How can DAC's have a SOUND SIGNATURE if they measure as transparent?

[pkane]

Thanks... so (at least) your resampler doesn't make a mess of things.

Of course, I wrote it It's really just a simple sinc resampler, nothing special. But I do test it with multiple repeated resamples to randomly picked rates to make sure it doesn't introduce much in the way of artifacts or distortions.

I'm not familiar with the algorithms, but the since the recovered file isn't exactly identical, my question is if the departure from the original increases if we run the process more than once.

Here's the resample repeated 10 times (44.1k to 192k and back to 44.1k). The noise floor did rise a bit, but still way, way below any audibility thresholds:

 

[killdozzer]

Ah... this mythical creature again...

Riding right next to this mythical creature:

Measurements don't always tell the whole story.

 

[khark]

"Puffery" is a long-standing tradition in marketing materials. The fact that Neumann is a great company with great products does not mean that they don't need to tailor their message for the intended audience.

This isn't really a marketing claim though, they don't actually mention in that article any super special Neumann topology in order to promote their products. It is just a general statement on what should be considered important. If anything, it isn't helpful from a marketing perspective, since it leads to more questions.

 

[xaviescacs]

Here's the resample repeated 10 times (44.1k to 192k and back to 44.1k). The noise floor did rise a bit, but still way, way below any audibility thresholds:

Nice! Now it just remains to see whether the SINAD has an upper bound as the number of resamples tend to infinity or not. I'm sure you don't have anything better do right now.

 

[killdozzer]

Yes, but apparently it's only the last that counts. At least that's what they must contend, which sorta maybe makes sense from the weird perspective of viewing all the prior harm as having been beaten into the bit stream that is now for the last time being converted into a continuous v(t). It's all voodoo to me.

Not quite so. @KozmoNaut has apoint there. If sound deteriorates when going back and forth through AD/DA processes and if you can hear that deterioration, you should be able to recognize that particular type of deterioration in recorded material. Like, you're listening to an album and a strange sound appears and you immediately know "aha, this sound was achieved by going multiple times AD/DA back and forth, I know what that sounds like", right?

 

[Killingbeans]

Neumann? I think you are taking it too far. They are not exactly the kind of manufacturer who needs to make stuff up for marketing purposes, their product measurements speak for themselves. Sometimes one need to be more respectful to the accomplished people in the industry and their views.

I don't necessarily think Neumann feel the need to make up stories. Some myths are simply just such an ingrown part of the industry that even the most well-grounded manufacturers let some of them slip through the net. Especially if your customer base consistently reports back the same BS, I'd imagine that it's nearly impossible to not just roll with it.

I honestly find it hard to name a single manufacturer that doesn't have any questionable claims at all.

 

[killdozzer]

This isn't really a marketing claim though, they don't actually mention in that article any super special Neumann topology in order to promote their products. It is just a general statement on what should be considered important. If anything, it isn't helpful from a marketing perspective, since it leads to more questions.

No marketing fluff ever leads to more questions and neither does this one. It simply translates to "don't consult specs, just take our word for it and our word is buy our product". No company is immune to the need of selling more and you sell more by spreading your market and while spreading you inevitably reach the uninformed and you address those with pink prose. It should suffice to see they wrote "these are subjective attributes". This is almost like a disclaimer, saying "you may not hear what we say you're supposed to". Everything one should understand about the term "subjective" is that it means two people can buy the exact same amp and one might say it's on the warm side, while the other might say it's on the cold side of sound "character". In other words, things like "grainy" have absolutely nothing to do with amp topology.

 

[killdozzer]

I don't necessarily think Neumann feel the need to make up stories. Some myths are simply just such an ingrown part of the industry that even the most well-grounded manufacturers let some of them slip through the net. Especially if your customer base consistently reports back the same BS, I'd imagine that it's nearly impossible to not just roll with it.

I honestly find it hard to name a single manufacturer that doesn't have any questionable claims at all.

Not just that, but every company has a marketing department. These departments don't base what they write on specs provided by the engineers. They use affectionate language that yields most results. Even if you give them some specs, your gut will wrench tomorrow when you see what they used it for.

