Nucleus / Mosaic

Two silly questions.

  1. Why on Mosaic am I seeing a track playing on Roon (if that track is actually playing on Roon)? Is the app just showing what the Upsampler is playing, even if the file is playing out of Roon? This is a file downloaded from Qobuz Hires.

  2. How can I play on the Mosaic app a file which physically resides on the Roon Nucleus?

for 2. I think that would only be possible if Roon Nucleus would allow you to install dlna/upnp services (that is imho not possible-I’m running a Roonn Nucleus as well, it’s a closed system)

1 Like

Yes exactly.

Roon Server itself is a relatively closed system, but Roon’s internal music file storage is open via SMB.

You can, but not easily. You need an external UPnP Server (like MinimServer), and mount Roon internal storage as an SMB fileshare as the UPnP’s music database.

But if you’re asking on the expectation that the music quality will be better than using Roon, the answer is no.

1 Like

@Anupc - I 100% agree with you on all points.

If you REALLY want to go down a rabbit hole you could move your music storage off of the Nucleus to a NAS that also ran UPnP. You could then use Mosaic or Roon. But this really would be more for the fun of it then any gain in SQ.


Thanks guys. So in reality Mosaic does not improve SQ per se if you already have Roon. Is this what you’re saying? what about adding a separate NAS and play files residing there using either Mosaic/UPnP or Roon?

On playing your same files over Mosaic…

My setup is as follows: A main computer (macpro) where I curate my local library. Here I rip CDs, scan album art (eg from physical CDs), dowload purchases, and curate my SACD rips. Once my library is how I want it I run Chronosync to update my NUC/ROCK library with the changes in this library. This works very well because I can make any changes I want in my library and then sync just those changes, which is fast. It also works great as I backup this library to a number of disks for safety.

Because I have this identical copy of my Roon ROCK library, I run Minimserver on this library which shows me all of my local files.

Although this is not quite your original question, I definitely recommend this workflow.


Hi Franco, on the recommendation of the DCS community, I first used a Melco N1Z/2EX H60 as a NAS server on which I stored all my music accessible by the DCS upsampler via Mosaic and Minimserver, then I bought a Roon Nucleus plus to test and see what it could give. At first the Nucleus could read the files stored on the Melco, then in a second time I added a big SSD inside the Nucleus and copied all the files stored on the Melco to the Nucleus from my home network. One is the safeguarding of the other and vice versa.

Thanks to Miguelito for Chronosync

Thanks to T38.45: for the clarifications on the internal storage of the Nucleus which is accessible.


Philippe, I strongly advise you to make at least one more backup. If you have only a single backup and your source ( either one in your case) fails then you have to replace it with a backup. If you do this you are then left with no backup and are vulnerable. I would suggest using portable USB HDD drives for this.

Ideally you should make a further backup which should be kept off site, perhaps with a friend. You then have protection against flood, fire and theft.

1 Like

Philippe, can you confirm that with your current setup you’re only using the Roon software to play hires files? If so, this is the same that I do.

I was wondering whether playing hires files using UPnP through the Mosaic app would improve SQ.

The disadvantage of digital is often the poor power supplies of computers and servers that end up straining listeners’ ears because of high frequency noise, hence my choice of Merlco with a Plixir power supply as a NAS server. This Melco has a network input and a player input (here Vivaldi upsampler) protected from these noises. In addition this Nas Melco is connected to the network through a Melco S100/1 switch also powered by a Plixir power supply. The Nucleus Plus, also powered by a Plixir power supply, is plugged into this switch. My next development is to insert optical fiber between my router and my Melco switch using an ADOT/Plixir combination. As a result, the insulation will in principle be total.
As for my audio files, they are almost all Aif or Wave (at least 16/44). Since switching to the much better APEX version of my Vivaldi dac, I use Roon more than Mosaic; But although the 2 are bit perfect, I can notice different tones.

You will find that there are two answers to this.

