Does High Resolution Audio sound better

[lamode]

Then DAC: Nearly all DAC chips will do the 8x oversampling in 2x steps: 44.1 -> 88.2 -> 176.4 -> 352.8. So that's three more conversions.

 

Sorry but that's rather pedantic and disingenuous to call 8x oversampling '3 conversions'. For one thing, the noise generated by the interpolation is tiny, less than -100dB, so the effect of the whole filter is still inaudible in the real world.

 

For the life of me, I can't understand the meaningless fixation on hi-res audio, when every transducer we are using is deeply flawed, not to mention listening environments. Upgrading a speaker or listening room will transform your music in ways which HD audio can only dream about.

[sandyk]

As long as it is an open forum, you will get people coming along from time to time and trying to reconnect you to reality.

 

It would appear that your "reality" is very different to that of many very talented EEs in the major corporations and enthusiastic consumers !

[Jud]

Sorry but that's rather pedantic and disingenuous to call 8x oversampling '3 conversions'. For one thing, the noise generated by the interpolation is tiny, less than -100dB, so the effect of the whole filter is still inaudible in the real world.

 

 

Interesting. Then all the engineers who designed these filters (without any of the limitations of time and resources when filtering in the DAC chip) must be really, really stupid not to have perfect conversions: SRC Comparisons

 

These major corporations need to hire a smart fellow like you to tell their stupid engineers how to do this trivial thing correctly.

[wakibaki]

I was actually trying to speak to the topic that holds some interest, rather than trying to "hide" the original issue. The answer to whether there's an audible difference between 16/44.1 and high res is trivial and was answered more than 25 years ago.

 

If you're saying that there was found to be a difference, where are the studies that showed that? Why aren't they quoted every time this argument comes up?

 

Oh, this was all settled 25 years ago... ...in your dreams.

 

This is not your area of expertise. A lawyer who has a lawyer for an engineer is a fool.

 

You state your conclusion, then go on to justify it by inference. The conclusions you draw reflect your incomplete grasp of the subject, there are more reasons to upsample than mere increased detail, and anyway the upsampled bitstream can contain no more information than the original.

[Jud]

If you're saying that there was found to be a difference, where are the studies that showed that? Why aren't they quoted every time this argument comes up?

 

Oh, this was all settled 25 years ago... ...in your dreams.

 

This is not your area of expertise. A lawyer who has a lawyer for an engineer is a fool.

 

You state your conclusion, then go on to justify it by inference. The conclusions you draw reflect your incomplete grasp of the subject, there are more reasons to upsample than mere increased detail, and anyway the upsampled bitstream can contain no more information than the original.

 

I said nothing about detail or information (and in any case, I believe that under the strict definition, involving the same Shannon who also lent his name to the sampling theorem, there actually is more information in an upsampled bitstream). What I talked about was distortion - aliasing creating harmonic distortion, to be more precise.

 

You're correct that this is not my area of expertise. My limited understanding is that interpolation allows the anti-aliasing filter to have a gentler slope and thus be more correct in the time domain (less ringing), while still cutting the signal sufficiently (edit: at audible frequencies) to avoid harmonic distortion.

[alfe]

Interesting. Then all the engineers who designed these filters (without any of the limitations of time and resources when filtering in the DAC chip) must be really, really stupid not to have perfect conversions: SRC Comparisons

 

These major corporations need to hire a smart fellow like you to tell their stupid engineers how to do this trivial thing correctly.

 

Jud,

 

As many you are still confusing oversampling and up sampling.

[Jud]

Jud,

 

As many you are still confusing oversampling and up sampling.

 

alfe, these are really marketing terms, and there is no fundamental difference in terms of the math or the design considerations that I know of. It's still sample rate conversion accompanied by the necessary filtering.

[esldude]

I was actually trying to speak to the topic that holds some interest, rather than trying to "hide" the original issue. The answer to whether there's an audible difference between 16/44.1 and high res is trivial and was answered more than 25 years ago. The reason 8x oversampling became a standard is that going direct to analog from 16/44.1, and to a lesser extent from 88.2 (2x oversampling was tried for a short time before 8x oversampling took over), creates audible distortion. So an entire industry - virtually all audio engineers working in the field at the time - agreed 16/44.1 was simply unworkable, 88.2 was no picnic either, and 352.8/384 was where we should be.

