Roon 1.6 AUDIO feedback thread

Sounds the same as 1.5 to me. I don’t like the UI changes to the DSP engine but I’m happy with the sound quality.

I hear differences. And there is a loss of accuracy in the filters of the new DSP interface.

Roon can tell you if any changes have been done on filters (in the implementation); if not - it is all in your head :smiley:

Thankfully I hear no differences and hopefully never will. Roon provides bit perfect delivery to my DAC so I cannot think of how the quality could possibly be changed unless via DSP changes initiated by the user.

Sound the same to me as well; but I don’t use filters at all.

It’s just an interface change not it’s implementation so how does that figure?

Psychology is a big field; people study it for years :smiley:
And our mind is easy to manipulate/trick. Happens to everyone daily.

3 Likes

There are already hundreds of posts on Roon 1.6 and almost all I read are related to the new interface radio, a few to the DSP but not on the point I mention and a few on the sound. This could be explained by having two broad categories in the population of Roon clients, with of course overlap:

  • Users primarily keen on integration of user interface and especially to clients like Tidal and Qobuz, with emphasis on streaming integration.

  • Others who don’t necessarily use much streaming but are attached to very fine rendering on their systems.
    The second category is typically with long experience; more frequently with carefully picked equipment, skills in fine tuning and (hopefully based on real acoustic music) an acute ear when it comes to timbric accuracy, musicality and ambience retrieval.

I fall in this second category and created this post because I have two AUDIO issues with 1.6. They are sufficiently prominent for me to have stepped back to 1.5 - and confirmed my assessment.

I use Roon in combination with HQplayer to feed a Neutron Star la Rosita DAC, in 32 bits float/384 kHz. I use DSP for fine tuning the bass response only, with two or three carefully adjusted filters at central frequencies all below 60 Hz. The system overall efficiency is 100 dB. I put these numbers only for reference as this is not a discussion point.

Issue 1: the new 1.6 sound in this ROON+HQplayer configuration, has subtly lost expressivity. It is less involving, delicately oversmoothed. The change is certainly subtle enough to go unnoticed on a less finely tuned system, It would even possibly sound ‘better’, should the system have some inner distorsion adding some unwanted rugosity to the sound. But on well-balanced low-distorsion system, very apt at reproducing familiar sounds as well as any kind of music, 1.6 sounds uninvolving. Full stop.

Issue 2: the new DSP parametric filter interface is a double regression. The first one is that one loses the simultaneous view of the response curve and the set of filter characteristics. In 1.6 they are visible one at a time or in a small pop-up window…
The most important one is that while this parametric filter and its 64 bits DSP filters are exceptionally transparent, the new interface reduces the accuracy of the settings to typically 2 digits. It uses cursors and with it one can define 0.5 dB but no longer 0.45 or 0.49. On my system the corrections are subtle and the difference is audible between 0.50 and 0.49 dB. The difference between 0.50 and 0.45 should be audible on most systems. Same for frequencies and Q factors. Two decimals are needed and cost nothing with 64 bits computations. So the interface is a step-back in accuracy.

For issue 2 the simplest fix would be to go back to the previous parametric filter interface. For me it was near-perfect apart for letting the name of the correction file visible in the same window.

I have only those two observations but I did not spend much time in 1.6 after hearing part of the previous ‘magic’ of music reproduction was gone. On a Mac with Time Capsule, reverting back to ROon 1.5 is trivial.

I did not try Roon without HQ player, as HQplayer also was a key improvement earlier on in my configuration.

As for me I will reinstall 1.6 once these problems will be addressed, because the primary objective of Roon (Roon/HQ in my case) is to provide the best audio quality for our digital music. That objective was better reached with Roon 1.5 than with Roon 1.6, or let’s call it 1.6.0…

I will be happy to discover the other functionalities of 1.6.x later on, issues 1 and 2 solved.

Lol. /10char

2 Likes

The team says they change the downsampling to linear phase in 1.6. May be this is an improvement on Roon used alone but this does not rule out a degradation its the interaction with HQplayer, consistent with my observations.

The team says they change the downsampling to linear phase in 1.6. May be this is an improvement on Roon used alone but this does not rule out a degradation its the interaction with HQplayer, consistent with my observations.

For the loss of accuracy due to only one decimal place it is a very obvious loss of accuracy, not in the algorithm but as a restriction in the choice of parameters.

See my other responses.

Do you use HQplayer ?

We are discussing DSP interface flaws and possibly a bug or a glitch, at least in the way Roon 1.6 interacts with the current version of HQplayer. This thread is not about psychology of perception.

Some people are more sensitive than others to subtle differences of timing as well as frequency (absolute ears being a known example). There is an excellent article about in in Physic’s Letters in a 2015 issue. And some systems are better tuned than others.

I’m not using HQPlayer

Not since I realised upsampling does nothing to improve or even noticeably change the sound at all - a simple waste of cpu cycles and power as far as I can see. I did upsampling everything to DSD for a long while due to some bizarre idea that DSD was closer to analog however having seen evidence to the contrary (Archimago and Audio Science Review measurements). These days a DAC with fantastic measurements and support for all commercially available bit rates can be purchased for a few hundred dollars (Topping, SMSL etc.) so I can see no reason to be downsampling anything.
You have introduced HQPlayer as a possible problem area well after your original post and many replies so I would hope you can understand why you have received the replies so far.

Christian, your points are valid regarding the loss in precision for Q and Gain. You can still enter them manually and not use the slider but they get rounded apparently.

On linear phase resampling : this is only for the resampling of the convolution filter. Do you use convolution? If not, there have been no change in the DSP engine that I am ware of between 1.5 and 1.6.

Does your setup allow to intercept the audio flux at the output of Roon, using a virtual device driver in Audacity? I yes, it’s easy to compare two versions of the same track, one from 1.5, one from 1.6.

The DACs you are mentioning are good ones, still they are doing filtering for volume change, upsampling and conversion to DSM, as they are delta-sigma chips (the most common today).
HQPlayer allows to do this with more advanced methods than what the DAC chip does, there is definitely still a case for HQPlayer, be it for volume control only.

At the end, it’s all a matter of doing the processing in software or hardware.

I was initially skeptical with HQPlayer and tried it. On a “cheap” 400€ ES9018 DAC, simple TDA SMSL-SA50 Amp and bookshelf monitors Diva 16, I was really blown away (and I am not easily impressed).

@Christian_DEPLANTE can you describe a bit more how you use Roon DSP, is it only for bass EQ ? Do you resample? Do you hear differences between 1.5 and 1.6 on all formats (44/16, 96/24…)? On which type of music?
FYI some weeks ago I challenged the PEQ of Roon and ran some tests:

  • defining a multiband PEQ filter, 5 bands, various gains and Q
  • implementing this filter in Rephase / minimum phase, generating associated convolution filter
  • Test 1: using Roon PEQ with my filter parameters
  • Test 2 : using Roon convolution engine with Rephase convolution file

There was no difference at all, correlation depth was maximal.
Running Diracs through the audio chain in the digital domain, I could reconstruct the response of the filters perfectly in REW, they were absolutely superposable.

This rounding up MUST be removed. It is an aberration. And what was wrong with the previous display which enabled to see both all filter settings and their effect ? It was much better IMO.

The Q filters are convolution filters, aren’t they ? So yes. As I use parametric filters with Q, I use convolutional filters. It now seems clear the subtle quality degradation I observe is associated with HQplayer and perhaps using convolutional filters.

PS: I use ClosedForm/ No dither as the best setup in HQplayer Roon 1.5.