Does vinyl sounds better than cd or not?

[Thad E Ginathom]

This is one of those so-obviously-true statements. But wait...

If 44 / 96/ 192 khz can be called as resolution of a digital content ...

But can it? Is it a genuine technical description, or is it an invention of the high-resolution marketing men.

If we get the same waveform out of 16/44.1 as is put in, because the theory behind digitisation says we do, then digital has infinite resolution too! Even at 16/44.1.

Is this true?


~

 

[asliarun]

With cars, it is quite straightforward. What came earlier is vintage and low performance while the new is high performance. It is automobile / mechanical engineering and that is the nature of the beast !!

On the other hand, the various storage methods in audio (digital, analogue etc) are developed to serve specific needs / trends of the industry at various point of time. If quality of sound reproduction is the criteria, it is very difficult to argue about ones superiority over the other although what best serves the industries needs better is anyones guess !!




Agree

In many cases, they were built for different specs and for different uses. For example, the monster big block engines of the 60s or 70s were built for straight line acceleration. So if you cherry pick the performance metric, I am sure we can make the old guy win as well.

Back to audio, if I say that clicks and pops and wow and flutter are utterly intolerable, I can easily say that a CD is superior.

So what exactly are we comparing when we say audio quality? Isn't noise of any kind highly objectionable too, the same way the analog camp calls CD cold and clinical or even less detailed. Our liking and not liking music is never because of the detail or loss of it but because of our subjective preferences for a certain sound. So why even talk about detail as if it were the holy grail especially if we are so ready to ignore the other flaws such as noise?

And this theoretical stuff about Nyquist and discrete sampling is really a red herring. It is simply marketing myth based on incomplete understanding of basic science - a strategy that audio marketeers have honed to a fine art. Heck, they even throw in quantum physics when they feel they need a plausible but irrefutable argument.

I also feel that we are puppets at the hands on the mastering engineer. They decide if this music should have the fat warm sound and what their audience would like - on their clients systems. Even if we setup a perfectly neutral sounding system, all we are doing is hearing what the mastering person intended, not what the actual music sounded like.

Another funny thing is that a lot of modern vinyl is targeted to hipsters who love the retroness of a vinyl player and have adopted it with a vengeance. This music is deliberately mixed to their preference and one of the big genres is actually low-fi music. They deliberately create low fidelity music and a lot of this music is ending up on vinyl. Oh, and portable vinyl players are all the rage, I see hipsters every once in a while with these devices, and they are sold in trendy jeans shops - right along with skinny dyed jeans and trendy hats.

 

[Santy]

This is one of those so-obviously-true statements. But wait...

But can it? Is it a genuine technical description, or is it an invention of the high-resolution marketing men.

If we get the same waveform out of 16/44.1 as is put in, because the theory behind digitisation says we do, then digital has infinite resolution too! Even at 16/44.1.

Is this true?

~

Hmmm why not. Technically, I believe sampling rate together with bit depth can be called as resolution of the digital music. Though the term "resolution" used in reviews to define sound quality is quite different.

Output of a low resolution content is still a sound wave . Irrespective of resolution, music that is churned out of DAC has infinite resolution but that is what DAC is supposed to do. Remember closing gaps? I think reconstruction of original waveform is easier and better when the content is recorded at a higher resolution.

Same example again
Low resolution music: 1024 samples per sec.
http://music.columbia.edu/cmc/musicandcomputers/sounds/chapter2/bragg.1024.mp3

Standard resolution music: 44,100 samples per sec
http://music.columbia.edu/cmc/musicandcomputers/sounds/chapter2/bragg.mp3

DTS-HD or DSD are high resolution formats purely because they support higher sampling rate (along with higher bit depth). I for one can tell the difference between DTS and DTS-HD so I wouldn't disregard it entirely as a marketing gimmick. May be I wouldn't be able to tell the difference between DSD64 and DSD128 (no experience yet) but atleast till that limit your ears could pick up the difference, resolution makes sense.

BUT. Just having a higher resolution doesn't necessarily mean superior SQ.
Music with high sampling rate can sound pathetic too. It only specifies the ability to carry more bandwidth of music. Just because our DAC is capable of 192 khz, it cannot make a 44.1 khz /16 music sound as good as that the one with a native 192khz/24 recording.

AND. Analog having infinite resolution is only theoretical and in reality, the grooves are not super smooth either, so the needle does skip "space" in between. It jumps at microscopic distance when it runs over dust. An LP doesn't sound good just because it has infinite resolution but it depends on quality of recording, the gear and the condition of the LP itself. These are my humble thoughts. Just brainstorming. Please correct me wherever I am wrong!