Answer 1 is that if set up correctly Roon delivers a bit perfect copy of the file to your DAC. When I say set up correctly in this instance I mean that no DSP functions have been applied which would, of course, change the sound in some way. If both Mosaic and Roon deliver bit perfect files to the DAC then there is no way that they can sound different , or so the argument goes.

Answer 2 is that despite the above Mosaic sounds better. This is similar to the argument that goes on regarding fancy network switches. Not possible in theory but significant numbers disagree.

There seems o be no rational reason why two bit perfect network sources can sound different , everything else being equal, (at least so far discovered , there is currently only unsupported speculation) . So either the belief in sonic differences is wrong or the cause lies elsewhere at least as far as digital distribution is concerned. I wonder if like is truly being compared with like? For example if different interfaces are being used for each. I don’t recall any details being posted.

If the comparisons are fair then this still does not provide any obvious cause for audible differences AFAIK. This , as so often in this hobby, leaves only a matter of preference which cannot be applied universally.

IMO, until there is documented proof one way or the other , the only way that you can be certain, at least for your own satisfaction, is to try it yourself.,


As a contribution to this thread, here’s a description of my “computer audio” — we used to call it so — setup. (It’s a rather long post. Maybe should I start another thread ?)

  • The audio files are stored on a volume (“Music”) on a Synology DS418 NAS.
    • The whole NAS is backed up daily to Backlaze B2 cloud storage with history access in case a file is inadvertently erased or modified.
  • MinimServer runs on the Synology NAS and serves the audio files for the Rossini via the Mosaic system.
    • This provides the utmost Sound Quality, endorsed by both dCS and fellow audiophiles.
    • Mosaic can be operated from many points on the network :
      • different generations of iPads
      • the M2 Mac Mini
      • an M1 MacBook
      • iPhone
  • Roon server runs on a moderately powerful 2018 Mac Mini. Locally, it accesses audio files stored on an attached 4TB SSD.
    • A Carbon Copy Cloner task makes a weekly mirror copy of the Music volume of the NAS.
    • I’ve made this design choice for speed reasons. Roon server is very responsive with this setup. Previously, I had Roon server running on a Silent Angel Rhein Z1. Sonically, I liked it. But Roon was soooo slow… sometimes almost unusable — searches taking tens and tens of seconds.
    • I dare say that the sound quality is at least equivalent to what I got from the Rhein Z1.
    • This Mac Mini is an “old” machine repurposed for music playing and it’s very satisfying. It does (almost) nothing else. The “almost” is because I also attached an external hard disk and, occasionally, it can also serve as a Mac file server. But this doesn’t interfere with music playing.
    • Roon server can be activated (i.e. used) from any machine that has access to the network : computers, tablets, phones
  • The M2 Mac Mini is my main (static) machine — the other one being an M1 MacBook Pro 14". I do everything on this machine. For what pertains to computer audio, here’s how it goes :
    • JRiver Media Center manages a library comprised of the audio files stored on the “Music” volume of the Synology NAS. With it, I can :
      • edit the metadata of the audio files — JRiver is the best and most complete metadata editor on MacOS. I was even able to add custom fields useful for organisational purposes, Roon tags, casual listening and mobile SD cards confection.
      • rip CDs directly to the NAS.
      • batch convert files— it’s very rarely needed, but it can be done. It did save many days in the past…
      • manage the organisation of the files on the NAS — it renders the “[by folder]” navigation in Mosaic actually useful. I can describe it further if someone is interested.
      • do casual listening, something I use a lot. To my audio files, I attached tags like “light”, “concentration”, “ballad”, “morning”, “evening”, “not for mobile”, “historic”, “monument”, etc. (NB : these tags are also usable by Roon) Thus I can make use of smart playlists like “music suitable for concentration, not played in the last 45 weeks”. It’s like Roon radio, but much more accurate to my tastes — Roon radio is still not good at all for classical music nowadays. It took many years to refine these tags, needless to say. :wink: These smart playlists rotate through my library. The additional BIG benefit of using JRiver for this task is volume leveling. When you have tracks ranging from –14 db to +14 db (at least), it’s an absolute must. Yes, there’s a toll taken from SQ, BUT, with this machine, it’s very little and it’s absolutely negligible when you are not mindfully listening. I wouldn’t have said that with my previous setup — a 2015 iMac — where the loss was significant for the piano, in transparency, etc. This speaks a lot about the M2 Pro processor.
      • play a smart playlist on my phone when I’m exercising .
      • prepare SD cards with selected files (through a smart playlist) for listening in the car.
    • JRiver on this machine can be handled on the MacBook by using screen sharing. It works pretty well.
  • Airplay does function satisfactorily — no hi res, not the top SQ and this dreaded 2 seconds lag… —, but it’s occasionally useful. I can discover classical music and interpreters with an IDAGIO subscription. Sometimes, I also look at a YouTube video with hifi sound.
  • What is not on the schema, because I didn’t want to brag about it, is a Bartók I have attached to my TV set and that serves as a source/hub in another room. No TV in the living room, only music !
    • The Bartók is wired to the same network and any music can be played indifferently on the Rossini or the Bartók — from MinimServer, Roon, JRiver, streamed from Qobuz, iDagio, etc. It’s a true multiroom system

