Downsampling as late as possible in signal path

Support
1

If I look at my signal path, downsampling from 96kHz to 48kHz is done before other filters (crossfeed and convolution). I wonder: since downsampling essentially trhows away information, shouldn’t this be done as the final step? Just like the conversion from 64bit to 24bit is done last?!

E8D4BF4E-C08C-4A45-A62D-19F37F708486
E8D4BF4E-C08C-4A45-A62D-19F37F708486.jpg2732×2048 855 KB

3

Hello @DeLub,

I have forwarded this question to the development team, I will be sure to let you know their thoughts once they have had a chance to take a look.

-John

4

Yes, this is intentional. There are several potential orderings here depending on whether one is upsampling, downsampling, dealing with DSD, etc. In this case, it doesn’t really make sense processing a region of frequency domain that we know will be removed later–we would be doubling the CPU usage of the crossfeed and convolution steps for no real benefit, so the downsampling happens relatively early.

5

Thanks for the answer. Makes perfect sense.

6

Is the downsampling to the max sample frequency of the output device? Or some other criterion?

7

By default Roon downsamples to the max rate of the device in order to maintain compatibility–it’s important to us that all content can play on all devices so people don’t feel like they have limited access to high res just because the zone they are playing on is not capable.

There are also configuration options that allow people to configure sample rate conversion rules manually, so it’s also possible that Roon would downsample because of how that configuration was done.

8

Thanks. So, in the case that DeLub describes in the thread starter, it’s downsampling to 48kHz because that’s the max rate of the output device?

I’m trying to understand whether, when crosstalk and convolution filters are applied, is it still the output device that decides the downsampling frequency? Or is there some property of the filters themselves that decides that?

Edit: I mean crossfeed, not crosstalk, sorry.

9

I’m trying to understand whether, when crosstalk and convolution filters are applied, is it still the output device that decides the downsampling frequency? Or is there some property of the filters themselves that decides that?

Output device + DSP settings only. The filters do not impose any limitations.

1 Like
10

This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.