Ending the Windows Audio Quality Debate

[edechamps]

no windows mixer involved, except maybe when I used JAVA driver in REW, I don't know what it uses

The Java output in REW behaves like a normal Windows application and will therefore use the normal path, including mixing and APOs.

 

[Offler]

I have been using "less professional" software called 'audt30d'

1. To remove the effect of limiter I had to set EqAPO or signal strenght to -0,14dBFS, else CAudiolimiter will trigger.

2. ASIO drivers or WASAPI Exclusive mode bypass all filters, and results were identical - much lower THD in general.

My measuring indicated that triggered CAudiolimiter causes THD+N to be around -60dB, and when its not triggered its at -80dB which is limit of my equipment. If you have even better equipment, its worth the effort even more.

Therefore ASIO or Wasapi exclusive are perfect for critical listening even on Windows.

3. Also when there is no other solution but to use Windows Mixer (like for gaming):
a) Use EqAPO at -0.14dBFS (or -6,14dBFS if mixing 2 channels)
b) Select at least 192KHz 24bit.
Windows apps and games are a mixed bag of 44.1 and 48KHz files and sources. Using 192KHz 24bit will move all artifacts from resampling above 20KHz and thus beyond range of hearing.

 

[RandomEar]

This is related to Windows audio quality and only affects Realtek HD codecs, so I thought I'd share it here: I recently learned that Realtek HD audio codecs do support 88.2 kHz output over S/PDIF under Windows 10 (and probably 7 and 11). However, the original Realtek driver doesn't offer this sample rate. You need to uninstall that driver and use the basic HD audio driver provided by Microsoft, which is automatically installed after a reboot. I originally found this info here and have succesfully tested it on my mainboard, which uses an older Realtek ALC892 codec.

My original motivation to enable 88.2 kHz was that I listen to mixed content (44.1 & 48 kHz) and wanted to avoid upsampling errors for the 44.1 kHz stuff, while also avoiding downsampling to that rate for the 48 kHz content. This left me looking for integer multiples of 44.1 kHz. And while 176.4 kHz isn't supported on my Realtek chip, 88.2 was supposed to be working but couldn't be enabled. Anyway - all this probaly won't make any audible difference and I could just have gone with 96 or 192 kHz and forgot about it, but 88.2 looked like the cleanest solution to me.

 

[tjtremor]

This is related to Windows audio quality and only affects Realtek HD codecs, so I thought I'd share it here: I recently learned that Realtek HD audio codecs do support 88.2 kHz output over S/PDIF under Windows 10 (and probably 7 and 11). However, the original Realtek driver doesn't offer this sample rate. You need to uninstall that driver and use the basic HD audio driver provided by Microsoft, which is automatically installed after a reboot. I originally found this info here and have succesfully tested it on my mainboard, which uses an older Realtek ALC892 codec.

My original motivation to enable 88.2 kHz was that I listen to mixed content (44.1 & 48 kHz) and wanted to avoid upsampling errors for the 44.1 kHz stuff, while also avoiding downsampling to that rate for the 48 kHz content. This left me looking for integer multiples of 44.1 kHz. And while 176.4 kHz isn't supported on my Realtek chip, 88.2 was supposed to be working but couldn't be enabled. Anyway - all this probaly won't make any audible difference and I could just have gone with 96 or 192 kHz and forgot about it, but 88.2 looked like the cleanest solution to me.

but why ? windows resample even with correct ratio has the chance of degrading the original quality & adding artifacts if the ratio is not perfect

From what I see on Amazon music & movies out there
cd quality officially stops at 16bit 44.1
ultra hd starts at 24bit 44.1 & I haven't seen anything use multiples of that
next is 24bit 48 & that's where multiple of that start, 96 & 192

I have shortcuts assigned for changing samples rates, since Amazon uses the windows mixer.
I kept 32bit to reduce these keyboard shortcuts.

Works good enough if you listen to albums where the sample rate doesn't change.

