"Best" Windows 10 Playback Sample Rate When Listening to Multiple Audio Sources at Once in Shared Mode

[gigatron]

Hello everyone - Long time reader and first time poster here.

I understand the idea of matching playback sample rate to source sample rate when listening to a single source, but what about when there are multiple sources and the Windows 10 resampler is doing sample rate conversions?

Is there a general consensus on what playback sample rate would be best while listening to multiple audio sources in shared mode (e.g. Spotify + Discord + Game audio)? Correct me if I'm wrong, but while choosing 44.1khz would be optimal for Spotify, it would likely be sacrificing Discord and Game audio to some degree since those are typically 48khz sources.

Since my DAC supports sample rates up to 384khz, would it be better to upsample to a higher rate? In my research so far, I have seen some evidence (e.g. https://www.audiosciencereview.com/...indows-resampling-not-actually-that-bad.9092/) that the Windows 10 resampler performs a cleaner conversion from 44.1khz to higher up multiples of 48khz rather than 48khz itself. Could this be a good middle-ground sort of solution, since a multiple of 48khz would work well for any 48khz source?

With the above in mind, it seems reasonable to conclude that choosing a sample rate of 384khz would optimize 48khz sources while also making best use of Windows 10 resampler for 44.1khz sources, and 352.8khz would do the reverse, in favor of 44.1khz. I could simply choose one of these top two sample rates depending on which sources I want to "optimize" for. Is there anything I'm missing here?

 

[infinitesymphony]

Food for thought: one of the takeaways from reading Benchmark's technical notes is that each DAC chip potentially has an operating frequency that yields the best SNR regardless of the source's sampling rate. In the case of the Analog Devices AD1853 DAC chip they chose for the DAC1, the highest performance was achieved when all material was asynchronously resampled to 110 KHz:

https://benchmarkmedia.com/blogs/application_notes/13127453-asynchronous-upsampling-to-110-khz

So, you may find that even though a DAC is capable of handling higher sampling rates, the performance level actually goes down as the sampling rate increases. It all depends on the chip and the implementation.

I would either not worry about this problem, or set everything to one sampling rate or the other (44.1 or 48) starting in Speakers Properties.

 

[gigatron]

Thanks for the reply infinite. That's a good point and I wasn't considering the possibility that DAC performance could potentially be degraded at higher sample rates. I'm wondering if there may be similar negative implications related to Windows 10 or application audio performance issues with higher sample rates as well.

I guess to put it simply, with all else being equal my primary focus was to determine whether there was any single playback sample rate which would have the least amount of issues (distortion, noise, artifacts, whatever) as a result of windows resampling multiple sources having different sample rates themselves.

I think what I might try as a next step is to see if I can create a setup which will allow me to perform the same experiment from the link in my original post and check to see how the latest Windows 10 version does when resampling from 44.1khz and 48khz to all other sample rates possible on my DAC. If I can get that working appropriately, I'll post the results here. I know there are some tricky caveats with Windows 10 though, one being the CAudioLimiter APO.

 

[infinitesymphony]

My thought was that if all applications were outputting the same sampling rate, no resampling would be required in shared mode. Perhaps someone with deeper understanding of the DS resampler can chime in.

 

[RayDunzl]

Is there a general consensus on what playback sample rate would be best while listening to multiple audio sources in shared mode (e.g. Spotify + Discord + Game audio)? Correct me if I'm wrong, but while choosing 44.1khz would be optimal for Spotify, it would likely be sacrificing Discord and Game audio to some degree since those are typically 48khz sources.

It's been observed:

https://www.audiosciencereview.com/...indows-resampling-not-actually-that-bad.9092/

https://archimago.blogspot.com/2015/11/measurements-windows-10-audio-stack.html

---

In my case, my DAC itself resamples (ASRC) whatever rate it receives to 211kHz.

So... I'm stuck with that in the end, for better or worse, till death do we part.

 

[bennetng]

Hello everyone - Long time reader and first time poster here.