Without any pretension of being either “right” or “correct” or “accurate” or “scientifically proven”, here’s what I would say about the sound quality of the different possibilities . On a scale of 0 to 1000, 1000 being the “epitome of what can be achieved with top cables, top Ethernet switches (read very expensive ones), top servers (read excessively expensive ones)”, I would rate my setup like this :

  • MinimServer : 995. There are people who say they got a “night and day” difference by upgrading their switches and servers in the many 10^4 USD price range. The placebo effect can’t be excluded and there are many respected people on this forum who doubt about this as much as I do. We have yet to read or hear someone from dCS endorsing such solutions. Furthermore, I can positively attest that the sound quality I get is absolutely immune to the network traffic and you can imagine there’s a lot, with all these connected machines : backups running (Backblaze), updates checking, syncing (Dropbox, iCloud, Obsidian, DEVONthink,…), email, etc. Hence the utmost doubts about the effect of switches or servers on the SQ, as long as the music is traveling through ethernet.
    Let’s save 5 points as room for improvement, just in case and for the perfectionists out there.
  • Roon server : at least 990. On very specific tracks — the beginning of the Hammerklavier Sonata by Beethoven, for example — I could occasionally hear a difference. But generally, I would say that SQ is on par with what I get from the MinimServer.
    In all honesty, when I positively, absolutely want the very best sound quality because I’m totally immersing myself in the whole listening experience of a great interpretation of a great musical masterpiece, I’ll go for MinimServer. The added benefit is the tranquility that I don’t have this little gremlin in my head saying that “maybe the sound could be better if only…”
    But otherwise, to discover music, to explore the musical world, especially in genres other than classical music, I totally like what Roon has to offer and I have absolutely no frustration regarding SQ. There is much more to be said about less than perfect recordings — but that’s an entirely different story…
  • JRiver with Volume Leveling : 950, in my actual setup on a MacMini with “M2 Pro” chip. On my previous setup (2015 iMac), I would’ve said 850, maybe even as low as 800.
  • Airplay : 900 — yes, even if it’s supposedly “bit perfect”, I hear a significant drop in sound quality. I know for a fact that the “bit perfect” is not 100 % guaranteed — where is the volume actually adjusted ? Isn’t it maybe in the core audio library on the Mac ? Actual listening tends to talk against bit perfect accuracy of Airplay. I’m not an engineer and I welcome any technical input on this matter.

What about Audirvāna ? Well, as much as I would like to love it, I still have mixed feelings about it, despite having tried and tested it for many years — from the very beginning, actually. Here’s what I have to say — please don’t throw eggs at me.

