systems just because digital symbols fed to them are identical.

The effects of jitter are not ill-defined. They can be PRECISELY described mathematically. Reconstructed signal is a sum of equidistant pulse responses in the shape a * sin(x) / x, where a is the amplitude of sample at a particular point - that's the value from our digital storage. The sum goes from the current moment to as far to the past as the beginning of audio track. When you write the formula down, you'll have a n * T factor somewhere where n goes from 1 to the number of samples used up so far. But if there is jitter, then you have to scrap the nice short formula and get rid of the sum symbol and add pulse responses one by one, centered at actual time they happened due to variations from jitter. You could possibly try to simplify it if you know parameters of the jitter. But you can always simulate it and see what you get. In practice this is even more complicated as the reconstruction formula requires Dirac pulses (infinitely narrow but infinitely tall, covering surface of 1) and all we have is usually DA converter holding the sample value on its output during its period. With capacitor based converters (unlike R2R networks) that value won't even be constant (although it will be constant enough).

Jitter is not necessarily random. Try to think from a different angle. I can't see either how a random fluctuation in jitter could cause a non-random effect. But according to one engineer who actually measured it as he was designing a reclocker, crystals have the tendency to modulate the fundamental with a very low frequency (a few Hz) sine. So your 12MHz would go a little faster and then a little slower in the course of a few seconds. Definitely not a random effect. In reality jitter on DA converter is caused by many factors, such as noise, EMI, interference reflections from the signal itself at the ends of the cable, and the variations in the clock. Some manufacturers give the graphs of variations in "phase noise" versus frequency. Obviously, some of the factors are random and some are not. So I can believe that a non-random effect such as audio sounding brighter are possible. I'm not saying that's what's happening, just that it seems possible. Formulas exist, whoever has time, knowledge and will to tackle it, can.

 

[Russ Arcuri]

Quote:

Originally posted by Joe Bloggs I have no doubt that bad digital copies of music CDs do not usually just sound brighter. However, I would say that it is much easier to make bad copies of music CDs than you think. Do you use EAC or some other rippers with secure modes? If not, then half of your copies may not be bit-exact

Alright, allow me to be as specific as possible, then.

Every data CD that I have burned, I have verified. Never a single error, and never a coaster, going all the way back to 1997, when I first obtained a CD burner. I tell people who have experience burning CDs this, and they assume I'm exaggerating or lying. I'm not. I don't know if it's because of the hardware I used (SCSI CD-R burners connected to SCSI-equipped Macs), the software (Toast in various incarnations) or simply my lucky nature. But the fact remains -- I've never made a single coaster or gotten a single byte error on any burns. As to audio CDs:

My software does not allow for verifies of audio CDs. Before you jump in to explain that these are two different situations, read on. I am not assuming that because I get no errors with data CDs, that I'm getting no errors with audio CDs. In fact, it bothered me that I could not optionally verify audio burns. So I decided to try an experiment. I have repeated the experiment two or three times since the first time I tried it, and the results are always identical. First, I extract several tracks from a given audio CD. Then I extract the same tracks from the CD-R copy. These files are then compared using data-comparison programs, such as gnu diff. It's worth noting here that it is entirely possible for both copies of each file to contain the same error, if the error copies "true" from the source to the CD-R, and diff would not indicate a problem. But the fact remains that the CD-R copy is indeed identical to the original, errors (if they exist) and all.

Still, while the fact that I get no errors on data CDs does not necessarily mean that I'm getting no errors on audio CDs, it does indicate that at least with my hardware/software/etc, the error rate is vanishingly low. This, coupled with the fact that my diff comparisons show no difference in audio tracks, indicate to my satisfaction that I'm getting copies with no errors. I don't know whether the type of drive matters, but the tests I have conducted were all done with CD-Rs made with CD-R drives, NOT CD-RW drives.

And yet still, even if my entire procedure is flawed, and I AM getting errors from source to copy, the fact remains that such errors would NOT manifest themselves simply as brighter copies of the source. I don't think Joe Bloggs is disagreeing with me on that point, right? As I've said several times already -- random errors in the data can NOT have non-random results in the music. "Brightness" is a non-random result. Quote:

So, even though the data is bit-perfect, the timing is consistently slightly off. This is a possible explanation for a CD-R sounding slightly brighter. Now, all this is conjecture on my part, and I don't know which timing error (faster or slower) would cause brightness, or even if it would. But it's food for thought.