SoundVolumeView.exe /SetDefaultFormat "Realtek(R) Audio\Device\Speakers\Render" 32 44100
SoundVolumeView.exe /SetDefaultFormat "Realtek(R) Audio\Device\Speakers\Render" 32 48000
SoundVolumeView.exe /SetDefaultFormat "Realtek(R) Audio\Device\Speakers\Render" 32 96000
SoundVolumeView.exe /SetDefaultFormat "Realtek(R) Audio\Device\Speakers\Render" 32 192000

 

[BeerBear]

but why ? windows resample even with correct ratio has the chance of degrading the original quality & adding artifacts if the ratio is not perfect

IMO:
1. The chance of that being audible is very very small. Any blind tests that show people can hear it?
2. If you're worried about Windows' resampling quality, the simplest solution is to use media players that can provide their own high quality resampling (such as Foobar, for instance). But sure, that might not work if you're forced to use some DRM/streaming app - but maybe it can still be solved with some virtual routing/EQAPO solution that includes a resampler, if you can apply that before the Windows' resampling kicks in.

 

[Grooved]

but why ? windows resample even with correct ratio has the chance of degrading the original quality & adding artifacts if the ratio is not perfect

From what I see on Amazon music & movies out there
cd quality officially stops at 16bit 44.1
ultra hd starts at 24bit 44.1 & I haven't seen anything use multiples of that
next is 24bit 48 & that's where multiple of that start, 96 & 192

I have shortcuts assigned for changing samples rates, since Amazon uses the windows mixer.
I kept 32bit to reduce these keyboard shortcuts.

Works good enough if you listen to albums where the sample rate doesn't change.

SoundVolumeView.exe /SetDefaultFormat "Realtek(R) Audio\Device\Speakers\Render" 32 44100
SoundVolumeView.exe /SetDefaultFormat "Realtek(R) Audio\Device\Speakers\Render" 32 48000
SoundVolumeView.exe /SetDefaultFormat "Realtek(R) Audio\Device\Speakers\Render" 32 96000
SoundVolumeView.exe /SetDefaultFormat "Realtek(R) Audio\Device\Speakers\Render" 32 192000

That's right that Amazon seems to not use 88.2, maybe because they know that a lot of drivers doesn't support it.
But if you use Qobuz, or Tidal with the MQA decoder on (by default on the app) or via Roon, you get 88.2 streams, and I also already tried the solution of @RandomEar to uninstall the Realtek driver to let Microsoft driver take care of 88.2

Note: I didn't cheched a lot on Amazon, but what made me think that they avoid 88.2 is that I saw their version of Random Access Memories is at 48kHz, while the original master is at 88.2kHz. They would use this version if they were allowin 88.2kHz.
Qobuz has it at 88.2, and Tidal had 16/44.1 FLAC only first, then 16/44.1 MQA, and latey added 24/88.2 (MQA 44.1 decoded in 88.2 stream)

 

[tjtremor]

IMO:
1. The chance of that being audible is very very small. Any blind tests that show people can hear it?
2. If you're worried about Windows' resampling quality, the simplest solution is to use media players that can provide their own high quality resampling (such as Foobar, for instance). But sure, that might not work if you're forced to use some DRM/streaming app - but maybe it can still be solved with some virtual routing/EQAPO solution that includes a resampler, if you can apply that before the Windows' resampling kicks in.

You don't need any blind test to pick-up inter sample artifacts as they are 100% audible.
Here's an example how it sounds, end of piano notes generates high glass like freq (12sec+); not on the master for this track (Amazon ultra hd).

I've encountered a few where artifacts sound like very brief broken or turning into robotic distortion at the end of certain sound patterns. I don't think it's the limiter since matching the sample rate fixes it. Could be realtek / windows software mixer too, maybe it got better with recent updates.
I didn't record or bookmark those songs.

 

[Eldus]

foobar2000 does it automatically out of the box, with the bundled SSRC resampler. If the source sample rate is different from the driver's sample rate, it will be automatically resampled to the driver's sample rate. That means, there is no need to add another resampler in foobar2000's DSP manager. If you have another resampler plugin installed, you can configure it here:
View attachment 107731

I've tested these resamplers. Search for "It has two bundled resamplers" in this article for the tests.
http://archimago.blogspot.com/2019/06/guest-post-why-we-should-use-software.html

Are you using a theme for Foobar or windows? Which one? I love the colors.

 

[BeerBear]

You don't need any blind test to pick-up inter sample artifacts as they are 100% audible.

