One has to remember that getting the info off of a CD is an analog process... subject to all sorts of analog issues. On a music CD, the amount of error correction redundancy was traded for playing time.
dave
Dave
Nevertheless, the checksums of both comparison CDs in these sets are identical.
Alex
Blu-spec CD BR Rock - BR 3M version
Checksums generated by ExactFile 1.0.0.15
; http://www.exactfile.com
; 7/06/2009 3:36:28 PM
e995b3b66693a0261bff3db248eb02b9 *Track01.wav
0d2770d07f5509c862522f5980e98050 *Track02.wav
f8f76a28cb2952429dfa01a0936d08d7 *Track03.wav
474df609920783f713acf668a78a69a5 *Track04.wav
6ea2bc30cf42b2d74cc43e34277f7514 *Track05.wav
aafca38d49cbe7786b1a339f8054db9d *Track06.wav
5306e6f23de8f19f0410682942fd5212 *Track07.wav
61f1e2e3808a957bb85768cf058b3b4f *Track08.wav
07c9e9a0ab95031a71ebca76e0f4346d *Track09.wav
0bd6fac2c6115f842f4d372d15164550 *Track10.wav
862c1def4ba09ffa26cb65c96c4f6974 *Track11.wav
f6601d3375a4526e0d4dd57a50aada3a *Track12.wav
4695a6e26370db32e9be839030e51fe3 *Track13.wav
61b69212e44e71b82f36c1fe788927e3 *Track14.wav
d1a9bc6f68454de661ff79315e957312 *Track15.wav
294b5494523e40a9262ba70f1eb60e34 *Track16.wav
8cc4c5e377c3446511a6b8f59a051769 *Track17.wav
; 17 files hashed.
Blu-spec CD Rock (CD version)
; Checksums generated by ExactFile 1.0.0.15
; http://www.exactfile.com
; 25/07/2009 6:51:47 PM
e995b3b66693a0261bff3db248eb02b9 *Track01.wav
0d2770d07f5509c862522f5980e98050 *Track02.wav
f8f76a28cb2952429dfa01a0936d08d7 *Track03.wav
474df609920783f713acf668a78a69a5 *Track04.wav
6ea2bc30cf42b2d74cc43e34277f7514 *Track05.wav
aafca38d49cbe7786b1a339f8054db9d *Track06.wav
5306e6f23de8f19f0410682942fd5212 *Track07.wav
61f1e2e3808a957bb85768cf058b3b4f *Track08.wav
07c9e9a0ab95031a71ebca76e0f4346d *Track09.wav
0bd6fac2c6115f842f4d372d15164550 *Track10.wav
862c1def4ba09ffa26cb65c96c4f6974 *Track11.wav
f6601d3375a4526e0d4dd57a50aada3a *Track12.wav
4695a6e26370db32e9be839030e51fe3 *Track13.wav
61b69212e44e71b82f36c1fe788927e3 *Track14.wav
d1a9bc6f68454de661ff79315e957312 *Track15.wav
294b5494523e40a9262ba70f1eb60e34 *Track16.wav
8cc4c5e377c3446511a6b8f59a051769 *Track17.wav
; 17 files hashed.
Last edited:
Correct.
A sampled system comprises of three processes in cascade:
-anti-alias filtering (limiting the spectrum of the input signal to less than half the sampling rate)
-sampling, and, upon playback
-reconstruction filtering, aka anti-imaging filtering, aka interpolation (limiting the spectrum of the output signal to less than half the sampling rate).
Note that the sampling theorem imposes a specific filter for the reconstruction: this filter has to be the Sinc(t) function, of which the Fourier Transform is a hat function with its transition at half the sample frequency.
If the above rules are observed then the theorem guarantees that the whole process is perfectly lossless from the output of the anti-alias filter on.
This is mathematics. How you implement it is engineering.
(Do note, however, that the theorem does not tell you how to make an anti-alias filter.)
The Sinc(t) function has infinite temporal extent and is acausal to boot: it has to start at the Big Bang to be in time for the notes you are going to play tomorrow. (Actually tomorrow I am going to play with a cyclotron but this is not quite related to audio.)
