Ending the Windows Audio Quality Debate

[DDF]

I'd like to address the several risks raised against using the digital loss.

Intersample Overs
Benchmark identified this as a generic problem within the ESS DAC chips, which are nearly ubiquitous. AKM provides only 2 dB headroom. Only Benchmark and RME have stated that they have a workaround. If there are others, it would be a useful list to compile. Go for it guys, we'd be thankful. I also don't think there are many 30 year old CD players that can stream.

Peak Music Levels as an Indication of Risk
Digital music is mastered to be very near 0dBFS, either exceeding CAudioLimiter (-0.12 dBFS) or if not, providing no where near enough headroom for intersample overs. Examples here, here, here, mastering advice to exceed CAudioLimiter here, "Many mastering engineers choose either -.1 dBFS to -.3 dBFS as the level at which the highest peaks should remain at or below" here....etc

System Gain and Noise
Using 4 dB digital loss could only be a system gain issue if the pre-amp gain needs to be operated within 4 dB of absolute maximum volume, without the 4dB digital loss present. That would be quite unusual. If that happens the unfortunate solution short of buying another pre-amp with gain properly matched to the system's needs is to remove as much digital loss as needed and then live with the clipping risks clearly laid out in the OP. For the other vast majority of us, its a non issue

Using 4 dB digital loss could only be a system noise issue if the pre-amp is so noisy and its noise is dominated by input stage noise or the DAC is so noisy that the extra 4 dB analog pre-amp gain needed to over-ride the 4 dB digital loss causes the extra preamp analog noise to suddenly exceed the audibility threshold in the room, where it didn't before. That would be very rare, requiring extremely efficient speakers, extremely quiet amplifiers and a meh pre-amp or DAC just on the edge of the needs of the system. The solution would be again to remove the digital loss and live with the risk, or to use a pre-amp properly matched to the high efficiency speakers. For the other vast_vast majority of us, its a non issue.

As for the supposed added noise that's created due to loss of less than 1 bit, I outlined the solution in the OP, run the DAC and Windows at 24 bits. Its a non issue.

 

[Vasr]

The solution would be again to remove the digital loss and live with the risk, or to use a pre-amp properly matched to the high efficiency speakers. For the other vast_vast majority of us, its a non issue.

If you were to properly quantify that risk, you may find that it is even more of a non-issue.

You want to upload a music sample here that when played without any attenuation at their listening volume, will make listeners gasp with horror on hearing the effects of inter-sample peaks? That would indeed be a convincing argument.

 

[Vasr]

I understand ASIO to be a sound option in audio interfaces-- is there a way to use it as a global system choice?

No. It is just a direct API to the sound device that is devoid of any mixing or resampling capability in the path (except for some fixed 44.1khz resampling in some implementations). So, players that can support ASIO need to be capable of either resampling/mixing on their own or the end device should be so capable or require the users to manually ensure the correct usage of sampling rates and exclusive usage.

As an example, web browsers (and even players like Amazon Music) are not able to do this. They can only send to audio devices in shared mode.

If you only used players with ASIO capability then you would point all of them to use the ASIO access to he sound card, nothing else needs to be done. Parallel access to the same device would fail when one app is using it.

 

[BrokenEnglishGuy]

is anyone hearing a improvement? im using my all day apo with music games and everything and i don't have any problem
i turn off the '' use original apo '' and im not sure what i did, but i don't know if i hear anything new or different xD
if i turn off the use original apo, im in bit perfect?

 

[lizhuoyin]

No. It is just a direct API to the sound device that is devoid of any mixing or resampling capability in the path (except for some fixed 44.1khz resampling in some implementations). So, players that can support ASIO need to be capable of either resampling/mixing on their own or the end device should be so capable or require the users to manually ensure the correct usage of sampling rates and exclusive usage.

As an example, web browsers (and even players like Amazon Music) are not able to do this. They can only send to audio devices in shared mode.

If you only used players with ASIO capability then you would point all of them to use the ASIO access to he sound card, nothing else needs to be done. Parallel access to the same device would fail when one app is using it.

Amazon Music HD in Windows support s exclusive mode.

 

[DDF]

If you were to properly quantify that risk, you may find that it is even more of a non-issue.

You want to upload a music sample here that when played without any attenuation at their listening volume, will make listeners gasp with horror on hearing the effects of inter-sample peaks? That would indeed be a convincing argument.

