Testing audiophile claims and myths

[Brahmsian]

What super audible frequencies are in a baby’s cry? It all sits right in the most sensitive range of human hearing 1 to 4kHz, not above the threshold of human hearing. It affects us because it’s strident and loud, not because babies make sound that only bats can hear.

Compressed audio isn’t comparable to drinking spit. That is a completely absurd analogy.

I was referring to the following study, not as proof of anything, only as hinting at the possibility that there might be ways that sound affects us that the scientific study of audio enjoyment hasn’t yet seen. A simple reminder against hubris.

https://pubmed.ncbi.nlm.nih.gov/31786800/

And spit/dirt is not that absurd an analogy considering the need for dithering and the consequent white noise added to the signal and how in general recording technology has always added noise (dirt) to the signal.

 

[MrHaelscheir]

As I’ve said before, even MP3 sounds good to me. But I’m also skeptical of claims that the science is settled. I posted a link way back to a study that found that ultrasonic frequencies in a baby’s cry can affect the mother physiologically. We may still not know all the byways of how audio affects us. I’m less concerned with whether we can hear the difference than I am with simply getting the most faithful reproduction, whether all of it is audible or not. If someone spits in your water and stirs it with a spoon, you probably won’t taste the difference. But you still don’t want spit in your water. Just the idea of it. Give me the most objectively accurate reproduction of the original. It shouldn’t bother anyone that I have high resolution files in my library. And if my imagination adds a little something to the experience, that’s all to the good. I don’t suffer from upgraditis and enjoy 16/44.1 just fine. I’m more concerned with finding headphones whose sound I like.

Heh. I mean, if you've seen my headphone measurements and "raving" about the Meze Elite's EQed distortion performance (e.g. https://audiosciencereview.com/foru...out-headphone-measurements.18451/post-1956895 (post #1,210)), I too am afflicted by a desire for that at least measurably "purest" sound, even if the distortion levels are already well below audible. There is a "feels good" element to measurable technical excellence.

 

[bigshot]

Ingesting spit can pass disease. You aren’t going to be infected by your headphones.

Noise is a natural part of sound reproduction. All recorded sound has noise. You just don’t want it at audible levels. The thresholds of perception are what matter, not some desire for absolute purity. Absolute purity doesn’t exist. If you chase numbers and worry about things you can’t hear, there’s no end to that rabbit hole.

There is such a thing as “good enough”, and the dividing line between “not good enough” and “good enough” is defined by audibility. Audiophiles in general are well versed on numbers and theoretical sound, and much too ignorant about what they can and can’t actually hear.

Focusing on inaudible differences in gear says more about OCD than it does quality of sound.

 

[71 dB]

If someone spits in your water and stirs it with a spoon, you probably won’t taste the difference. But you still don’t want spit in your water. Just the idea of it.

Yes, the idea of having spit in our water is repulsive, because spit contains bacteria. Better analogy would be adding harmless substance (e.g. flour) to the water, but so little we can't taste it.

Give me the most objectively accurate reproduction of the original.

Perceptually accurate is enough for me. Going from that to objectively accurate doesn't provide anything perceptually, but increases the technical demands unnecessarily. I don't have unlimited resources to spend on audio/music. I have to make compromises and this is an area where making a "compromise" is perceptually harmless.

It shouldn’t bother anyone that I have high resolution files in my library.

I'm not bothered at all what kind of files you have in your library. That's your business. What I am bothered with is the messaging of people "needing" hi-rez. That's nonsense. If the music is to your taste, pretty low quality reproduction allows enjoyment. Hi-rez is for those who think about spit in their water. Please don't spit into the water glasses of those who have been happily drinking their water just to justify your own choices.

And if my imagination adds a little something to the experience, that’s all to the good. I don’t suffer from upgraditis and enjoy 16/44.1 just fine. I’m more concerned with finding headphones whose sound I like.

That's all good. It is good you acknowledge 16/44.1 is just fine.

 

[Brahmsian]

I simply detect a subtle tendency in this thread to shame people who pursue high resolution audio, and I’m not going to engage in that. Nor will I dispute people’s preference for DSD or what have you. That’s really the only issue I have with the thread. For the rest, I think it’s an important thread as far as head-fi goes, and I’m glad it exists. I know next to nothing about the technical aspect of audio, and this thread has been informative to me. Indeed, I have more questions than answers to provide here.