It, the perfect and divine interpolator Sinc(t), cannot be realised in the analogue domain, so much is clear.
While an analogue filter with near-sufficiently fast/steep suppression can be imagined, and even built, it would be extremely complex (think tens of opamps), expensive, crappy (if you believe the signal should not pass through tens of opamps), and suboptimal as, being causal, it would not be what the doctor ordered and impart serious non-linear phase distortion (of which the audibility can be debated). Incidentally, does anyone of you happen to have the circuit diagram for the Sony LPF101 filter modules used in the PCM1600-series and/or F1?)
Above sentence is misleading. The theorem-imposed requirement of the reconstruction filter being a Sinc(t)/brickwall at 22.05kHz can never be relaxed, if correct and complete reconstruction is the aim.
However, when the sampled data stream is oversampled/upsampled, the frequency space in which it 'exists' is increased from half the original rate (i.e. 22.05kHz) to half the oversampled rate (i.e. 88.2kHz if we adopt 4x oversampling).
If you do the oversampling by inserting three zeroes after every original sample, then this new frequency space contains the original spectrum (limited to 22.05kHz) and its images up to 88.2kHz. (You can insert something else than zeroes, but this is a bad idea, really.)
Since we are in the digital domain now, we can have a digital filter that implements (part of) the required reconstruction. Indeed, in the newly-alloted frequency space we can now digitally suppress the range from 22.05kHz to 88.2kHz, and we have to rely on an external post-DAC analogue filter only for the further suppression of the remaining images (which will start at 154kHz). Here a shallow-slope filter suffices.
The bulk of this hybrid reconstruction strategy is now defined by the digital filter.
Digital filters are trivial to make acausal. Just store the samples in a memory to delay them, and give the filter access to the whole memory, so that it sees past, present, and future at the same instant.
Over the window afforded by that memory the sample train is convolved with a filter impulse response that is essentially a numerical approximation of the required Sinc(t) function.
The accuracy of that approximation is entirely a matter of engineering and available budgets. Suffice to say that, with software-based oversampling/upsampling, one can reach,to all intents and purposes, near-perfection (I have built 16000-tap oversamplers with over 200dB rejection at 22.05kHz+). With hardware it is a bit tougher, but sound compromises do exist.
Digital filtering for reconstructors is linear phase (as is Sinc(t) itself: it is acausal and symmetric). The filter impulse response rings like hell, but this is utterly immaterial as this ringing disappears entirely for each and every correctly-bandlimited input signal.
Numerical accuracy of these filters is an issue. The output of such a filter is the convolution of the 16 bit signal samples and the 24, 32, 48, ... bit filter coefficients. The output, in other words, has an accuracy far exceeding 16 bit. Therefore the output of an oversampler should be converted in a DAC chip with a resolution in excess of 16 bits, i.e. 24 bits with today's silicon (which really attain 20-22bit performance at the best, but that's another story).
That's it in a nutshell. They are many more interesting things to tell, such as the whys of Meridian's minimum phase filter, the errors of Wadia, how NONOS can work, the idiocy of worshipping '96', and so on.
As said before, and investigated numerous times since 1983, CD error correction is not an issue. Hard, uncorrected errors are a rarity on a well cared-for CD.
And even if such errors were 10x, 100x more frequent, they would amount to transient upset, and as such are incapable of affecting sound quality averaged over time.
However, the physical properties of disc media may differ in the sense of pit edges and reflectivity, leading to different operating conditions for the optical head and the transport's controlling servoes. A harder working transport likely draws more and dirtier current from the power supply, and this may be felt in the player's clock, DAC, and output amplifier circuits.
Again, a matter of engineering, no voodoo required.
I am aware of this. I have however never seen any controlled test or repeatable data that confirms this. I can only conclude that there is no objectively confirmed proof that such an audible difference exists.
It's never black and white. If I say that blue caps sound harsher then red ones, and you say no they don't, one of us is wrong. Does that make you or me a liar? That's unnecessarily harsh and not necessarily true.
