Hi guys,
In simple terms, when Roon is downsampling from 24/192 to 24/96 what is happening?
Is the sound that is above 48kHz essentially cut off? And is anything happening below 48kHz?
Yes I know we probably can’t hear above 48kHz (or 20kHz) but that’s not what this question is asking so let’s not go there please 
Or have I got this downsampling method wrong? Specifically for 24/192 down to 24/96.
Many thanks in advance
Hi, it is to do with digital data rates. Nothing to do with frequency of sound. It would be down sampling to meet the requirements of a DAC or endpoint that only works up to 96k.
In the simplest terms Roon is throwing away samples to go from 192Ks/sec to 96Ks/sec. Although not this simple the best way to think of it is that every other sample in the data stream is discarded.
If the base sample frequency changes (say from 96K to 44.1K) then the process is more complex as the data stream needs to be interpolated into a multiple of 44.1 and then decimated.
As for frequencies there is a direct correlation, but not in a completely intuitive way.
At 192K the data stream technically contains frequency information at 96KHz, but this isn’t something that can be reproduced by any audio system and is typically going to get chopped off by the analog reconstruction filter at a much lower frequency. The point of high sample rates is to move the frequency at which the digital filtering takes place well into the ultrasonics so that the filters can be far less agressive and, therefore, have a much lower impact on phase and group delay. The closer the cutoff frequency gets to 20KHz the more chances there are of the filter artifacts folding down into the audible range.
In your example at 96KHz the data stream can only contain frequency information up to 48KHz. That doesn’t mean that the output of the DAC will contain that information as the analog reconstruction filter is going to be chopping that down into something much closer to 20KHz. This doesn’t take into account the upsampling (typically to 384KHz) that takes place in most modern DACs.
Thanks guys, so while downsampling 24/192 to 24/96 is a destructive process, it most likely won’t have an audible impact, compared with downsampling 24/96 to 16/44 for example (much closer to the audible range and Dac’s output range)?
And when you say “In the simplest terms Roon is throwing away samples to go from 192Ks/sec to 96Ks/sec. Although not this simple the best way to think of it is that every other sample in the data stream is discarded.”
Since 24/192 music has frequency information up to 96kHz and 24/96 has music up to 48kHz, this is why I was asking in my OP if we can think of this particular downsampling as Roon cutting off everything above 48kHz. Is that a fair way to look at it?
You could look at it that way, but it’s not the right way of looking at it. The frequency content is related to the sample rate, but changing the frequency content is a side-effect of changing the sample rate.
Saying that Roon is “cutting off” everything above 48KHz implies filtering (which is done in the frequency domain) which by its very nature creates issues with impulse response or phase (or both). Neither of these things are especially good and very careful filter design is required to minimize the impact of filtering on the actual signal.
Downsampling doesn’t need to worry about these things as the entire operation is conducted within the time domain.
Another way of looking at it is that you could filter out everything above 48KHz but still end up with a 192K sample frequency. In fact that’s exactly what you’d end up with as the filtering process doesn’t have an impact on sample rate. Downsampling from 192KHz to 96KHz makes it impossible to reconstruct anything above 48KHz, but that’s due to the mathematical relationship between sample rate and the output frequency not because frequencies above 48KHz have been selectively removed.
This is going to be very dependent on the DAC and it’s DSP / Filtering implementation. Some are far better than others here. I will say that throwing away bits is going to be far more sonically destructive than throwing out samples.
Higher sampling rates are mathematically useful for the filtering that the DAC must do before decoding. They’re also very, very useful for the processing that’s done in the studio. Higher frequencies provide more mathematical headroom and that can be a real benefit.
Thanks Andrew. Really nice explanations. I was definitely over simplifying things so this helps to understand that it’s a little more complicated.
Sorry to revive an old thread but I have the following question:
My DAC can only supports up to 24/48. I was wondering how will Roon down-sample a 88.2 (or 176.4)? Will it automatically down-sample it to 24/44.1 or it will just look for the highest resolution my DAC supports and convert it to 24/48? I can only hope 24/96 and 24/192 are down-sampled to 24/48.
Many thanks,
Rudy
PS. Sorry if this was already answered.
You can control this. Under Sample Rate Conversion, Choose Custom. Then you can specify the target output resolution for each source resolution. That is what I do to ensure it stays in the same “family”.
This sounds great! Thanks a lot for taking the time to answer.