 

[JRS]

Not quite so. @KozmoNaut has apoint there. If sound deteriorates when going back and forth through AD/DA processes and if you can hear that deterioration, you should be able to recognize that particular type of deterioration in recorded material. Like, you're listening to an album and a strange sound appears and you immediately know "aha, this sound was achieved by going multiple times AD/DA back and forth, I know what that sounds like", right?

Point taken--the recording practices for many commercial offerings are abhorrent to say the least, and could easily be improved,especially with respect to compression, hard panning and use of colored monitors, along with unnecessary conversions and bad EQ. But wrt DAC's, my point is either it's audible or its not, and if we assume that DAC's are perfect, EDIT (snipped reference to resampling) , then arguably should not inject artefact into the recording. This leaves only the last DAC and whatever sonic flavoring to alter what we here. I'm here to learn and just trying to understand the arguments, and have no dog in the fight at all. I do have some appreciation for what was said with respect to the analog filters corrupting a potentially deep null, but otherwise am clueless. BTW I'm a scientist or at least once was and am in no way defending the Moku Moku afficianados I hear over at DIY for instance.

What is it that you hear, and perhaps I can become a more discriminating listener?

 

[JRS]

The issue here is not that a bit can't change due to some electromagnetic interaction or any kind of hardware error, the point is that computers protect against errors when copying files. Therefore, if there is more than one copy of the same file, distributed among different places on earth, as all major platform do, and unless there is a major global catastrophe, it's very (very) unlikely that those files loose information with time. Of course, all supports suffer from degradation, but the strategy here is to keep coping the files indefinitely. It's like if you ask HBO if there is a possibility that they loose Game of Thrones. They will tell you that only if all earth is obliterated simultaneously. In the case of storage at home, the support degradation time is far more than a human life span in a general case, so no worries about it.

My point about entropy was meant tongue in cheek. On a happy note, I see that scientists have recently discovered an entropy free clock and I sleep better as a result. At least we can keep time while the universe propels towards a heat death.

 

[killdozzer]

Point taken--the recording practices for many commercial offerings are abhorrent to say the least, and could easily be improved,especially with respect to compression, hard panning and use of colored monitors, along with unnecessary conversions and bad EQ. But wrt DAC's, my point is either it's audible or its not, and if we assume that DAC's are perfect, absent a bunch of resampling interfaces, then arguably should not inject artefact into the recording. This leaves only the last DAC and whatever sonic flavoring to alter what we here. I'm here to learn and just trying to understand the arguments, and have no dog in the fight at all. I do have some appreciation for what was said with respect to the analog filters corrupting a potentially deep null, but otherwise am clueless. BTW I'm a scientist or at least once was and am in no way defending the Moku Moku afficianados I hear over at DIY for instance.

What is it that you hear, and perhaps I can become a more discriminating listener?

No, no. There has been some noise in our channel of communication. I failed to explain my point, sorry. I'll try again. I don't hear any differences and my bet is always on "if it can't be measured, it can't be heard". I simply think that KozmoNaut had a point when he said that not only the last one should matter. But I said this only for the purpose of challenging those robust beliefs that some audiophiles hold true, that they hear differences or deterioration in sound after a lot of converting or among well measuring DACs.

And it goes like this; if you claim you can hear a sound signature of a DAC, you should be able to recognize it as a piece equipment used by a studio only by listening to a recording. Much like you can hear a good Stratocaster. If it has a sound signature and you can hear it, you can also recognize it. It should be distinct.

So if audiophiles claim they can hear deterioration in sound quality after multiple analogue to digital and back to analogue conversions, it shouldn't matter if it's recorded on an album. They should be able to detect it while listening to a recording.

KozmoNaut said: "they moan about minute SNR differences all the while most of their music was converted from analogue to digital and back a hundred times while it was recorded in a studio" (I paraphrase). Meaning if they can detect deterioration in digital signal being copied many times, they should easily pick up on that on an original recording from the studio. The same deterioration should be present since the studio did dozens of conversions of the signal.