Dave, I considered this possibility as well (in my deleted essay). The problem is that if the timing errors are consistent, then they aren't errors at all. Most DACs derive their clocks from the data; it's not compared to an outside clock. So if each bit is some number of pS too late OR early compared to an imaginary standard clock, and this error is consistent, the DAC simply accepts the data's clock, and each sample is following 1/44,100 second after the previous one. "Jitter," so to speak, implies non-consistency in timing in its very name -- a jittery hand shakes back and forth slightly -- if it's "shaking" back, but not forward, it's moving, not jittering. Get what I'm saying here? Jitter is, by definition, inconsistent. Besides, data is read off of audio CDs at a fixed rate -- seventy-five 2352 byte sectors per second. You can't deviate from this even a little bit and still have it work.

I hope I'm getting this across clearly, but I suspect I'm not. That's part of the reason I deleted the original essay -- this part, in particular, was a mess.

 

[Russ Arcuri]

Quote:

Originally posted by aos So your 12MHz would go a little faster and then a little slower in the course of a few seconds. [...] Obviously, some of the factors are random and some are not. So I can believe that a non-random effect such as audio sounding brighter are possible.

Okay, I think I see what you're saying here. The problem then is that the audibility of the effects you're describing here are questionable, and even if they could be heard, they would fluctuate over the course of a few seconds. Rather than so-called "jittery" CD-Rs sounding consistently brighter than the source, they would sound brighter for a few seconds, darker for a few seconds, brighter, etc. Note that this is assuming the effect was brightening/darkening after all -- it might just as easily be fluctuations between normal dynamic range and reduced dynamic range, or fluctuations between a noise floor of -100 dB and -90 dB, for example. There are many other more plausible (IMO) effects than a simple brightening or darkening of the music. Try "muddying" vs. "clarifying," if you get my sense.

I'm still not convinced this is the case anyway. Jitter acts on sound in a micro-, not macro- fashion. I still say distortion of the audio due to jitter would muddy the sound more than any other effect. Even the articles I have read on jitter written by "true believers" in its audible effects, no one has said the music sounded "brighter" in blind testing. In fact, one of the articles referenced earlier in this thread said that the Sony pressings made with the low-jitter process sounded "more true to reality" than the jittery pressing; not brighter or darker. It seems to me that any report on clearly audible differences due to jitter would make specific mention of brightness (harshness?) or darkness if it was heard.

 

[Joe Bloggs]

Oh Jesus, come on. The guy who raised this **** in the first place later stated that he didn't even have the original to compare to. What's all the fuss about? It's not as if someone had consistently picked one CD in a double-blind test and called it 'brighter'...

 

[gaineso]

Quote:

Originally posted by Joe Bloggs Oh Jesus, come on. The guy who raised this **** in the first place later stated that he didn't even have the original to compare to. What's all the fuss about? It's not as if someone had consistently picked one CD in a double-blind test and called it 'brighter'...

I've got a feeling that we're way into theoretical differences and there probably isn't a bleeding edge playback system, set up with two transports feeding the same DAC and on to whatever final transducer to produce sound that would reliably produce an audible difference that someone would reliably be able to identify as Original or Copy.

Is that a run on sentence?

 

[Tim D]

Brighter is a subjective term. It is NOT an objective term describing the rise of higher frequencies. Brighter *can* be a subjective term to describe an objective consequence of a rise of higher frequencies but not always nor interchangable.

Another possibility is, errors which introduce themselves in higher frequencies, perhaps something like phase errors or other distortions, become ANNOYING, and stand out, and hence higher frequencies that stand out regardless of not being "EQ'd" elicit a subjective response of brighter. Jan's x-feed does not EQ bass to be lower, but the subjective opinion typically is that the bass does diminish. Jan's explanation being that when bass is more even on both channels, it does not stand out as much as when bass is uneven between channels.

D-D copies are usually bit perfect and jitter is not an issue if all you are doing is D-D. However since we are talking about a persons experience with CDR's played back through his system, there are things that DO matter. It would be the last media for which he is playing. The original jitter on the original CD does not matter. So in D-D-D-D-D-D-D-D-A, the only thing that matters would be the last D-A...but that does not mean the jitter that exists on the last transcription does not matter. The jitter values of the prior dubs do not matter, but the last one before the conversion does(because there are now new constraints on real-time performance).