 

[jls001]

I also feel that we are puppets at the hands on the mastering engineer. They decide if this music should have the fat warm sound and what their audience would like - on their clients systems. Even if we setup a perfectly neutral sounding system, all we are doing is hearing what the mastering person intended, not what the actual music sounded like.

I would respectfully disagree with this part of your post. Musicians themselves are mainly responsible for the sound of their music.

Musicians are very particular about the tonality of their instruments. For example, George Benson's guitar tone can be classified as warm, and even fuzzy. Just the kind of rolled-off sound that comes from a well-chosen, vintage tube guitar amp.

Or take the music of The Alan Parsons Project - their highs have a sense of being rolled off a wee bit; they lack that sparkle. He is the musician and (a very reputed) producer so may be that's what he envisages his sound should be. I say this from extensive hearings of both CD and LP of their music.

If you read about modern guitarists, they have very specific preferences about their guitars (eg: a 1959 Les Paul, or a Fender Stratocaster or an Ibanez or a Gibson of a particular vintage; and most of them stay loyal to their choice of brand throughout their lives), guitar processors, guitar amps (whether tubed or solid state or hybrid), and speaker stacks (Marshall, Peavey, etc). They even have preferences about the brand of strings they use (nickel or steel, etc). All these for a careful synergy that leads to their sound signature. Most guitarists have their own sound signature and the cognoscenti will immediately recognise them.

The mastering engineer's job is more in the nature of how to assemble the various parts of the music into a what we finally hear. His job is to decide the relative amplitudes and placements of "voices" in the mix. Of course I am sure he will take into consideration what the musicians and producer want him to do.

More than the mastering engineer, the producer would have more say in what the final product should sound like. Example: Robert "Mutt" Lange was highly sought after by hard rockers as a producer as he was reputed to be able to bring out "hardness" or the "punch" in their rock.

 

[Thad E Ginathom]

And this theoretical stuff about Nyquist and discrete sampling is really a red herring. It is simply marketing myth based on incomplete understanding of basic science - a strategy that audio marketeers have honed to a fine art. Heck, they even throw in quantum physics when they feel they need a plausible but irrefutable argument.

I do feel that there is a lot of marketing behind the high-sampling-rate movement. I prefer to call it that, because, as I've said before, the rolls-off-the-tongue, audiophile-beloved word resolution is just so much more saleable. However, the theory is not a red herring. As one of the papers I read recently says (paraphrased): the theory is not an attempt to explain the music, the theory came first, we would not have digital music without that theory.

See the way these things work on our psychology and get moulded in our language. Notice recent references to 16/44 as "low resolution?" Why would we call it that? Because someone wants it to become unpopular? OK, maybe its a bit far fetched to call that a marketing-led conspiracy theory, but, even if we are doing it to ourselves, the resultant attitudes and expectations are the same.

Technically, isn't resolution as such more to do with bit rate than sampling rate? ie, the more bits per sample then, potentially, the more accurate is each sample?

<Later>

In fact, come to think of it, I have an addin/plugin/something that allows one to reduce the bit depth of the music. There are two things to doing this: one is that it is amazing how low one can go and still get music, and the other is finding out what what happens when you do go too far, and the music goes like a JPG picture when the percentage is reduced too far. That, I suppose, is "low resolution music."

 

[Santy]

What also matters is the quality of the recording gear, no wonder there are different studios for different budgets. They charge from $20 to $120 an hour (googled) so the equipment must be making some difference.

Notice recent references to 16/44 as "low resolution?" Why would we call it that?

I don't think that can be called low or poor resolution. It is standard resolution. But may be higher sample rates aids the DAC in quantization? Not sure though.

Technically, isn't resolution as such more to do with bit rate than sampling rate? ie, the more bits per sample then, potentially, the more accurate is each sample?

It looks like but then would someone refer to DSD which has 1 bit depth as pathetic in resolution?

 

[asliarun]

I do feel that there is a lot of marketing behind the high-sampling-rate movement. I prefer to call it that, because, as I've said before, the rolls-off-the-tongue, audiophile-beloved word resolution is just so much more saleable. However, the theory is not a red herring. As one of the papers I read recently says (paraphrased): the theory is not an attempt to explain the music, the theory came first, we would not have digital music without that theory.

See the way these things work on our psychology and get moulded in our language. Notice recent references to 16/44 as "low resolution?" Why would we call it that? Because someone wants it to become unpopular? OK, maybe its a bit far fetched to call that a marketing-led conspiracy theory, but, even if we are doing it to ourselves, the resultant attitudes and expectations are the same.

