Ending the Windows Audio Quality Debate

[edechamps]

If there is no "disable sound enhancements" check box, it means there are no sAPOs and therefore nothing to disable.

Installing EAPO and then unchecking the box is a somewhat ridiculous roundabout way of ending up in the exact same place you started.

Any sound difference you think you hear by doing this is placebo.

You can double-check that there are no sAPOs installed by looking them up in the registry. IIRC, the case where there is no checkbox corresponds to the case where there is not even a FxProperties key. (If you have EAPO installed on the device, then you will see entries there, of course.)

 

[Aerith Gainsborough]

So I can consequently assume that sound enhancements are disabled by default(since they miss from the driver) and installing EQ apo so to make them available and disable them is an unnecessary step I don't have to take for my future win installations.
I will just need to install windows+topping driver and I can be sure windows enhancements are disabled.

It is unnecessary unless you want to start using PEQ.

If you want maximum quality you use bit perfect transmission methods anyway like WASABI exclusive or ASIO. Both should bypass windows enhancements even if the driver supports them.

 

[Frank2]

I bumped into an interesting experiment: https://hydrogenaud.io/index.php/topic,121315.0.html?PHPSESSID=gu7lbir1rh6gihpgugccql7g5e
It seems that CAudioLimiter is not engaged if the output endpoint is a 'digital destination' such as HDMI, USB, or SPDIF.
The statement on the site is: "So if you want to avoid the limiter on shared output, use a digital destination (HDMI or S/PDIF). Or as it turns out here, a USB device".

So I decided to do a small experiment on our laptop (which has no E-APO installed whatsoever). I installed the latest versions of Foobar and Audacity on this Win11 machine, and:

1. Used Audacity to generate a 96/24 stereo wav file with a 1 kHz tone at max amplitude. (The generated tone is mono, make sure to export it as a stereo wav file).
2. Then used Foobar to play this wav file towards a Musical Fidelity V-Link connected to USB, in shared WASAPI mode,
3. and used Audacity to record the digital signal sent by Foobar (via 96/24 loopback recording on the Musical Fidelity, using the WASAPI service).

Audacity should record the exact digital signal as it is send by the Windows 11 shared WASAPI audio chain to the Musical Fidelity endpoint.
If CAaudioLimiter is engaged along this chain, the max. amplitude should be 0,14 dB lower.

Instead, the recorded signal was exactly the same as the original wav file. Same spectrum plot, same amplitude (checked visually and by checking the amplification headroom indicated by Audacity - 0 dB in both cases). Also, same waveform with same amplitude when zooming in up to having the digital samples visualized. It had all the hallmarks of an exact digital copy.

It makes me wonder if the "E-APO trickery" as proposed by the OP is necessary in shared WASAPI mode towards a digital endpoint.

 

[MaxwellsEq]

Rehydrating the thread for people who have not seen it.

I had an "event" and had to reset Windows11 (all my files retained, but all applications deleted and settings returned to default). I put most of the settings back and reinstalled the software. I decided to listen to some music on headphones but it sounded poor. I realised that Foobar was using standard Windows audio - switching to exclusive mode fixed the sound.

Then I realised I'd not done the steps in this post. Once I had done it, there's almost no audible difference between Windows standard audio paths and exclusive mode!

 

[Blew]

Then I realised I'd not done the steps in this post. Once I had done it, there's almost no audible difference between Windows standard audio paths and exclusive mode!

Almost...

 

[Emlin]

Almost...

So what are the audible differences, then...

 

[NTK]

Since it is logically impossible to prove the absence of difference, and if there exists a difference, it seems it is extremely elusive, thus the obligatory "almost"

 

[Emlin]

Since it is logically impossible to prove the absence of difference, and if there exists a difference, it seems it is extremely elusive, thus the obligatory "almost"

Not with an audio waveform it isn't.

 

[NTK]

Not with an audio waveform it isn't.

You have a formal proof that the Windows audio stack is bug free?

 

[Emlin]

You have a formal proof that the Windows audio stack is bug free?

Different topic, surely. One can compare waveforms with way beyond the accuracy needed for audible differences

 

[Aerith Gainsborough]

You have a formal proof that the Windows audio stack is bug free?

