Sampling Rate Theory As It Pertains to the XMC-1 and Dirac

[chipless]

I went down a DEEP rabbit hole this evening, and as usual I'm left with more questions than answers.

I'm in the process of designing a dedicated computer for audio playback, and it turns out there are lots of options for media players. This led me to reading all about sampling rates/upsampling. From my understanding, there is some consensus that upsampling via the computer/media player in order to match the DACs highest accepted sampling rate is a good thing given an acceptable method/algorithm of doing so.

My question is relatively "simple": In the case of the XMC-1 using Dirac, if I'm going to upsample from the computer, should I do so to 48kHz (assuming mostly 44.1kHz source material)? Or, should I do so to 192kHz?

As I understand it, Dirac downsamples to 48kHz if the incoming sample rate is higher than that, so it seems that it wouldn't make sense to "feed" the XMC-1 a rate higher than 48kHz even though it can accept 192kHz. Then again, I have no knowledge as to Dirac's relationship with the DACs of the XMC-1, and perhaps Dirac's downsampling from a 192kHz source would be better than having an incoming source of 48kHz to begin with.

Please feel free to set me straight. I've read so many engineering white papers that I might vomit.

Thank you so much for your knowledge.

[wilburthegoose]

Easy - compare the audio from non-Dirac to Dirac and pick a winner.

[Chuck Elliot]

I don't think it is just Dirac.

I believe all processing in the DSP is done at 48kHz!

Could be wrong, but that was my understanding!

[chipless]

Dec 26, 2016 9:36:01 GMT -5 wilburthegoose said:

Easy - compare the audio from non-Dirac to Dirac and pick a winner.

Thanks. I will definitely compare feeding the XMC-1 a fixed 48kHz signal to feeding it a fixed 192kHz signal to see if I detect any differences. I will also try the native 44.1kHz of the source material. I did want to clarify that I will be using Dirac no matter what, and my inquiry is more of a theoretical "which is best" based on upsampling via PC and the interaction of the digital signal with the XMC-1 and Dirac. I am hoping to learn more about this from those that have knowledge of both digital signals and the XMC-1 itself.

[Lsc]

In my experience, I felt things sounded best when I playback the music in its original sample rate. Upsampling tends to dull the sound a bit so only you can decide if that's what you prefer.

Dirac is always on for me as well.

[fbczar]

Dec 26, 2016 10:26:02 GMT -5 chipless said:

Dec 26, 2016 9:36:01 GMT -5 wilburthegoose said:

Easy - compare the audio from non-Dirac to Dirac and pick a winner.

Thanks. I will definitely compare feeding the XMC-1 a fixed 48kHz signal to feeding it a fixed 192kHz signal to see if I detect any differences. I will also try the native 44.1kHz of the source material. I did want to clarify that I will be using Dirac no matter what, and my inquiry is more of a theoretical "which is best" based on upsampling via PC and the interaction of the digital signal with the XMC-1 and Dirac. I am hoping to learn more about this from those that have knowledge of both digital signals and the XMC-1 itself.

Wilberthegoose has the right idea. In my experience using the XMC-1 with the Audirvana software upsampling along with the associated setting modifications followed by Dirac Full with a custom room curve is the way to go. That said, I am pleased the upcoming XMC-2 will have Dirac 96/24 and the RMC-1 will have Dirac with 192/24 capability. That, and the RMC-1's DSD capability are the reason I will upgrade.

[igorzep]

Considering the XMC-1 have ASRC in it any incoming content, including the same (seemingly, but actually a little bit different) sample rate will be resampled by ASRC anyway. But if you do 192kHz (or probably even better 188.2 kHz) quality upsampling externally then the job of quality resampling it to 48kHz for the DSP will be easier task for the ASRC chip, I think. But it might depend on on the ASRC chip. Anyway some odd, not directly related to the input frequency instead of 48 kHz would be better for internal processing, but it's done how it's done.