High Resolution Audio?

[daftcombo]

Techno and ambient

Great! I'm even more into electro, synthwave & italo disco, genres for which hi-res is absolutely pointless!

 

[flipflop]

@andreasmaaan
Did you make this?

 

[andreasmaaan]

@andreasmaaan
Did you make this?

I did actually A first attempt at producing music.

 

[maty]

Today, Highly Resolving Redux by Mark Waldrep

http://www.realhd-audio.com/?p=6545

wrote by John Siau

[ There are exactly three potential advantages provided by high-resolution audio:

1: An increased high frequency limit

2: An increased immunity to the clipping of intersample peaks

3: An increased SNR

Items 2 and 3 are audible under the right circumstances. Item 1 may never or almost never be audible. Here is my reasoning...

2: Clipping of intersample peaks

The worst-case clipping is caused by a tone at 1/4 of the sample rate that is shifted 45 degrees relative to the sampling clock. This tone can reach 3.01 dB over 0 dBFS before the maximum digital codes are reached. This 0 dBFS + condition can happen 1000s of times on a single CD track. When upsampled in a sigma-delta D/A converter, these intersample peaks can cause clipping in the D/A converter. This often causes a DSP overload that creates a burst of IMD. This artifact is audible, but completely avoidable. The Benchmark DAC2 and DAC3 converters do not have this artifact at any sample rate. Most other D/A converters have this artifact. This will tend to make high sample rates sound better unless you are using Benchmark DAC2 or DAC3 converters. Here is a case where a better system makes all sample rates sound good. Most systems will make 44.1 sound worse than it should.

3: SNR:

A few quick calculations will show that the listeners will need playback systems with at least an 87 dB SNR and they will need to play the audio at peak levels exceeding 93 dB SPL. Otherwise, it is mathematically impossible to hear the noise advantage of anything beyond 16 bits...

Here is a chart that I created. It sums the digital channel with the playback system noise to calculate the SNR that will be delivered to the listener. I included bit depths of 8 through 24 bits. Find your playback SNR or peak playback SPL (whichever is lowest) on the X axis and then find the y value for each bit-depth curve. If the lines are separated at your SNR or SPL, then you will be able to hear a change in the noise floor. ]

 

[maty]

Maybe the foobar2000 at least has a problem with the intersample peaks like the DACs.

That is why the software is so important, its optimization, the instructions set, priorities, access modes ... something I have been tinkering with for some time, achieving an obvious improvement in sound. What problems do I minimize?

Clipping of intersample peaks, jitter? I do not know. I am only aware of the sound improvement.

 

[daftcombo]

The worst-case clipping is caused by a tone at 1/4 of the sample rate that is shifted 45 degrees relative to the sampling clock. This tone can reach 3.01 dB over 0 dBFS before the maximum digital codes are reached. This 0 dBFS + condition can happen 1000s of times on a single CD track. When upsampled in a sigma-delta D/A converter, these intersample peaks can cause clipping in the D/A converter. This often causes a DSP overload that creates a burst of IMD. This artifact is audible, but completely avoidable. The Benchmark DAC2 and DAC3 converters do not have this artifact at any sample rate. Most other D/A converters have this artifact. This will tend to make high sample rates sound better unless you are using Benchmark DAC2 or DAC3 converters. Here is a case where a better system makes all sample rates sound good. Most systems will make 44.1 sound worse than it should.

This really puzzles me. Why don't you expect more IMD when you have more info (here: over 22.1 kHz)?

 

[daftcombo]

Maybe the foobar2000 at least has a problem with the intersample peaks like the DACs.

That's why the software is so important, its optimization, the set of instructions, priorities, access modes ... something I have been tinkering with for some time, achieving an obvious improvement in sound. What problems do I minimize?

Clipping of intersample peaks, jitter? I do not know. I am only aware of the sound improvement.

Can you list the configuration you set in JRMC?

 

[maty]

foobar2000, JRMC

* Output: Kernel Streaming.

* Buffer: 50 ms

Some less important things.

Oh, and I force the players to use the AVX2 instruction set to play audio. I do not have it automated and sometimes I forget it and then I notice the difference!

