Ending the Windows Audio Quality Debate

[Offler]

oh 88.2 is just so that I don't have to change every time while not have major effect since majority of my contents are 44.1kHz
and I don't plan on using exclusive mode (too inconvenient when switching to another audio stream)

for the distortion, it certainly seems that most amp nowadays are able to do that, not sure about speaker/headphones though
anyway, this overall does seem beneficial for my specific system/usage

Fair enough.

For me exclusive mode represents best way to listen high-quality CD, DVD and BluRay content on Windows. Especially if its uncompressed (or lossless compression), its worth all the effort and money. Media players are usually smart enough to switch to correct samplerate on their own.

Windows desktop, browsers and common streamed videos run on 44.1KHz, as I checked today. This is actually reason why i will keep 44.1 as a default, so i can mix music (usually of CD Audio quality) with browsing.

And then there is the issue with games. Half of them certainly uses 48Khz, the other half 44.1Khz.... And mostly titles with really good soundtrack use 48KHz...

Now... As simple as it might be, windows is incapable to switch output samplerate depending on Application preference. I can set 44.1 applications to use one audio endpoint, and 48Khz applications to another, manually configure them and then switch between sources on amplifier.

So I am looking how to remove the inconvenience of manually switching Default output samplerate, but its not much of an issue.

 

[garbz]

and re-sampling in Windows has poor quality.

Other than the link from 6 years ago in the first post is there any other reference to tests on this? The link in the first post was explored extensively by the guys over at hydrogen audio and most couldn't reproduce it. There was a lot of thought that this was related to a bug in a specific version of windows. They did however identify that CAudioLimiter was causing a lot of grief with full range signals, but with limited signals most shows graphs with distortion components around -150dB.

Maybe if I get a moment on the weekend I'll see if I can investigate this further.

 

[Offler]

Other than the link from 6 years ago in the first post is there any other reference to tests on this? The link in the first post was explored extensively by the guys over at hydrogen audio and most couldn't reproduce it. There was a lot of thought that this was related to a bug in a specific version of windows. They did however identify that CAudioLimiter was causing a lot of grief with full range signals, but with limited signals most shows graphs with distortion components around -150dB.

Maybe if I get a moment on the weekend I'll see if I can investigate this further.

I tested few of the claims and workaround made in the first posts of this thread, once I got my amplifier and learned how to measure things in FFT graph using soundcard as an input. I confirmed the claims about CAudioLimiter about 5 months ago (-0.14dB to fix the distortion it causes for single sinewave, - found about that when measuring for THD. -6.3dB for two sinewaves when I tried to measure IMD.)

I just assume that resampler is quite bad, because when I used 44,1KHz output and 48KHz signal or vice-versa, there was some more spikes at the noisefloor.

 

[MKreroo]

Fair enough.

For me exclusive mode represents best way to listen high-quality CD, DVD and BluRay content on Windows. Especially if its uncompressed (or lossless compression), its worth all the effort and money. Media players are usually smart enough to switch to correct samplerate on their own.

Windows desktop, browsers and common streamed videos run on 44.1KHz, as I checked today. This is actually reason why i will keep 44.1 as a default, so i can mix music (usually of CD Audio quality) with browsing.

And then there is the issue with games. Half of them certainly uses 48Khz, the other half 44.1Khz.... And mostly titles with really good soundtrack use 48KHz...

Now... As simple as it might be, windows is incapable to switch output samplerate depending on Application preference. I can set 44.1 applications to use one audio endpoint, and 48Khz applications to another, manually configure them and then switch between sources on amplifier.

So I am looking how to remove the inconvenience of manually switching Default output samplerate, but its not much of an issue.

That got me thinking, if it's possible to have something that works similar to WASAPI exclusive, where it follows sampling rate of audio stream coming from a selected software (music player for example), and any other audio stream simply get re-sampled into the current rate. though this does seem to be too much for such a small problem.

 

[garbz]

I just assume that resampler is quite bad, because when I used 44,1KHz output and 48KHz signal or vice-versa, there was some more spikes at the noisefloor.

