We can argue which sounds better but in my system there is definitely a difference.
Always hard to put words on these things but for me the difference between my NUC using RRAT through USB to a Chord Hugo TT sounds quite different than using the NUC as a HQPLAYER connection to my DAC. This difference is greater than all the permeation within HQPLAYER itself ( to my ears).
why is this?
SJB.
Hi SJB,
I understand you are saying that leaving all connections and gear the same, you hear a difference when using your NUC as a Roon Bridge and HQP’s NAA. What operating system are you running on the NUC ?
The first place I would look for an explanation is in the settings for HQP. I understand you are using no filters. Is the volume control set to 0 dB ? Do you hear the same difference for PCM and DSD inputs ?
volume is set to -3 which seems to be what is recommended.
It’s all PCM for me.
Win 10.
I don’t have an issue with it, just wondering if others notice this and why it might be?
I don’t have any absolute preference for one over the other, but notice a clear difference.
SJB
I think the the volume setting may be one reason.
In HQP -3 dB is recommended in order to avoid limiting when converting or filtering. If there is no converting or filtering occurring then it could be set to 0 dB (the Limited field in the info pane tells us if playback is being limited). I have found that I need to set mine to -7 dB to avoid limiting on very dynamic passages when converting from 44.1 kHz to 128DSD.
If Roon is playing bit perfect through Roon Bridge, then it won’t be making any adjustments to volume.
One of the unusual features of audio perception is that small differences in volume can be heard and distinguished, but are not necessarily perceived as changes in loudness. Sometimes we perceive such differences as changes in timbre or tonal balance.
I would suggest comparing HQP with volume set to 0 dB (check the info pane to see if this results in limiting) with bit perfect Roon.
If you are still hearing differences with volume equalised between the apps, then I think we are looking at how the OS runs the two different applications. If that is the case then although you may not be able to control what processes are needed there are Win 10 optimisation programs (I think Audiophile Optimiser now does Win 10) which will strip down the OS on the NUC to the bare minimum. You might also try a minimal Linux on the NUC (can be installed in parallel with Win 10 depending on disk space).
At that point it would also be interesting to compare a microRendu where the NUC is. It is heavily optimised to run only the bare minimum of processes necessary to transmit audio from Ethernet to USB.
So Andy, if you compare Roon RAAT and HQP unfiltered they sound the same?
SJB
I haven’t tried it. I will tonight and let you know.
There is a short delay changing the mR from NAA to Roon Ready, so it’s not A/B, more like A … B.
Ahem, I may have had convolution turned on, which may explain a lot.
Must check again this evening.
And there I was thinking “ah yes it’s the bit perfect brigade again blaming my settings”
… I’ll get my coat…
SJB
The -3dB is an interesting one.
A few years ago, some of the smart folks over on the Meridian forum, discovered that in certain DSP stereo presets a -3dB offset was also employed (in surround processors).
This explained why some members, including myself, found that feeding a digital source direct to the speaker was better than routing via the SSP when a direct 2.0 source was all that was needed.
Later, the same clever folks found a way to “tweak” the config files to restore the missing 3dB. This isn’t just about the volume being different. Digital volume controls do affect the signal.
A year or two later, Meridian acknowledged the findings and adjusted presets to only take 3dB away on presets that absolutely needed it. No need to then take a dual-route to the DSP.
So, the learning here is not to under-estimate the effect that these things can have…either in product design or modern day tools/apps setup.
Wow, luckily they got right on it otherwise things could really have dragged out… 
Indeed…that was quick. Trifield width parameter was a whole other story 