[jagwap]

I am talking about the analogue bits of reproduction.

Then you can get 10s of uS difference between channels with modest tolerance out of band. Do the maths and you may be surprised.

Digital will not have this.

 

[pinnahertz]

Sony now makes headphones that go up to 100 kHz.

MDR-Z1. "Specifications" (yes, in quotes) says 4Hz - 120kHz. While I couldn't find actual measurements that go higher than 25kHz.

This review sums it up:

"Sound Quality
At first blush the MDR-Z1R sounds like a warm warm headphone with some treble emphasis—a bit smilie face, but not bad. May brain never did get around to accommodating though.

The bass is a bit too strong and remains emphasized too far into the midrange giving it a thick character. A moderately withdrawn presence region adds some veil to the thick bass. A lack of energy 4-6kHz leaves cymbals lacking body, and a big peak at 10kHz add too much zing to everything and, while not piercing as would a 5kHz be, becomes significantly fatiguing over time.

....I was regularly struck with how the others provided a much better sense of an integrated whole; the Sony in comparison was disjointed and incoherent. Not once did it make me tap my feet.

I really don't know what more to say. When I'd rather listen to the Audio Technica ATH-M50x we've got problems. I don't want to misslead here, it's an okay sounding headphone, but 'okay' just doesn't cut it at $2299."

So you can have "response" (actual level or deviation unstated for anywhere in the total FR... wow, thanks Sony) up to 120kHz, and still have rather unimpressive sounding headphones. Sounds like they threw the baby out with the bathwater.

 

[pinnahertz]

I am talking about the analogue bits of reproduction.

Please define "analog bits".

Then you can get 10s of uS difference between channels with modest tolerance out of band. Do the maths and you may be surprised.

Maximum possible ITD is about 600uS, that's a source facing one ear, and on the opposite side of the head from the other. Considering 0 degrees azimuth (directly in front of you) has 0uS ITD, and 90 degrees has 600uS...now show me where 10uS at 20kHz (and not mid-band...this is non-flat group delay, right?) where humans use exclusively intensity cues to localize, makes any difference.

Digital will not have this.

I'm sorry I didn't track the jump away from "digital" to "analog bits" (whatever that is).

 

[bigshot]

Tweeters that only go to 20kHz can have very severe ringing at a resonence a few kHz above that.

Super audible frequencies can certainly be problematic especially for young ears, and headroom is always a good thing. But you shouldn't get carried away with doubling down on more headroom expecting it to make more of an improvement. If you push ultrasonic ringing up a full octave beyond the range of hearing, that's probably more than enough to make it not problematic.

 

[jagwap]

Super audible frequencies can certainly be problematic especially for young ears, and headroom is always a good thing. But you shouldn't get carried away with doubling down on more headroom expecting it to make more of an improvement. If you push ultrasonic ringing up a full octave beyond the range of hearing, that's probably more than enough to make it not problematic.

I am not saying the ringing frequency is audible (Although at an audio demonstration a very young girl once heard 26kHz, but again that would be seen as irrelevant here). The high Q of the filter itself is problematic in the audio-band. If it were a perfectly linear system maybe it wouldn't be, but transducers are not very linear. The in-band phase distortion, and intermodulation is not benign.

 

[bigshot]

26kHz is only two notes higher than 20kHz on the musical scale. It's not very much. The number sounds bigger than it is. I'm sure she wasn't hearing 26kHz in music because anything that high would be completely masked and besides, I can't think of more than a handful of recordings that might actually contain signal in that range. Also, she might not have been actually hearing it- it's possible to feel sound pressure from frequencies you can't actually hear. What you say is true though. If you can push the problem safely above 20kHz at a volume level that isn't excessively loud, it's nothing to worry about any more. Problems in the audible range are much more important.

 

[jagwap]

Generally I agree that there is not absolute evidence that we can hear or perceive audio above 20kHz, but I am illustrating why there are reasons for widening the bandwidth beyond 20kHz in order to improve the in-band performance. We may not need "DC to light" frequency response, but -3dB at 20Hz and 20kHz is nowhere near good enough. Most of these reasons are in the analogue domain, as this is where things are less defined.

 

[pinnahertz]

I am not saying the ringing frequency is audible (Although at an audio demonstration a very young girl once heard 26kHz, but again that would be seen as irrelevant here). The high Q of the filter itself is problematic in the audio-band. If it were a perfectly linear system maybe it wouldn't be, but transducers are not very linear. The in-band phase distortion, and intermodulation is not benign.

