I’m about 2 weeks into ownership of my new NDX 2 and SN3, coming from a Nova. The run-in period has been a little hit and miss but overall I am really happy with the direction things are going in.
As a Roon user, I have a Core installed on an Innuos Zen MK III so all of my control is done within Roon’s app. This week I started experimenting with different upsampling in Roon via DSP and found that on much of the content, upsampling to Max PCM Rate (2x) offered good results, with a little more detail and space in the notes especially on 44/16 content.
But then I remembered that the Nova did internal upsampling via its own onboard DSP and I did some more research and found the NDX 2 does the same thing:
“All incoming sample rates are increased to one of two elevated base rates – 768kHz (for 48k/96k/192kHz media) and 705.6kHz (for 44.1k/88.2k/176.4kHz media and DSD64 files)”
So I did some experimenting with custom upsampling in Roon based on different sample rates in hopes of getting the cleanest signals into the NDX 2 while allowing max sample rate of 768kHz, and set my 44 and 88 kHz files to be upsampled to 96kHz and the 172kHz to be upsampled to 192kHz, leaving 48 & 96 source files to run lossless into he NDX 2 for the Naim DSP to handle.
I don’t know in theory what difference this makes, but it sounds clear, spacious, and more engaging to my ears than running the raw 44 & 88 kHz files I have on my server into the NDX 2.
I wonder if anyone else out there has experimented with upsampling in Roon.
I did for a while but found no real advantage with the Atom which handles all the same freq as the rest of the range. Why waste cycles which is power with no gain. I think it only is advantageous if your DACs internal rate matches that of the incoming signal thus reducing extra processing in the DAC. This is the argument some use any how and they claim its better. Personally I have decide to let my DAC do what its designed to do and let it handle the processing.
The odd thing is that I can definitely detect a difference between lossless 44kHz streaming and upscaling it to 96kHz for example. So if everything in the NDX2 is being upscaled to 768 or 705kHz, meaning there should be no difference in sound, I find it curious that I notice a difference when upscaling from Roon first (the NDX2 is upscaling a higher original file?). I am confident many will attribute this to placebo effect on my part but there is a slight but undeniable uptick in space and clarity on piano and vocals (easy to hear when comparing just a piano and vocal track versus a busy track). Could this be the difference between 768 and 705, or the difference between less computing power needed to upscale samples that are rounded up on the front end in Roon? As a final disclaimer I am sure if I came in blind I would not be able to distinguish between an upscaled file and a non-upscaled file.
On my ND5XS (transport) > nDAC/XPS Roon’s ‘upsampling’ is inferior to what the nDAC can do. I’ve tried on several occasions and the native stream always sounds better. The nDAC has 16x oversampling so that’s 705.6 kHz for a CD quality file.
I only use upsampling if listening to low-quality radio e.g AAC/MP3 etc where it helps a little.
I’m a little surprised at this as the NDX2 “only” supports up to 32/384 WAV or 24/384 FLAC.
The NDAC can only oversample and not upsample internally. Roon from memory can do both, so for it to oversample only you need to set it appropriately. Upsampling can be inferior to oversampling.
Also there are different methods to oversampling, although many do simply zero value sample insertion, not all do such as the Chord mScalar for example.
The other consideration is the transport data rate side effects … handing a higher bandwidth data stream will produce a different set of digital noise to a lower data rate transport and locally oversampling… almost certainly this will be audible, just as different network switches can be, flac vs wav and different firmware versions.
I wasn’t sure about whether it upsampled or oversampled hence the quotes. I’ve not seen anything in the Sample Rate Conversion section in Roon.
I know the Naim DAC does zero value insertion from reading the White Paper, I don’t know how this affects sound quality though.
It was worth a try though and of course free!
Yes in Roon ensure ‘power of two’ is selected for oversample if you wish to use that.
Aha! That’s very interesting as I did find the ‘power of 2’ option better than the maximum sample rate (192kHz) for 44.1kHz material, I expected this was because of the mathematics.
Makes sense now thanks
Here’s what seems to be working very well for me…
Please let me know if this works better for you as I have a similar setup.
When I first got Roon I tried all combinations of sample rate changes but none of them are as good as the nDAC to me.
The examples you have shown, including the sample rate filter, were the worst sounding. The least worst was an oversample from 44.1 to 88.2 and a Precise Linear phase filter. As @Simon-in-Suffolk has educated me on the correct names, although I was already aware that of the maths, 44 to 96 is an upsample rather than an oversample and would require different construction.
I also tried oversampling (power of 2x) on my RME DAC and much preferred it native it sounded a bit off using Roon, where same settings on Atom could not tell. Maybe different systems exhibit it more. My Ifi DAC internally processes everything to 384/24 on the GTO firmware with exception of MQA as its doing it’s own thing. Tried upsampling to this and again found no real benefit or detriment so just let the DAC do its thing.
I tried the oversampling(power of 2x) with my nDAC last night and it wasn’t too bad at all. I used precise, minimum phase(I’ll try linear tonight) with a bunch of 44.1 tracks. In general using the nd5xs2/nDAC there isn’t really a need to do anything.
Thanks for the feedback, there’s clearly a lot to unpack and consider here. I have done some further research on oversampling (2x etc) vs upsampling (44 to 96 etc) and I’d say it seems there’s a lot of case-dependency in how the tech is implemented. From a Roon standpoint, my hope was to upsample the source files to a sampling rate that could then be upsampled to the highest fidelity in the NDX2 (768) in the spirit of doing the upsampling “closest to the source.” But as I have read here there are many who have experimented and so it doesn’t seem to be that clear. Regardless I will continue to explore this - probably ultimately bypassing the Roon DSP in the end if I’m being honest. Thanks for all the feedback!
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.