Sound Quality journey

Constructive discussion on general Sound Quality topic:
Observations, experiments, findings, tips, ideas, guides…

Please make sure you have good understanding basic principles as described in Roon KB articles and so the discussion starting point is aligned with basic assumptions.

This topic is created as follow-up on thread which had unfortunate destiny, IMO because it was specifically referring to Roon v1.8 and also the starting point of discussion was not set properly, not to mention insensitive personal attacks.

Here we should stay constructive and not be limited to Roon specific SQ aspects and respect each other.

It is normal we have different experiences because we have different setups, different ears and personal preferences and there is no reason why one should try to convince others about his truth as it might be different in the others setup.

Ideally we should be able to isolate and better understand impact of different approaches to different parts of signal chain and find a way how to properly quantify the results.
However it is nothing unusual that we would hit limit of our understanding during the exploration of this topic. It is not to shame or ignore that change, we should accept that for now we do not understand how it works, but we might be able to articulate or even better measure the results.
Sometimes we are just looking to wrong place for explanation, as my limited experience suggests - virtually everything can matter to some extent and our change might impact place or a way no one would expect.

As my introduction to topic from the Roon 1.8 thread i’ve put together few pictures to help myself and folks with genuine interest to better grasp areas of discussion and set “prerequisites” (e.g. eliminate sound quality detractors) for their own experiments. Please note this is not some final ultimate bible, it is limited to previous discussion and areas I felt i should cover based on my experience.

Happy chatting and getting most pleasure of our systems!

1 Like

Whenever a Sound Quality is mentioned, I wonder who first will bring this up, that books could also be High End - do you know how many discussions will be on HighRes faithful font representation in new Louise Glück poetry volume? And could you imagine, how much more interesting the written in HighRes stories will be then… :wink:

The biggest jump in sound I have achieved with my system was adding the Audioquest Niagara 5000

The biggest issue with these kind of discussion seems to be strong held and often incorrect belief about how stuff in this domain actually works and those who seem to take offence when such is attempted to be explained or corrected or when it is suggested that something is not needed when someone else simply wants it without understanding whether such is actually useful or not.

And then add in resistance to any kind of actual concrete data gathering.

Its a shame as these kind of discussions are what make a forum such as this interesting to at least some of us.

OK, that is the cynical bit over with :slight_smile:

In terms of measuring, well, the folks over at audio science review appear to have a process for testing a DAC. Do more or less the same, but with the test tones being wav files sourced from Roon and/or other players. Repeat each DAC test with different things done to the chain (one at a time) - 3k USB cleaners, 15k servers vs cheap NUC etc, different player etc, direct USB connect to core vs via endpoint. Different endpoints etc.

The problem is someone probably has the test gear, lots of other people have the equipment needed to be tested, so we can talk, but that would be about it. Shame - it could be an interesting journey if someone like ASR were to be involved. I guess talking may hopefully lead to education which often is the big part of the battle.

Also one needs to look at the audio equipment chain of DAC -> pre-amp -> AMP to sanity check that the gain staging is sensible and not set up such that what is normally way below the threshold of hearing is now instead above it thus making normally inaudible defects very audible (on DAC digital volume control comes to mind). This could explain why some people are really sensitive to player (and other) differences while others are blissfully ignorant of them.

1 Like

I thought thats what pop-up books were? no? :wink:

To be fair to hi-res - I do prefer to read a book on my iPhone XS vs my old iPhone 6 :slight_smile:

1 Like

@Gary_Foux nice, having clean and stable power is way to go on SQ journey :+1:

Only thing to significantly change sq in my journey has been changes to amplification, DAC and the biggest change being speakers or headphone changes. Not all of the improvements have come by spending more or about upgrading to a newer model, just about finding what works best together the synergy of the combined components. Focus on the fundamentals and it all falls in to place.

The trend at the moment seems to try and fix things that really should not and in my experience don’t matter or make a difference worth pursuing.


One of the most effective upgrades to my sound was the time spent shuffling the speakers around a few inches each way to give them the space they needed and then using REW to do some sweep measurements while I setup an EQ in Roon to smooth out the bass.

Cost nothing but a little time (I already had a suitable measurement mic), but significantly transformed my sound.