 

Back then storage and transmission wouldn't allow distributing 352.8/384, so we got the present-day kludge of downsampling at the recording end, then upsampling at the consumer end. That's been pretty well set in stone until now just from sheer economics: due to the fact everyone's doing it so you have economies of scale, it's a hell of a lot cheaper to stay with the same distribution model. But if you were starting with a clean sheet of paper, these days you'd have the music companies sending us DSD or DXD (352.8/384) instead of fiddling with all these conversions.

 

This and your previous post Jud contain much revisionist history.

 

The reason for oversampling, sigma delta and such changes were not because the question of audibility was answered 25 years ago. It was for less expensive ADC and DAC chips and with higher performance (higher measured performance). The chips became cheaper and better (closer to theoretical perfection). Whether that was audible was not conclusively determined and in fact was not a factor. The marketing of the changes of course were claiming improvements in sound. But hey, we get that all the time from things that don't matter.

[Jud]

This and your previous post Jud contain much revisionist history.

 

The reason for oversampling, sigma delta and such changes were not because the question of audibility was answered 25 years ago. It was for less expensive ADC and DAC chips and with higher performance (higher measured performance). The chips became cheaper and better (closer to theoretical perfection). Whether that was audible was not conclusively determined and in fact was not a factor. The marketing of the changes of course were claiming improvements in sound. But hey, we get that all the time from things that don't matter.

 

Sigma-delta chips were cheaper. 8x oversampling preceded them. Did 8x oversampling allow good filtering to be done more easily and cheaply? Sure. Another way to say this is that it's damn difficult and expensive to do good analog to digital conversion with acceptable distortion levels at 44.1kHz.

[alfe]

alfe, these are really marketing terms, and there is no fundamental difference in terms of the math or the design considerations that I know of. It's still sample rate conversion accompanied by the necessary filtering.

 

Then why you put a link to SRC comparison?

[Jud]

Then why you put a link to SRC comparison?

 

Because I wanted to illustrate visually that sample rate conversion has got tradeoffs. This is what Tony, JonP, and Barry were talking about - doing a conversion that is completely transparent is mathematically impossible, and according to people with a lot of experience, difficult to impossible to do inaudibly.

 

Re the two terms, "oversampling" is most often used to describe sample rate conversion by integer multiples, "upsampling" to describe conversion to a rate that's a non-integer multiple. For the computer or chip on which the algorithm runs, there is no fundamental difference, given that the multiple, though not necessarily an integer, is a rational number.

[esldude]

Sigma-delta chips were cheaper. 8x oversampling preceded them. Did 8x oversampling allow good filtering to be done more easily and cheaply? Sure. Another way to say this is that it's damn difficult and expensive to do good analog to digital conversion with acceptable distortion levels at 44.1kHz.

 

The very 1st Sony CDP 101 had THD of less than .004%. Is that not an acceptable distortion level?

[beetlemania]

Yet no-one has ever passed a 16/44 v HD audio ABX test by a statistically meaningful margin.

 

I did my own testing (with Foobar comparator) and don't give damn about your opinion or, even, peer-reviewed "ABX test by a statistically meaningful margin". I found that I was able to discern 16 from 24 bit but not 44 kHz from 96 or 192 kHz. As a result, I try to buy 24 bit files when available but almost never pay the premium for 192 kHz.

[Jud]

The very 1st Sony CDP 101 had THD of less than .004%. Is that not an acceptable distortion level?

 

http://www.research.philips.com/technologies/projects/cd/technology.html

[alfe]

Because I wanted to illustrate visually that sample rate conversion has got tradeoffs. This is what Tony, JonP, and Barry were talking about - doing a conversion that is completely transparent is mathematically impossible, and according to people with a lot of experience, difficult to impossible to do inaudibly.