 

[daftcombo]

foobar2000, JRMC

* Output: Kernel Streaming.

* Buffer: 50 ms

Some less important things.

Oh, and I force the players to use the AVX2 instruction set to play audio. I do not have it automated and sometimes I forget it and then I notice the difference!

Thanks.
How do you do the AVX2 thingy? I want to try to be sure.
And why would buffer length change something in the sound quality except preventing clicks & pops?

 

[maty]

I do not know all the answers!

I have been testing things that others claimed to work and also ideas of mine.

The thing about the instruction set is something more complex since it affects the OS, like other optimizations that I do. Neither is the original idea mine.

I would have to write a long text with many images to explain the whole process. Then criticism would come and it would be the story of never ending. More than fifteen years ago, when I wrote about computer security, I decided that I would not do it again, much less on the open Internet. I feel much more comfortable away from the current mass and bad manners. Then of the Spaniards and now in any language!

I promised myself that I would only comment on Amirm's amplification reviews. I failed.

 

[daftcombo]

I do not know all the answers!

I have been testing things that others claimed to work and also ideas of mine.

The thing about the instruction set is something more complex since it affects the OS, like other optimizations that I do. Neither is the original idea mine.

I would have to write a long text with many images to explain the whole process. Then criticism would come and it would be the story of never ending. More than fifteen years ago, when I wrote about computer security, I decided that I would not do it again, much less on the open Internet. I feel much more comfortable away from the current mass and bad manners. Then of the Spaniards and now in any language!

I promised myself that I would only comment on Amirm's amplification reviews. I failed.

I understand. You can write in Spanish if you want, I can read it. I want to try before criticizing, even if there shouldn't be any difference.

 

[mansr]

2: Clipping of intersample peaks

The worst-case clipping is caused by a tone at 1/4 of the sample rate that is shifted 45 degrees relative to the sampling clock. This tone can reach 3.01 dB over 0 dBFS before the maximum digital codes are reached.

It is possible to create signals with far higher inter-sample peaks, but that's beside the point.

This 0 dBFS + condition can happen 1000s of times on a single CD track. When upsampled in a sigma-delta D/A converter, these intersample peaks can cause clipping in the D/A converter.

That is true at any sample rate. When such clipping occurs, it is simply a case of bad production. If the digital signal from the ADC exhibits inter-sample overs, it means the recording gain was set too high, and an analogue clipping alert would have noticed it. If inter-sample clipping is introduced during mixing or mastering, someone has boosted the level too much.

You are correct that such clipping does occur in practice. However, it is caused by poor choices during production, not by the format as such.

This often causes a DSP overload that creates a burst of IMD. This artifact is audible, but completely avoidable. The Benchmark DAC2 and DAC3 converters do not have this artifact at any sample rate. Most other D/A converters have this artifact. This will tend to make high sample rates sound better unless you are using Benchmark DAC2 or DAC3 converters. Here is a case where a better system makes all sample rates sound good. Most systems will make 44.1 sound worse than it should.

Dealing with inter-sample overs is as simple as adding a few bits of headroom in any processing. I guess this is what Benchmark do. If your DAC lacks this headroom (and yes, many do), reducing the volume a few dB in software is enough to avoid the problems.

 

[edechamps]

2: Clipping of intersample peaks

The worst-case clipping is caused by a tone at 1/4 of the sample rate that is shifted 45 degrees relative to the sampling clock. This tone can reach 3.01 dB over 0 dBFS before the maximum digital codes are reached. This 0 dBFS + condition can happen 1000s of times on a single CD track. When upsampled in a sigma-delta D/A converter, these intersample peaks can cause clipping in the D/A converter. This often causes a DSP overload that creates a burst of IMD. This artifact is audible, but completely avoidable. The Benchmark DAC2 and DAC3 converters do not have this artifact at any sample rate. Most other D/A converters have this artifact. This will tend to make high sample rates sound better unless you are using Benchmark DAC2 or DAC3 converters. Here is a case where a better system makes all sample rates sound good. Most systems will make 44.1 sound worse than it should.

