[Niouke]

Gents,

I've read countless times that reading music files via ASIO improves the sound quality as the music is sent "bit-perfect" to the DAC. That statement does not match my experience with rendering amateur music projects, is there any truth in it? Thanks.

 

[Roseval]

All OS work more or less the same audio wise.
They support multiple streams hence there must be a mixer.
As you cannot mix streams of different sample rate, all streams must be converted to the same sample rate as set in the audio panel.
Mixing is calculating so converting to float, calculate, add dither, convert back to integer.
I do think it is possible to hear the dither when using a 16 bit DAC, in case of a 24 bit it is way down the noise floor of the DAC.

Likewise sampler rate conversion is a calculation.
The WIN src has a bad reputation, at least there is measurable distortion.
http://archimago.blogspot.nl/2015/11/measurements-windows-10-audio-stack.html
If you bypass the WIN audio stack you won’t have these problems.

 

[Niouke]

I agree, but...

The link you posted is testing upsampling, and I'm not touching that. I use redbook audio settings and redbook files. And I'm also suspicious of the conclusions of the article, the rollof they claim measuring would be obvious to the listener (althougt the graph is not very clear avec 10Khz), I guess I'll have to try it!

 

[castleofargh]

asio drivers are device specific, some are cool, some a buggy crap where you can't even find out where to change the settings you need. some result in the exact same thing as using wasapi, some come with a different streaming solution for the device, being a different driver. there cannot be some rule about asio always being better.
now if the asio settings are set for lowest latency(as is historically the use for pro gears and asio), and you start throwing VSTs and heaving processing at it, you can end up with some buffering fun and a sound close to

but fooling around with sample size can solve or ruin anything, ASIO isn't special that way.

sometimes ASIO drivers will also oversample everything to the max sample rate the device can handle. that could probably lead to small audible differences in some cases compared to good guy wasapi, that will stick to whatever it is you had before. once everything is set up the same way, chances are that the remaining differences will be limited to the least significant bits and nobody should care.

otherwise when we're lazy and we have alternatives, the cool thing to do is to use what works and leave it at that ^_^ (this advice is brought to you by "captain obvious").

 

[Niouke]

I use ASIO with the minimum buffer possible in order to have a decent delay on my MIDI devices and I know how it sound when the buffer underruns , but I had no idea it could be have an impact on bit depth and sample rate. I didn't doesn't change much for me as my main source of music is spotify, and the app doesn't let you use ASIO. I was wondering about the theoretical aspects tho.

 

[Roseval]

Strictly spoken ASIO (or its Win equivalent WASAPI) have nothing to do with “bit perfect”.
Bit perfect is defined as delivering the audio to the DAC without any change in channels, samples or sample rate.

All ASIO & co do is delivering the output of a media player unaltered to an audio device.
If the media player applies DSP e.g. volume control, you are no longer “bit perfect”.
Likewise an audio device can do DSP like upsampling, volume control, etc.

The protocol itself is transparent and does what almost all protocols do, transferring data from one device to another without alteration.
Like all protocols it is not in control of what happens up- or down-stream