Of course, you still need a source that gives you the clarity you desire and a pre-amp/amp that will preserve it and an amp that will control your speakers properly with the headroom you need, and speakers you are happy to live with in the room.

I do think far too many people upgrade things in an odd order and/or neglect the room and so never really get the most out of what they have.


Yes speaker placement makes a world of difference to how they work in the room. Another thing to fall under synergy in my book. But in some cases the speaker is just wrong for the room. It’s why I have ended up stepping down to smaller stand mounts from Full size floorstanders.


Lots of time spent wading through the Roon community, but on reading the background information and the few picture slides in the introduction - maybe this is the right place !

In pursuit of sound quality, I am planning to replace my existing PC which is dedicated to running Roon core within my music system. Although there are now a number of high end solutions such as Taiko Extreme and TotalDac d1-player , few of us have the chance to experiment with or compare different computer host configurations, so I am looking for additional input based on peoples listening experience.

A direct question to Roon Support referred me to this from 5 years ago Suggested hardware Roon also said they are working on a better article about minimum requirements.

I am not looking for an “appliance” solution but rather to optimize sound quality in what Paul McGowan of PSAudio would call a highly resolving system, even though that will mean more maintenance tasks for me.

Here are my thoughts so far:

My assumptions about Roon (running as a headless core)

Roon is designed to “work” across many system configurations and will deliver a musical output once the minimum hardware requirements are met, but sound quality can be improved with the investment of time and money.

The Roon core is running several tasks including:

  • Outputting digital music information to audio devices
  • Receiving control signals from Roon apps
  • Possibly doing some digital signal processing
  • Dynamically managing its music information database (which can be very large as seen from backup sizes)
  • Outputting information to Roon apps so that they display the engaging Roon interface

Hopefully Roon is designed to run at least some of these as different processor threads so they don’t interfere with each other , but Roon is reported as using few threads and Roon themselves provide no real information about this.

My assumptions about digital:

  • Error correction is happening throughout the digital domain - it is beneficial to minimize the need for this. (Leaving aside Linus Torvalds comments on ECC memory in the consumer space as probably too expensive to address given the low level of errors. Memories of CD players delivering improved sound quality when error correction was operating at a minimum).

  • Digital transmission involves an identification of bits of digital information in what is really an analogue waveform. Electromagnetic noise here has a chance of causing distortion or of hiding fine detail in the subsequently decoded digital signal.

So what does this mean should be driving my decision on a new dedicated server for Roon core?

For the database processing – high levels of CPU cache, fast RAM (DDR4 or DDR5) and fast SSD (with NVMe PCIe4) will be important. CPU speed itself is probably not a limiting factor once a certain level of performance is reached. Higher CPU models from AMD now seem comparable with Intel in how they use on-chip cache. SSD definitely makes an audible difference for me.

For DSP – CPU speed and possibly multiple cores and threads are going to be important if this is used. But lower CPU speed and less cores and threads needed if no DSP used.

For graphics – there will be minimal server requirements if Roon core is used in a headless configuration.

A “lighter” operating system should help focus resources on the music. (suggested by Roon ROCK and people’s experience of Audiophile Optimizer).

Minimize potential electromagnetic noise sources: computer power supply, cooling fans, LAN card and its power supply/ optical isolation, final LAN cable from computer to server/DAC.

Perhaps consider lower power CPU so that computer cooling fans are not needed. (more for the reduction in potential electromagnetic noise as it will be in a different room than the rest of my music system).

Interested to hear what others think I have missed or anything written here which doesn’t match with their practical listening experience.


Before I dive into some sentences I’ll leave here just to give you more to think about… a lot of your points simply go away and don’t matter by keeping the Core far away in a closet and focusing on the network infrastructure and streamer. The whole point of the core / zone / remote relationship is to keep the stuff that matters, streamer and dac, skinny to the larger leg work that is library management, dip, etc.

ok, lets get into…

Yes, and no… Generically I’d say yes but you’re dealing with a lot of network and application layers + digital formats. Not all of them employ error correction.

This isn’t really a concern… if you’re running RAAT to a well respected endpoint. Although, I’d argue audio is “real time” so jitter and propagation delay matter… in reality… you can have high jitter and propagation delay because the streamer is reclocking everything anyway. Keep the buffer full. That’s all you need to do. Keep the buffer on the streamer full and the core is keeping up just fine. RAAT does this very well and will terminate any session to a zone that can’t keep-up.

