LP with better dynamic range than digital

[ack]

I just read the Benchmark blog... what he says is basically what Xiph/Monty show on that video we have seen a number of times in the past, regarding noise. However, I also got stuck on that "Properly-dithered digital systems have infinite amplitude resolution" claim... I thought the blog post was mostly coherent, but some conclusions are non-sequitur. After the following opening statement...

While it is true that digital systems quantize the amplitude of the audio signal to the nearest step in the digital encoding system, this does not necessarily mean that digital systems cannot have infinite resolution. Contrary to popular belief, digital systems can provide infinite amplitude resolution if they are properly dithered.

...I felt he failed to successfully argue the underlined conclusion.

Regarding the original post: a) from what I have seen, no one has claimed to have an LP with 120dB dynamic range or that their analog rig is capable of such, only that we have LPs that sound more dynamic than any 16-bit based material; and I think we all acknowledge that our analog rigs are the primary limiting factor; b) indeed, don't use a cartridge's reference output as an indication of its *lowest* output voltage, when calculating its dynamic range.

 

[esldude]

I just read the Benchmark blog... what he says is basically what Xiph/Monty show on that video we have seen a number of times in the past, regarding noise. However, I also got stuck on that "Properly-dithered digital systems have infinite amplitude resolution" claim...

snip

Many people get hung up on the infinite resolution idea. Let us imagine we had a fantastically wonderful, and perfect recording and playback system. Let us imagine it is distortionless in any manner as well as having 180 db of dynamic range and the appropriate resolution to go with it. Right out to the speakers themselves. For obvious reasons loudspeaker output is limited to 120 db in our imaginary system. It has the ability to resolve down -180 db below this which means it is some -60 db below the threshold of hearing. Plus most listening rooms will have at least 30 db of noise or so. Now humans can hear into noise perhaps some 20 db. So you might perceive down to 10 db. But our fantastic imaginary system will resolve still lower by -70 db.

Now there may be some signal simply so low though resolved by this system it is too faint for humans to hear in our listening room environment due to the noise of it. The fineness of the resolution could be so good a note barely heard could be played back a near infinitely smaller amount louder though we wouldn't hear the difference so small it is. Though it would be resolved though lost mostly to us in the noise level.

This is somewhat what happens with properly dithered digital audio at low levels. The fineness of the resolution is actually a fraction of the LSB in how much or how little a given barely audible note can be increased or decreased in loudness. This is due to how dither works. The resolution of levels below the LSB works, but is obscured eventually by the wideband noise floor of the digital process itself. 24 bit will have lower noise than 16 bit, but both can resolve the smallest level differences at near infinitely small degrees of fineness.

Listening to the musical portion of Don Hill's link with both dithered and unditherd 4 bit audio should make that reasonably clear.

 

[ack]

"at near infinitely small degrees of fineness" yes, but still discrete. I don't see anyone presenting any proof that "infinitely small" is the same as non-discrete. I saw Don's link to tnt-audio, and most links in there are broken, so I must have missed something.

 

[esldude]

"at near infinitely small degrees of fineness" yes, but still discrete. I don't see anyone presenting any proof that "infinitely small" is the same as non-discrete. I saw Don's link to tnt-audio, and most links in there are broken, so I must have missed something.

All the links were working two days ago. I just checked and only the lower links to the music work. So maybe try again later.

I said near infinite though the main point is the degree of fineness is well, well below the LSB.

As for proof what would you accept as proof that levels of signal can be portrayed at differences below one LSB?

 

[ack]

All the links were working two days ago. I just checked and only the lower links to the music work. So maybe try again later.

I said near infinite though the main point is the degree of fineness is well, well below the LSB.

As for proof what would you accept as proof that levels of signal can be portrayed at differences below one LSB?

Hmm, in fact all the links to the music were still broken when I responded, and are still broken now?!? I am getting a 400 - Bad request?!??? I am looking for mathematical proof, and not only involving the LSB.