People that maintain that gold plated CDs sound better may genuinely be convinced that it is so. It's not possible to proof they are wrong. It IS possible to proof they are right. Someone should try that.
jan didden
Jan
I do not believe that is unnecessarily harsh, when a major International Record company, and co developer of the CD system makes such a claim.
If the claim is false , then I feel sure that they would have been taken to task long ago for making false claims. As a technical journalist, perhaps you are in a better position than most to get further information from Sony about the claims made by them ?
SandyK
Thank you. Most interresting and convincing post ... although I was already convinced 😉
To an extent, yes. Many of the MFSL are remastered, so of course they'll sound different. But it's GOLD. That's the 1990s equivalent of the extra-thick vinyl they sold in the 1980s with the heavily bumped up bass.
There's all kinds of marketing spin to try to move stuff from catalogs. Given the demographics of CD buyers (older- younger people skew toward digital downloads), it's totally unsurprising that Sony is trying to monetize already-existing software inventory- it's certainly more profitable to sell music that Sony already has paid for, saving costs of musicians, studios, engineering, and promotion.
But really, with a Chrysalis "Pink Island" pressing, a Reprise pressing, and a "normal" 1985-vintage CD all on hand, do I need yet one more version of Stand Up? 😀
Thanks for trying to elaborate.I got it,as it is.My experience by using various filters is the same as Jan said.But being sneaky and open to opinions and knowledge,I want to learn more.
B.L.
If the checksums are the same but the files sound different, then the problem is elsewhere. There is no other property of the file other than the bits (by definition) and they are more than adequately verified as identical by the md5 checksum.
If you are using computer playback of the files, for example, the Windows sound kernel is well known to not be bit-perfect, meaning that the bitstream that goes into it is not identical to the bitstream that comes out the other end. That's why ASIO was developed.
Come on, this is audio! Cables open up de soundstage, green pens give you tight bass, racks pinpoint the artists, whatever. You want to take all those to task? It's advertising, just like BMW claims their cars 'cling to the road', for instance. Of course they don't, but hey it's marketing.
Sony has the money, clout and expertise to do the mother of all controlled tests to once and for all prove that, for instance, gold CDs sound better. They don't. Why should they? They can only lose; so far everybody that counts believes them anyway 😉
jan didden
Audio retailing is tied to the chariots of the specialized press.Underground or not.Their scriblings,are like the Bible to the ignorant buyer. Just follow basic mechanics and electronics ,and you are there.Thanks to some sane engineers ,those who seek,get enough knowledge to avoid the tar pit.
Speaking of Sony,the "perfect sound forever",was a pure marketing hype too?
I like the, racks pinpoint the artists 🙄
B.L.
I completely agree and have said so elsewhere. It ain't the files. The files are the same, I've even gone to the trouble to check them. If there is a difference in the final audio, the trouble has to be in playback.
For those who do hear a difference, don't you want to know why? Really? I would certainly look in the direction indicated by Werner below, it's a very good place to start. Measure noise and jitter caused by playback of the same file on different media. Is that where the problem lies? Cause it sure ain't the files.
I once heard a claim that a valve rectified PSU for the transport created an audible improvement. I was naturally somewhat incredulous to say the least. Does anyone here think this is indeed possible?
There is no reason for disbelief.If the valve PSU,was producing measurably cleaner juice,for the transport,it was simply better,than the one that was compared to.
B.L
Actually,I should have said that the comparison was with an already pretty fancy SS power supply.
Either valve or solid state,if the recipe has the ingredients, done right,there's no black magic or occult.Pure engineering only. 😎
B.L
I have seen sidebands consistent with servo current demand at DVD rotation rate when testing a oppo multi format player analog out into Julia@ soundcard @24/96 - I don't believe these were digtial errors - rather unacceptable power supply or common gnd impedance coupling
still orders of magnitude lower than turntable wow and flutter, disc centering error, or stamper physical stretching/distortion on being pulled from the master
still orders of magnitude lower than turntable wow and flutter, disc centering error, or stamper physical stretching/distortion on being pulled from the master
Last edited:
- Status
- Not open for further replies.