I thought I did a pretty good job of framing the risk in the OP, clearly stating that its possible, and I heard and measured issues but YMMV depending on the system (I stated 2 of my 3 systems had no issues). It's intent was clear, to ease the minds of those that think Windows audio isn't up to snuff so that they can use it and bypass the risks and save money, and to provide objective proof of the issues and how to get around them to placate the doubters who think Windows audio problems are universally figments of the imagination. I also provided links to more material that explain when these risks evidence themselves and how, and practices in the recording industry that exacerbate the risk. People can decide for themselves, but the data is the data. No one including you has a study of the extent of the risk but at least I identified the mechanisms and resolutions..

It's fair to highlight the risk again as you are doing, but please read my replies. If you feel these aren't a risk, great. However you're doing no one any help by blanket saying they don't exist while blindly ignoring the data presented.

Many of these points were far from common knowledge, my intent was just to share this. It's not clear what your intent is? You aren't addressing anything I wrote concretely, and some of your responses are now bordering on trolling.

 

[bennetng]

is anyone hearing a improvement? im using my all day apo with music games and everything and i don't have any problem
i turn off the '' use original apo '' and im not sure what i did, but i don't know if i hear anything new or different xD
if i turn off the use original apo, im in bit perfect?
View attachment 107242

Well, within the mainstream consumer formats, it is unsurprising that there is no audible difference, as the limiter is not always activated, and when activated it could just be a brief moment. However, for a situation like this:

https://hydrogenaud.io/index.php?topic=104051.msg854152#msg854152

I knew the limiter is there a decade ago, not because someone discovered it in some Microsoft doc, or in some measurements, but because I heard it, clearly, right after I moved from Windows XP to 7. So things like 0.2dB or 4dB solved nothing.

Also, for interactive applications (e.g. games), the audio engines used by games are not too different from a DAW, like this:
https://www.criware.com/en/products/adx2.html
In a DAW, users move faders and knobs around, creating automation event to change the audio data over time. In games, for example, when the player is physically close to something, (e.g. someone who talks), the voice is louder and brighter. At a distance, or when there are obstacles in-between, the voice is weaker and filtered in some ways. Things like background music and other sound effects are separately mixed. The player controls the automation events interactively. So games also have options to adjust these things as the mixed output is "clippable". People don't control the volume of a DAW outside of the DAW as well. Also, being interactive, games require low latency as well. So unless there is an absolute need to EQ these apps, if the apps themselves offer internal adjustment, adjusting volume within the app will achieve the lowest possible latency.

Digital music is mastered to be very near 0dBFS, either exceeding CAudioLimiter (-0.12 dBFS) or if not, providing no where near enough headroom for intersample overs. Examples here, here, here, mastering advice to exceed CAudioLimiter here, "Many mastering engineers choose either -.1 dBFS to -.3 dBFS as the level at which the highest peaks should remain at or below" here....etc

For music players, if the concern is not about losing DNR, volume management within the music player also provides a more consistent playback level without using any dynamic compression, due to the fact that the overall level and highest peak of individual audio files and albums are known before playback. So either 0.2dB, 4dB at the playback side, or whatever "best practice" in music production would become irrelevant. If the mastered music have heavy dynamic processing, even if the level is lowered, the sound quality will still be ruined. Like this:
https://www.audiosciencereview.com/...lity-in-windows-using-wasapi.5272/post-117708
https://www.audiosciencereview.com/...ot-a-psychoacoustic-process.11169/post-327625

Instead of pointing to some "best practice" mastering tutorials, it is possible to scan the audio files you own for potential intersample overs, and absolute overs in formats that decodes to floating point. As mentioned previously, this article shows how to do this, with some real-life examples.
http://archimago.blogspot.com/2019/06/guest-post-why-we-should-use-software.html

As a matter of fact, it is impossible for an album released in for example, pre-2005, to address CAudioLimiter. Also, how the audio files are being mastered are not controllable by end users, so these "best practices" have little meaning to end user anyway, they have what they have.

 

[cursive]

Thanks for posting this DDF, very informative thread. I do have a couple questions if anyone can elaborate.

1. I noticed my "use original apo" options are both greyed out, is this due to using an external USB dac?
2. I am using the equalizer in APO to equalize my headphones, currently I'm at -5.5 on the preamp as this is the level I'm boosting a specific frequency. Does this mean I should lower the preamp an additional amount to compensate for the limiter?

 

[diyaudnut]

Amazon Music HD in Windows support s exclusive mode.

It is not true exclusive. The music stream is resampled to the shared rate setting in windows.
As far I know, Amazon HD desktop app has no way of passing the stream untouched to the ASIO driver of your DAC.