Inter sample artifacts?
Anyway, I'm not convinced, but I am willing to test this.
Which media players guaranteed do not apply their own resampling and leave it to Windows? I thought VLC would work, but it seems not (if I disable the resampler in the settings it just doesn't produce any sound anymore). MPV maybe?

EDIT: OP says:

Playing a 44.1kHz file with upsampling to 96kHz, 24 bits, and EAPO set to -0.2 dB to avoid CAudioLimiter, the highest distortion component was at -134 dB

Which makes me think that it would not be audible. But it's not guaranteed that every sound will have such low artifacts.

 

[RandomEar]

but why ? windows resample even with correct ratio has the chance of degrading the original quality & adding artifacts if the ratio is not perfect

From what I see on Amazon music & movies out there
cd quality officially stops at 16bit 44.1
ultra hd starts at 24bit 44.1 & I haven't seen anything use multiples of that
next is 24bit 48 & that's where multiple of that start, 96 & 192

I have shortcuts assigned for changing samples rates, since Amazon uses the windows mixer.
I kept 32bit to reduce these keyboard shortcuts.

Works good enough if you listen to albums where the sample rate doesn't change.

SoundVolumeView.exe /SetDefaultFormat "Realtek(R) Audio\Device\Speakers\Render" 32 44100
SoundVolumeView.exe /SetDefaultFormat "Realtek(R) Audio\Device\Speakers\Render" 32 48000
SoundVolumeView.exe /SetDefaultFormat "Realtek(R) Audio\Device\Speakers\Render" 32 96000
SoundVolumeView.exe /SetDefaultFormat "Realtek(R) Audio\Device\Speakers\Render" 32 192000

Honestly: I thought about not giving any reason and just putting the info about the driver into the reply, because I already expected the random criticism. The post was mostly about letting interested people know about that driver issue/solution, because it was new to me.

There is no perfect solution for everybody. I don't want to manually switch sample rates every 30 minutes - I don't want to ever do that. I don't care about distortion components at -135 dB or so for the 48 kHz stuff and I assume that the 44.1 kHz content performs better than that when upsampled to 88.2. And yes, I know: To have absolute peace of mind, I would have to measure and compare my current solution to the alternatives. But I currently lack the time and patience to do so. Therefore, I'll leave it as is for now.

My assumption, that integer multiples of the sample rate would lead to less artifacts during upsampling sounds plausible to me, but is yet to be proven. Feel free to check it if you have the means to, I'd be thankful to know if it is correct.

 

[Offler]

Honestly: I thought about not giving any reason and just putting the info about the driver into the reply, because I already expected the random criticism. The post was mostly about letting interested people know about that driver issue/solution, because it was new to me.

There is no perfect solution for everybody. I don't want to manually switch sample rates every 30 minutes - I don't want to ever do that. I don't care about distortion components at -135 dB or so for the 48 kHz stuff and I assume that the 44.1 kHz content performs better than that when upsampled to 88.2. And yes, I know: To have absolute peace of mind, I would have to measure and compare my current solution to the alternatives. But I currently lack the time and patience to do so. Therefore, I'll leave it as is for now.

My assumption, that integer multiples of the sample rate would lead to less artifacts during upsampling sounds plausible to me, but is yet to be proven. Feel free to check it if you have the means to, I'd be thankful to know if it is correct.

1. Use the driver provided by the manufacturer of your mainboard.
Unfortunately, all Realtek onboard soundcards are customized, and they require customized driver.

Driver provided by Microsoft would work, however I was experiencing some strange errors with it.

2. You can use WASAPI Exclusive mode where audio player bypasses all windows components and filters.
In most situations the player automatically switches to "native" sample and bitrate and sends it to output "as is".

 

[digitalfrost]

no, I just quickly tested the digital loopback for difference between 0 and -1dBFS, but not in the same conditions than the first post (no EqAPO and no windows mixer involved, except maybe when I used JAVA driver in REW, I don't know what it uses)

How do I enable this internal loopback in REW? I have only my normal audio interfaces for selection.

 

[Grooved]

How do I enable this internal loopback in REW? I have only my normal audio interfaces for selection.