For instance, when it comes to the PCM versus PDM debate, I simply see them as two ingenious ways of arriving at the same place. Both of them fascinate me. Now my understanding is that both produce noise, but their way of handling it is different. In PCM, the white noise produced by dithering is made inaudible by pushing it down so far that you can’t hear it. With DSD/PDM, the noise is flung so far up that you can’t hear it either. My two-part question is whether with DSD the noise remains part of the signal or does the signal remain pure once the high frequency noise is filtered out. Correct me if I’m wrong, but in dithering PCM the white noise, while inaudible, can’t be separated from the music. I have even read that if you turn up the volume of 16/44.1 you will actually be able to hear the white noise and that it sounds a lot like tape hiss. In fact, I think it’s in that now famous Monty YouTube video. So the second, practical part of my question is, if I can play both DSD and PCM files just as easily on my gear, why I shouldn’t I prefer DSD over PCM for playback? Or, to put it differently, why should I prefer 16/44.1 over a 256 DSD file? I guess some here might say play either one because there’s no audible difference. But if I wanted to play the most objectively faithful reproduction irrespective of audibility which would it be?

 

[castleofargh]

I simply detect a subtle tendency in this thread to shame people who pursue high resolution audio, and I’m not going to engage in that. Nor will I dispute people’s preference for DSD or what have you. That’s really the only issue I have with the thread. For the rest, I think it’s an important thread as far as head-fi goes, and I’m glad it exists. I know next to nothing about the technical aspect of audio, and this thread has been informative to me. Indeed, I have more questions than answers to provide here.

For instance, when it comes to the PCM versus PDM debate, I simply see them as two ingenious ways of arriving at the same place. Both of them fascinate me. Now my understanding is that both produce noise, but their way of handling it is different. In PCM, the white noise produced by dithering is made inaudible by pushing it down so far that you can’t hear it. With DSD/PDM, the noise is flung so far up that you can’t hear it either. My two-part question is whether with DSD the noise remains part of the signal or does the signal remain pure once the high frequency noise is filtered out. Correct me if I’m wrong, but in dithering PCM the white noise, while inaudible, can’t be separated from the music. I have even read that if you turn up the volume of 16/44.1 you will actually be able to hear the white noise and that it sounds a lot like tape hiss. In fact, I think it’s in that now famous Monty YouTube video. So the second, practical part of my question is, if I can play both DSD and PCM files just as easily on my gear, why I shouldn’t I prefer DSD over PCM for playback? Or, to put it differently, why should I prefer 16/44.1 over a 256 DSD file? I guess some here might say play either one because there’s no audible difference. But if I wanted to play the most objectively faithful reproduction irrespective of audibility which would it be?

You start with DSD VS PCM but end with 256 DSD VS 16/44.
You can find some tentative estimate for PCM equivalence in DSD, but that's all it is and even then, it's only about the resolution output for what's on the file and there is the question of the usually remaining important noise in the ultrasound range that arguably could end up creating a lot more 'spit in your water' if it's allowed to run into the analog elements of the playback chain.
But considering that DSD is more often than not, mixed and mastered in PCM, Is there even a point in making them fight? Personally, I never cared for DSD. I never saw the point in a stereo system, and I would rather have a DVD or actually Blu-ray for multichannel if possible. The times when I was interested in a master they had kept exclusive to DSD, my pure hatred for those marketing methods kept me from getting it. I don't believe all that much in voting, but I sure believe in voting with my wallet, so I get stubborn like that sometimes. ^_^
My opinion, and that's all it is because I don't have the means to properly test, never will, and I don't know if anybody will care enough to bother trying, I don't think DSD is the solution for those who aim strictly for the best fidelity. At some point, as DACs clearly tend to have embraced a solution of a handful of bits and high sample rate, I wonder if the files shouldn't also move in that direction for optimal fidelity. For my listening experience, it's all the same to me in the sense that I'd rather put my money in the things that do have a significant impact instead of going crazy over the ultimate extra resolution no playback system or environment will ever achieve. Even as a purely objective purpose, there is such a thing as meaningless effort.

But if you're going to make it face 16/44, then DSD can achieve higher fidelity, and so can any higher PCM sample rate(if there actually is extra music data to put in those resolutions).

 

[Brahmsian]

You start with DSD VS PCM but end with 256 DSD VS 16/44.
You can find some tentative estimate for PCM equivalence in DSD, but that's all it is and even then, it's only about the resolution output for what's on the file and there is the question of the usually remaining important noise in the ultrasound range that arguably could end up creating a lot more 'spit in your water' if it's allowed to run into the analog elements of the playback chain.
But considering that DSD is more often than not, mixed and mastered in PCM, Is there even a point in making them fight? Personally, I never cared for DSD. I never saw the point in a stereo system, and I would rather have a DVD or actually Blu-ray for multichannel if possible. The times when I was interested in a master they had kept exclusive to DSD, my pure hatred for those marketing methods kept me from getting it. I don't believe all that much in voting, but I sure believe in voting with my wallet, so I get stubborn like that sometimes. ^_^
My opinion, and that's all it is because I don't have the means to properly test, never will, and I don't know if anybody will care enough to bother trying, I don't think DSD is the solution for those who aim strictly for the best fidelity. At some point, as DACs clearly tend to have embraced a solution of a handful of bits and high sample rate, I wonder if the files shouldn't also move in that direction. For optimal fidelity, for my listening experience, it's all the same to me in the sense that I'd rather put my money in the things that do have a significant impact instead of going crazy over the ultimate extra resolution no playback system or environment will ever achieve. Even as a purely objective purpose, there is such a thing as meaningless effort.

