CAN we re-create the "vinyl sound"

Topic: CAN we re-create the "vinyl sound" (Read 35431 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.

CAN we re-create the "vinyl sound"

Reply #25 – 2003-01-24 12:28:35

Only few best DACs (veeeery expensive) can achieve 19bit resolution... oh well,
best ADCs are even worse!
And another thing is thermal noise in connections, it's level is about 20bits...
You can lower it only by deep-freezing (few Kelvins) connections, which is unreal.

To compare, well dithered 16bit sound has dynamic range close to 18bit.

<edit>
typos...
</edit>

CAN we re-create the "vinyl sound"

Reply #26 – 2003-01-24 12:34:00

X1-X2-X3 is
X1: recording
X2: editing
X3: support

So for a CD, of course the last letter is always D. DDD is fully digital, without any analogic step.

CAN we re-create the "vinyl sound"

Reply #27 – 2003-01-24 12:36:39

Lame question...
Could anyone say, what formats are currently supported for MPC?
someone told, that it's possible to encode up to 8 channels with 48KHz samplerate...
Is it so?

CAN we re-create the "vinyl sound"

Reply #28 – 2003-01-24 13:12:39

Quote

- what is a D-D-D CD ? does it mean, recorded without passage through the analog domain ?

It just means that digital technology was used during the recording.

Personally I think that many late analog recordings are better than those of the very early digital era (beginning of the 80's). There's often some rumbling noise in the background.

CAN we re-create the "vinyl sound"

Reply #29 – 2003-01-24 13:15:15

It's good that you said late and early... because early digital was mostly 12bit.
That's apple and oranges case again.

CAN we re-create the "vinyl sound"

Reply #30 – 2003-01-24 13:28:23

Quote

X1-X2-X3 is
X1: recording
X2: editing
X3: support

So for a CD, of course the last letter is always D. DDD is fully digital, without any analogic step.

Oh. I just learned something, thanks !

Continuum:
I agree that early cd's - and consumer level players - were a bit rough sounding..

SacRat:
IIRC, the current MPC format (SV7) and encoder/decoder currently support two channels, at a maximum sampling rate of 48 kHz.
The upcoming new bitstream format, SV8, should remove most of these barriers.

CAN we re-create the "vinyl sound"

Reply #31 – 2003-01-24 14:47:55

Quote

It's good that you said late and early... because early digital was mostly 12bit.
That's apple and oranges case again.

Interesting. I might check a few CDs, with said problem, for this.

But actually I don't think that the problems I spoke of are caused by too low resolution. I assume the technical machinery used wasn't fully developed in many points and that sound engineers weren't experienced with it.

About apples: I said that because I think that the sound quality of a CD is far more dependent on other factors than on a DDD, ADD or even AAD label.

CAN we re-create the "vinyl sound"

Reply #32 – 2003-01-24 14:58:08

Quote

X1-X2-X3 is
X1: recording
X2: editing
X3: support

I almost agree

1) recorded 2) mixed / edited 3) mastered

In the 80s it was common for some record labels to put this three letter info on the CD (as the audiophiles were fed up with analog )

[span style='font-size:8pt;line-height:100%']No really, audiophiles were rightly skeptical and some CDs are badly mastered indeed. But a good mastered CD on a good player it convinces me every time
[/span]
--
Ge Someone

CAN we re-create the "vinyl sound"

Reply #33 – 2003-01-24 15:00:38

Some, even professional equipment
had ADCs with only 12bit effective resolution.
That's bad quality, but that was best of digital of that time,
because DACs of that time were too 12bit

CAN we re-create the "vinyl sound"

Reply #34 – 2003-01-24 16:46:46

The very first CD player had a 14-bit DAC. It maintained an effective 16-bit noise floor by 4x oversampling. I kidd you not.

I believe in 96kHz sampling. And I'm sure 24-bits have their uses.

To answer the original poster's question: you can feed a 24-bit .wav file into mppenc, and decode the resulting mpc file at 24-bit resolution using the correct paramters (and version of) mppdec - it's all on Frank's site...

http://www.personal.uni-jena.de/~pfk/mpp/

As noted though, you can only do this at a maximum of 48kHZ sampling. If you want to maintain 96kHz, use lossless. Though there's a DTS 24/96 codec (which is lossy), there really is no use to using lossy coding at 96kHz: no one really understands why it sounds better, so what pschocaoustic model can you apply to the top half of the frequency range at 96k?!

