Intersample-overs, software up-sampling and IMD

[TwoEyedJack]

Hello, I am trying to reach a conclusion regarding what settings should one use in windows audio settings (I understand that most of the things I am going to ask will probably be inaudible to me but i just want to know for the sake of knowledge). DAC's are know to produce intersample-overs and clipping when trying to upsample 44.1khz signals.
Do DAC's always try to upsample their input signal before converting it to analog? Is this just their operating principle while trying to interpolate the rest of the "analog samples" in-between the digital ones?
Does windows/software up sampling also suffer from this besides the rest of the IMD/Aliasing distortion it introduces? Apparently it used to although testing done here suggest things have gotten better.
Are the Intersample-overs affected by the sample rate of the input signal? Meaning are there fewer cases of clipping if the signal is getting up sampled from a higher base sample rate? If so would it beneficial to use software upsampling to bring it closer to the target upsample signal of the dac? This comes from the debate about whether you should go back and forth to change the sample rates of windows depended on the sample rate of the audio file you are playing. The advice given here was to just set it to 96khz since the windows upsampling was improved in the past years.
But I am skeptical about this since it's a given that you wont get any benefit from playing back higher sample rate files (in contrast to mixing an producing) , and usually-depending on your equipment- you might hurt performance . In my case while my dac is capable of 24/96khz I can hear loud intermodulation distortion if i use the 24/96 setting on windows and playback the test tones (supersonic tones).

Lastly do you get the same issues when downsampling? Going from a 96 source file (or a 48 movie) down to a 44.1 windows output, could this hurt performance -not loss of hypothetical supersonic fidelity- in a meaningful way like distortion, clipping or aliasing?

 

[BeerBear]

Regarding intersample overs: with a higher sample rate, the sample peaks are closer to the actual peaks of the audio signal. The screenshot here should make that clear. That's useful for estimating actual peak values. But to prevent IS overs you just lower the digital volume by a few dB.

Upsampling can be useful to avoid the "weak" anti-imaging filter of a DAC, which, at 44.1 or 48k, is close to the audible range. By upsampling to 88.2k or more, that filter is pushed up in the range where it's unlikely to affect the sound. Whether that results in an audible difference is debatable.

Windows resampling quality is probably fine. I posted some audio examples and you can do a blind test to see if you can hear a difference. And some media players (such as foobar2000) provide their own resampling, which you can use instead of leaving it to Windows.

 

[TwoEyedJack]

Regarding intersample overs: with a higher sample rate, the sample peaks are closer to the actual peaks of the audio signal. The screenshot here should make that clear. That's useful for estimating actual peak values. But to prevent IS overs you just lower the digital volume by a few dB.

Upsampling can be useful to avoid the "weak" anti-imaging filter of a DAC, which, at 44.1 or 48k, is close to the audible range. By upsampling to 88.2k or more, that filter is pushed up in the range where it's unlikely to affect the sound. Whether that results in an audible difference is debatable.

Windows resampling quality is probably fine. I posted some audio examples and you can do a blind test to see if you can hear a difference. And some media players (such as foobar2000) provide their own resampling, which you can use instead of leaving it to Windows.

So the DAC wont try to up sample the signal regardless of the window's sampling option? Meaning if my dac can do 24/192 and i feed it a 24/44.1 signal it wont up sample it and it will engage an imaging filter for 44.1 sample rate?
This also eliminates intrasample overs.

Is this considered a "weak" reconstruction filter?

I couldn't find a difference in the resampling test with abx, looking at your graphs it's obvious why . Is the performance of the resampler affected by the distance the base signal has from the target sample rate ? i.e can the performance deteriorate further if you insterad of 44 to 48 back to 44 you where to go to 96 or beyond?

 

[BeerBear]

So the DAC wont try to up sample the signal regardless of the window's sampling option? Meaning if my dac can do 24/192 and i feed it a 24/44.1 signal it wont up sample it and it will engage an imaging filter for 44.1 sample rate?

The DAC has its own internal oversampling, but that's besides the point. The key here is that if you run it at 44.1k, the filter is going to be at ~22k and if you run it at 192k, it's going to be at ~96k. (What the DAC can do doesn't really matter.)

This also eliminates intrasample overs.

Not sure what you mean by that. But to eliminate clipping you need to lower the volume.

Is this considered a "weak" reconstruction filter?

You could say that, seeing how it doesn't filter everything above 22.05k. It's not necessarily bad for a DAC, but a software upsampling filter can do a better job (at the cost of more resource usage and latency).

Is the performance of the resampler affected by the distance the base signal has from the target sample rate ? i.e can the performance deteriorate further if you insterad of 44 to 48 back to 44 you where to go to 96 or beyond?

I don't think it would change much.

 

[TwoEyedJack]

The DAC has its own internal oversampling, but that's besides the point. The key here is that if you run it at 44.1k, the filter is going to be at ~22k and if you run it at 192k, it's going to be at ~96k. (What the DAC can do doesn't really matter.)


Not sure what you mean by that. But to eliminate clipping you need to lower the volume.


You could say that, seeing how it doesn't filter everything above 22.05k. It's not necessarily bad for a DAC, but a software upsampling filter can do a better job (at the cost of more resource usage and latency).


I don't think it would change much.

Well, for me at least, up sampling doesn't seem to be worth it, especially since I get IMD when playing ultra sonic sounds. Don't know why even though my dac should be able to handle it.

Never the less, if you don't have high sample rate files I dont see the benefit of upsampling. Except if down sampling(a 48khz or the occasional high sample rate file) can be even more harmful?