 

[Orb]

Yes so it is actually infinite precision rather than infinite amplitude resolution (two different things), and that is mostly theoretical rather than real world IMO.

Cheers
Orb

 

[esldude]

Hmm, in fact all the links to the music were still broken when I responded, and are still broken now?!? I am getting a 400 - Bad request?!??? I am looking for mathematical proof, and not only involving the LSB.

http://www.robertwannamaker.com/writings/rw_phd.pdf

You can try this. I have only read a few select sections related to this topic. The conclusion may be enough or not.

http://www.users.qwest.net/~volt42/cadenzarecording/DitherExplained.pdf

Shorter version with some illustrations though no real math.

 

[Don Hills]

<sigh>*There's* something they don't teach in school... </sigh> I'm sorry but that statement is pretty funny.

We can pop a hole in that one pretty quick. Let's limit the word length to 4 bits (since there is "no advantage to long word lengths", we will prove or disprove that by using a short word length). Please show how a 4 bit word can describe infinite values. (Hint: 16 values are possible with 4 bits. With 16 bit you get 65536 values.) ...

A properly implemented 4 bit digital system (includes dither) can resolve an infinite number of input values (between 0 and FS.)
For example, let's say that the 4 bit system represents 1 volt of input per step. Input 2 volts, get digital value 2. Input 3 volts, get digital value 3.
Apply 2.5 volts. What do we get? Without dither, we get either value 2 or 3 depending on the ADC design.

Now apply dither, which is a random value voltage between +0.5 and -0.5 volt with a certain probability function - that is, the likelihood that a given value will occur. The most common is TPDF, or Triangular Probability Density Function. If you plot the likelihood of a given dither voltage occurring, the curve will be shaped like a triangle: 0 will occur most often, with a linearly decreasing likelihood as you approach +0.5 or -0.5 volts.

Therefore, the digital value for our 2.5 volt input will vary between 2 and 3 for any given sample period. Because our input is 2.5 volts, and the most probable value for the dither voltage is 0 volts, half the time the dither voltage will be 0 or negative and half the time 0 or positive. Adding this to the input voltage means that half of the time the input will be 2 to 2.5 volts and the other half the input will be 2.5 to 3 volts. The result will be a series of digital values of either 2 or 3, with a random distribution but over time averaging out to 50% probability of either.

Now we feed this to the DAC. It will produce a string of output voltages of either 2 or 3 volts, with an equal probability of either value. Over time, it will average to 2.5 volts, with +-0.5 volt of noise.

If the input is, for example, 2.1 volts, the same rules apply except that the probabilities shift for the voltage including dither. 10% of the time the total will exceed 2.5 volts, and the other 90% of the time it will be less than 2.5 volts. Therefore, the digital stream will contain 10% values of 3 and 90% values of 2. After the DAC, the output will be 2.1 volts with +-0.5 volts noise.

You can accurately resolve a value lower than the lowest bit value, provided you allow enough time to average the output. You can see this in the 4-bit music examples pointed to earlier - undithered, the "decay/reverb tails" become distorted and then disappear suddenly as their values become less than 1 bit. For the dithered examples, they fade smoothly into the noise. Our ears are quite good at resolving signals in noise. We do it all the time when listening to LPs...

"But", you might say, "at any given sample the value can still only be one of 16 values." This is true, but irrelevant. Let's say we listen to (and measure by any means you choose) the input analogue signal with the dither signal added. We then listen to (or measure) the output signal after ADC - DAC. They sound the same, and they measure the same. "But the original signal didn't have noise", you might say. But it will have noise. In many cases, it will have more noise than the dither signal. So if the dither noise is lower than the noise in the input signal, and / or is so low in level as to be inaudible under almost all conditions (16 bit) or all conditions (24 bit), does it matter?

 

[Don Hills]

Yes so it is actually infinite precision rather than infinite amplitude resolution (two different things), and that is mostly theoretical rather than real world IMO. ...