If you really hear an advantage of 24-bits compared to 16-bits, I'd be interested to know if that survives lossy compression. I think it might. This statement sounds like nonesense because lossy compression will change the signal DRAMATICALLY more than converting 24-bits to 16-bits. However, you still get a 24-bit output from the process. If the audible advantage is simply using more bits in the DAC, using all those extra levels could help with the sound. Accepting the DACs are imperfect, usually non-linear, and sometimes even non-monotonic, this could give an audible effect. A simlar process is used in comunications DACs to linearise them (i.e. adding noise to spread the signal over more quantiser levels, even though those levels are themselves not accurate) - there's even an Analog Devices data sheet on it - you used to be able to find it on their site by searching under "dither".

So, I'm not saying you'll maintain the true "24-bit advantage", whatever that may be (and it's certainly not noise floor in this case) - but you might gain some advantage compared to 16-bit before and after mpc. btw, mpc, like other lossy codecs, doesn't really have an internal bit For example, even mp3 will handle 24-bit data correctly - i.e. it can handle a signal at -120dB below digital full scale.

Have fun!

Cheers,
David.

P.S. guys - give him an easy time - he asked a straight technical question which could be answered easily - no need jump on him!

CAN we re-create the "vinyl sound"

Reply #35 – 2003-01-24 16:48:56

What bothers me the most about this thread is the number of users with cats in their avatar. Coincidence?

CAN we re-create the "vinyl sound"

Reply #36 – 2003-01-24 17:11:29

Quote

If you really hear an advantage of 24-bits compared to 16-bits, I'd be interested to know if that survives lossy compression. I think it might. This statement sounds like nonesense because lossy compression will change the signal DRAMATICALLY more than converting 24-bits to 16-bits. However, you still get a 24-bit output from the process. If the audible advantage is simply using more bits in the DAC, using all those extra levels could help with the sound.

But wouldn't it be sufficient to use 24 bit precision at the decoding stage?

CAN we re-create the "vinyl sound"

Reply #37 – 2003-01-24 17:37:04

Oh god, I am afraid I didn’t mean to cause such a stir. I am not trying to convince anyone, and I am not trying to start a fight, which I seem to have, and I apologize for that.

I see many technical explanations around here, but if anyone still wants to listen to my opinion, sound is not a completely technical matter. Sometimes something that sounds better to one person may sound less good to others. I for instance don’t mind having some of that analog hiss, that is indeed bothersome, and indeed disappeared in the digital era, if I get some warmth back in the sound, I don’t say its necessarily "better". I suppose it has allot to do with the original material.
Allot of music today is original digital as we hear more and more electronically enhanced instruments, that obviously suffer less from the A/D process.

I play the cello since I was 10 years old and have yet heard something that sounds the same as the instrument itself. I suppose it has to do with harmonic vibrations in the air surrounding the instrument. And that I can tell you for sure, after 15 years of listening to strings vibrate, no sound system sounds quite the same.

The reason I was thinking of 24/96 is because the dynamic range of 16/44.1 I clearly not enough to reproduce classical instruments, it sounds dull and lifeless (to me anyway, please don’t argue with it, its just an opinion). I dunno maybe 24/96 is not enough either and maybe the rest of the circuitry is as much a bottleneck as the sampling.

I suppose the only interesting thing that came out of all the discussion around here is that I should give up mpc and go lossless as I clearly don’t care how big the file will be.
Am I right to assume that 1200kb/sec should be enough t to losslessly compress a 24/96 signal? I understand FLAC can go as low as 800kb/sec for 16/44.1 and as the 24bit sample is 50% bigger so should the final size… am I missing anything (technical).

Again, I apologize for steering up such a storm, its just music people…

<edit> I want to think everyone who replied this post... I appreciate the effort </edit>

CAN we re-create the "vinyl sound"

Reply #38 – 2003-01-24 18:46:06

Quote

Am I right to assume that 1200kb/sec should be enough t to losslessly compress a 24/96 signal?

No. Because the lower 8 bits of 24-bit audio are basically noise (at least mathematically) they don't compress at all. So, the maximum compression you can usually expect for 24/96 material is about 1/3 off (compared to about 1/2 off for 16/44). Since raw 24/96 is about 4600 kbps, you're going to end up at about 3000 kbps. Converting to 20-bit will help some and probably won't be too detrimental.

The reason I suggested WavPack's lossy mode is because I specifically designed it for this application. It preserves the entire dynamic range of the original (because it's logarithmic like all lossy audio codecs) and obviously preserves the bandwidth (because it encodes every sample).

When I use this mode at 1280 kbps on 24/96 material and use CoolEdit to invert-paste the original with the lossy version to derive the noise that was actually added, the noise is completely inaudible at normal listening levels even when the music is not playing, so no psychoacoustic masking is even required.

