• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

ASIO, WASAPI, Direct Sound... is there any difference in sound quality?

Monstieur

Member
Joined
Feb 3, 2020
Messages
89
Likes
31
Digital clipping with EQ and even without EQ may be more common than people think. I ran a 0 dB sweep on my system (WASAPI Exclusive with Foobar200) and was surprised to hear "hard clipping" when using room EQ which was supposed to allow for that, I had to back off an additional 4 dB to eliminated it. Even without EQ there was some digital clipping with the 0 dB sweep. I am not sure how accurate of a test a 0 dB sweep is... it might be overkill? I do notice that some songs that previously had come "clicks" in them during loud passages no longer do after the -4 dB adjustment. What is the "best" way to check your system digital clipping?
Technically, tracks can clip by over 1000 dB when decoded as they are floating point. Run gain analysis on all of your albums and adjust them accordingly. Many albums from the iTunes Store and Apple Music clip by several dB. The Mastered for iTunes certification guarantees no digital clipping.

For other content you have to actively monitor your system. This may be impossible on Windows with 24-bit Integer interfaces and CAudioLimiter as it will auto-limit below 0 dBFS. On macOS the full playback chain is 32-bit float with some interfaces like RME's, so you can see the output exceeding 0 dBFS in your playback device if it has a meter.
 
Last edited:

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
1,445
Likes
2,075
Location
San Diego
Technically, tracks can clip by over 1000 dB when decoded. Run gain analysis on all of your albums and adjust them accordingly.
Thank you for reply but I am not sure I understand. What does "clip by 1000 dB" mean? Do you mean that some music can have "+1000 dB" peaks when decoded? When you say run analysis on all albums and adjust accordingly what does that mean? I have never see anything "over" 0dB in for example "DR Meter" . Is there a "safety margin" of 1 or 2 dB that should be allowed for i.e. lower 0 dB peaks to -2 dB? Thanks.
 

Monstieur

Member
Joined
Feb 3, 2020
Messages
89
Likes
31
Thank you for reply but I am not sure I understand. What does "clip by 1000 dB" mean? Do you mean that some music can have "+1000 dB" peaks when decoded? When you say run analysis on all albums and adjust accordingly what does that mean? I have never see anything "over" 0dB in for example "DR Meter" . Is there a "safety margin" of 1 or 2 dB that should be allowed for i.e. lower 0 dB peaks to -2 dB? Thanks.
Integer audio ranges from Negative Infinity to 0 dBFS. Floating point audio can losslessly store positive values over 0 dBFS. Audio is typically mastered in floating point and exported directly to MP3 or AAC which can store these positive values without clipping. The clipping occurs only when converting the output to integer due to the output format on the playback device.

The mastering engineers are also at fault for making the floating point audio too hot. Before they export the audio, they should check that no peaks exceed 0 dBFS even in floating point. Apple's workflow does this when mastering for iTunes.
 

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
1,445
Likes
2,075
Location
San Diego
Integer audio ranges from Negative Infinity to 0 dBFS. Floating point audio can losslessly store positive values over 0 dBFS. Audio is typically mastered in floating point and exported directly to MP3 or AAC which can store these positive values without clipping. The clipping occurs only when converting the output to integer due to the output format on the playback device.

The mastering engineers are also at fault for making the floating point audio too hot. Before they export the audio, they should check that no peaks exceed 0 dBFS even in floating point. Apple's workflow does this when mastering for iTunes.
So if my music library is primarily ripped CD's (to WAV/Flac) created on a Windows machines do I need to be concerned with their gain? Seems like they should be limited to 0 dB?
 

Monstieur

Member
Joined
Feb 3, 2020
Messages
89
Likes
31
So if my music library is primarily ripped CD's (to WAV/Flac) created on a Windows machines do I need to be concerned with their gain? Seems like they should be limited to 0 dB?
They will be an exact representation of the disc. There may be intersample overs in the DAC if the samples are close to 0 dBFS, but most DACs have an internal headroom of a few dB to handle this.

However, if the disc has intersample overs, it's likely the audio is also clipped or compressed. Since the disc is integer, the clipped samples are lost permanently. This is one way streaming services are superior to CDs - the clipped audio can be recovered losslessly by applying negative gain since it's floating point. The distortion from clipping is far worse than the compression to AAC.
 
Last edited:

BeerBear

Active Member
Joined
Mar 9, 2020
Messages
121
Likes
121
The mastering engineers are also at fault for making the floating point audio too hot. Before they export the audio, they should check that no peaks exceed 0 dBFS even in floating point.
Even if the original signal is under 0 dBFS, converting it to MP3 can make it go above 0.
 

Katji

Major Contributor
Joined
Sep 26, 2017
Messages
3,002
Likes
2,212
How does it relate to the clipping values shown in audio file analysis programs like Similarity and Fakin' The Funk? (Or vice versa.)

I think it's the main criterion they use in rating a track OK / not OK. I was looking at some yesterday - a specific producer, mixes including tracks from wherever, and [if I remember right] the incidence of clipping stopped when it got to some later version of Logic Pro. (I didn't save it, although that's where i started, seeing that I'd added a lot more since I last checked them. ...It takes quite long, I wouldn't like to be doing it with anything less than this old i7.)
 

Monstieur

Member
Joined
Feb 3, 2020
Messages
89
Likes
31
Even if the original signal is under 0 dBFS, converting it to MP3 can make it go above 0.
Yeah, but those errors are a fraction of a dB and are trivially solved with a small negative gain. The primary cause is the floating point master being too hot and already above 0 dB by an unknown amount per track.
 