 

Re the two terms, "oversampling" is most often used to describe sample rate conversion by integer multiples, "upsampling" to describe conversion to a rate that's a non-integer multiple. For the computer or chip on which the algorithm runs, there is no fundamental difference, given that the multiple, though not necessarily an integer, is a rational number.

 

You will find here a public domain resampling algorithm:

https://ccrma.stanford.edu/~jos/resample/resample.pdf

[jabbr]

Very good Jud, but none of this speaks to the superiority or otherwise of RB as a distribution system. It all hinges on whether the differences are audible, not whether there's a possibility that skipping 2 steps might make a difference.

 

I wish you'd confine yourself to less prolix posts, but I suppose that's all part of trying to hide the concept that you're trying to palm.

 

Maybe that's a bit unfair, but it's a trait of lawyers to become entranced by a good argument, to the point of considering the truth irrelevant, but I guess in a crowd of audiophiles you're not going to stick out.

 

And you are trying to make what point exactly? Argument by personal attack is appropriate to what field?

[esldude]

Philips Research - History of the CD - Technology

 

Yes, I owned one of those early Magnavox players (same internally as Philips). The stuff about oversampling and filtering is true. The real reason Philips went that route was they had planned all along on a 14 bit system. Had 14 bit chips ready, the involvement of Sony, with Sony insisting on 16 bit, meant that was the standard agreed upon for CD. Philips didn't have 16 bit chips yet, but with oversampling could get closer to 16 bit performance. Wasn't too long until they had 4x oversampled 16 bit chips. None of this was about lowering distortion. It was about hardware possible at the time, and making things perform better and doing so more cheaply.

[wakibaki]

there actually is more information in an upsampled bitstream

 

You're dead wrong, where would it come from?

 

Oversampled, at the same bit depth, yes.

 

These are not trivial errors. These are vital distinctions. If I said that, I'd get ripped to shreds.

 

If you're dismissing them as marketing jargon, you really don't understand the process.

[esldude]

You're dead wrong, where would it come from?

 

 

The extra info comes from the invisible bit gnomes. These gnomes (much like Keebler elves) add goodness in every extra bit. Results in obvious smoothness, more there there (after all the bit gnomes put it there), a greater sense of relaxation and ease, along with more good of everything.

[wakibaki]

And you are trying to make what point exactly? Argument by personal attack is appropriate to what field?

 

1. Jud's posts are too long.

2. He appreciates a good quote, and he's almost certainly read Catch 22. Ou sont les Neigedons d'antan?

3. I prefer to use a carrot, but some mules are stubborn.

[sandyk]

The extra info comes from the invisible bit gnomes. These gnomes (much like Keebler elves) add goodness in every extra bit. Results in obvious smoothness, more there there (after all the bit gnomes put it there), a greater sense of relaxation and ease, along with more good of everything.

 

And you wonder why you needed to start that self indulgent Poll recently ?

[esldude]

And you wonder why you needed to start that self indulgent Poll recently ?

 

The bit gnomes are meant to be entertaining make believe. Just like the idea upsampling adds information.

[lamode]

there actually is more information in an upsampled bitstream

 

It is technically possible, but I don't think anyone is doing this with current algorithms. I have often wondered how well it would work with digital audio. I have worked a bit with image interpolation, which solves similar problems. Have a look at these photos and click on the different upscale buttons to see what's possible.

 

http://www.wisdom.weizmann.ac.il/~vision/SingleImageSR.html

 

CDP 101

 

Wow, haven't heard that name for a looong time. Brings back memories of my first audition

[wakibaki]

The extra info comes from the invisible bit gnomes. These gnomes (much like Keebler elves) add goodness in every extra bit. Results in obvious smoothness, more there there (after all the bit gnomes put it there), a greater sense of relaxation and ease, along with more good of everything.

 

Nobody likes a smartass...

 

 

 

...oh. I just had a horrible moment of personal insight.

[alfe]

And you wonder why you needed to start that self indulgent Poll recently ?

 

Be indulgent with Dennis he had to read all the crap from jonP without reacting:)