Just for fun I played a bit with an audio driver loop to see if I can replicate something measurable. I set the audio loop out to 24/96. I set the audio loop in to 24/96. I then tried to play a bit with various settings. I should really go through and do something more definitive but I was tired and didn't see much difference between the figures. Not until I got above -0.1dB. I forced Foobar to generate a sine at 44.1kHz (I know the filename says 48kHz but ... as I said... tired) in DS mode meaning that Windows was resampling to 96kHz.

I'm curious now that I'll do something more detailed.

To be clear this wasn't always the case. There are documented issues with Window's resampler over the years, but the other problem is it really is a moving target. MS tweaks and changes things all the time. Even the CAudioLimiter has changed in response (there was some test software posted on github a long time ago and that has shown me different results over the years).

That got me thinking, if it's possible to have something that works similar to WASAPI exclusive, where it follows sampling rate of audio stream coming from a selected software (music player for example), and any other audio stream simply get re-sampled into the current rate. though this does seem to be too much for such a small problem.

To what end? The issue with bypassing windows is something needs to take control of volume. The last thing you want is your application controlling the volume playing music, and then another random app coming in playing something at full blast. Then there's the issue of mixing. The issues which cause quality loss like the CAudioLimiter are there for a reason. It's not just MS screwing with things. You either have bit perfect high res audio or you introduce a mixer which by its nature messes with things (though I don't disagree with messing with things, people put far too much value in "bitperfect").

 

[dasdoing]

people put far too much value in "bitperfect"

is there any value at all? never seen a modern resampler meassuring bad to the point of creating a real diference

 

[MKreroo]

Just for fun I played a bit with an audio driver loop to see if I can replicate something measurable. I set the audio loop out to 24/96. I set the audio loop in to 24/96. I then tried to play a bit with various settings. I should really go through and do something more definitive but I was tired and didn't see much difference between the figures. Not until I got above -0.1dB. I forced Foobar to generate a sine at 44.1kHz (I know the filename says 48kHz but ... as I said... tired) in DS mode meaning that Windows was resampling to 96kHz.

I'm curious now that I'll do something more detailed.

To be clear this wasn't always the case. There are documented issues with Window's resampler over the years, but the other problem is it really is a moving target. MS tweaks and changes things all the time. Even the CAudioLimiter has changed in response (there was some test software posted on github a long time ago and that has shown me different results over the years).


To what end? The issue with bypassing windows is something needs to take control of volume. The last thing you want is your application controlling the volume playing music, and then another random app coming in playing something at full blast. Then there's the issue of mixing. The issues which cause quality loss like the CAudioLimiter are there for a reason. It's not just MS screwing with things. You either have bit perfect high res audio or you introduce a mixer which by its nature messes with things (though I don't disagree with messing with things, people put far too much value in "bitperfect").

yeah, I really don't see the point of chasing after bit-perfect, I mean I certainly don't hear any audible difference, but since we are here, might as well have some fun with it.

 

[Offler]

is there any value at all? never seen a modern resampler meassuring bad to the point of creating a real diference

My amplifier is able to play sound at -80dB/0.005% THD+N. Under normal circumstances Windows adds more distortion at level of -60dB/0.1%. In case you invested in good equipment, you simply dont want Windows to reduce the quality. That is a result of measurings I did few months ago.

If you ask whether this is audible difference, my answer to that question is Yes. Not every amp/speaker combo would make the difference or not everyone will hear it.

There are basically two workarounds, either EAPO and to use Lower preamp, or WASAPI Exlusive/ASIO.

Reclocking from 44.1 to 48 KHz is not that audible and I mostly dont have any issues with it in terms of sound. SOME games might have issues with audio-image sync if those are not properly matched, you might suffer framedrops or microstuttering or there would not be any issue at all.

To me it looks like when the game IS port from a console to PC or it was developed for multiple platforms at the time, 48KHz might be the more default option and the game may run in more optimal fashion. But that is a question of performance, not sound quality.

But if you are playing CD, DVD or BD media, simply go for bit perfect via WASAPI or ASIO. Its easy to configure and the results are always great.

Edit: Some interesting stuff about Windows audio APOs:

http://optimalsound.com/Documents/Updated%20Chapters%20to%20High%20Definition%20Audio%20for%20the%20Digital%20Home%20(12-5-12).pdf

 

[dasdoing]

My amplifier is able to play sound at -80dB/0.005% THD+N

Windows adds more distortion at level of -60dB/0.1%

what about your speakers?