You say potayto, I say potahto...

For those getting 400 erors when accessing the samples, I saw the same thing the first couple of times. Retrying usually worked. I've seen other examples of 4-bit sampling from time to time, but I didn't bookmark them.

 

[Orb]

You say potayto, I say potahto...

For those getting 400 erors when accessing the samples, I saw the same thing the first couple of times. Retrying usually worked. I've seen other examples of 4-bit sampling from time to time, but I didn't bookmark them.

Bah being an audio forum and for me just to be an argumentative so and so... I say tomato
Cheers
Orb

 

[Atmasphere]

Now we feed this to the DAC. It will produce a string of output voltages of either 2 or 3 volts, with an equal probability of either value. Over time, it will average to 2.5 volts, with +-0.5 volt of noise.


"But", you might say, "at any given sample the value can still only be one of 16 values." This is true, but irrelevant. Let's say we listen to (and measure by any means you choose) the input analogue signal with the dither signal added. We then listen to (or measure) the output signal after ADC - DAC. They sound the same, and they measure the same. "But the original signal didn't have noise", you might say. But it will have noise. In many cases, it will have more noise than the dither signal. So if the dither noise is lower than the noise in the input signal, and / or is so low in level as to be inaudible under almost all conditions (16 bit) or all conditions (24 bit), does it matter?

Thanks for taking the time with your explanation!

I think it must matter- *something* does, that's for sure.... As Mr. Wannamaker points out, more study is needed (see conclusions in his thesis). If there is noise in the input signal, and the output signal contains noise but not the same noise as the input, is there a difference? Is is audible? Would that be considered a distortion?

In a nutshell, something is wrong with this picture: with every digital system out there you can still hear that the LP is superior. The most sophisticated and musical DAC I have heard (being the Stahltek system, which retailed at $72,000, and I did not just hear it, I lived with it as I know the designer and we have done shows together) to date; simply put, you could hear how the LP was superior and it only took 5 seconds to hear the difference. Its one thing for me to hear it, its a bit of another when you have the designer of the gear with you to hear it at the same time (he simply turned to me and said 'Digital has such a long way to go.."). Now I have to qualify that the analog system we were playing was excellent and I do find that many who espouse digital particularly heavily often have not got a state of the art analog system to which they can compare (in some cases, have never even heard a state of the art analog system). In the example above, both the analog and digital systems were or were very nearly state of the art in each camp. I can list the gear if needed.

I highlighted on bold in your explanation this issue of time. Its just a theory of mine, but I think the timing thing is a problem. I agree we can hear into the noise floor (a pet theory of mine is this has something to do with wind sounds in our environment). It might be simply that while the DAC will eventually get it right, that in practice this does not happen because there is not enough time for it to do so.

Where I get a little testy is the idea of infinite. My GF hates that I take things so literal, but when we are talking infinite, I take that to mean any kind of signal; perhaps a poor choice of words??

 

[esldude]

Thanks for taking the time with your explanation!

I think it must matter- *something* does, that's for sure.... As Mr. Wannamaker points out, more study is needed (see conclusions in his thesis). If there is noise in the input signal, and the output signal contains noise but not the same noise as the input, is there a difference? Is is audible? Would that be considered a distortion?

In a nutshell, something is wrong with this picture: with every digital system out there you can still hear that the LP is superior. The most sophisticated and musical DAC I have heard (being the Stahltek system, which retailed at $72,000, and I did not just hear it, I lived with it as I know the designer and we have done shows together) to date; simply put, you could hear how the LP was superior and it only took 5 seconds to hear the difference. Its one thing for me to hear it, its a bit of another when you have the designer of the gear with you to hear it at the same time (he simply turned to me and said 'Digital has such a long way to go.."). Now I have to qualify that the analog system we were playing was excellent and I do find that many who espouse digital particularly heavily often have not got a state of the art analog system to which they can compare (in some cases, have never even heard a state of the art analog system). In the example above, both the analog and digital systems were or were very nearly state of the art in each camp. I can list the gear if needed.