But if you're going to make it face 16/44, then DSD can achieve higher fidelity, and so can any higher PCM sample rate(if there actually is extra music data to put in those resolutions).

I have DXD files as well. I guess what I’m trying to get at is whether even in high resolution PCM you can filter out those impurities or whether they remain mixed up with the signal, even if you’ll never be able to hear them. I love PCM as much as PDM and am interested in the elegance of both solutions.

I have also read recommendations on the Audirvana forum that you should upsample all your music to the highest rate possible (or convert to the highest DSD equivalent that your equipment is capable of) because the external DAC is going to upsample anyway, and it’s better to do it on the relatively powerful computer than in the DAC. Anybody want to take a crack at that claim?

 

[Ryokan]

I'm going to guess an external dac can easily convert and it's a non issue.

 

[Ryokan]

Ingesting spit can pass disease. You aren’t going to be infected by your headphones.

Noise is a natural part of sound reproduction. All recorded sound has noise. You just don’t want it at audible levels. The thresholds of perception are what matter, not some desire for absolute purity. Absolute purity doesn’t exist. If you chase numbers and worry about things you can’t hear, there’s no end to that rabbit hole.

There is such a thing as “good enough”, and the dividing line between “not good enough” and “good enough” is defined by audibility. Audiophiles in general are well versed on numbers and theoretical sound, and much too ignorant about what they can and can’t actually hear.

Focusing on inaudible differences in gear says more about OCD than it does quality of sound.

Can numbers always have purity in computing, e.g SD card storage size differing from usable size - SD 128gb = 119gb.

 

[danadam]

In PCM, the white noise produced by dithering is made inaudible by pushing it down so far that you can’t hear it. With DSD/PDM, the noise is flung so far up that you can’t hear it either.

In DSD some noise is moved from 0-20k band to ultrasonic range. The amount to move is such that what remains is at acceptably low level. It is possible that this level is lower than 16-bit dither but probably not lower than 24-bit dither. In any case, there's still some noise left.

I have even read that if you turn up the volume of 16/44.1 you will actually be able to hear the white noise and that it sounds a lot like tape hiss.

The same will happen with DSD, maybe you'll have to turn up the volume more.

That's of course if you can find any music that goes quiet enough for that to matter

 

[sander99]

SD card storage size differing from usable size - SD 128gb = 119gb.

That is not a difference between size and usable size, but different definitions for kilo, mega, giga, tera, etc. are used by storage manufacturers versus by your computer.
Decimal based (powers of 1000) versus binary based (powers of 1024).
128*(1000^3) =~ 119*(1024^3).

 

[bigshot]

Can numbers always have purity in computing, e.g SD card storage size differing from usable size - SD 128gb = 119gb.

For numbers to have context with audible sound, they need to be related to perceptual thresholds. A lot of audiophiles think better numbers always mean better sound, and that just isn’t the case. Human ears hear up to a point and they stop. Super audible frequencies are inaudible. Compressed audio at high data rates filters out inaudible sound. Neither of these is better than the other. For the purposes of listening to recorded music in your living room, they are identical.

An old timer once told me that when the Burwen broadband noise reduction unit was released, it had a huge range- all the way from no real difference all the way to completely muffled. Sound engineers were shocked that a device would be sold that could be set high enough to degrade the sound. But Burwen said he didn’t know all the conditions it might be used, so he wanted to allow the maximum latitude. Later, he admitted that his device got a bad name because of the maximum intensity of the filter. In retrospect, if he had to do it over, he would have limited it significantly. Not because it made the device any better… just to assuage the fears of engineers.

I think the same is true of compressed audio. When it was first released to the public files were usually Fraunhofer MP3 128 CBR. Today’s codecs and data rates sound perfect, but people remember those old crappy MP3s. If it had been introduced with AAC 192 VBR, audiophiles would have much fewer objections to it, and 320 VBR would be considered an audiophile format, because it’s overkill.

 

[mbphotox]

Is the DAC in a computer really "better" for upsampling?!
I feel like there's sooo much stuff going on in a mainboard with a huge power supply and all those other things.. I would really rather trust the little dedicated chip in a small form factor desktop DAC/AMP combo.

