The transport debate
no debate for me
it is all about 0s and 1s: all transports are the same!
have you a problem?
no debate for me
it is all about 0s and 1s: all transports are the same!
have you a problem?
how can digital be different
- Thread starter ChrisD
- Start date
Are you talking to me Sir?Originally posted by voodoo
You'd be a fool to think there would be anything but a difference.
If yes, I am Dr. prof. Rai please. Also, do not contradict me and stick in that tongue too.
Thank you
, twas merely a 'general comment' to be applied across the board 
PeteH
Natural Blue
Thanks for that link Graham - very interesting, the system is actually cleverer than I thought (not that I'd ever thought about it that much). The myth about the error correction "working harder" and causing "distortion" is a remarkably widespread and extremely irritating one - it must come up half a dozen times in every issue of What HiFi alone. I guess the misconceptions just arise because people are used to thinking in analogue terms and it seems intuitively obvious that if there are small errors in reading the data there must be some small distortion in the analogue output - whereas, of course, the whole point of a properly designed digital system is that nothing of the sort happens as the errors are either corrected exactly or not at all.
Ok.. lets move on from the Paper Theory...
A very good friend of mine..works in the industry.. as part of his job he has had to assess various transports/DACs/complete players to make a decision as to which one to base his companies offering on .. I've had this conversation with him more than once.. and his answer is always the same and runs along the lines of...
It's all well and good reading the theory about how CD works and coming to conclusions that the whole system is perfect .. but real life measurements of real life mass produced transports/DACs etc show real life errors.. they exist.. they can be measured.. and they vary from transport to transport and DAC to DAC. At the end of the day.. real life electronics is NOT perfect for all sorts of reasons.. budget/manufacturing variation etc etc etc..
For me, (as someone with a scientific background taught to believe in the empirical evidence of experimentation NOT the theory !), .. that makes the whole debate from a purely theoretical point of view pointless.
GTM
A very good friend of mine..works in the industry.. as part of his job he has had to assess various transports/DACs/complete players to make a decision as to which one to base his companies offering on .. I've had this conversation with him more than once.. and his answer is always the same and runs along the lines of...
It's all well and good reading the theory about how CD works and coming to conclusions that the whole system is perfect .. but real life measurements of real life mass produced transports/DACs etc show real life errors.. they exist.. they can be measured.. and they vary from transport to transport and DAC to DAC. At the end of the day.. real life electronics is NOT perfect for all sorts of reasons.. budget/manufacturing variation etc etc etc..
For me, (as someone with a scientific background taught to believe in the empirical evidence of experimentation NOT the theory !), .. that makes the whole debate from a purely theoretical point of view pointless.
GTM
PeteH
Natural Blue
So what's different between audio transports, which are apparently subject to bit errors, and CD-ROM computer transports which self-evidently are not (ie. you can read the same document any number of times and it'll have exactly the same content every time)? Not trying to have a go or anything BTW, just interested as I don't really know anything about it.
CD-ROM transports in your PC can re-read the data in the disc when errors occur..it only fails when it is impossible for the system to rectify the error given multiple attempts.. even then however.. I've recorded CD-ROMs on PCs that have been corrupted and the next PC can not read.. CD transports can't re-read the disc .. they can only rectify them in real time. ANY and EVERY "real time" error correction system has limits on the level of errors it can correct. There comes a time when there simply isn't enough information available for the system to accurately decide what the output should be.
If you want proof that the system is not infalable.. and that units are variable in their ability to handle errors.. get two CD players..(different models/makes/price range).. and an old CD that you don't like.. and make progressively more damage to the disc.. you will almost certainly find that one of the players skips before the other one does.. there is only ONE reason for this .. variation in the quality of the error correction system leading to a different ability of that system to work effectively.