Ultrasonic IMD is, I believe, one area in which more attention should be given and research should be done. There are several means of evaluating the resulting intermodulation products created by ultrasonic frequency components that fall done into the very audible range. While I cannot prove this, I think this is what people are hearing (or not hearing) when they listen to high bandwidth systems and recordings.

Generally I agree that there is not absolute evidence that we can hear or perceive audio above 20kHz, but I am illustrating why there are reasons for widening the bandwidth beyond 20kHz in order to improve the in-band performance. We may not need "DC to light" frequency response, but -3dB at 20Hz and 20kHz is nowhere near good enough.

-3dB @ 20kHz is a very analog specification, you won't find that in any current high quality digital audio system, nor any back many decades. However, analog recorders and older analog equipment could easily be -3dB @20khz. Also, -3dB @ 50Hz.

Most of these reasons are in the analogue domain, as this is where things are less defined.

I'm curious as to what you think these reasons are, and what their audible threshold might be. (I know this might sound like a bait, it's not meant to be, I'm actually curious).

 

[jagwap]

Ultrasonic IMD is, I believe, one area in which more attention should be given and research should be done. There are several means of evaluating the resulting intermodulation products created by ultrasonic frequency components that fall done into the very audible range. While I cannot prove this, I think this is what people are hearing (or not hearing) when they listen to high bandwidth systems and recordings.

I agree. It may not be the only issue, but it seems a likely one, and people have found this to be the issue before.

-3dB @ 20kHz is a very analog specification, you won't find that in any current high quality digital audio system, nor any back many decades. However, analog recorders and older analog equipment could easily be -3dB @20khz. Also, -3dB @ 50Hz.

Not ideal. All this talk of group delay, at low frequencies it is very real and cumulative. Ironically: mathematically you can show that -3dB at 20Hz is worse for group delay than 50Hz.

I'm curious as to what you think these reasons are, and what their audible threshold might be. (I know this might sound like a bait, it's not meant to be, I'm actually curious).

Sure. I am still curious what you think an absolute peak SPL of a drum kit would be... But I have answered this above: IMD and ringing are 2 issues. A high Q system doesn't need exciting at its resonant frequency to start ringing. Near is good enough. So 26kHz 12dB peaks on transducers are not a good thing.

 

[jagwap]

So 26kHz 12dB peaks on transducers are not a good thing.

Oh, and you should see what happens when an open loop class D amplifier meets a tweeter! Huge potential high Q peaks.

 

[gregorio]

While I don't disagree that the noise floor of 16bits is practically inaudible, it can be heard on silent passages if turned up a little and the room is reasonably quiet. I know I can and so could most of the subjects in the Meyer and Moran study.

Sure you can hear noise, it's the noise floor of the recording, not the digital noise floor though! The Meyer and Moran study just truncated to 16bit, not even triangular dither was applied, let alone noise-shaped dither!

(1) A loud SPL is 96dBA, yes, but that is an average not the peaks of the waveform. A good rock recording has a DNR of say 16dB, so the peaks are 112dB if you don't clip (this is from the loudness database. I am not convinced of the workings of the DNR meter they use from FOOBAR). That leaves the noise floor of CD at 16dBA (OK 13dBA as there are 2 channels to make the 96dB). Fine, thats more than 14dB below the 30dBA people are talking about here.
(2) Remember the 30dBA talked about isn't white or pink noise in the background. It is general domestic interference.
(2a) It is entirely possible that the 16dBA background is audible, especially if near field monitoring.
(2b) However more dynamic music exists, and people listen at higher levels for the short peaks.
(3) Also the noise floor of the CD player is going to have other noise added by the system.

(4) I didn't ask for MORE than 150dB! I'm not crazy!

(5) By the same token, low frequencies need more power to be perceived the same, so as most music has all the energy at low frequencies (as MQA papers show all the way into ultrasonic - see: back on topic), more dynamic range is needed to reproduce it. The bottom end at say 40Hz needs 34.5dB extra, so that loses you 34dB - 8.5dB dynamic range for unclipped middle C (ish).
(5a) Presence band is where we are so sensitive so we will hear the hiss. Unfortunately PCM does not compensate for this, apart from a token amount at the top end if someone bothers to noise shape the dither.

1. Complete nonsense! You are comparing the DR database (which is similar to a crest factor measurement), with dBSPL peaks and dBA of the noise floor of CD. All of this is nonsense, you are comparing apples with oranges and from that comparison, coming up with the conclusion that aircarft shouldn't be made of concrete??!

2. Huh? "General domestic interference" in a quiet environment is random noise, not mathematically identical to pink or white noise but audibly extremely similar.
2a. Nonsense! In a very quiet room, how are you going to hear "A" weighted random noise that's 14dB below "A" weighted random noise? But your 14dBA figure is nonsense anyway: Read the responses given to you and if there's something you don't understand ask, DON'T just keep making up more nonsense!

3. And that's going to be the same amount of noise whether it's 16bit or 300bits!

4. Either you are asking for more than 150dB, in which case you're crazy, or you're making up nonsense facts and conclusions to argue for more than 150dB without even realising it, in which case you're crazy!!!

5. Again, more nonsense! How many recordings can you name with more than 96dB of dynamic range? There's only a handful or so with more than 60dB!

5a. Congrats, a complete full house of nonsense! 1, ~20dB is not a "token amount"! 2, Noise-shaping has been industry standard practice for nearly 20 years.

(1) Back to MQA from a computing viewpoint, I can't bend my head around the concept of a lossy codec that would beat a non lossy one...
(2) as I understand the creators of the codec want to control the whole chain from the microphone to the speaker does that mean the ADC ...

1. There is no getting one's head around that!
2. That's what they want you to understand but it's a marketing lie. It's a lie because it misses out by far the most significant part of the chain, the mixing/processing! The ADC and DAC are completely insignificant compared to the mixing.

(1) A handful is still some, and the argument is was that 16b is enough. MQA when decoded offers more ...
(2) It is not hard to offer 20 bit or more performance ...

1. No it doesn't, where did you get that from? According to the patent application, MQA offers 13-17bit.
2. Correct it's not hard, it's impossible!

(1) He measured instruments and found that there's life to the sound they make above 20kHz.
(2) He also leaves open the possibility that bone conduction can affect our enjoyment of music. He points to a study on that as well as to one showing electrical activity in the brain produced by frequencies above 26kHz.

1. Did you actually look at what you posted? Fractions of a percent at freqs which only young people can hear and even then, only if you blast them with pure tones! So, if you're young, have a collection of recordings of nothing but pure >20kHz sine waves and play them back at >100dB, then yes, >44.1kHz sample rates would be worth it for you!
2. He does leave it open, as a marketing ploy, because there is ZERO evidence to support it! Brain activity was measured BUT none of the study participants were actually aware of it. So, if you mechanically bolt your speakers to your skull, you're "enjoyment of music" will be unaffected UNLESS you enjoy watching a real time EEG while you're listening to your music!

(1) Throughout this thread when talking about timing error in the uS size, the ear's sensitivity is highest between arrival between the ears. This is below 20kHz. Also note the first arrival of the sound it the one that give those queues.
(2) The reflections of the room tend to be masked out when it comes to spacial awareness.

1. Only the marketing BS and those repeating it talk about uS size! Positional cues are in the ranges of milli-seconds, not micro-seconds, as has been explained in this thread on several occasions and to you personally! It's trivial to run a test and see for yourself, so instead of keep repeating the marketing lies, why don't you find out for yourself if you don't want to believe what we're telling you?
2. That's completely backwards! It's the reflections of the room/environment which gives spatial awareness.

G

 

[jagwap]

1. Complete nonsense! You are comparing the DR database (which is similar to a crest factor measurement), with dBSPL peaks and dBA of the noise floor of CD. All of this is nonsense, you are comparing apples with oranges and from that comparison, coming up with the conclusion that aircarft shouldn't be made of concrete??!

Then please, put in your numbers: Maximum average SPL to peak ratio ever encountered please, so we can have a discussion rather than mud slinging.

2. Huh? "General domestic interference" in a quiet environment is random noise, not mathematically identical to pink or white noise but audibly extremely similar.
2a. Nonsense! In a very quiet room, how are you going to hear "A" weighted random noise that's 14dB below "A" weighted random noise? But your 14dBA figure is nonsense anyway: Read the responses given to you and if there's something you don't understand ask, DON'T just keep making up more nonsense!

Not necessarily. Hum from electrical and air conditioning is not broadband noise. Many other sources are tones and their harmonics.

3. And that's going to be the same amount of noise whether it's 16bit or 300bits!

Yes. But if it is anything more than say around -12dB from the next largest it is going to add to it. http://www.sengpielaudio.com/calculator-leveladding.htm

4. Either you are asking for more than 150dB, in which case you're crazy, or you're making up nonsense facts and conclusions to argue for more than 150dB without even realising it, in which case you're crazy!!!

I am NOT asking for more than 150dB. I was citing another professional and how much headroom he wanted for the audio chain to be blameless.

5. Again, more nonsense! How many recordings can you name with more than 96dB of dynamic range? There's only a handful or so with more than 60dB!

I didn't say that. I think you like the word nonsense when you can apply it to others and not yourself. I was showing that middle C generally needs EDIT: less dynamic range than lower notes in typical program material, so perhaps middle C is not the ideal example.

5a. Congrats, a complete full house of nonsense! 1, ~20dB is not a "token amount"! 2, Noise-shaping has been industry standard practice for nearly 20 years.



1. There is no getting one's head around that!
2. That's what they want you to understand but it's a marketing lie. It's a lie because it misses out by far the most significant part of the chain, the mixing/processing! The ADC and DAC are completely insignificant compared to the mixing.



1. No it doesn't, where did you get that from? According to the patent application, MQA offers 13-17bit.

It depend on the source material but 24bit source has 23bit decoded by MQA. Enough? Unless you mean undecoded then yes it can be less than 16bit.
https://www.stereophile.com/content/mqa-questions-and-answers-bit-depth-mqa

2. Correct it's not hard, it's impossible!

I meant in the digital domain.

1. Did you actually look at what you posted? Fractions of a percent at freqs which only young people can hear and even then, only if you blast them with pure tones! So, if you're young, have a collection of recordings of nothing but pure >20kHz sine waves and play them back at >100dB, then yes, >44.1kHz sample rates would be worth it for you!
2. He does leave it open, as a marketing ploy, because there is ZERO evidence to support it! Brain activity was measured BUT none of the study participants were actually aware of it. So, if you mechanically bolt your speakers to your skull, you're "enjoyment of music" will be unaffected UNLESS you enjoy watching a real time EEG while you're listening to your music!



1. Only the marketing BS and those repeating it talk about uS size! Positional cues are in the ranges of milli-seconds, not micro-seconds, as has been explained in this thread on several occasions and to you personally! It's trivial to run a test and see for yourself, so instead of keep repeating the marketing lies, why don't you find out for yourself if you don't want to believe what we're telling you?

No, there are papers showing studies that 15uS are detectable between the ears in arrival time. It has to less than 1mS as our ears are too close together for that to be the difference for spacial cues as sound travels 340 mm per mS. Dare I say it.... NONSENSE!!! Oh you're right... feels quite good!

That is not the same as the temporal stuff MQA is talking about, and you guys torn a new one on those other papers.

2. That's completely backwards! It's the reflections of the room/environment which gives spatial awareness.

G

See above: NONSENSE...OK, the room contributes, but for arrival time the early arrival wavefront is the spatial cue. It has to be or we wouldn't figure out where anything came from. Edit: Spacial cues come from relative arrival time, relative level and head transfer function effects, no?

Nonsense?

 

[jagwap]

Let me clarify my position as the mob is getting tetchy:

I have at no point said I believe all the special sauce in MQA. But I know a little about the people behind it and they are experts in their field, and care passionately about audio reproduction, so I do not think it should be written off or the people involved insulted before it is proved to be nonsense. By all means say it isn't proven true, but that does not mean they are liars or cheats, just not forthcoming with the information.

I think that 96dB MAY not be enough for all cases, and more is a good idea if it is available, and it is. It is an interesting discussion.

I am not convinced that more than 20kHz is useful, but I remain open minded. However I strongly believe that many parts of the reproduction chain should be considerably wider bandwidth than 20Hz to 20kHz at both ends when it is at all practicable, as it gets in-band audible errors reduced when you push the limits away.

 

[pinnahertz]

I am still curious what you think an absolute peak SPL of a drum kit would be...

I don't know, but I would guess between 90dB SPL and 140dB SPL. Depends on what drums, who's playing, and most importantly, where the mic is. When recording drums we use mics that can handle high SPL and pad the preamp.

 

[pinnahertz]

I think that 96dB MAY not be enough for all cases, and more is a good idea if it is available, and it is. It is an interesting discussion.

Perhaps not every single case, but the cases where 96dB isn't enough are ridiculously rare. 24 bit data is available but the 4-6 LSBs are all just noise, we don't have any real 24 bit DR recordings at all. Most so-called hi-res is resampled standard res or worse, analog originals. So what would those with the "need" for more than 96dB actually be playing?

I am not convinced that more than 20kHz is useful, but I remain open minded. However I strongly believe that many parts of the reproduction chain should be considerably wider bandwidth than 20Hz to 20kHz at both ends when it is at all practicable, as it gets in-band audible errors reduced when you push the limits away.

I would agree that any distortion mechanism that results in in-band products should be carefully examined. However, specifying reproduction systems as having FR "up to XkHz" is highly misleading, but will be a by-product of increased bandwidth systems.