If you use a software player such as foobar2000 it's very easy to avoid this problem: just set the software volume control (such as the volume control of foobar2000 itself) to -3 dB. This will get rid of all (or almost all) potential intersample peaks, and then it doesn't matter what sample rate you use. This will reduce overall SNR by 3 dB but that's benign in comparison.

When I use audio players, I always set their internal software volume control to around -3 dB precisely for that reason. It don't leave it at 100%.

 

[daftcombo]

If you use a software player such as foobar2000 it's very easy to avoid this problem: just set the software volume control (such as the volume control of foobar2000 itself) to -3 dB. This will get rid of all (or almost all) potential intersample peaks, and then it doesn't matter what sample rate you use. This will reduce overall SNR by 3 dB but that's benign in comparison.

When I use audio players, I always set their internal software volume control to around -3 dB precisely for that reason. It don't leave it at 100%.

When I create an impulse in RePhase, I check at the max impulse level.
I had read that it should be less then 0dB so I used to play with the curve height cursor until the max impulse was around -2dB.

But I have a few tracks which, played in JRiver with that impulse maxing at -2dB (by convolution), peak at 100% (as JRiver reports).
It've found out that having a max impulse at around -8dB prevents me from reaching 100% - at least with all the tracks I have played yet.



In Foobar2000, the "auto level adjust" box should be unckecked, otherwise I hear clipping on those sames tracks. Which I do not hear in JRiver even when "100%" is displayed. But I don't want to see that value, just in case...

 

[edechamps]

But I have a few tracks which, played in JRiver with that impulse maxing at -2dB (by convolution), peak at 100% (as JRiver reports).
It've found out that having a max impulse at around -8dB prevents me from reaching 100% - at least with all the tracks I have played yet.

Yes, in general, avoiding clipping when using such filters can be tricky, because it's often possible to come up with a pathological input signal that will trigger the impulse just right and drive it over the line. What I usually do when I generate such filters is that I try them on a few music files (preferably loud, highly compressed ones for a worst case scenario) and I make sure that I have at least a few dB of headroom in the resulting peak level of all of them.

 

[eliash]

foobar2000, JRMC

* Output: Kernel Streaming.

* Buffer: 50 ms

Some less important things.

Oh, and I force the players to use the AVX2 instruction set to play audio. I do not have it automated and sometimes I forget it and then I notice the difference!

Foobar 3.1.9 driving ASIO for C-MEDIA device (CM6631a), buffer 3s (HD sleep) - the trick to make it work perfectly on an 1.6 GHz N270 ATOM (Samsung N10 Netbook) is running ASIO as realtime process (via PRIO on WIN XP, no foobar high prio set!), fed by stable LAN or local HD, WLAN should be deactivated!

No issues any more, no drop-outs, no clicks, also for 192KHz files...see below:

Followed by ground loop isolation from Netbook (via LAN shield) to Sabre ESS9018s DAC/Amp via S/PDIF transformer...no hum, no digital noise, can hear bit #24 on full volume, w ear next to speaker...

Eventually 88/96K sounds a bit better than 44/48K, depending source material, when spectrum above 20KHz existent...try downsampling 192K live ´78 "Thin Lizzy -Live and Dangerous" to 48K (used SOX, best quality setting), it´s astonishing!

96K is therefore my choice, 24bits help because of increased headroom (adjusted digital clipping limit of app. 650Wpeak amp chain/speakers at 2h volume position is my practical limit)!

Biggest current problem for me is digital clipping on almost all digital media (as bad as mostly associated poor Peak-to-Loudness),
True Peak (on tc-electronic´s Clarity M audio analyser) hitting +0.7dB over FS (even measured on analog audio output)...