I run ECC memory in my core. I purposely kept with a Xeon proc / architecture and motherboard for this reason. I have no idea if it makes a difference for Roon. But do know it’s not just the cost of the memory to use ECC as the normal motherboards don’t support. You have to step-up to server grade hardware and that’s usually a generation behind the raw performance of the consumer stuff (at least last time I looked). And multi CPU / huge number of cores / high HT count has no benefit to Roon.

To your own point… don’t do the D to A conversion near the core and you no longer need to worry about this when building your core.

There is a theory that higher clock speeds create additional “noise”. This is partially true as higher clocks require a far more careful design to keep those higher frequencies from interacting with each other (shielding, spacing, etc.). But designed well, higher clock speeds won’t actually result in higher “noise”. Why do I quote your “fast RAM” because, generally in computing we get speed by running higher clock speeds. This is counter to the noise argument. So, if you are truly concerned with noise, and don’t want to go with an industrial design, then build to the minimum requirement.

A note on graphics… there are 0 requirements here unless Roon decides to implement DSP algro on the GPU… but I think those solutions are better left to HQPlayer and the others.

Yes, absolutely, but now you’re creeping into “appliance” territory.

Again, just keep it away from the stuff doing the D/A conversion. If you’re really worried about it go full electrical isolation with Fiber or Wifi between Core and Streamer and put the streamer side on dedicated power and / or conditioner. In fact, I joked once I was going to ask my neighbor if I could leave the core plugged in at his house. That’s really good electrical isolation right there!

and in summary… isolate it and don’t worry about it. Roon allows you to do this easily so take advantage of that in the architecture and your build.

Many thanks for your comments - most of these align with my thoughts and I do realize that we are into fairly esoteric territory here. The joy of high-end audio as a hobby and as a listening pleasure :blush:

A few comments from me:

I do agree with you on the importance of the power supply (I personally use a PSAudio regenerator plus very high quality power supplies to individual components and all this does make a clear audible difference.

My Streamer and DAC are already linked using TotalDac’s Gigafilter which works very well.

Discussion of LAN cables always generates many opinions but in a sufficiently resolving system the LAN cable linking to the streamer does seem to be important and there is an audible difference when a cable in this specific position is good at rejecting electromagnetic noise. So reducing such noise coming from the computer music server seems to be important. Reviews of the JCAT Net card and the adoption of an optical connection by some high end computer server producers, seem to indicate there is a music quality opportunity here.

To the question of memory – cache, RAM, drives:

I believe Roon when they say that they are manipulating the (large) core database in real time. Such large database manipulation probably involves processor, cache, RAM and SSD. With insufficient cache you certainly get horrible dropouts, but these are easily avoided once adequate cache is in place (Roon recommend higher spec Intel CPUs for this reason).

Recently I have found there is a very audible difference with the Roon core and its database on a fast SSD. It doesn’t seem to matter if the audio files themselves are on a slower drive.

That leads me to believe that Roon’s database manipulation does have an effect on sound quality and therefore cache, RAM and SSD speed may all be important factors – they are just a bit difficult to experiment with. I am happy to swap cables about to see if sound quality is affected, but I don’t think I would want to be swapping RAM and SSDs in the same way !

This is an area where it would be great to have a more detailed input by technical designers and experts from Roon itself.

1 Like

Regarding streamer connection to LAN/RoonServer, here is the link to some “measurements” I did recently to compare 100mbps , 1gbps , EtherRegen and wifi on the RPi4 transport: Audiophile Switch Experiment Results

From my perspective, galvanically separating streamer from LAN/RoonServer brought quite significant improvement in SQ. I assume this is system depended as some folks might have streamer already galvanically isolated from LAN, or maybe they don’t have many noisy machines connected to LAN.
I have started years ago with cheap 100mbps TP-Link LAN optical convertors which did great job. Now i’m experimenting with EtherRegen and OpticalModule , but although i have 250+ hours on the ER+OM I think the SQ is still not fully consistent so I hesitate to compare it to the old TP yet. However when ER+OM has good moment it sounds crazily damn good :slight_smile: (realism and better tonal balance - more even and better integrated all frequencies together (I can’t grasp how’s that possible yet)).