Science behind claim of MS OS Mixer bypass necessity?

[Ron Kuper]

I mainly refer to shared mode / exclusive mode. But also integrated in WASPI vs. System output (Direct Sound?) vs. ASIO drivers.

Is there a difference if no oversampling occure (e.g. content is equal to configured in device settings)?

Is the OS mixer volume implementation good?

Were things measured? Audible?


Thanks,
Ron

 

[amirm]

Hi Ron. I plan to do a detailed measurements of it.

I will also do some controlled listening tests if I can get a fixture going for it that allows AB testing.

 

[Soniclife]

This previous thread has some discussion of this.
https://audiosciencereview.com/foru...ty-volume-control-in-windows.2031/#post-54035

 

[Ron Kuper]

plan to do a detailed measurements of it.

*Thumbs up*!

This previous thread has some discussion of this.

Thanks!

Are there any conditions under which the sound quality is not affected?
I ask since I want to use the amazing free Equalizer APO (seriously impressive project!) but the one thing it is clearly lacking is in its reliance on windows audio enhancements (in itself a clean design choice, disregarding the SQ compromise) so that means - no "exclusive mode" in Roon/JRiver since it is bypassed there.

If only one app (i.e. Roon) is playing through the OS Mixer, and the sample rates are aligned, will it still degrade the quality?

 

[amirm]

The operating system mixer will always convert samples to floating point, then convert back to integer PCM samples with dither. It has no advance knowledge of only one app running as to bypass this. So the samples are always transformed. Whether that is an audible concern is another matter but they objectively change.

 

[Ron Kuper]

Hi Ron. I plan to do a detailed measurement of it.

Would be interesting to compare it to VoiceMeeter which sort of replaces the OS Mixer while also allowing input from multiple sources but in the form of multiple virtual audio interfaces.

I use it to get the amazing EqualizerAPO without going through the OS Mixer -

Roon/JRiver --> VoiceMeeter Virtual Interface 1 (WDM Exclusive mode), with EqualizerAPO*--> DIRAC DAP** (WDM Exclusive mode).
Roon is showing the "lossless" signal path indicator (relevant only to the first stage and I verified that VoiceMeeter is opening DAP in Exclusive mode).

I set VoiceMeeter Virtual Interface 2 as the default device for the system, that way I don't lose audio from other software sources and able to easily control them separately and having EqualizerAPO on both (Hi there bass management for Netflix )

*EqualizerAPO has implemented a specific support for VoiceMeeter API so that in this case it is not reliant on the Windows APO framework and not going through the OS Mixer.

**I use Dirac DAP since I don't have access to the VST plugins. Waiting for the new version. Would prefer to use Dirac as a VST under EqualizerAPO and simplify the chain.

***It is going through the VoiceMeeter Mixer which I hope is better implemented than the Windows OS Mixer, otherwise, I'm just fooling myself. Thus the desire to compare

****The newer VoiceMeeter BANANA is required to use with DAP if anyone is interested.

*****The main context is having BM with Dirac, done properly while maintaining a software-only approach for anything DSP - no reliance on vendors grace

And lastly, I can't stress enough how fantastic the opensource EqualizerAPO project is. Truly remarkable functionality. SQ-Wise the code is open. I've read the calculations could benefit a move from 32bit to 64bit floating point. Any C++ devs in the crowd?

 

[Ron Kuper]

Here is a link for a more detailed overview of the setup -
https://community.roonlabs.com/t/ho...-together-oh-and-dirac-is-there-too-8/41949/6

 

[Vincent Kars]

As Amir pointed out, almost all OS are using the same principle.
They expect multiple streams, convert to float, dither and convert back to integer.
As almost all DAC are 24 bit today, I don’t expect the dither (at -144 dBFS) audible.

The SRC is a different matter.
Windows, probably because it runs everywhere including under powered boxes, does it very efficient hence dirty.
OSX and Linux do a good job.

Some measurements:
http://archimago.blogspot.nl/2015/11/measurements-apple-mac-os-x-yosemite.html
http://archimago.blogspot.nl/2015/11/measurements-windows-10-audio-stack.html

 

[amirm]

Would be interesting to compare it to VoiceMeeter which sort of replaces the OS Mixer while also allowing input from multiple sources but in the form of multiple virtual audio interfaces.

Interesting. These VB-Audio guys make pretty cool stuff.