 

[Mnyb]

This tread got me really interested in EAPO I'm in for new computer so i might move my headphone rig to the computer , so that i actually can use EQ for the phones , I'll just get an USB DAC.

For the music streamer i use LMS server with a PiCore player with a digital output board, this has everything resampled to 24/96 with SoX and the level is set to -3dB ! The (squeezebox streamers always outputs 24bit btw regardless of source file) then my old MeridianGG8J pre processer does not have to do the up sampling, the processer then does its room correction 5.1 upmix and slight treble tilt .
This is then feed digitally to the DSP5200 speakers (LCR) and DSP3000 (RrLr) via MHR link ( and an analog Rhythmic 15" sub ). The DSP speakers have their internal digital volume control controlled by the processor . so volume is done in the speakers.

So in my system for typical 16 bit material there is no loss , it's basically the same as different volume control positions .

For 24bit material you could say that i lose a bit or two out of 24 (do i have 22bit now) . but it's just random noise anyway not many recordings challenge even 16 bit..
And no practical DAC ever has real 24bit 144dB sinad . the best ones are practically 19-21 bits .

I actually have no clue to what DAC there is inside my active DSP speakers . The xover is digital so there is one per driver .

 

[ninetylol]

Would setting Windows volume to -0.3dB have the same effect as setting it in EqualizerApo?

 

[bennetng]

Would setting Windows volume to -0.3dB have the same effect as setting it in EqualizerApo?
View attachment 107348

https://www.audiosciencereview.com/...y-of-software-volume-control.5922/post-172865
In your case, try the volume control in both exclusive and shared modes, of course.

 

[ninetylol]

https://www.audiosciencereview.com/...y-of-software-volume-control.5922/post-172865
In your case, try the volume control in both exclusive and shared modes, of course.

Thanks but I cant use windows volume control when in WASAPI mode so i cant test if the distortion is still there. Distortion is already gone at -0 dBFS when not in WASAPI mode.

 

[bennetng]

Thanks but I cant use windows volume control when in WASAPI mode so i cant test if the distortion is still there.

Then use a volume control that works with WASAPI exclusive mode, the one within the music player most likely can do this. Or don't use WASAPI exclusive mode when testing that volume control.

 

[ninetylol]

Then use a volume control that works with WASAPI exclusive mode. The one within the music player most likely can do this.

Yeh that works of course (foobar) but thats not what im trying to test, as i want to test windows volume control.

But i guess windows volume control is working then even at 100 (-0 dBFS) cause its not distorting like WASAPI is.

 

[bennetng]

Yeh that works of course (foobar) but thats not what im trying to test, as i want to test windows volume control.

That's because the Windows mixer volume control is bypased when using WASAPI exclusive mode, and your audio device driver does not have any other usable system-wide volume control. If you want to test the Windows volume control, then don't use WASAPI exclusive mode.

 

[ninetylol]

That's because the Windows mixer volume control is bypased when using WASAPI exclusive mode, and your audio device driver does not have any other usable system-wide volume control. If you want to test the Windows volume control, then don't use WASAPI exclusive mode.

I understand what you are saying. The problem is how can i test if windows volume achieves the same effect as EAPO when i cant change windows volume with your method? Theres no way to test it this way.

What i can say though is that Peace shows clipping in its UI and of course file "a" shows clipping there. And it still shows clipping even when changing windows volume control. See below:

When using preamp in EAPO the clipping dissappears so i guess I will just use that.

By the way you can also test the artist Scarlxrd on spotify since like all his songs are clipping because of the most ****** mastering ever lol.

 

[bennetng]

Theres no way to test it this way

Use your ears. Since it is a rather extreme example (12dB over), if it still sounds fine to you, then there is little chance that your approach, in your specific setup, is audibly problematic, to your ears.

 

[ninetylol]

Use your ears. Since it is a rather extreme example (12dB over), if it still sounds fine to you, then there is little chance that your approach, in your specific setup, is audibly problematic, to your ears.

Even at 100% windows volume theres no audible distortion (only when I put WASAPI on its there). Probably because CAudioLimiter is already active? Still shows as clipping in Peace though.

 

[bennetng]

Even at 100% windows volume theres no audible distortion (only when I put WASAPI on its there). Probably because CAudioLimiter is already active? Still shows as clipping in Peace though.

If you ask a question like this, then CAudioLimiter is either inactive, or not as bad as others described. I can clearly hear the limiting in my specific setup. If you insist, the only way to verify it is to install EQ APO and try it yourself.