I understand the idea of matching playback sample rate to source sample rate when listening to a single source, but what about when there are multiple sources and the Windows 10 resampler is doing sample rate conversions?

Is there a general consensus on what playback sample rate would be best while listening to multiple audio sources in shared mode (e.g. Spotify + Discord + Game audio)? Correct me if I'm wrong, but while choosing 44.1khz would be optimal for Spotify, it would likely be sacrificing Discord and Game audio to some degree since those are typically 48khz sources.

Since my DAC supports sample rates up to 384khz, would it be better to upsample to a higher rate? In my research so far, I have seen some evidence (e.g. https://www.audiosciencereview.com/...indows-resampling-not-actually-that-bad.9092/) that the Windows 10 resampler performs a cleaner conversion from 44.1khz to higher up multiples of 48khz rather than 48khz itself. Could this be a good middle-ground sort of solution, since a multiple of 48khz would work well for any 48khz source?

With the above in mind, it seems reasonable to conclude that choosing a sample rate of 384khz would optimize 48khz sources while also making best use of Windows 10 resampler for 44.1khz sources, and 352.8khz would do the reverse, in favor of 44.1khz. I could simply choose one of these top two sample rates depending on which sources I want to "optimize" for. Is there anything I'm missing here?

ASRC on the same PCB as the DAC chip, or within the DAC chip (e.g. ESS) are often used to reduce jitter. ASRC produces both imaging and aliasing artifacts, but when the quality is high enough, both jitter and resampling artifacts can be reduced to negligible levels.

OS resamplers are used to combine different audio sources with different sample rates, and convert sources with unsupported sample rates to the supported ones. So ASRC and OS resamplers have completely different purposes.

Games and other interactive apps often use lossy codecs, sample-based synthesis, and non-sample-based synthesis. These processes often create artifacts way beyond what the OS resampler does and it is a waste of time to worry about resampler quality of these applications. Some games allow configurable output sample rates, but it only means the result of the game's audio output can be set to a certain sample rate, it doesn't change all the "dirty" audio synthesis within the game engine. So in this case, if you know a specific game has a specific output sample rate, matching it with the OS sample rate may achieve better latency. Also, make sure the output level of apps/games using shared output is low enough to avoid triggering CAudioLimiter. 0.2dB in my illustration is not always enough, it depends on the sources.
https://hydrogenaud.io/index.php?topic=104051.msg854152#msg854152

In short, only worry about this issue when playing "pristine 24-bit lossless audiophile grade demo tracks", but in this case self-proclaimed audiophiles will use WASAPI exclusive mode or ASIO anyway.

 

[abdo123]

Games and other interactive apps often use lossy codecs, sample-based synthesis, and non-sample-based synthesis.

From the few games I extracted they mostly use .WAV depending on the game engine, they don't pay royalties for codecs.

 

[bennetng]

From the few games I extracted they mostly use .WAV depending on the game engine, they don't pay royalties for codecs.

Codecs like vorbis and opus are also free. Also, mp3 patents are expired and also free now. Some games use ADPCM (lossy, but can also have .wav extension) as well. Developers can also use third party libraries so the usable formats are not strictly limited by the game engine. The fact is games are interactive and they usually apply effects interactively rather than simply mixing sounds together. So even when the source audio files are lossless, unless you rip and play them with an external audio player, the files can still subject to additional processing.

Actually 30+ years ago, NEC PC-Engine/TG-16 CD-Rom games use CDDA audio, it is even possible to play those CDs on a traditional CD player. So of course, audio data in games are not necessarily lossy.

 

[outerspace]

Also, make sure the output level of apps/games using shared output is low enough to avoid triggering CAudioLimiter. 0.2dB in my illustration is not always enough, it depends on the sources.

Maybe you know, is -1 dB preamplification in Equalizer APO do the trick for all apps?

 

[bennetng]

Maybe you know, is -1 dB preamplification in Equalizer APO do the trick for all apps?

Someone asked a similar question. The most reliable way is to try and listen for yourself. Don' completely rely on other people's test results.
https://www.audiosciencereview.com/...indows-audio-quality-debate.19438/post-643490