Option for optimized performance for a single core and single zone playback [Not on Roadmap]

Hi Mark, my DAC’s USB input is isolated (power, ground and data lines). On top of that Rob Watts will tell you he filters RF heavily.

So yes I agree, as I mentioned above, airborne RF is possibly one of the major culprits.

While I’ve known this already (I already heard the sonic improvement that comes with moving an i7 Core out of the listening space) I guess I didn’t expect to hear a difference between RFI (due to CPU/GPU loads as Danny and Brian hinted) between 2 apps… if that is the culprit here.

Danny, what kind evidence are you looking for when it comes to sound quality of Roon/RAAT?

The users gave the input can be experienced on. Have you compared A+/UPnP vs. Roon/RAAT? Do you claim that they have equivalent sound quality in your experience?

If the UPnP implementation is not terrible and the device is Roon Ready, it should be identical. If you hear otherwise, I’d be interested in a statistically relevant blind test that removes the chances of bias.


If you did, did they sound identical?

I’ve not done this test for all the RAAT devices we have that we have certified. We normally leave the audio parts of the certifications to the hardware manufacturer. They have all done the tests to verify the SQ is there.

However, we’ve actually gone one step further, and tapped the actual bits coming out of the other side… and they are identical (obviously). Both are pull protocols, so timing is driven by the clock in the device, and not the sender.

Danny, thank you again promptly getting back to me.

Your answer - diverting to third-party SQ tests - implies that you do not have internal experience comparing Roon/RAAT with other software, such as A+/UPnP. (Correct me if this is not the case).

Your response “However, we’ve actually gone one step further, and tapped the actual bits coming out of the other side… and they are identical (obviously)” also implies that you and/or Roon think that the “identical bits” would exclude the possibility of Roon having inferior sound quality in “networked stuff” compared to other products.

This thread is about sound quality which naturally involves listening to the sound. With all due respect, what is the basis of your statements referring to identical SQ in “networked stuff” in view that you have shared no information with us on your recent listening experience on the subject?

Let me share the relevant part of mines: I am user of Roon for about 9 months, I like numerous features and the convenience of your product. Have tested A+, JRiver, mConnect, they all sound noticeable different on my setup - Mac Mini (I7, SSD); PS Audio DirectSream (connected with Purist Audio ethernet cable through Apple router, WiFi off), VTL MB750, JM Lab Utopia speakers, Purist Audio Neptune Cables, PS Audio P3 regenerator, dedicated circuits for each amps and for the front end. I find JRiver and mConnect the most similar sounding and Roon is on the other end. Without attempting to rank the SQ I find that Roon has the widest and deepest soundstage, however it sounds “thinner” has “emphasized highs”, more overhangs and how another user (woodford) described: “brasher, splashy, hashy, spitty” see Sound Quality change with build 259.

Ironically, what drove me to these tests was Roon 259 build which sounded different compared to the previous release and was not able to revert back to verify what I was hearing Sound quality change and re-installing older builds . After these tests, I can now spot the differences between Roon, A+, JRiver in less than 30s on tracks I am familiar with.

I extend my invitation to you and two colleges of your choice to listen and decide on your own. I live in Austin Texas.

Please allow me to point it out that I find it disrespectful to tell users/clients in this thread: “you can keep your antecedal evidence to yourself until you have an explanation that can be reproduced and experimented on”. (i) first as mentioned above you have not shared with us “a statistically relevant blind test that removes the chances of bias” to prove that Roon has identical SQ in “networked stuff” compared to other products. So, on what ground would you ask a rebuttal evidence?; (ii) the users in this thread did point to A+/UPnP vs. Roon/RAAT which can be experimented on; (iii) asking us to have an explanation does not fit to the posture of us being the user of the product and clients of Roon; lastly you do not need to know why a rotten egg stinks to be able to determine that is rotten.

My reply could sound blunt may be edgy, but I want to get my point through clear: I do not think that Roon has the appropriate focus on sound quality

To be fair, it’s worth referencing these two quotes from the Roon team:


I think I would sum up Roon’s position as this: the data coming out of our system is bit-identical to the data coming out of other product XYZ, but it’s possible that the cpu intensive Roon UI is generating higher cpu/gpu load and causing downstream RFI/electrical USB interference at the D/A process in a DAC. Moving the Roon core/UI away from the USB connection will mitigate this.

it would be interesting to see if anyone can validate/nullify @danny’s hypothesis - does loading up the cpu while using A+, via Roon UI or otherwise, cause the SQ to level out??

I haven’t done that specific test (yet) but I have monitored cpu activity on my mac mini and while the Roon UI does use additional cpu cycles, I find it a stretch to believe that this would cause sufficient RFI/electrical USB noise to the extent that it’s so obviously audible. That said, I would be very happy to have my doubt removed with good evidence.

FYI only (related to the subject being discussed, but not directly related to Roon) :

For something more related to Roon: If we look at Sound Galleries (a Roon partner) music server, it also states “Fully optimising the operating system to only run the processes necessary for HQPlayer and roon to work.”

If you are talking about single computer then Roon has clearly stated that it has made a decision to offer a graphics rich heavy UI that is not as light or optimised for audio as other products. That isn’t going to change. Roon’s UI is at the heart of the product.

Roon reccomends separating the Core, Control and Output functions into different devices. When people compare that architecture to other products, they don’t hear a difference.

Single computer SQ is important to the Roon devs, but it is not the reccomended architecture for maximising SQ.

Hi Andy, I know what’s you’re saying there but we need to be careful with generalisations (I’ve been lazy sometimes also, with choosing my words).

I know of 3 people that only have their Sonore ultraRendu in their listening room (Core is outside) and still prefer A+ over Roon.