 

[maty]

Thin Lizzy - Live And Dangerous (1978), Vinyl x2, Vertigo, UK 1st Pressing

https://www.discogs.com/Thin-Lizzy-Live-And-Dangerous/release/1321817

DR Peak RMS Filename
———————————————————————————————-
DR12 -0.45 dB -14.96 dB A1 Jailbreak
DR11 -1.24 dB -13.68 dB A2 Emerald
DR12 -2.12 dB -15.28 dB A3 Southbound
DR11 -3.47 dB -16.25 dB A4 Rosalie-Cowgirl’s Song
DR12 -1.69 dB -16.55 dB B1 Dancing In The Moonlight (It’s Caught Me In It’s Spotlight)
DR11 -1.63 dB -14.11 dB B2 Massacre
DR12 -1.51 dB -16.00 dB B3 Still In Love With You
DR13 -2.26 dB -18.18 dB B4 Johnny The Fox Meets Jimmy The Weed
DR12 -0.42 dB -14.97 dB C1 Cowboy Song
DR12 -0.93 dB -14.46 dB C2 The Boys Are Back In Town
DR12 -1.44 dB -14.88 dB C3 Don’t Believe A Word
DR11 -2.12 dB -14.75 dB C4 Warrior
DR11 -3.30 dB -16.17 dB C5 Are You Ready
DR12 -1.56 dB -14.85 dB D1 Suicide
DR12 -1.34 dB -15.29 dB D2 Sha La La
DR12 -1.09 dB -16.29 dB D3 Baby Drives Me Crazy
DR13 -2.51 dB -17.05 dB D4 The Rocker
———————————————————————————————-
Number of files: 17
Official DR value: DR12

but... dr.loudness-war.info Thin Lizzy - Live And Dangerous

 

[eliash]

Thin Lizzy - Live And Dangerous (1978), Vinyl x2, Vertigo, UK 1st Pressing

https://www.discogs.com/Thin-Lizzy-Live-And-Dangerous/release/1321817

DR Peak RMS Filename
———————————————————————————————-
DR12 -0.45 dB -14.96 dB A1 Jailbreak
DR11 -1.24 dB -13.68 dB A2 Emerald
DR12 -2.12 dB -15.28 dB A3 Southbound
DR11 -3.47 dB -16.25 dB A4 Rosalie-Cowgirl’s Song
DR12 -1.69 dB -16.55 dB B1 Dancing In The Moonlight (It’s Caught Me In It’s Spotlight)
DR11 -1.63 dB -14.11 dB B2 Massacre
DR12 -1.51 dB -16.00 dB B3 Still In Love With You
DR13 -2.26 dB -18.18 dB B4 Johnny The Fox Meets Jimmy The Weed
DR12 -0.42 dB -14.97 dB C1 Cowboy Song
DR12 -0.93 dB -14.46 dB C2 The Boys Are Back In Town
DR12 -1.44 dB -14.88 dB C3 Don’t Believe A Word
DR11 -2.12 dB -14.75 dB C4 Warrior
DR11 -3.30 dB -16.17 dB C5 Are You Ready
DR12 -1.56 dB -14.85 dB D1 Suicide
DR12 -1.34 dB -15.29 dB D2 Sha La La
DR12 -1.09 dB -16.29 dB D3 Baby Drives Me Crazy
DR13 -2.51 dB -17.05 dB D4 The Rocker
———————————————————————————————-
Number of files: 17
Official DR value: DR12

but... dr.loudness-war.info Thin Lizzy - Live And Dangerous

I have this downsampled to 96K:

foobar2000 1.3.9 / Dynamic Range Meter 1.1.1
--------------------------------------------------------------------------------
Analyzed: Thin Lizzy / Live And Dangerous (HD)
--------------------------------------------------------------------------------