WavPack also has lossless compression, so it would be very easy to compare the modes and determine what is really acceptable to you. Of course, if you really mean that storage space doesn't matter then 3000 kbps shouldn't be an issue.

BTW, don't worry about causing such a stir. This crowd is easily aroused.

CAN we re-create the "vinyl sound"

Reply #39 – 2003-01-24 19:03:36

Thanks bryant, i will have a look at WavPack.

its interesting what you say about the lower 8 bits being mostly noise, it actually sounds reasonable that this is where the harmonic vibrations I am missing are. When you listen to string vibrate you can hear allot of things that come from other vibrating objects around the instrument and this is not necessarily a bad thing. You can also "feel" the air vibrating around the string, and it does sound like kind of a "dust" around the sound, I suppose classical music has completely different requirements then other kinds of music. Maybe it simply requires a different psychoacoustic model?

BTW i have a ruther musical cat, maybe i should put up a picture of him too...

CAN we re-create the "vinyl sound"

Reply #40 – 2003-01-24 19:03:55

Here's another way to think about it. Let's say we have an original 24/96 source recording. If we convert this to 16/44 we are essentially doing a lossy compression of just over 3:1. However, this is a particularly stupid compression because it both limits the dynamic range and the bandwidth. In addition, the result can be further compressed losslessly, showing it is not even efficient from an pure informational standpoint.

All WavPack does is perform a different sort of lossy compression that makes no assumptions about what can and cannot be heard and simply tries to reproduce the waveform, leaving the dynamic range and bandwidth untouched. And WavPack files are not further compressible, showing that they are information efficient as well.

CAN we re-create the "vinyl sound"

Reply #41 – 2003-01-24 19:06:43

Hello,
I think that Meridian's MLP Losslessly Packs 24/192 material at about half the size in DVD Audio disks.
How ?.

CAN we re-create the "vinyl sound"

Reply #42 – 2003-01-24 19:10:35

This 24/96 vs 16/44.1 reminds me of the what John Carmac said about color resolutions.

24 bits of colors i enogh for the final picture (32 bits color is 24 bits color + 8 alpha bits)
BUT when you are rendering the colors and use many transparant textures the resolution is degraded.
the many recalculations done on the color values degrades the color resoluton due to rounding errors and a finite calculation resolution.

That's why we know have 3d accelerators with 128bits color resolutions
not because we can se the difference but because editing & altering decreases the resolutions.
The average grey and the "one bit more blue" grey is total alike to the naked eye.

THATS probaly why you have 24/96 in your studio
beacuse the samples might later be reused and mixed again.

However the final product might no be better if produced in 16/44 or 24/96 in the final stage.
and thats why people would like you to ABX the 16/44 agains 24/96 so prove to you that in the final stage there is no audiable differenc to 16/44 and 24/96

CAN we re-create the "vinyl sound"

Reply #43 – 2003-01-24 19:19:31

Quote

its interesting what you say about the lower 8 bits being mostly noise, it actually sounds reasonable that this is where the harmonic vibrations I am missing are.

The lower 8 bits are noise from a mathematical viewpoint, but I didn't mean to imply that there wasn't information there also. And between transients and during reverb tails (when the upper bits aren't being used) these lower bits become more significant.

And yes, by all means, all cat pictures are welcome!

CAN we re-create the "vinyl sound"

Reply #44 – 2003-01-24 19:30:10

Welcome to the beautiful world of real instruments....
I've played piano, started at the same age

[span style='font-size:14pt;line-height:100%']Theory:[/span]
64 kHz is the maximum measured harmonic frequency coefficient generated by any instrument.
(mostly brass) (I've read it in some old book... can't quite remember which was it, so don't quote me)

High frequencies might interfere with lower frequencies...
but do you have an equipment to generate 32 kHz tone?
I guess not. Even most 'adiophilistic' equipment
has a frequency range of maximum 24 kHz.
So there is no perfect reproduction of instruments.

Additionally, PCM method might not be the best for our ears to represent analog audio.
All sound recordings (not only digital) use properties of our hearing to conserve space.

[span style='font-size:14pt;line-height:100%']Practics:[/span]
If 44,1 kHz is not enough, 48 kHz might be...
96 kHz is an overkill, as is 24bit audio. This only adds margin to avoid rounding errors in processing.
16bit might be too little, so record as 24 bit/48000Hz,
then add dithering&downsample to 16bit to keep most of additional dynamics.
Then you should be able to compress it well.