If you ask whether this is audible difference, my answer to that question is Yes.

in a blind test? have you made the Klippel test?

 

[Offler]

what about your speakers?



in a blind test? have you made the Klippel test?

NAD d3020v2 + Elac DBR 62s. I trust the Klippel tests done by Amirn here in case of speakers. In case of amp, i had to do my own measurements and I take them with grain of salt.

I did a blind test when i was switching playback of uncompressed audio over a player, tied Exclusive mode to a hotkey, kept pressing until I had no idea what is selected, and I could tell when Exclusive and Shared modes are used. But this is still quite subjective and i trust the graphs more than to my ears.

Frequency vs distortion Without correction in EAPO:

and with EAPO correction (also identical to Exclusive mode)

 

[dasdoing]

Elac DBR 62s

I mean, it is strange to care about 0.1% when the system is way above that

but if you hear it, you are right

 

[garbz]

is there any value at all? never seen a modern resampler meassuring bad to the point of creating a real diference

Personally, I'd argue no. Not just in resampling but also digital volume control and DSP. I'd much prefer listening to a DSP corrected system compensating for some of the weaknesses in equipment / room acoustics than the ridiculous notion of "purity".

 

[bennetng]

Just for fun I played a bit with an audio driver loop to see if I can replicate something measurable. I set the audio loop out to 24/96. I set the audio loop in to 24/96. I then tried to play a bit with various settings. I should really go through and do something more definitive but I was tired and didn't see much difference between the figures. Not until I got above -0.1dB. I forced Foobar to generate a sine at 44.1kHz (I know the filename says 48kHz but ... as I said... tired) in DS mode meaning that Windows was resampling to 96kHz.

In my tests running the ALC892 through the Windows resampler actually improved measurements. Can't say it is applicable to everyone but people who don't know how to test things tend to believe running anything through Windows Mixer must severely degrade things.

DAC resampling through Windows mixer:

DacMagicPlus Filter Mirror Images - Short Study

I agree, resampling from 44.1kHz to 96kHz with a proper re-sampler and resampling filters is the way to get better results and usually better sound as well ;). The challenge here is that there is no off-the-shelf ASRC with a stopband that begins at Fs/2. Every single chip I looked at...

www.audiosciencereview.com

ADC resampling through Windows mixer:

Interface Mystery

Inspired by ASR, I've started looking into the performance of various bits of equipment around the house and have come across a bit of a mystery. A certain 'audio recorder' (which shall go nameless for now in case this is all a misunderstanding on my part) is said by the manufacturer to be a...

www.audiosciencereview.com

 

[Offler]

I mean, it is strange to care about 0.1% when the system is way above that

but if you hear it, you are right

That was probably measured at +96dBspl, i am probably listening to much lower value.

Anyway this was what I got when i tried to mix sinewave at 50Hz and 7KHz. After testing i found out that CaudioLimiter is the culprit.

 

[bennetng]

As mentioned before, if you are mixing things and the result is above 0dBFS, even without a limiter the signal will clip anyway, just with a different distortion profile.

 

[Offler]

As mentioned before, if you are mixing things and the result is above 0dBFS, even without a limiter the signal will clip anyway, just with a different distortion profile.

And i keep repeating that it was distorted even when the result was at -6 dBFS.

 

[bennetng]

And i keep repeating that it was distorted even when the result was at -6 dBFS.

Did you match the channel count in the playback software, recording software, windows settings, and third party driver settings? Just want to tell you I don't have this issue, and therefore not universal to everyone.

 

[dasdoing]

That was probably measured at +96dBspl

86dB

 

[dasdoing]

And i keep repeating that it was distorted even when the result was at -6 dBFS.

where did you atenuate? the cliping could have acoured in the pre-mixing stage

 

[bennetng]

Okay, just recorded the ALC892 analog output, using stock Windows driver with 50Hz and 7kHz, using MPC-HC with SaneAR shared mode. 99% player volume when playing a single file and 70% player volume when playing both files in mix with two separate player instances. Without using EAPO as it is not needed.

FLAC attached in the zip file.