  • The interface is gorgeous…
    • but slow and impractical for me — the searches are easier, more powerful and faster in JRiver.
  • The scanning of the library takes very long, even when there are incremental changes.
  • The sound quality is in no way better than what I can get with the other solutions mentioned above.
  • It can’t read the tags I so carefully associated with my music. So…
  • Although you can set up smart playlists, they are not that smart for me.
  • The metadata editor is no match for JRiver’s . It’s adequate for correcting typos, but that’s about it. Whereas in JRiver you can do glorious batch editing and that’s what I need !

So… I’m still testing, version after version, but haven’t yet made it my player of choice.

I hope I have provided some interesting ideas and facts. Sorry for the length of the post.

Olivier :-{)

NB : the schema describes the topology of the network, not the actual wiring. There are some switches here and there…

1 Like

Nicely done. A master’s class!

Hi Olivier, your post is very interesting and superbly presented. You must have spent time (and money) making it all work because in my experience, in computer audio, there is always something that does not work. I too used to use JRiver among others with a Paganini stack and a Sony laptop. This software was also the one used by the DCS team at this time. I also remember memorable listenings with the simple Foobar for example. After my gradual switch to the VIVALDI stack, I was even tempted for a while to buy a QNAP HS 453 DX as a NAS server. Mais non (but no).
Among my immediate entourage audiophile and musicophile (electronics friends, “Grande Ecole” engineers, resellers known for a long time, …), a very large part of hifi journalists of all nationalities and manufacturers themselves evoke this question of these high frequency interference emitted by devices intended primarily for computing and which are likely to lower the Signal/Noise ratio.
Of course DCS products are very strongly or even nearly perfectly filtered against these noises but I still wanted to check this out.
The CD -a simple 16/44- that served as comparison test is as following:

Henri Dutilleux/Métaboles – Ochestre de Paris – Paavo Jarvi at ERATO.
This CD has a very high dynamic of about 19 according to ROON.
Track 1 (Incantatory) and 5 (Flamboyant) were used. This last track #5 has a very strong increase in dynamics (like the BOLERO of RAVEL for example).
The listening was done at an identical sound level (whatever the source) but realistic (configured on the track #5).
The sources tested are as follows:

    1. Vivaldi CD/SACD player,
    1. Macbook Pro early/mid 2013 (16 GB) with Audirvana up to date for OS HIGHSIERRA (so it is not the latest version of AUDIRVANA STUDIO or ORIGIN),
    1. Roon Nucleus Plus with a 8TB SSD (with a copy of the CD used here),

I replaced the MELCO S100/1/PLIXIR switch combination with a SILENT ANGEL N8 with its Forester linear power supply.
The NAS server for MINIMSERVER and AUDIRVANA is the Melco N1Z/2EX H60 HD (6 TB) which also contains the copy of the CD.
The cables are original AUDIOQUEST, MIT, CRYSTAL, TRANSPARENT. The VIVALDI dac is APEX and broken in (very strong improvement with this evolution).
Context :
the preamp is a SOLO VIOLA set, the amp a CELLO RHAPSODY MATTHEW JAMES and the speakers the MARTIN LOGAN ESL 15A. A PS AUDIO P10 is used as a current generator. The listening room is dedicated solely to listening to music. There is no annoying parasitic resonance.
Listening to the CD is worth 100.
With all the products connected to the SWITCH and according to my opinion:

    1. MINIMSERVER 2 is 97,
    1. ROON NUCLEUS is 96,
    1. AUDIRVANA is worth 95.

If I plug the upsampler back into the MELCO player jack, all the listening on the MELCO NAS improves:

  • MINIMSERVER 2 is 98,
  • AUDIRVANA is worth 96.

If I go back to the MELCO switch to replace the SILENT ANGEL, all the listening still progresses slightly IMO.
The notes are there to give a feeling, what counts is the hierarchy that is established.
With streaming, things get worse. In any case, as you also say, listening quality is related to recordings and as audio systems are more and more solving, they do not forgive bad recordings.
I am very long but I spent three afternoons making these comparisons however very pleasant and instructive.