 

[JRS]

No, no. There has been some noise in our channel of communication. I failed to explain my point, sorry. I'll try again. I don't hear any differences and my bet is always on "if it can't be measured, it can't be heard". I simply think that KozmoNaut had a point when he said that not only the last one should matter. But I said this only for the purpose of challenging those robust beliefs that some audiophiles hold true, that they hear differences or deterioration in sound after a lot of converting or among well measuring DACs.

And it goes like this; if you claim you can hear a sound signature of a DAC, you should be able to recognize it as a piece equipment used by a studio only by listening to a recording. Much like you can hear a good Stratocaster. If it has a sound signature and you can hear it, you can also recognize it. It should be distinct.

So if audiophiles claim they can hear deterioration in sound quality after multiple analogue to digital and back to analogue conversions, it shouldn't matter if it's recorded on an album. They should be able to detect it while listening to a recording.

KozmoNaut said: "they moan about minute SNR differences all the while most of their music was converted from analogue to digital and back a hundred times while it was recorded in a studio" (I paraphrase). Meaning if they can detect deterioration in digital signal being copied many times, they should easily pick up on that on an original recording from the studio. The same deterioration should be present since the studio did dozens of conversions of the signal.

Thanks, now I get it. So we are on the same side.

 

[Jimbob54]

I'm not familiar with the algorithms, but the since the recovered file isn't exactly identical, my question is if the departure from the original increases if we run the process more than once.

I'm not stating that as fact, I'm paraphrasing pkanes post for killdozer.

 

[pkane]

Nice! Now it just remains to see whether the SINAD has an upper bound as the number of resamples tend to infinity or not. I'm sure you don't have anything better do right now.

SINAD isn't the issue, quantization noise eventually starts to dominate as the number of resamples approaches infinity

 

[xaviescacs]

SINAD isn't the issue, quantization noise eventually starts to dominate as the number of resamples approaches infinity

Quantization noise sounds like something introduced by an algorithm, by which of them? Or by both? There will be a point where this noise becomes audible?

Edit: how does the noise increases with each resample? It's proportional to the original noise? Or its just a constant that only depends on the wave characteristics (sample rates)? In the absurd extreme on infinite resamples, the signal will be pure noise?

 

[pkane]

Quantization noise sounds like something introduced by an algorithm, by which of them? Or by both? There will be a point where this noise becomes audible?

Quantization noise is generated by any DSP algorithm due to the finite number of digits used to represent sample values in a computer. With proper algorithm design, this noise can be kept very far below any measurable or audible levels.

Such noise can become potentially audible when it gets to a level above -120dB. As you can see in my example, even after 20 resamples DeltaWave resampler quantization noise is still far below any audibility or measurability, well below -170dB.

 

[xaviescacs]

Quantization noise is generated by any DSP algorithm due to the finite number of digits used to represent sample values in a computer. With proper algorithm design, this noise can be kept very far below any measurable or audible levels.

Such noise can become potentially audible when it gets to a level above -120dB. As you can see in my example, even after 20 resamples DeltaWave resampler quantization noise is still far below any audibility or measurability, well below -170dB.

I see, of course, I wasn't considering that the signal it's being converted from analog to digital every time and that necessarily involves some loss of information as you explain. Thanks. The takeout of your simulation is that there is no reason to be afraid of DSP at all.

 

[pkane]

I see. Thanks. The takeout is that there is no reason to be afraid of DSP at all.

Certainly not of properly implemented DSP

 

[audio2design]

But all they say is that "attractive sound" can't be fully explained by frequency response and distortion. I think amir has stated more than once that his personal tastes don't always correlate with measurements and that this deserves some investigation. But they are talking about sound, which is far more complex than a signal on a wire.

There is no inherent reason why "accurate" = "preferred". It does not for many things. People like over saturated colors too.

BUT, also very important, "preferred" != "accurate", something that subjective audio people often confuse.

 

[MaxBuck]

I don't agree that the analog output stage is not a part of the DAC. It's the job of a DAC to take a digital signal and convert it to a usable analog signal. And it's clear to me that, to the extent a DAC has a "sound signature," the output stage is largely responsible.