One clarification: I meant to say that using Nyquist and Shannon theorems in a debate like this is a red herring. The theorems are themselves utterly sound, and have been vetted and proven by a scientific process. The burden of proof is on the party that wants to disprove it, and the proof has to be equally sound, not subjective. If you claim that the digital version has losses in it of any nature, you need to show waveform analysis or other such measures and get the scientific community or a journal of repute to vet and proof your disproof.

I would expect the same from someone who stands up and tries to disprove any other basic scientific law or established theorem.

Otherwise, the argument goes the same way as popular debates such as the ones creationists love, when they try to disprove the theory of evolution. And neer the twain shall meet.

 

[Thad E Ginathom]

OK, understood.

I like this stuff because I think we do make a bit of new ground every time we dig in (Mixed metaphor alert!).

It looks like but then would someone refer to DSD which has 1 bit depth as pathetic in resolution?

DSD is still a new idea to me. I have no clue whatever what "1-bit" means in this context. How can one get any information (apart from on or off) into a single bit? I just don't know what it means. I guess I have a lot of catching up to do!

But may be higher sample rates aids the DAC in quantization?

If there is only one waveform that "connect the dots" in 16/44, and it is the right one, then how can it be any more right in 96 or 192, connecting more dots? This is one the things that puzzles me.

Oh well... I have some new 24/96 music to listen to tonight

But even if I think it sounds wonderful, it won't be relevant to these conversations, because I have no 16/44 copy to compare it to.

 

[Santy]

I finally got the image I was looking for (that helps in easy visualisation of both bit depth and sampling rate in a single illustration).

I understand from this graph that sampling rate and bit depth work in tandem.
They are simply the resolution at x-axis and y-axis respectively.
Both help in better recreation of original analog waveform.

@asliarun
The unshaded boxes in the above graph below the red wave is what according to me is "loss of music". With high resolution music, these losses are minimised to a great extent.


With 1-bit audio, I think we are surely drifting away from the topic (apologies to others) but this one is interesting.
DSD and PCM seem to be two entirely different concept as it is explained here:

How dense are you? | PS Audio

 

[jls001]

@Santy: the unshaded parts are quantisation errors, I think. The higher the number of bits, the lesser the error. Also, the graph is just an amplitude (Y) versus Freq (X) in the classic Gaussian Distribution, to show the differences in fineness of quantisation.

More bits equals greater S/N and dynamic range. I can't seem to think of other usefulness of greater bit depth. Anyone?

Higher sampling rate is useful to push unwanted image frequencies further away from the audio band. I can't think of other usefulness. Anyone?

 

[Thad E Ginathom]

Doesn't Santy's diagram prove my [suggested] point? That the red wave form, the actual sound created, is the same in each case?

It's totally offtopic, but I just had a not-very-satisfactory experience with my 24/96 purchases of yesterday. One has some rather strange noises [with my software/hardware chain] and the other was hard and fatiguing, rather like the low-quality MP3 experience. Disappointing. I'm refreshing my ears with a plain CD version (but this is not a comparison: it is a completely different recording) that has none of those faults.

Sadly, I cannot bring vinyl into the picture, as, in one instance, the LP is scratched to hell, and, in the other, it is the DDA recording that I never liked. But again, different orchestras, places, conductors etc.

 

[asliarun]

I finally got the image I was looking for (that helps in easy visualisation of both bit depth and sampling rate in a single illustration).

I understand from this graph that sampling rate and bit depth work in tandem.
They are simply the resolution at x-axis and y-axis respectively.
Both help in better recreation of original analog waveform.

@asliarun
The unshaded boxes in the above graph below the red wave is what according to me is "loss of music". With high resolution music, these losses are minimised to a great extent.

With 1-bit audio, I think we are surely drifting away from the topic (apologies to others) but this one is interesting.
DSD and PCM seem to be two entirely different concept as it is explained here:

How dense are you? | PS Audio

Santy that is a nice visualization. I respectfully disagree on the conclusion though. The unshaded parts are not loss of music or gaps in any way. Please note that the final output of the digital format is also an analog wave and the output wave will perfectly overlap your red wave. Perfectly. Because the Shannon and nyquist theorem guarantees this to be so. The intermediate quantizations are completely irrelevant as long as the dictats of the theorem are met.

The reason why digital actually sounds inferior is because of technology not the theorem or the sample rate. In other words, filters and caps etc that are in the encoding circuit don't do a perfect job of either capping the audio spectrum to the limits of the theorem and nor are they fast enough for accurate quantization.