You need to add virtual input/output, like with VB Audio HiFi Cable. Then you can select it in ASIO4ALL like any other audio device

 

[GunGrave87]

All this sounds very nice and well, except for the fact that I have not been able to get EAPO working on 3 different machines (one PC, 2 laptops) over the past 2 years.
The latest iteration of EAPO installed on a Windows 10 Dell Latitude 5521 shows the checkboxes for "Use Original APO" unchecked and disabled for onboard audio. At there are some serious signal processing shenanigans as by default I have some serious stereo crosstalk.
Disabling "audio enhancements" in control panel fixes this, but completely brakes EAPO.

 

[Deleted member 46664]

A simple formula for good sound from windows...

1) Install Media Player Classic Black Edition ... Go through the settings and get it to play your music files normally.
2) In the Options/Audio tab ... Select MPC Audio Renderer as your renderer and hit "Properties". Set it like this...

Now when you play a music file, the normal Windows mixer and effects engine is bypassed and the music player owns the DAC or sound chip. Files will play at their native speeds and bit depths .... for much better sound. (If your modem displays sample rates you will see it change with each song)

For streaming services like Amazon (etc) install their appys and look in their settings for Exclusive mode, too.

Almost all of the "crappy sound" problems happen in the mixer and effects engines... Exclusive Mode bypasses them completely.

 

[GunGrave87]

Well I'll be. Allowing the apps through the firewall, deleting the registry entries and installing as experimental did the trick.
The equalizer is now working and producing an audible result. Sadly, even with "Use Original APOs" disabled, I get stereo cross talk.

I added the -4db amp in the editor, that did nothing.
Bonus, now I have mad clipping in my audio and some distortion.

 

[edechamps]

The equalizer is now working and producing an audible result. Sadly, even with "Use Original APOs" disabled, I get stereo cross talk.

I had some issues like that with EAPO and recent versions of Windows. Some third-party audio processing still appeared to be taking place even though I told EAPO to disable original APOs. I managed to fix it using this procedure.

 

[Aerith Gainsborough]

You need to add virtual input/output, like with VB Audio HiFi Cable. Then you can select it in ASIO4ALL like any other audio device

Thank you, that was the last piece I needed to recreate the measurements.

TLR: You've got to be shitting me. How crappy are my ears, that garbage such as this isn't audible in actual content. (I still can't hear a lick of difference between ASIO and WASAPI shared).

You buy state of the art RME gear and then Windoof™ clobbers you down to a measly 55dB to the next harmonic. :'D

1KHz test signal.

19KHz + 20KHz ... oO

 

[Offler]

Thank you, that was the last piece I needed to recreate the measurements.

TLR: You've got to be shitting me. How crappy are my ears, that garbage such as this isn't audible in actual content. (I still can't hear a lick of difference between ASIO and WASAPI shared).

You buy state of the art RME gear and then Windoof™ clobbers you down to a measly 55dB to the next harmonic. :'D

View attachment 203141
1KHz test signal.
View attachment 203142
19KHz + 20KHz ... oO

Set both test signals to -6.3dB in EAPO preamp or in signal generator. Both IMD and THD should start behave better.

I stopped to dabble with any tweaks at point where i could not tell difference between WASAPI: Shared, WASAPI: Exlusive and ASIO.

 

[Aerith Gainsborough]

Set both test signals to -6.3dB in EAPO preamp or in signal generator. Both IMD and THD should start behave better.

I stopped to dabble with any tweaks at point where i could not tell difference between WASAPI: Shared, WASAPI: Exlusive and ASIO.

Amplitude has no effect, even at -10dB down in the signal generator I still get the same spectrum of distortions. This is definitely some processing going on.

No biggie though. Since I use the rather excellent ASIO driver of my RME for my VSTi anyway, I just set Foobar to ASIO for peace of mind. This way I get bit-perfect playback (verified) and can still mix in other sounds from other programs if need be. Best of both worlds, I guess.

The difference is close to impossible to hear for me. I put on a track ( a loud one, lots of high frequency content) at my usual 90dB Z-peak listening volume and turned it down by 55dB. I could hear faint high frequencies just at the edge of my perception. There is -0- chance I would ever hear that while actual music is blasting into my ears.