Are they the only 3 people on the planet? Nobody knows but I wouldn’t say everyone that follows the Roon Recommended Core/Endpoint setup (I know it well and agree with it whole heartedly) agrees that they don’t hear a difference.

That is interesting. I suppose a better way of putting it is that the Roon devs don’t hear a difference. That is the comparison where Danny would be interested in a statistically relevant blind test.

It’s always possible. I know a few highly regarded DAC designers who say all the inputs on their DACs measure and sound the same but there are tens or possibly hundreds of customers who disagree and say a certain input sounds better than the others.

And that’s not to be rude or insulting to anyone. These things happen.

Something I still haven’t gotten around to trying yet but definitely will do this weekend with a friend.

I’ve been using A+ even more this past week and it feels like I’ve gone backwards 10 years in terms of UX. But this thread isn’t about that - this focus is sound quality.

Be sure the levels are matched and double blind trial if you can. Audiophiles are notorious for falling victim to placebo effect. You’ll prefer the louder of 2 sources. If you expect something to sound better, it will. In the case of playback to a USB DAC, there should be less transport effects like those I’m about to describe.

When I notice a difference, it is usually in air and space. Voicing is unchanged but I’m 69 and have some codger’s high frequency attenuation.

This is most apparent for me when delivering bits to an Apple TV 3 vs a ChromeCast audio, both TOSlink connected to a Parasound P5’s internal DAC. It is subtle. To give a couple of examples, Joey Alexander’s Countdown record and Chris Thile and Brad Mehldau jazz/newgrass mashup. Both were ALAC transfers I made using iTunes converters.

Today, I have a HiFiBerry Digi+ schlepping bits and it sounds much like the Chromecast Audio. I’d be hard pressed to tell the difference. Again, both are TOSlink and SPDIF interconnects.

With the non-apple endpoints, piano play sparkles and the notes dance in space. With the ATV, the openness and sparkle was diminished.

This was pretty repeatable across all of my good material but is most obvious on small ensemble jazz. Amazingly, Snarky Puppy’s Sylva record has that same feel. This record is Metropole Orkest plus Snarky Puppy playing big band funk/jazz. It is a close miked, multi-miked monster. Each instrument had at least 1 mic and some two. The mix down is amazing. Engineer was wizard in placing everything in space. And it is damned good modern orchestral music.

The ATV was fed 44.1 RedBook data. The ChromeCast/Roon transport was up-sampled to 192/24. The upsampling transformation appears to help significantly. So we can attribute changes to the different playback signal processing chains in use.

I first noticed this behavior with Plex and it persisted across the migration to Roon. There may have been some gain difference in the two paths as a result of different design decisions with regard to headroom and overflow prevention or loss of significance prevention in the floating point. When you hear a difference, it is likely a result of implementation decisions made to ensure that the output is really well behaved.

I’m going to play both of these disks in the Cambridge 851. It also upsamples to 192/24 before conversion. It can provide balanced and unbalanced audio and TOSlink. Both would be affected by the upsampling in the CD player. The P5 also upsamples to 192/24 using Burr Brown parts and Burr Brown converters. John Curl did the analog bits. My recollection is that the 851 sounds very much like Plex/Roon to Chromecast or HiFiBerry.

Associated equipment: GAS Ampzilla (original) and Dhalquist DQ-10 mirrored. Seriously spacious. I find that the presentation differs significantly with playback level. For the magic to happen, piano level needs to be similar to live unamplified play. Dialing back for casual listening looses the magic.

Roon Core runs on a Debian instance in a BHYVE VM under FreeNAS 11. Three switches are in the digital path, 2 Netgear unmanaged rack switches (8 port) and a Ubiquity 8 port POE switch in the distribution cabinet. The way in/out is through a Netgear Orbi and Netgear cabel modem.

I’ll play these records with Roon on the Mac. I’ve only used Mac Roon for local listening with the internal iMac speakers. These are OK but the sound is definitely muggle. No magic this way.

David, Apple TV sample rate converts everything to 48 kHz. My understanding is that it never outputs bit perfect audio. So, that factor plays a role in any comparison.


I figured out this too, SQ of roon does not much anymore a simple Minim Server on a NAS (Synology), connected UPNP to my DSD PS Audio.
I hope this won’t be the case in the long haul. Using Linn Kazoo to manage my music (with Minimserver) on my IOS and OSX is a nightmare. I want to go back to Roon :slight_smile:

I have changed the install on my Mac mini from Roon to Roon Server. It makes noticeable improvement on SQ. I find the undesirable SQ characteristic (splashy, hashy, spitty see Sound Quality change with build 259) noticed with and after build 259 reduced by running Roon Server only. I find the notes also airer with less overhang. The CPU utilization is about the same, around 5% (when only music is paying with Roon and without touching the UI). Will do a more thorough tests and will post the findings.

Can you provide info on what this means, on the Mac?

Roon Server is is a “bare bone” appliance without Control (UI and related functions) https://roonlabs.com/downloads.html. I think it is a better option than to move the computer having Roon on it into another room, because the control part of the software will still “talk” to the essential functions stripped to the Roon Server if those functions are on the same computer. In the setup I am using the control of the server is strictly though iPad or iPhone.

I yet to sort out how to co-exist Roon and Roon Server on the same Mac to be able to do controlled tests.

1 Like

You can install Roon and RoonServer on the same machine. Roon will ask you if you want to use the RoonServer Core (meaning it would operate as a Remote for the RoonServer Core). Say no and set it up as a separate Core.

Unless you have multiple licences you won’t be able to run two Cores simultaneously. You can, however, swap your licence between the Cores. This adds a step to A/B comparison, but you would have to close the unused application each time anyway in order to properly test.

Running Roon Server separates the Control function from the Core. An additional significant improvement in SQ can be obtained by separating the Output from the Core.