I highlighted on bold in your explanation this issue of time. Its just a theory of mine, but I think the timing thing is a problem. I agree we can hear into the noise floor (a pet theory of mine is this has something to do with wind sounds in our environment). It might be simply that while the DAC will eventually get it right, that in practice this does not happen because there is not enough time for it to do so.

Where I get a little testy is the idea of infinite. My GF hates that I take things so literal, but when we are talking infinite, I take that to mean any kind of signal; perhaps a poor choice of words??

I am not alone in thinking the LP is not superior. Different yes. Superior no. Equal in quality no. But you already have stated your excuse for disagreement. I must not have heard this in a state of the art analog system by your definition, blah, blah, blah.

The LP upon playback is inferior in fidelity. It may still be subjectively preferred. The best course of action isn't to assume it is superior instead, but to figure out why it might be preferred though of lower fidelity.

 

[rockitman]

I am not alone in thinking the LP is not superior. Different yes. Superior no. Equal in quality no. But you already have stated your excuse for disagreement. I must not have heard this in a state of the art analog system by your definition, blah, blah, blah.

The LP upon playback is inferior in fidelity. It may still be subjectively preferred. The best course of action isn't to assume it is superior instead, but to figure out why it might be preferred though of lower fidelity.

^^^^hmmm, analog mastered greats from the 50's-70's sound far better on well sorted vinyl and tape rigs. I feel bad for those who rely on record labels to give you their best digital approximations of master analog material that is compromised in terms of micro detail and dynamic range via compression, normalization ect.

 

[esldude]

^^^^hmmm, analog mastered greats from the 50's-70's sound far better on well sorted vinyl and tape rigs. I feel bad for those who rely on record labels to give you their best digital approximations of master analog material that is compromised in terms of micro detail and dynamic range via compression, normalization ect.

In my own comparisons of such material side by side using reel tape, CD and LP it was always LP that was the odd man out in terms of sound. Was surprised when I first did this how close some reel and CD releases were. Others were further apart, but the LP was in each case much more different.

 

[microstrip]

I am not alone in thinking the LP is not superior. Different yes. Superior no. Equal in quality no. But you already have stated your excuse for disagreement. I must not have heard this in a state of the art analog system by your definition, blah, blah, blah.

The LP upon playback is inferior in fidelity. It may still be subjectively preferred. The best course of action isn't to assume it is superior instead, but to figure out why it might be preferred though of lower fidelity.

Esldude,

We know the systems of people who express the opinion that LP playback is superior. Just to enlighten us, can we know the complete LP system that you have used to build your opinions?

IMHO the main interest of the debate is not the preference opinion, but the detailed reasons why people have this preference. Some of these people have since long written extensive comparisons, reviews, hundreds of posts and messages explaining their subjective preference with great detail. All of them suggest a more realistic perception of the recording using the LP. For me this is the ultimate fidelity and makes it a superior media, although I recognize that in some aspects digital wins. Unfortunately I also valuate a lot the LP winning qualities, I would be a much happier man if I could have a completely digital system.

 

[Mike Lavigne]

In my own comparisons of such material side by side using reel tape, CD and LP it was always LP that was the odd man out in terms of sound. Was surprised when I first did this how close some reel and CD releases were. Others were further apart, but the LP was in each case much more different.

maybe since your perception on this subject is so contrary to quite a few of us folks, and clearly the system listed in your profile does not include any analog, you could describe the system and gear and media context of.....

my own comparisons of such material side by side using reel tape, CD and LP

.....that you hold up as evidence....for us to better understand why you hold that opinion.