GTM
PeteH
Natural Blue
Absolutely, and I'm quite aware that you can sometimes make pristine copies of scratched discs due to the ability of the computer to re-read data. And I don't think anyone doubts that bits are sometimes dropped, which is why there's an error-correction system in the first place - but due to redundancy in the word-length and other factors it's not necessary for every bit to be read correctly in order for the data to be transferred perfectly. The crux of the debate is whether or not CD players can carry on playing without a pop or skip (or other obvious anomaly) if the error-correction can't cope - which is only the case if the system starts "guessing" what the output should be as julian2002 suggested, and which FWIW I don't believe to be the case (but am still quite prepared to be proven wrong, as it'd actually be more interesting if the system can keep going when it can't read the data
) When you say there are measured error rates, is that in terms of dropped bits (which don't matter up to a certain point, as long as the 'words' are read correctly) or whole dropped words (which clearly do matter)?
Last edited by a moderator:
julian2002
Muper Soderator
i think the thing here is that people are thinking that there is only 1 error correction mechanism at work.
rs is the front line defence against errors but when it breaks (which it will, given the right / wrong? circumstances) you have 2 options.
1) re read the data until there are no errors - i know for a fact that this is the case as i've written load functions for pc, playstation, x-box, amiga, atari st, playstation 2, etc. and have had to cater for data being re-read if it's not read correctly first time. if the disk is completely shafted then the code will have to re try a number of times and then inform the user that the disk is corrupt. this is MANDATORY for most consoles and your app will not pass the manufacturers aproval process if your code does not cater for this circumstance. if transports are always 100% why is this the case?
2) for audio where absolute integrity of the data is not 100% necesary some companies also try re-reading (meridian is the one that springs to mind) however it is also possible to interpolate based on what came before (and after too, presumably using what is read correctly from the current data frame). in this case you'll get a close aproximation of the waveform. however it will not be exact, hence distortion will be created.
whether this distortion is audiable or matters is perhaps the real question but saying that a drive will always recover data from a disk 100% of the time is absolutely not true.
cheers
julian
rs is the front line defence against errors but when it breaks (which it will, given the right / wrong? circumstances) you have 2 options.
1) re read the data until there are no errors - i know for a fact that this is the case as i've written load functions for pc, playstation, x-box, amiga, atari st, playstation 2, etc. and have had to cater for data being re-read if it's not read correctly first time. if the disk is completely shafted then the code will have to re try a number of times and then inform the user that the disk is corrupt. this is MANDATORY for most consoles and your app will not pass the manufacturers aproval process if your code does not cater for this circumstance. if transports are always 100% why is this the case?
2) for audio where absolute integrity of the data is not 100% necesary some companies also try re-reading (meridian is the one that springs to mind) however it is also possible to interpolate based on what came before (and after too, presumably using what is read correctly from the current data frame). in this case you'll get a close aproximation of the waveform. however it will not be exact, hence distortion will be created.
whether this distortion is audiable or matters is perhaps the real question but saying that a drive will always recover data from a disk 100% of the time is absolutely not true.
cheers
julian
Still going I see
Well let's say that the CD/DVD Rom with large buffer CAN eliminate all errors and perfectly reproduce what's on the polycarbonate.
Have you listened to a Meridian 800:SLEEP: :SLEEP:
If that's perfection, I think I'll be trawling the record fairs for years to come and get myself a Vpi Record cleaner.
Perfection is in the ear of the beholder.
Well let's say that the CD/DVD Rom with large buffer CAN eliminate all errors and perfectly reproduce what's on the polycarbonate.
Have you listened to a Meridian 800:SLEEP: :SLEEP:
If that's perfection, I think I'll be trawling the record fairs for years to come and get myself a Vpi Record cleaner.
Perfection is in the ear of the beholder.
Antonio,
The Tact corrects for the room anomolies and yes, I do miss that when listening critically so to speak.
If only people realised that the aberrations caused by the listening enviroment far exceed any minor errors coming from the source, then maybe they would stop worrying about the scientific aspect of the hobby and enjoy the end product, namely the tunes!
Editted to add. I don't recall ever saying anything was perfect, nothing is. What the RCS will give you, IMHO, is the closest approximation to the original data recording at the listening position. Even then, tweaking for personal preference is essential, and many prefer to listen through colouration anyway.