Last edited:

Monstieur

Member
Joined
Feb 3, 2020
Messages
89
Likes
31
How does it relate to the clipping values shown in audio file analysis programs like Similarity and Fakin' The Funk? (Or vice versa.)
Opening the file in an audio editor will show the floating point values above 0 dBFS. These values would become clipping when converted to integer for playback, unless you apply negative gain.
 

Katji

Major Contributor
Joined
Sep 26, 2017
Messages
3,002
Likes
2,212
...Waiting for Similarity....seems to be processing the entire folder again...


1642619904946.png
 

Monstieur

Member
Joined
Feb 3, 2020
Messages
89
Likes
31
At which point does it happen? ...The Windows [/whatever] output device, or the DAC...?
On Windows, each application is responsible for resampling audio to match the WASAPI Shared mode output format or any one of the formats supported by the hardware in WASAPI Exclusive mode. The quality of the resampling is now application-specific as it's not done by the Windows mixer. ReplayGain etc. must be applied by the application to avoid clipping.

On macOS, CoreAudio accepts 32-bit float from applications and resamples at the end. If the playback device supports 32-bit float then positive values above 0 dBFS are passed to the device. You can then apply negative gain in the hardware if it has such controls. With 32-bit float output, reducing the volume in macOS is sufficient to prevent clipping across all applications.
 
Last edited:

Rottmannash

Major Contributor
Forum Donor
Joined
Nov 11, 2020
Messages
2,037
Likes
1,657
Location
Nashville
The driver package is actually two drivers - WDM and ASIO. The WDM driver shows the device in Sound settings and the mixer. The ASIO driver is visible only in ASIO-aware applications. The drivers can support simultaneous playback of ASIO and WASAPI (Exclusive or Shared) if designed to do so.

WASAPI Exclusive silences only other WASAPI streams when in use - it can still be mixed with native ASIO. This does not apply to ASIO4ALL as it uses WASAPI Exclusive or Kernel Streaming as the backend.

If the WDM driver uses hardware volume control, there may be odd interactions like the ASIO stream being muted by the Windows volume control, but the ASIO stream is still being mixed in. On Windows 11, the mute button or 0 volume does not silence WASAPI Exclusive streams anymore, at least on devices with software volume control.
Audirvana now offers "kernel streaming" along with WASAPI and ASIO. What is the advantage of using that over WASAPI exclusive?
 

Monstieur

Member
Joined
Feb 3, 2020
Messages
89
Likes
31
Audirvana now offers "kernel streaming" along with WASAPI and ASIO. What is the advantage of using that over WASAPI exclusive?
None in practice. It’s a lower level API and is less consistent from an application perspective due to driver variations. Before WASAPI supported low latency audio, Kernel Streaming provided the lowest latency and exclusive access.
 

Rottmannash

Major Contributor
Forum Donor
Joined
Nov 11, 2020
Messages
2,037
Likes
1,657
Location
Nashville
None in practice. It’s a lower level API and is less consistent from an application perspective due to driver variations. Before WASAPI supported low latency audio, Kernel Streaming provided the lowest latency and exclusive access.
Interesting...they intimated it was a sonic improvement when it was announced in their "Studio" version.
 

Trell

Major Contributor
Joined
May 13, 2021
Messages
1,403
Likes
1,456
Interesting...they intimated it was a sonic improvement when it was announced in their "Studio" version.

Does that mean that their WASAPI and ASIO implementation is somehow defective? :oops:
 

Rottmannash

Major Contributor
Forum Donor
Joined
Nov 11, 2020
Messages
2,037
Likes
1,657
Location
Nashville
Does that mean that their WASAPI and ASIO implementation is somehow defective? :oops:
No clue-they all sound the same but each uses varying levels of CPU load and RAM being utilized with a display showing in real time the usage.
 

Trell

Major Contributor
Joined
May 13, 2021
Messages
1,403
Likes
1,456
No clue-they all sound the same but each uses varying levels of CPU load and RAM being utilized with a display showing in real time the usage.

My comment was tongue-in-cheek as I find it very unlikely that their “intimation” of sonic improvements is true, at least in general, unless they have bad implementation of ASIO and WASAPI. If there are edge cases they should be transparent about them.
 

Rottmannash

Major Contributor
Forum Donor
Joined
Nov 11, 2020
Messages
2,037
Likes
1,657
Location
Nashville
My comment was tongue-in-cheek as I find it very unlikely that their “intimation” of sonic improvements is true, at least in general. If there are edge cases they should be transparent about them.
I was literally asking if it made a sonic difference. I know nothing about CPU usage and RAM usage so was truly curious.
 

MIOM

Member
Joined
Jun 26, 2022
Messages
7
Likes
3
Hi, new member here :)

Anyway, I registered here out of curiosity of the post below, as I'm not sure what DJuan2022 means....
UPDATE:
RadioBoss tech support informed me that Windows Mixer doesn't support ASIO, that's why you don't see your app in the mixer.
as I can definitively say Windows Mixer most definitely supports ASIO and has done so since at least 2016 when I bought my Teac UD-503 USB DAC. It's also supported in Windows 11 (shown below) as has been in Windows 10. Teac ASIO driver used in JRiver Media Center...

Teac ASIO driver in JRMC.jpg


and seen and recognized in Windows 11 Sound Mixer (was also there in Windows 10)

Windows 11 Sound Mixer.jpg


Also, in "Sound"...

ASIO drivers in sound mixer.jpg


Apologies if I misunderstand and ask for clarification. Thanks

All that said, I've tried both ASIO and WASPI, but to my ears AISO sounds a tad smoother and less "digital" to me. Probably because I play through USB.
 
Last edited:
Top Bottom