Nothing is ever "always guaranteed" bug free.
Your fancy audiophile ASIO driver or even player can also have signal altering bugs.

People should really stop straining, projecting, and convincing themselves that they can hear some kind of difference between digital processing methods and listen to the actual music.

In the digital realm, it's really easy: if something is amiss, the effects are clearly audible w/o straining at all (e.g.: buggy ASIO, runaway EQ, etc).

 

[MaxwellsEq]

Almost...

Honestly, level matched I can't tell the difference into my DAC with my IEMs. Sometimes I think I can, but I don't trust my perception.

 

[MaxwellsEq]

So what are the audible differences, then...

Level matched I can't hear any difference. There may be measurable but inaudible differences of course. Sometimes, though I convince myself I can hear a difference, but I think it's just bias etc.

 

[Sokel]

The whole thing about the Windows quality are not the traditional "audible difference"
Is about having an uninterrupted,solid OS who doesn't messes up when we have a listen and follows whatever we intend to do.
That's all.

 

[Aerith Gainsborough]

The whole thing about the Windows quality are not the traditional "audible difference"
Is about having an uninterrupted,solid OS who doesn't messes up when we have a listen and follows whatever we intend to do.
That's all.

Unless you do real time stuff for music production, in which case a low latency w/o dropouts can be challenging to achieve, mere playback has been solved on PC's decades ago.

On modern systems, even if my Windows does god know what or I play video games on the side, my Foobar playback never falters.

 

[herr.jemine]

One note: With Windows 10/11, it makes a difference whether the virtualization functions are activated in the BIOS/UEFI of the computer and whether, for example, Windows Hyper-V has been installed. With Windows 11, kernel isolation is automatically activated if the system supports virtualization functions. This certainly has an impact on timing and latency.

I haven't read through everything, but if you have audio problems you can use the software LatencyMon to check your system for audio latency problems

Resplendence Software - LatencyMon: suitability checker for real-time audio and other tasks

LatencyMon: suitability checker for real-time audio and other tasks

www.resplendence.com

 

[Blew]

So what are the audible differences, then...

It sounds less detailed to me. I guess it's due to the artifacts from the resampling such as phase distortion and aliasing which were not measured here.

 

[frgmnt]

I have most of music in 44.1kHz, but most of the time when I listen to it, I am playing games or in a Discord call, which plays audio in 48kHz usually. For my situation, is it better to set my sample rate to 44.1, 48, 96, or 192? I saw the original post suggested to set it to 96k and another suggestion by someone else to set it to 192k. Amongst my collection of ~3000 songs, less than 20 of them are 96-192kHz, a few hundred in 48kHz, and the rest in 44.1.

 

[AnalogSteph]

For considerations along those lines, I am running 48 kHz on my main PC, with an upsampler DSP in my player of choice for extra peace of mind (SoX resampler DSP for Foobar2000 in "Best" quality, which causes negligible CPU load even on an old Core 2 Duo thanks to SSE3 optimization and thus will bother a modern one even less). The volume on everything else has been reduced to roughly match my levels with RG including some negative pre-gain, so there is plenty of headroom to go around. (My default YouTube volume is 35% these days.)

There should be little need to go higher unless you've got a case of wonky digital fliter, which is why my laptop is getting 192 kHz instead.

 

[frgmnt]

For considerations along those lines, I am running 48 kHz on my main PC, with an upsampler DSP in my player of choice for extra peace of mind (SoX resampler DSP for Foobar2000 in "Best" quality, which causes negligible CPU load even on an old Core 2 Duo thanks to SSE3 optimization and thus will bother a modern one even less). The volume on everything else has been reduced to roughly match my levels with RG including some negative pre-gain, so there is plenty of headroom to go around. (My default YouTube volume is 35% these days.)

There should be little need to go higher unless you've got a case of wonky digital fliter, which is why my laptop is getting 192 kHz instead.

It's been a while since I messed with Foobar2000. Are the stock settings fine? I installed SoX resampler and have it set like in the picture, or should I have different settings to upsample and downsample.