The Tact corrects for the room anomolies and yes, I do miss that when listening critically so to speak.
If only people realised that the aberrations caused by the listening enviroment far exceed any minor errors coming from the source, then maybe they would stop worrying about the scientific aspect of the hobby and enjoy the end product, namely the tunes!
Editted to add. I don't recall ever saying anything was perfect, nothing is. What the RCS will give you, IMHO, is the closest approximation to the original data recording at the listening position. Even then, tweaking for personal preference is essential, and many prefer to listen through colouration anyway.
Last edited by a moderator:
penance
Arrogant Cock
i see digital data re-reads and data degredation at work everyday
PeteH
Natural Blue
Of course transports aren't always 100% - but all I'm saying is that I'm under the impression that when they're not you hear a pop or it skips / repeats, ie. a clearly audible discontinuity. If that's not the case - if it's possible to approximate the data as you say - I'd be very interested to do some further reading on the matter and I'd be grateful if you could point me in the correct direction
For intelligent and knowlegable people it seems to me Gerard and Julian are managing to grab the wrong end of the stick quite spectacularly.
Of course nothing can read 100% perfect data in all possible scenarios, but that's completely missing the point. The question was not about how different transports can read scratched etc and otherwise marginal discs, but about how cables and transports (and green pen - which is getting equally moronic comments elsewhere) can affect the sound for normal discs.
For misreading data to affect the normal sound (i.e. not obvious pops/clicks/track-repeats etc), misreads uncorrectable by the CIRC would have to occur at the level of 10s or 100s per second. Are you saying Julian that this is the level at which you've had experience of having to re-read discs? I would find this extremely difficult to believe, as it's certainly not mine, but if that's what your experience is, then so be it.
For bit errors in transmissions along the cables as cited by GTM (which I agree are uncorrectable - there's insufficient parity information in the datastream), again if this is were the widespread problem he thinks it is I'd be surprised if we'd hear anything, and I doubt such an obviously flawed standard would have lasted the 20 years it has. I'd be interested in any information on bit error rates obtained over SPDIF cables. I guess there's a simple experiment: connect the digi-out from a CDP to a PC input, dump a track to disk several times and compare. Obviously you'd have to take a bit of care to synchronise the bytestream for the comparison to work, but that's not difficult (I did something similar when trying to see any difference between a production CD and CD-ROM: of course no differences seen). If there are differences, but they only appear at the level of the LSB or thereabouts it would seem to me that there's some DAC-ADC process going on inside the capture card, and that's just acoustic noise getting into the measurement through the sound-card.
The errors in digital transmision that could account for transport and cable differences are IMO purely to do with extracting the timing information from the digital stream, which it why spearate clock-links or on-demand flow control is a sensible idea (and I'm staggered the audio industry has done little or nothing about standardising on it). There seem to be a variety of possible causes for this: GTM's point about cable capacitance is a fair one here, combined with crass clock-extraction methods at the DAC end resulting in data-induced jitter; varying power demands from the transport servos, combined with inadequate digital/analogue power-supply isolation, causing output signal voltage variations, pick-up of external noise.....
Of course nothing can read 100% perfect data in all possible scenarios, but that's completely missing the point. The question was not about how different transports can read scratched etc and otherwise marginal discs, but about how cables and transports (and green pen - which is getting equally moronic comments elsewhere) can affect the sound for normal discs.
For misreading data to affect the normal sound (i.e. not obvious pops/clicks/track-repeats etc), misreads uncorrectable by the CIRC would have to occur at the level of 10s or 100s per second. Are you saying Julian that this is the level at which you've had experience of having to re-read discs? I would find this extremely difficult to believe, as it's certainly not mine, but if that's what your experience is, then so be it.