 

[FrantzM]

Then we get back to a matter of preferences. I find interesting that people are asked to explain their preferences .. Interesting

This said .. I am still waiting to be shown how an LP can have 120 dB of dynamic range (Ok We can try 100 dB to make things a little bit easier) .. One thing I haven't approached yet are the mechanics of such LP. The groove modulation that would be necessary to produce 120 dB and what that would mean for the poor cartridge and arm trying to reproduce those ...

 

[Don Hills]

Thanks for taking the time with your explanation!

I think it must matter- *something* does, that's for sure.... As Mr. Wannamaker points out, more study is needed (see conclusions in his thesis). If there is noise in the input signal, and the output signal contains noise but not the same noise as the input, is there a difference? Is is audible? Would that be considered a distortion?

There is a difference, but it may not be audible. I don't think it matters provided it isn't audible, and there is no mechanism for it to cause a problem (for example, the huge amount of ultrasonic noise in the output of a DSD DAC chip can cause problems if not filtered, although it is in theory inaudible). There's also a significant amount of noise in the playback of an LP, but it doesn't seem to matter for vinyl enthusiasts.

... In a nutshell, something is wrong with this picture: with every digital system out there you can still hear that the LP is superior.

"Superior" is a matter of preference. The word you were perhaps looking for is "different", which is a matter of fact.

... I highlighted on bold in your explanation this issue of time. Its just a theory of mine, but I think the timing thing is a problem. I agree we can hear into the noise floor (a pet theory of mine is this has something to do with wind sounds in our environment). It might be simply that while the DAC will eventually get it right, that in practice this does not happen because there is not enough time for it to do so.

I can see the reasoning behind that theory, but in practice the digital system gets it close enough for rock'n'roll. Remember that each sample represents a very small moment in time. Even at 16 bits, the error on any one sample is at or below the limits of audibility. If the input is silent, you might just hear a slight hiss from the dither under optimum conditions. As soon as you add other input, program or noise, masking effects make it impractical to hear the dither. Remember that the dither removes any program-related error due to finite quantisation and replaces it with uncorrelated noise.

Where I get a little testy is the idea of infinite. My GF hates that I take things so literal, but when we are talking infinite, I take that to mean any kind of signal; perhaps a poor choice of words??

They were John Siau's words, I bowed to his significant expertise on the subject. I did qualify it by saying that infinite resolution requires infinite time.

Returning to the "superiority" of vinyl reproduction, I used to perform a party trick which gave many people food for thought. On their highly regarded turntable setup, I would play a simple 1 KHz tone from a test record. Then I would play the same tone from a test CD. The differences were not subtle. Then I would point out that the same problems audible in the LP playback of the tone were present in all music played on that turntable. If it couldn't get a single tone right, how could it hope to do justice to the music? I did, and still do, derive a great deal of enjoyment from LP playback. And I'm not blind to the shortcomings of digital systems. But I know which I prefer, on balance.


One more point:
There is a widely held understanding that a "needledrop" of an LP, if done with care, sounds very much like the original LP playback. If the digital chain can do this accurately, it can likewise accurately capture / reproduce the original signal that was used to cut the LP. Any differences are therefore due to the LP cutting / reproduction chain and not to the digital chain. So if they sound different, which one is getting it wrong?

 

[rockitman]

In my own comparisons of such material side by side using reel tape, CD and LP it was always LP that was the odd man out in terms of sound. Was surprised when I first did this how close some reel and CD releases were. Others were further apart, but the LP was in each case much more different.

It appears to me your only reference is cd. Please share with us your vinyl and tape playback devices you used to come up with your cd is superior conclusion. I fear I will hear the sound of crickets...

 

[microstrip]

Then we get back to a matter of preferences. I find interesting that people are asked to explain their preferences .. Interesting
(...) .

Surely. Analyzing preferences is the ultimate tool of sound reproduction evaluation. It is why we enjoy debating the reasons of our preferences. IMHO we should analyze preferences using our experience, statistical knowledge and credibility of posters - it is why details enhance the value of opinions.