Looking at the abomination that is USB, I cannot imagine the computer doing any better at all.

I send my audio via TOSLINK to my SMSL DL200, no upsampling applied.
Although I'm still torn between using Tidal in the "exclusive" mode or applying the EQ via Peace interface. (which doesn't work when exclusive mode is enabled.. the computer cannot affect the signal in any way, not with volume, nothing..)

 

[71 dB]

For instance, when it comes to the PCM versus PDM debate, I simply see them as two ingenious ways of arriving at the same place.

Kind of true.

Now my understanding is that both produce noise,

Reality is noisy. The second someone records guitar playing in studio for a new music production noise is introduced. The studio has a noise floor, hopefully very low but the noise is there. A tiny amount of noise gets induced into the microphone cable The mic preamp generates some noise etc. The guitar playing gets recorded probably at 24 bit. The first 3 bits are probably dynamic safety margin (not used and remain zero). The last 8 bits are probably noise. Despite of this, the recording sounds hopefully extremely clean and ready for mixing. Totally noise free music is impossible, but fortunately we can push the noise low enough in the 21st century. Even before that.

but their way of handling it is different. In PCM, the white noise produced by dithering is made inaudible by pushing it down so far that you can’t hear it.

The level of dither is the lowest amount that is able to remove all noise modulation. More than that just increases noise level and less wouldn't achieve what dither is supposed to do, remove the correlation between signal and noise.

With DSD/PDM, the noise is flung so far up that you can’t hear it either.

In fact 1 bit DSD can't be properly dithered, because the correct dither level is more than 1 bit, but there is not enough bits for that. Fortunately the distortion energy caused by under-dithering is mostly in the ultrasound region and inaudible.

My two-part question is whether with DSD the noise remains part of the signal or does the signal remain pure once the high frequency noise is filtered out.

Some of the dither energy is in the audible range and remains "part of the signal" after filtering. However, the level of this noise is very low. Practically it is beyond inaudible.

Correct me if I’m wrong, but in dithering PCM the white noise, while inaudible, can’t be separated from the music.

Yes, but white noise is not the only option. All kind of shaped dither can be used to make the dither perceptually even more inaudible pushing the noise energy more into the frequencies human ears are insensitive (above 15 kHz).

I have even read that if you turn up the volume of 16/44.1 you will actually be able to hear the white noise and that it sounds a lot like tape hiss.

Yes, but why would you do so? This only works if there is hardly any music (silence between tracks). If there is music (next track starts), the sound pressure level is very high. Loudspeakers pop, hearing damage etc.

Or, to put it differently, why should I prefer 16/44.1 over a 256 DSD file?

Three reasons in my opinion:

- Smaller files.
- Much more music available in 16/44.1. If you want your music to be hi-rez, you are likely to ignore all the great music in the world that is/can't be available in hi-rez.
- From the perspective of perceived sound quality it doesn't matter whether you listen to CD quality or hi-rez (assuming we are talking about the same master or course). Prefering hi-rez because of "better specs" is about not having an insight about what matters in audio and what doesn't. It can even be considered intellectual laziness of a sort. Real sound quality that matters comes from good production, mixing and mastering plus the headphones/loudspeakers/room acoustics+correction/listening position/speaker placement. That's the stuff that matters. The production/mixing/mastering side we can't do much about as consumers except vote with our wallets, but the other stuff we can do a lot about.

I guess some here might say play either one because there’s no audible difference. But if I wanted to play the most objectively faithful reproduction irrespective of audibility which would it be?

It depends. In the case of old recordings, the most faithful reproduction would be to play the original studio tape with the same tape machine they used back in the day and of course in the same studio! Yes, ridiculous, but this illustrates how sometimes the most faithful reproduction is a tall order and perhaps even unnecessary from the point of music enjoyment. How faithful does it need to be? CD quality is about having audible (actually more than that) part of faithfulness in a form that is easy for consumers.

My advice to you is to concentrate more on the music and think about less if the dither level at 15 kHz is 45 or 55 dB below audibility. If there is something wrong with the sound in a way that matter, you are likely to hear it and you can figure out what to do about it. No harm listening to DSD 256 files if you have those. Just do not believe you NEED them to enjoy music.

 

[Ryokan]

That is not a difference between size and usable size, but different definitions for kilo, mega, giga, tera, etc. are used by storage manufacturers versus by your computer.
Decimal based (powers of 1000) versus binary based (powers of 1024).
128*(1000^3) =~ 119*(1024^3).

Yeah, thanks, I realised soon after it bore no relation to 'perfect numbers' for audio.