What matters, is when playing two different CD's, their differing jitter qualities can matter. Jitter matters at the DAC, from which point all the qualities of jitter in the chain down to the source media can affect playback quality. The transports ability to transcribe data, without bothering the DAC process matters. Since in many CD players power is shared, there can be issues. If you want to burn a new CD by first ripping onto your HD, than burning a new disc, than it does not matter which CD you use as long as there are no extreme unrecoverable errors. Since the person in question is not saying that the disc is brighter as it lays in bits on his hard-drive, but during playback, than jitter from the media to the transport, to the DAC, does matter. Once you hit a DAC process, jitter matters. Do we know if jitter caused his subjective response? No. We do know however that jitter is one variable of many which affects playback quality which in turn is one variable of many that affects subjective response.

Plextor just came out with a new CD burner that addresses jitter with its ability to vary laser strength.

But again I say that it is a possibility instead of an impossibility, but proof of jitter affecting audio does not automatically qualify everyone's subjective response. Just as discovering that there may have been life on mars does not all of the sudden qualify every farmer's statement of UFO citings. However it is a possibility as opposed to an impossibility.

 

[Dusty Chalk]

Quote:

Originally posted by Russ Arcuri My software does not allow for verifies of audio CDs...First, I extract several tracks from a given audio CD. Then I extract the same tracks from the CD-R copy. These files are then compared using data-comparison programs...

Holy crap, that's not a test! A - it's not realtime, as CD playback is. B - it's after the CD redbook playback error correction techniques -- yes, rippers use that too (they would have to -- .WAV files do not contain the entire redbook information).

Sorry, you've just completely undermined my confidence in your statements that you have never had an error. The error rates that contribute to jitter can just as well be before these error correction techniques as after. Quote:

Originally posted by Russ Arcuri Most DACs derive their clocks from the data...

Huh?!?!? No they don't. Why is it so important that CD players and DAC's have nice clocking chips, such as those sold by lcaudio.com? Why do we have jitter reducers, which buffer and reclock the incoming data? Why is the input buffer so important on a standalone DAC? There is a lot of misinformation in that one statement.

Now, I admit that badly designed DAC's can be affected by badly clocked data -- that's the crux of my argument, but the converse is also true. You can have all the jitter you want in the digital domain, as long as (a) you don't lose data and (b) you "fix" it all in the last stage, then the rest of it is unimportant. Unfortunately, (b) is rarely true. Quote:

Originally posted by Russ Arcuri Jitter is a property of the digital bitstream...

The definition of jitter is "a deviation of the digitized signal from its optimal value" -- in this case, the actual signal its supposed to represent. Seems to me that that has everything to do with the A/D and D/A processes. Quote:

Originally posted by me That's just wrong. Jitter -- inasmuch as it matters -- is a property of the A/D and D/A processes.

Quote:

Originally posted by Russ Arcuri This statement is simply not true. Jitter is a property of the digital bitstream...

I stand by my original statement. Please note: "inasmuch as it matters"...meaning to audiophile quality recording and playback. Quote:

Originally posted by Russ Arcuri Let's use aos's LCD screen analogy. Is a random perturbation of the spacing of the individual pixels going to result in a screen that looks normal, except that it's simply brighter? This would require that all pixels smaller than the standard size represent dark colors, and all pixels larger than the standard size represent brighter colors. But there's nothing inherent in the 'screen jitter' which would cause this to be so. i.e. We'd fully expect that some of the shortened pixels be bright colors and some of them be dark. The analogy to music is apt -- we'd expect the 'screen jitter' to muddy the picture somewhat, depending on the severity, just as I would expect jitter in a digital audio bitstream to muddy the sound if it were severe enough to be heard at all.

Horse hockey, I say! This is complete jibberish. The only way that jitter is analogous to screen brightness is if there is some sort of analog reconstruction, which there isn't. The timing problems -- and effects -- caused by jitter would be more analogous to streaming video, and the effects of timing problems on the motion described by that video. Quote:

Originally posted by Russ Arcuri If you want to participate in this thread in a meaningful manner...

Don't discount what Joe has to say just because he hasn't read the entire thread. I do believe his experience with CD burning is more valid than this entire academic discussion (including mine!).