If you can ABX between 24bit audio and dithered 16bit,
then only choices left are WavPack lossy or any lossless codec.

<edit>
Slight modifications to interpunction.
</edit>

CAN we re-create the "vinyl sound"

Reply #45 – 2003-01-24 19:32:10

Quote

and thats why people would like you to ABX the 16/44 agains 24/96 so prove to you that in the final stage there is no audiable differenc to 16/44 and 24/96

This has been discussed many times around here, but I keep seeing it come up. ABX testing can only prove when a difference is audible. It cannot prove that something is inaudible. A negative ABX test proves nothing!

About MLP, I have read that when MLP is unable to meet the target bitrate it drops into a lossy mode that is probably identical to WavPack's in principle. But, of course, since it's a closed standard there's no way to know how often this is. And I can tell you that there's no way that they compress 24/96 losslessly 2:1 like they claim, although 24/192 is actually a little easier to do that with because there's more correlation between samples. But we're starting with over 9000 kbps now!

<edit> typo

CAN we re-create the "vinyl sound"

Reply #46 – 2003-01-24 19:45:21

Quote

Quote
and thats why people would like you to ABX the 16/44 agains 24/96 so prove to you that in the final stage there is no audiable differenc to 16/44 and 24/96

This has been discussed many times around here, but I keep seeing it come up. ABX testing can only prove when a difference is audible. It cannot prove that something is inaudible. A negative ABX test proves nothing!

No, actually it proves that person who took the test can't hear the difference
between the samples on a given setup!

CAN we re-create the "vinyl sound"

Reply #47 – 2003-01-24 21:05:37

Nope. If a listener gives up after, say, 4 trials because they were all wrong (thinking they couldn't hear a difference), that doesn't mean he wouldn't have been able to score 20 good trials afterwards. (Does that make any sense? Hope so. )

ff123, our Master of Statistics, has posted a lot about these issues, you might want to try searching around a bit.

CAN we re-create the "vinyl sound"

Reply #48 – 2003-01-24 21:35:30

Hello..

All these posts about dynamic range yet it hasn't been quantified well.. so here we go:

Dynamic Range

For 16-bit sampling: 2^(16) = 65,536 levels

10 * LOG [ 2^(16) ] = 48 dB

20 * LOG [ 2^(16) ] = 96 dB

For 24-bit sampling: 2^(24) = 16,777,216 levels

10 * LOG [ 2^(24) ] = 72 dB

20 * LOG [ 2^(24) ] = 144 dB

This is how I would calculate the dynamic range.

Shannons Law

According to Shannons Law.. in order to digitize in an alias-free way.. a sample rate of TWICE the highest frequency is required (plus a margin factor).. hence 44 kHz for audio. This is how I understand the theory. Comments on dynamic range vs bit depth and the subject of sampling rate are welcome. Note: for discussion purposes only.. content subject to change.

CAN we re-create the "vinyl sound"

Reply #49 – 2003-01-24 22:09:28

Quote

Quote
Quote
and thats why people would like you to ABX the 16/44 agains 24/96 so prove to you that in the final stage there is no audiable differenc to 16/44 and 24/96

This has been discussed many times around here, but I keep seeing it come up. ABX testing can only prove when a difference is audible. It cannot prove that something is inaudible. A negative ABX test proves nothing!

No, actually it proves that person who took the test can't hear the difference
between the samples on a given setup!

Really? What if I am taking an ABX test and I can clearly hear a difference in the samples. But instead of answering honestly, I secretly flip a coin in my pocket and use that to make my choices. The test is obviously not proving anything here.

This is an exaggerated case, but the fact that the test can be fooled makes it scientifically invalid. While a negative result has some practical value, only a positive result proves anything. Donald Rumsfeld uses this logic often when talking about Iraq's banned weapons when he says "absence of evidence is not evidence of absence."

People might not like hearing this, but the fact is that from a strict scientific viewpoint it is impossible to ever prove that there is no perceptual difference between 16/44 and 24/96 audio reproduction. People can choose to believe it based on everything they know about perception of audio and their own experiences, but they should be aware their belief is based on faith, not scientific proof.

Now, I am not saying that I believe that everything people say they hear is real or that there isn't a lot of money being made on worthless audiophile "snake oil". But I think that there is sometimes an attitude of superiority and condescension around here that (although supposedly based on scientific "enlightenment") isn't really justified by the facts. The truth is, like so many things, we really don't know the answer and I certainly don't think anyone deserves to be insulted for deciding to err on the side of a little safety margin.

<edit> typo

Notice