For bit errors in transmissions along the cables as cited by GTM (which I agree are uncorrectable - there's insufficient parity information in the datastream), again if this is were the widespread problem he thinks it is I'd be surprised if we'd hear anything, and I doubt such an obviously flawed standard would have lasted the 20 years it has. I'd be interested in any information on bit error rates obtained over SPDIF cables. I guess there's a simple experiment: connect the digi-out from a CDP to a PC input, dump a track to disk several times and compare. Obviously you'd have to take a bit of care to synchronise the bytestream for the comparison to work, but that's not difficult (I did something similar when trying to see any difference between a production CD and CD-ROM: of course no differences seen). If there are differences, but they only appear at the level of the LSB or thereabouts it would seem to me that there's some DAC-ADC process going on inside the capture card, and that's just acoustic noise getting into the measurement through the sound-card.
The errors in digital transmision that could account for transport and cable differences are IMO purely to do with extracting the timing information from the digital stream, which it why spearate clock-links or on-demand flow control is a sensible idea (and I'm staggered the audio industry has done little or nothing about standardising on it). There seem to be a variety of possible causes for this: GTM's point about cable capacitance is a fair one here, combined with crass clock-extraction methods at the DAC end resulting in data-induced jitter; varying power demands from the transport servos, combined with inadequate digital/analogue power-supply isolation, causing output signal voltage variations, pick-up of external noise.....
Last edited by a moderator:
julian2002
Muper Soderator
peteh,
how do you think upsampling works? in a way the procedure of upsampling from 44.1 to 192 is in effect correcting an error frequency of 192 / 44.1 i.e. for every 192 samples only 44.1 are produced correctly the intermediatary samples are created through either a simple linear interpolation basicly an average of the previous and next samples or by more complex interpolation methods such as wadias digimaster or the algorythm used in the perpetual tech upsampler.
have a look here:
http://home.mira.net/~gnb/mac-cdis/cd9.html
graham,
it had been implied during this thread that a transport can recover data with 100% accuracy every time (and if it doesn;t it gives up the ghost completely and starts making nasty noises)... all i'm saying is that transports are not 100% infalible even with CIRC and that CIRC is not the only error correction scheme at work. the origional question was (to paraphrase) why does digital sound different? i still say that errors and overworked error correction play a part. by overworked i mean that the error rate can exceed the limits of CIRC.
as for the loaders etc. i've dealt with, i don;t have any exact retry frequency data but i do remember that it didn;t take long before the code hit a breakpoint in my retry code on a test cd. also another interesting point is that ALL of the hardware cd emulators for consoles allow you to produce test read errors to see how your game copes with these kind of issues, i hardly think that this would be a feature if it could never happen.
cheers
julian
how do you think upsampling works? in a way the procedure of upsampling from 44.1 to 192 is in effect correcting an error frequency of 192 / 44.1 i.e. for every 192 samples only 44.1 are produced correctly the intermediatary samples are created through either a simple linear interpolation basicly an average of the previous and next samples or by more complex interpolation methods such as wadias digimaster or the algorythm used in the perpetual tech upsampler.
have a look here:
http://home.mira.net/~gnb/mac-cdis/cd9.html
graham,
it had been implied during this thread that a transport can recover data with 100% accuracy every time (and if it doesn;t it gives up the ghost completely and starts making nasty noises)... all i'm saying is that transports are not 100% infalible even with CIRC and that CIRC is not the only error correction scheme at work. the origional question was (to paraphrase) why does digital sound different? i still say that errors and overworked error correction play a part. by overworked i mean that the error rate can exceed the limits of CIRC.
as for the loaders etc. i've dealt with, i don;t have any exact retry frequency data but i do remember that it didn;t take long before the code hit a breakpoint in my retry code on a test cd. also another interesting point is that ALL of the hardware cd emulators for consoles allow you to produce test read errors to see how your game copes with these kind of issues, i hardly think that this would be a feature if it could never happen.
cheers
julian
PeteH
Natural Blue
No it hasn't. It's been stated that because of redundancy in the data that it doesn't matter if it's not 100% accurate - you can randomly lose a lot of 1's and 0's with no effect because of the failsafe mechanisms built into the system.