And so trickery is resorted to, in the form of oversampling and greater than 44.1khz sampling. It lets the playback circuit deal better with errors in the original sample, but it does not improve the music itself in any way.

Or at least this is what I know about this subject. I may be wrong too.

On a side note I love the simplicity and elegance of the analog vinyl solution. The mic diaphragm converts sound into analog electrical signals which are then literally etched into the vinyl record. The player and speaker driver then do the exact opposite to produce sound. This is zen like.

 

[manoj.p]

On a side note I love the simplicity and elegance of the analog vinyl solution. The mic diaphragm converts sound into analog electrical signals which are then literally etched into the vinyl record. The player and speaker driver then do the exact opposite to produce sound. This is zen like.

The only problem is that medium is not accurate anymore further. Vinyl is a degrading medium which changes after every playback due to wear, dirt, scratches, time etc. You never get the same data which was literally etched on it ever. Even in the replicating factory, none of the vinyl is exact replica of master. But we can ignore all that because vinyl just sounds so romantic.

 

[Santy]

@Thad
The red wave is not the output but the input and the pic only shows how the digitization traces it.
The measured output wave when zoomed into could show aberrations like this.
This is the measurement of TEAC's "reference" series DAC doing its duty at its best, but without any digital filtering.

Even with digital filtering, you could see it is still not smooth but jaggered to a fine degree.

@Joshua
Yes they are quantization errors. But what do errors mean?
When I see this pic, it makes me believe that errors are nothing but "deviations" from the original signal.

Now how well it is taken care by the DAC is beyond the scope of the discussion but errors do exist and have to be dealt with meticulously to achieve best possible output. I also think not all DACs deal with errors with same efficiency.

The unshaded parts are not loss of music or gaps in any way. Please note that the final output of the digital format is also an analog wave and the output wave will perfectly overlap your red wave. Perfectly. Because the Shannon and nyquist theorem guarantees this to be so. The intermediate quantizations are completely irrelevant as long as the dictats of the theorem are met.

By loss of music, I do not mean loss of frequency. It is loss of detail. When we say DAC 'A' has more detail than DAC 'B', what do we actually mean here. It likely means that DAC 'A' is able to perform better in reconstruction of the original wave and thereby retaining the finer details of music.

The theories are guiding principles to program the semiconductors but the output as you have mentioned in your own post, depends on how well we are able to implement it. It guarantees the process not necessarily the output. If merely applying the theorem is all it takes for recovering the original waveform 'perfectly', then all DAC's would sound the same isn't it? The theorem is in the DAC chip while the quality of output depends on various other components apart from the chip itself. They do help to a great extent in overcoming the limitations of reconstructing the original signal through technologies like oversampling, noise cancellations, jitter management, over clocking, digital filtering etc.

And so trickery is resorted to, in the form of oversampling and greater than 44.1khz sampling. It lets the playback circuit deal better with errors in the original sample, but it does not improve the music itself in any way.

Not sure if I got it clear but if errors are minimized how that will not result in improvement in sound?

Ofcourse I will be the last man on earth to conclude that digital is inferior to analog because it has to deal with the "approximations" but I am merely reiterating that analog does not face this challenge.

 

[prem]

Hi Santy

Music signal is a sine wave. Digital generates a square wave. The challenge is to get it as close to a sine wave as possible. Up sampling is one way of getting closer to a sine wave.

Because digital generates a square wave, it sounds different from vinyl. The high end cd players through their algorithms approximate a sine wave much closely. Hence they sound closer to analogue and tonally more dense.

 

[Santy]

The challenge is to get it as close to a sine wave as possible. Up sampling is one way of getting closer to a sine wave.

Because digital generates a square wave, it sounds different from vinyl. The high end cd players through their algorithms approximate a sine wave much closely. Hence they sound closer to analogue and tonally more dense.

Exactly. It pretty much summarizes what I wanted to stress on.

That getting closer to the original analog signal is a challenge which is not an issue with Vinyls.
Upsampling helps in this challenge so does content with denser recording
High end DAC/CDP have better ways of dealing with approximations to get smoother waves

 

[prem]

Hi Santy

Because digital has lesser weight and tonal density, it sounds more open with higher separation. Analogue on the other hand has better tonal qualities but tends to sound closed in. As you go up higher the chain, differences reduce but the portrayal is still a bit different.