 

[Dusty Chalk]

Quote:

Originally posted by gaineso I've got a feeling that we're way into theoretical differences and there probably isn't a bleeding edge playback system, set up with two transports feeding the same DAC and on to whatever final transducer to produce sound that would reliably produce an audible difference that someone would reliably be able to identify as Original or Copy.

I disagree. I do agree that the system is very important, but I don't think enough attention is paid to the D/A stage in all cases as to make that all-important. I think that there are some mid-fi systems with a very good D/A stage -- mine, for example, the MSB Tech Link DAC III has a very good reclocking digital input stage that does wonders to, for example, a DVD player -- that could outdo a high-fi stage, and vice versa.

But I would also agree that there are other things to consider, and some of them may render jitter as inaudible.

 

[Joe Bloggs]

Yep, as I said, *all* CDRs have error rates, it's just that newly-burnt CDRs almost always have error rates low enough to be corrected completely by the C1 error correction layer. Now it is my understanding that as long as the errors can be corrected by C1 and (maybe) C2, the music CD can be played back flawlessly in real time (not so sure if this is the case with most players if C2 is required...)

NOW, it is my experience that on many CDs that I have ripped using EAC, there is at least one or two trouble spots that fall through both C1 and C2 error correction, meaning that C1 and C2 put together can only indicate that there is are errors there, but not how to correct it, so serious is the error.

IF such discs are copied in burst mode, (as I would argue they have to be if you are copying from one drive to another on the fly) (which I never do since I've only got one drive

) these trouble spots can't possibly be copied correctly.

(Even so, the copied CDR may well sound pretty much the same as the original CD, even at the trouble spot, because real-time playback is just as unable to make sense of the trouble spot on the original CD as it is of the flawed copy of the trouble spot on the CDR.)

 

[Hirsch]

It is possible to produce an error-free CD-R, just not likely unless you are very careful indeed, which includes careful matching of media to your writer. CD-R Diagnostic has the capability of disabling error correction so that you can see the actual error rate of a CD recording.

I live with the error correction, but I know one person who uses this program to analyze disks he prepares for others, and he won't release a disk unless it is completely error-free. It can be done...

 

[Russ Arcuri]

Quote:

Originally posted by DustyChalk Holy crap, that's not a test! A - it's not realtime, as CD playback is. [...] Sorry, you've just completely undermined my confidence in your statements that you have never had an error.

We're NOT talking about CD playback here -- we're talking about the integrity of the data in a CD-R as compared to the original. What is it that you don't understand here? I'm comparing the original to the source. No difference in the bits. Period.

Your concern about whether this is done on-the-fly is irrelevant to the question: Is the data the same? The answer is "yes." Now, the CD-R may or may not have more or less jitter than the source, but the data -- the ones and zeros -- are the same. Remember that when the CD-R is burned, there's no going back to fix questionable sectors. Quote:

I said: Most DACs derive their clocks from the data... to which DustyChalk replied: Huh?!?!? No they don't...

No, you've got this wrong too. The VAST MAJORITY of DACs on the market do exactly that. Most articles about jitter point this out early on. For example, let's look at the very first article Tim D posted a link to, which contains the following statement: "A typical D to A converter derives its system clock (the clock that controls the sample and hold circuit) from the incoming digital signal." (Emphasis mine.) Here's a link if you want to check it yourself: http://www.digido.com/jitteressay.html

Note too that jitter could be made a non-issue if only manufacturers would re-clock the data before feeding it to the DAC. As you correctly pointed out, there are manufacturers that make the correct parts to do this, but sadly they are seldom used except in high-end gear. Of course, you confuse the issue with your very next paragraph, in which you basically agree with what I said: Quote:

Now, I admit that badly designed DAC's can be affected by badly clocked data -- that's the crux of my argument, but the converse is also true. You can have all the jitter you want in the digital domain, as long as (a) you don't lose data and (b) you "fix" it all in the last stage, then the rest of it is unimportant. Unfortunately, (b) is rarely true.

See? How is that any different from what I said? Quote:

The definition of jitter is "a deviation of the digitized signal from its optimal value" -- in this case, the actual signal its supposed to represent. Seems to me that that has everything to do with the A/D and D/A processes.