DR Peak RMS Duration Track
--------------------------------------------------------------------------------
DR10 -0.84 dB -13.53 dB 4:44 01-Jailbreak
DR10 -1.27 dB -12.18 dB 4:23 02-Emerald
DR11 -1.53 dB -13.72 dB 4:43 03-Southbound
DR10 -2.26 dB -14.79 dB 4:14 04-RosalieCowgirl's Song
DR11 -1.97 dB -15.74 dB 3:54 05-Dancing In The Moonlight (It's Caught Me In It's Spotlight)
DR10 -1.08 dB -12.87 dB 2:56 06-Massacre
DR11 -1.24 dB -14.57 dB 7:41 07-Still In Love With You
DR12 -0.79 dB -16.59 dB 3:47 08-Johnny The Fox Meets Jimmy The Weed
DR10 -0.95 dB -13.45 dB 4:54 09-Cowboy Song
DR10 -1.10 dB -12.88 dB 4:41 10-The Boys Are Back In Town
DR11 -0.21 dB -12.86 dB 2:19 11-Don't Believe A Word
DR10 -0.59 dB -12.75 dB 4:01 12-Warrior
DR10 -1.26 dB -13.68 dB 2:50 13-Are You Ready
DR11 -0.80 dB -13.40 dB 5:13 14-Suicide
DR11 -0.15 dB -13.53 dB 5:34 15-Sha-La-La
DR11 -0.79 dB -14.41 dB 7:10 16-Baby Drives Me Crazy
DR11 -1.10 dB -14.58 dB 3:49 17-The Rocker
--------------------------------------------------------------------------------

Number of tracks: 17
Official DR value: DR11

Samplerate: 96000 Hz
Channels: 2
Bits per sample: 24
Bitrate: 3161 kbps
Codec: FLAC
================================================================================

 

[eliash]

Foobar 3.1.9 driving ASIO for C-MEDIA device (CM6631a), buffer 3s (HD sleep) - the trick to make it work perfectly on an 1.6 GHz N270 ATOM (Samsung N10 Netbook) is running ASIO as realtime process (via PRIO on WIN XP, no foobar high prio set!), fed by stable LAN or local HD, WLAN should be deactivated!

No issues any more, no drop-outs, no clicks, also for 192KHz files...see below:View attachment 27319

Followed by ground loop isolation from Netbook (via LAN shield) to Sabre ESS9018s DAC/Amp via S/PDIF transformer...no hum, no digital noise, can hear bit #24 on full volume, w ear next to speaker...

Eventually 88/96K sounds a bit better than 44/48K, depending source material, when spectrum above 20KHz existent...try downsampling 192K live ´78 "Thin Lizzy -Live and Dangerous" to 48K (used SOX, best quality setting), it´s astonishing!

96K is therefore my choice, 24bits help because of increased headroom (adjusted digital clipping limit of app. 650Wpeak amp chain/speakers at 2h volume position is my practical limit)!

Biggest current problem for me is digital clipping on almost all digital media (as bad as mostly associated poor Peak-to-Loudness),
True Peak (on tc-electronic´s Clarity M audio analyser) hitting +0.7dB over FS (even measured on analog audio output)...

One addendum related to the original lecture in the beginning of the thread:

Max SPL and therefore possible dynamic range (a la milkmaid, comments welcome!)

Dynaudio speaker (x2): 87dB @2W/1m (claimed)
Amp 650W (pulse total): +25dB over the above
Listening distance 2.5m: -4dB wrt to 1m
= SPL (pulse): 108dB (no clipping)

Ear sensivity: 0dB (hopefully, at least when listening to a highly dynamic recording, no compressing)

= Theoretical listening dynamic range: 108dB

CD: 93dB (incl. dithering)
HiRes 24/96 (incl. power amp SNR): app. 110dB

= The answer is somehow clear to me...

 

[solderdude]

The trouble is that when listening to peaks of 108dB and average level of around 90dB is present.
You cannot possibly hear 0dB faint sounds.
For this you need to be acclimatised in a acoustically dead room (which no one has) so while the 0dB is theoretically correct it is not of practical use.
Consider 20dB SPL what you could possibly hear in a room.
That leaves about 90dB of DR which fits inside the CD format.

I have found that my dynamic range is about 70dB when playing music very loud till hearing absolutely nothing (late evening listening).

I use the following device to determine what I could hear (instead of reading research about not relevant limits):

When listening to comfortably very loud music I could switch in a preset attenuation.
Turned out that when setting -70dB attenuation I could hear absolutely nothing any more.


These days it would be easier to create a file with music and then attenuate the music certain dB's and playing that file to find out where ones personal limits are on their system.

EDIT: Just made a 48/24 FLAC file with 0dB > -50dB >0dB > -60dB > 0dB > -70dB > 0dB > -80dB >0dB - 90dB > that -90dB part normalized to 0dB again to show what the -90dB file actually represented.