 

[asliarun]

@Thad
The red wave is not the output but the input and the pic only shows how the digitization traces it.
The measured output wave when zoomed into could show aberrations like this.
This is the measurement of TEAC's "reference" series DAC doing its duty at its best, but without any digital filtering.

Even with digital filtering, you could see it is still not smooth but jaggered to a fine degree.

@Joshua
Yes they are quantization errors. But what do errors mean?
When I see this pic, it makes me believe that errors are nothing but "deviations" from the original signal.

Now how well it is taken care by the DAC is beyond the scope of the discussion but errors do exist and have to be dealt with meticulously to achieve best possible output. I also think not all DACs deal with errors with same efficiency.


By loss of music, I do not mean loss of frequency. It is loss of detail. When we say DAC 'A' has more detail than DAC 'B', what do we actually mean here. It likely means that DAC 'A' is able to perform better in reconstruction of the original wave and thereby retaining the finer details of music.

The theories are guiding principles to program the semiconductors but the output as you have mentioned in your own post, depends on how well we are able to implement it. It guarantees the process not necessarily the output. If merely applying the theorem is all it takes for recovering the original waveform 'perfectly', then all DAC's would sound the same isn't it? The theorem is in the DAC chip while the quality of output depends on various other components apart from the chip itself. They do help to a great extent in overcoming the limitations of reconstructing the original signal through technologies like oversampling, noise cancellations, jitter management, over clocking, digital filtering etc.



Not sure if I got it clear but if errors are minimized how that will not result in improvement in sound?

Ofcourse I will be the last man on earth to conclude that digital is inferior to analog because it has to deal with the "approximations" but I am merely reiterating that analog does not face this challenge.

The Nyquist theory is not a guiding principle - it is an exact theorem that states that an an analog signal can be perfectly reproduced in a specific bandwidth as long as the sample rate is at least twice the frequency.

To put it bluntly, if you follow these strictures, and convert your audio music from digital to analog back and forth a few thousand times, you will still get back the perfect signal or waveform you started with. This is fact. You want to disagree, fine, then disprove it with scientific proof.

There's no saw tooth waveform. That is a figment of your imagination. There is only a digital encoding of the analog waveform and the guarantee that it can be perfectly codified. Meaning, if the digital signal is translated into analog it will perfectly represent the original signal. Not an approximation but perfection.

For that matter, your brain itself is a discrete sampling device. The audio frequencies that are pucked up in your ear drum are translated into discrete neural pulses. There's no analog equivalent of the brain signals. And before you object, it is not binary either. But it is indeed discrete pulses and most certainly not a seamless waveform.

And when you so easily say, merely applying the theorem to practice, it is not that easy. And that is where the whole issue arises from. It is also equally easy to say that all an analog recording medium needs is to be able to perfectly record and translate the full audio signal it receives. Easy to say. Not so easy to implement. At least not perfectly.

So all we are arguing is the compromises we are willing to make.

 

[Thad E Ginathom]

I wonder at those generalisations --- although you have probably heard more of both than I have. It's some months since I did any digitising, but I do not see loss of weight or density in the result. I still say that, so far as home experiment is concerned, digitisation of one's own vinyl is the closest that we can practically get to comparing an orange with an orange. And it has been known to floor listeners in tests far better conceived and set up than any informal and unscientific trial I have ever done.

Because digital generates a square wave

Watch the Xiph video (I'll come back with the link this afternoon, I have to go out now) --- it very clearly shows sine-wave in, sine wave out. Or maybe you talking about something else? EDIT: Departure delayed, Xiph videos. I know they have been criticised for over-simplifying, but for an audience of people like me, that's just fine.

There's no saw tooth waveform. That is a figment of your imagination.

True, but a little harsh, because we have all seen that stepped graph in almost all explanations of digital sound that we have ever seen. It is only a matter of weeks since the Xiph video put me right on that one --- and even they say that, although it is not the best representation, even they use that graph in certain cases. It seems, by the way (from my memory of the same source) that early DACs may indeed have produced such a wave form, essentially playing each sample as a continuous sound until they receive the next one. It would be interesting to compare such a DAC with a modern one.

 

[Santy]

to put it bluntly, if you follow these strictures, and convert your audio music from digital to analog back and forth a few thousand times, you will still get back the perfect signal or waveform you started with. this is fact. you want to disagree, fine, then disprove it with scientific proof.

There's no saw tooth waveform. That is a figment of your imagination. There is only a digital encoding of the analog waveform and the guarantee that it can be perfectly codified. Meaning, if the digital signal is translated into analog it will perfectly represent the original signal. Not an approximation but perfection.

Hmmm. OK...