Why do you insist on interpreting the definition? "A deviation of the digitized signal..." -- i.e., the bitstream. Yes, it will affect the sound once converted -- I never said it wouldn't -- but "jittery" is a description of the mistimings of the 1s and 0s, not a description of the reconstructed waveform on the output side of the DAC. If jitter was a property of the sound itself, reclocking the data before D/A conversion could not possibly fix the problem. Ergo it is a property of the digital bitstream. We keep going back and forth on this particular point, and I have no idea why. Quote:

Horse hockey, I say! This is complete jibberish. [Referring to the "computer screen" analogy.] The only way that jitter is analogous to screen brightness is if there is some sort of analog reconstruction, which there isn't. The timing problems -- and effects -- caused by jitter would be more analogous to streaming video, and the effects of timing problems on the motion described by that video.

Look, it wasn't my analogy, it was aos's. And unless I'm mistaken, he's arguing the same side of this as you are. I was working with what I was given. Besides, the statement I highlighted in blue above is completely wrong anyway -- a standard television screen is absolutely an analog reconstruction of (in some cases, such as with DVD) digital data. I don't think the analogy is worth pursuing further anyway. But your criticism of it as "not applicable" should be aimed at aos, not me. Quote:

Don't discount what Joe has to say just because he hasn't read the entire thread. I do believe his experience with CD burning is more valid than this entire academic discussion (including mine!).

I wasn't discounting what he had to say -- I was simply pointing out that his concerns had already been discussed. Heck, he actually supported my main point -- that random errors in the data can't have a non-random result after conversion. I'm not discounting him! Speaking of which, I would like to quote another point from the first jitter reference: "The audible effect of this jitter is a possible loss of low level resolution caused by added noise, spurious (phantom) tones, or distortion added to the signal." Again, no mention of "normal" changes in EQ, such as brightening compared with the source.

Once again, I find myself awake fairly late (hey, I'm a father of three young kids). So I'm going to address one more point and then head off to bed. Tim D wrote (in part): Quote:

Brighter is a subjective term. It is NOT an objective term describing the rise of higher frequencies.

Sorry, Tim -- but it is EXACTLY a term describing a rise of frequencies in the treble range. Whether you find this objectionable or not is a subjective determination, but "brighter" describes (and has for as long as I can remember) a rise in treble energy. If you don't like that definition, then kindly substitute "increased treble energy" for "brightened" everywhere I used the term. The points are still valid.

And with that, I bid you all "goodnight."

 

[gaineso]

Quote:

Originally posted by DustyChalk I disagree. I do agree that the system is very important, but I don't think enough attention is paid to the D/A stage in all cases as to make that all-important. I think that there are some mid-fi systems with a very good D/A stage -- mine, for example, the MSB Tech Link DAC III has a very good reclocking digital input stage that does wonders to, for example, a DVD player -- that could outdo a high-fi stage, and vice versa. But I would also agree that there are other things to consider, and some of them may render jitter as inaudible.

Disagree all you want, but you haven't addressed my statement.

An original, factory CD and a copy, made with some care, played on two identical transports, your choice, through the same DAC, again your choice, switched however you would choose and played through the same downstream chain. I do not believe that anyone will be able to reliably and cosistently pick the original from the copy with blind switching. I doubt if you would even be able to hear a difference at all, in no small part to what you have stated about reclocking and other error correction.

The copy may not be an absolutely perfect duplicate of the original. So what. The primary issue is, how does it sound when played back? No matter what system, lo, mid, hi or beyond the bleeding edge. If it sounds good, what's the issue? If it sounds bad, how did the original sound? Generally bad also. Unless there was a real screw-up in the copy process.

This is an interesting thread, and I have enjoyed reading it. There have been a tremendous number of OPINIONS expressed here. However, remember that some of the very best audio engineers in the world also disagree on some of these very issues. We all know that all errors must be reduced where possible. We should all agree also that the only thing that really matters in the end is HOW DOES THE MUSIC SOUND?

Unfortunately, in this type of pursuit of audio, or any other, perfection, we seem to lose the beauty in the technology. If it sounds good to you, how it sounds to someone else doesn't matter. If the only difference you can detect is, "I know it's a copy" pyschology is getting in the way of the enjoyment.

Meanwhile, I'm listening to "Showdown" when I should be sleeping so, enjoy the music.