I'll have a look at your link though, thanks
julian2002
Muper Soderator
peteh,
sorry that's semantics, you're saying you can loose data and it doesn;t matter. i.e. the drive will always produce a 100% correct version of the data at it's output.
the correct statement is the drive will error correct using CIRC up to a limit. beyond this limit CIRC will be compromised. at this point a pc data only drive will have to be told to retry or announce that the data cannot be loaded due to a corrupt disk.
for audio when this case occurrs interpolation of the 'missing' or incorrect data takes place to aproximate the audio waveform. therefore distorting the music. the drive cannot retry as data is pulled off the disk in real time. some players such as the meridian use a cd-rom drive running faster than normal so that they can pull off data into a buffer, thus if unrecoverable errors occur then a second, third or more chance at reading the data can be taken.
cheers
julian
sorry that's semantics, you're saying you can loose data and it doesn;t matter. i.e. the drive will always produce a 100% correct version of the data at it's output.
the correct statement is the drive will error correct using CIRC up to a limit. beyond this limit CIRC will be compromised. at this point a pc data only drive will have to be told to retry or announce that the data cannot be loaded due to a corrupt disk.
for audio when this case occurrs interpolation of the 'missing' or incorrect data takes place to aproximate the audio waveform. therefore distorting the music. the drive cannot retry as data is pulled off the disk in real time. some players such as the meridian use a cd-rom drive running faster than normal so that they can pull off data into a buffer, thus if unrecoverable errors occur then a second, third or more chance at reading the data can be taken.
cheers
julian
PeteH
Natural Blue
Noone's arguing with that I think
OK, having done a bit more reading I see that it is indeed the case that CD players will attempt to "mask" missing data, so I've learnt something today which is all good
From the link you posted above:All along, I have been talking about the effect on the BER [block error rate] of physical things, and unless interpolations occur, all errors are completely corrected and the digital sample reproduction is perfect. Mild doses of interpolation are inaudible, but if it does become chronic, a grainy caste begins to appear over the music.
What he doesn't say in that quotation though is that it appears uncorrectable errors are extremely rare in practice. http://pauillac.inria.fr/~lang/hotlist/cdrom/Documents/tech-summary.html provides another useful discussion about CD encoding, and from which I quote:
This first level error correction (the only type used for CD Audio data) is extremely powerful. The CD specification allows for discs to have up to 220 raw errors per second. Every one of these errors is (almost always) perfectly corrected by the CIRC scheme for a net error rate of zero. For example, our tests using Apple's CD-ROM drive (which also plays audio) show that raw error rates are around 50-100 per second these days. Of course, these are perfectly corrected, meaning that the original data is perfectly recovered. We have tested flawed discs with raw rates up to 300 per second. Net errors on all of these discs? Zero!
In explaining the "almost always" in the above quotation, he goes on to state that he had encountered two occasions (ever) where an uncorrectable error had arisen, in both cases from a heavy fingerprint - and wiping it off again resulted in zero net errors. And this is using a computer CD-ROM drive in CD-audio transport mode (ie. without re-reading), not an audiophile-grade transport.
So while I accept your statement that CD players will try to "mask" errors where the CIRC error-correction fails, I still don't believe that that accounts for any difference between transports - in the worst-case scenario, in real-world cases interpolation will occur for a few hundred microseconds total at most in the course of playing a disc, hence the issue is totally irrelevant to perceived sound quality. Going back to your link, the guy says he found one disc ever where there was a large amount of interpolation happening, and you could see pinholes in the disc when you held it up to the light - presumably a dodgy production problem or something. And in any case, it seems to me that the interpolation is more likely to be performed in the analogue domain as by its nature you can't interpolate digital information, in which case it would be a job for the DAC and again have nothing to do with the transport - although I'm blagging this now so it might be nonsense.
And BTW I've now learnt that besides the possibility of re-reading the data there are additional error-checking mechanisms in the CD-ROM specification, so that sort of answers my earlier question. Wish the weekend wasn't nearly over, I'm enjoying finding out about CD-encoding algorithms and the like
Last edited by a moderator:
