High Resolution Audio: Does It Matter?

[j_j]

Stereo can't hack it. Stereo can not reproduce the soundfield in a concert hall. Yes, I know, "we only have two ears" but this was established definitively in 1933 by Steinburg and Snow, you need at least 3 front channels just to capture the front part of the soundfield. The center speaker is key, must be identical to L and R and full range. There is no question. What's more, a proper 3-channel capture will have a nice, wide stereo spread, even though there is more energy in the center than in either L or R. What's more, you need two more side channels positioned in front of the Pinna shadow, and two more back channels BEHIND the pinna shadow. That's the minimum for establishing the sensation of a concert hall.

Yes, you can do something decent with binaural for ONE LISTENING POSITION ONLY. That's not how listeners listen. Heads move in concerts. All the time.

As to mics, DAC's, ADC's, sorry, not the problem here. Good electronics? Not a problem, either.

When you record something to two channels, you have at most 2/8 ths of the soundfield at two points, regardless of how you've actually recorded, mixed, produced it.

I think I touch briefly on this in this talk: http://www.aes-media.org/sections/pnw/pnwrecaps/2013/apr_jj/ and more so in this talk, but there's no recording I know of:

http://www.aes-media.org/sections/pnw/ppt/jj/jj_aes04_ts1.pdf I've given that talk literally all over the world.

 

[M00ndancer]

Stereo can't hack it. Stereo can not reproduce the soundfield in a concert hall.

True, live music is live music. Even the old Eagles, Hell freezes over DVD with DTS soundtrack is superior to the CD version. Not in absolute quality but in immersion and placement. Although you can hear that they did master it afterwards. The best song is the non-video version of Seven Bridges Road on the same DVD.

 

[amirm]

Even the old Eagles, Hell freezes over DVD with DTS soundtrack is superior to the CD version.

Indeed. Anyone who thinks 2-channel can do justice to live music should play such DVDs and switch between PCM stereo and DTS surround. The difference is quite dramatic.

 

[Hypnotoad]

If I want to listen to High Resolution music I go to a concert.

I did that and they had those big arrays of speakers on either side of the stage. Do they even have un-amplified concerts anymore?

 

[SIY]

I did that and they had those big arrays of speakers on either side of the stage. Do they even have un-amplified concerts anymore?

Yes. Not so much with rock, a bit more (a bit!) with jazz, but you can hear a lot of folk, bluegrass, and classical live and unamplified.

 

[Sal1950]

Indeed. Anyone who thinks 2-channel can do justice to live music should play such DVDs and switch between PCM stereo and DTS surround. The difference is quite dramatic.

Apologies for cross posting,
https://www.audiosciencereview.com/...hat-is-wrong-with-this-hobby.6948/post-156098
"A few of the newer BluRay' releases of older discreet multich recordings are including the original 4.0 master from the quad days plus the modern remaster in 5.1. I have BD's of Pink Floyd's WYWH & DSOTHM that include 2.0 stereo, 4.0 quad and 5.1 masters on the same disc. You can compare the different mastering with just a push of the Audio button on your disc's remote..
Then with an additional button push on my Pre/Pro's remote, I can upmix the multch source to include immersive playback with the 4 overhead speakers using either Dolby Surround, DTS Master X Neural, or Auro 3D DSP processing.
Chooses are good!"

 

[yummy]

i no give no shit about dsd since day zero, all dsd edited in pcm

 

[MRC01]

...
We've talked around it for pages and pages, but when it comes to playback (as distinct from processing), nobody has been able to offer any reasonable theory or experimental evidence as to how a low-pass filter which is completely outside the audio band could possibly generate audible artefacts, nor how it could audibly exacerbate artefacts already present in an analogue recording from which it was derived.

I can think of a "reasonable theory" for the belief that low-pass filters might generate audible (or at least measurable) artifacts. Whittaker-Shannon gives the mathematical formula for the theoretically ideal D-A reconstruction filter. Essentially, it plants the function sinc(x) at every sampling point x, scaled to that sample's amplitude. Then you sum them all up to create the analog wave. Each sinc(x) ripples out to infinity both forward & backward in time with diminishing amplitude. This ripple is the "pre-echo" that Amir mentioned in post #1 of this thread, sometimes called the Gibbs phenomenon or "ringing". It cannot be eliminated, even in a theoretically ideal DA reconstruction. Ripple is inherent to bandwidth limited transients.

These ripples are based on samples, not time. So at higher sampling rates, the ripples decay faster in real time, which could be more transparent even if you don't care about capturing high frequencies.

Now whether anyone can actually hear this ripple is a different question! But my point is just that there is a reasonable theory why low-pass filters might be more transparent at higher sampling frequencies. Especially when you consider that real-world DACs don't use sinc(x), they use methods like delta-sigma that are easier to engineer. So their results, while very good, must fall somewhat short of the theoretical ideal.

PS The pragmatic engineer asks, "how far away do you have to be for the ripple to be negligible?" I made a Whittaker-Shannon spreadsheet and found that the ripple of each sample drops by 96 dB at a distance of about 44,000 samples away. So for CD, each sample "ripples" or affects the signal (construction of the analog wave) above the 16-bit noise floor for 2 full seconds (1 second before & after). At 176 kHz, that ripple still decays to -96 dB in 40k samples, which is now only 1/4 second before & after instead of 1 full second. Whatever level you think the ripple might be audible, say -30 dB, or whatever, it decays to that level 4 times faster, shortening both pre-echo and post-echo.

 

[March Audio]

Question. What frequency are these ripples at?

 

[Blumlein 88]

J_J has said having a 65 khz sample rate with response to 25 khz and the wider transition band to create the filter would have been his preference given what he knows of human hearing for a fully blameless system in audible terms. A few ADC and DAC makers (very few) have implemented this idea by starting a filter at 30 or 35 khz with a slow wide transition zone for their higher sample rates. All but a small number of microphones also pretty much die in the mid 30 khz range. So it seems like a good idea. So for instance at 192 khz instead of flat to 80 khz (or at 96 khz rates instead of 40 khz) they'll begin the filter at 30 khz. The main company who does this is Lavry.

So what do you think of this? Would it be a good idea to start the filter roll off at 30 khz while feeding a DAC at 768 khz sample rate?

 

[MRC01]

Question. What frequency are these ripples at?

Sinc(x) ripples at Nyquist, so it should be inaudible even at CD sampling rates. But that is the theoretical ideal. Real-world DAC implementations ring louder, longer, and at lower frequencies, depending on the filter implementation.

It should be a dual effect. Higher sampling frequencies shorten the ripple in real time, and they also permit the use of more gradual filters that induce less ripple to begin with.

 

[March Audio]

J_J has said having a 65 khz sample rate with response to 25 khz and the wider transition band to create the filter would have been his preference given what he knows of human hearing for a fully blameless system in audible terms. A few ADC and DAC makers (very few) have implemented this idea by starting a filter at 30 or 35 khz with a slow wide transition zone for their higher sample rates. All but a small number of microphones also pretty much die in the mid 30 khz range. So it seems like a good idea. So for instance at 192 khz instead of flat to 80 khz (or at 96 khz rates instead of 40 khz) they'll begin the filter at 30 khz. The main company who does this is Lavry.

So what do you think of this? Would it be a good idea to start the filter roll off at 30 khz while feeding a DAC at 768 khz sample rate?

I certainly agree with JJs proposition, as you say this will provide a "blameless" path with that bit of extra "padding".

So filtering at a lower frequency even in high bandwidth recordings holds no fear for me . There is nothing to be gained from recording signals which, assuming they are actually musically derived (not noise), are inaudibly above the hearing range and cannot be reproduced by most speakers in any case. Recording the noise that may exist in the ultrasonic band equally has no benefit - possibly even detriment..

 

[March Audio]

Sinc(x) ripples at Nyquist, so it should be inaudible even at CD sampling rates. But that is the theoretical ideal. Real-world DAC implementations ring louder, longer, and at lower frequencies, depending on the filter implementation.

It should be a dual effect. Higher sampling frequencies shorten the ripple in real time, and they also permit the use of more gradual filters that induce ripple to begin with.

So, those ripples are at a frequency we cant hear - above 22kHz. Do you (anybody) have plots which show the difference between the "theoretical" ringing amplitude level and what DACs actually do and what ends up in band?

You can probably see where I am leading with this. A lot of fuss has been made in recent years regarding filter types, pre/post ringing issues, whilst missing some pretty fundamental points regarding the nature of the ringing.

Im sure Im not the only one who has endlessly gone round in circles with a dac trying to decide which filter type sounds best, or even if they sound different at all. The ones where I have been fairly confident about a difference have been those that actually impact the audible band (ie roll off in it)

 

[MRC01]

...
So what do you think of this? Would it be a good idea to start the filter roll off at 30 khz while feeding a DAC at 768 khz sample rate?

Yes, I think so.
For perspective, I think CD is transparent for all practical purposes. But just barely transparent; people with exceptional hearing and training can differentiate it from higher rates. I speculate that they're not hearing super-high frequencies, instead they're hearing subtle passband distortions caused by steep DA filters (or AD filters used during recording). So a higher sampling frequency with a more gradual filter should make it completely transparent. I don't think you'd need 768 kHz either. 88 or 96 gives you twice as wide a transition band as CD; if your goal is a more gradual filter, that should be plenty.

 

[MRC01]

... Do you (anybody) have plots which show the difference between the "theoretical" ringing amplitude level and what DACs actually do and what ends up in band?
...
Im sure Im not the only one who has endlessly gone round in circles with a dac trying to decide which filter type sounds best, or even if they sound different at all. The ones where I have been fairly confident about a difference have been those that actually impact the audible band (ie roll off)

Archimago measured some here: http://archimago.blogspot.com/2013/06/measurements-digital-filters-and.html
Some of them - like the minimum phase - ring much louder and longer than sinc(x)!

Keith Howard wrote about this for Stereophile a few years ago: https://www.stereophile.com/reference/106ringing
I love their experimental attitude. But when they talk about how hard it was to tell the filters apart, it is kinda funny thinking about a bunch of middle-age guys wondering why they can’t hear a supersonic ripple, especially when masked by music in which it occurs. Especially when most of them understand math & engineering well enough to know why.

 

[March Audio]

Archimago measured some here: http://archimago.blogspot.com/2013/06/measurements-digital-filters-and.html
Some of them - like the minimum phase - ring much louder and longer than sinc(x)!

Keith Howard wrote about this for Stereophile a few years ago: https://www.stereophile.com/reference/106ringing
I love their experimental attitude. But when they talk about how hard it was to tell the filters apart, it is kinda funny thinking about a bunch of middle-age guys wondering why they can’t hear a supersonic ripple, especially when masked by music in which it occurs. Especially when most of them understand math & engineering well enough to know why.

Just pulling the scope out of the cupboard and getting adobe in action.

 

[March Audio]

OK, dirac pulse generated in adobe played back on my DAC1and measured on my Keysight scope.

Spot the difference

 

[March Audio]

Motu 8A

 

[RayDunzl]

Let's see it in-room...

 

[March Audio]

Let's see it in-room...

Im going to do that just cos I can.