Qobuz stream versus local server using Mosaic?

Hello this is my first post. I own a Bartok (without headphone amp) and combine this with a Rossini clock. I use a Innuos Zenith MKIII server directly connected to my highspeed fibre optic connected hub/router via network isolator (EMOSAFE EN-1005+)* which aside from protecting from voltage surges also serves as galvanic isolation from the upstream connections. The ethernet cable run from the Innuos server to the Bartok is 1m of CAT 6 (unshielded). The whole system (except for the hub/router which is on the main house mains ring) is served from a dedicated 16A supply with noise filtering.

*The ethernet isolator does not have an audible impact when added or removed from the server input.

I have enjoyed many of the entries in the community over the last few months since getting into DCS. Notable contributors from whom I have gained real insight include Pale Rider, Ermos, PAR and all2ofme - many thanks for your time.

To concur with many others on this forum, I find that:

-I much prefer using the Bartok with a pre-amp - especially for low level listening.
-6V out is more dynamic (faster transients) and musical than 2v - if your system is compatible?
-Ethernet input is more coherent and in my view more authentic to reality than USB - however USB when sent via a Innuos Pheonix re-clocker can render more micro detail, air, decay and apparent speed but overall is less focussed.
-The addition of the Rossini clock brings about a dramatic uplift to the sound (literally) with overall focus and instrument separation /depth improved while the sound stage in my case literally lifted upwards on my speaker with more treble detail becoming apparent and the bass less diffuse.
-Clock cables by ‘Design cable’ (£15 a go) with 75Ohm Neutrik connectors are a decent improvement on the ones supplied with the clock.
-The Bartok works fine with a Puccini clock - set for either 44.1 or 48 with Bartok automatically taking care of the other sample rate for each file played - a decent improvement on the Bartok alone.
-The Rossini clock performance is audibly better than the Puccini clock.
-The Innous Zenith MKIII server now works properly with Mosaic since the recent software update.

So in the context of the above I wanted to see if other members also experience the following:

The sound quality of the same albums played (streamed) from Qobuz sound inferior to those stored locally. By 'inferior 'I mean that Qobuz files, while good to listen too are less detailed, less fluid / free flowing with marginally muted attack and decays compared to the locally streamed version. (Note: I have Qobuz set to maximum resolution as the default). I have tried this comparison with a number of albums with varying bit depth and sample rates and always find this to be the case? In all instances to date, the file I have purchased from Qobuz (.WAV or AIFF format), sounds better when played direct from my server than when streamed from Qobuz. In both scenarios I am using Mosaic control the stream.

Do you people think think the loss of ‘quality’ is a function of the extra processing that the Bartok /Mosaic platforms needs to complete when decoding a Qobuz stream or is that Qobuz’ file transmission /compression regime is having a negative effect? Either way I do not obtain the same pleasure listening to internet streamed files as I do from locally stored versions. My ‘perception’ that there is a difference would appear to be at odds with the idea these files when rendered by a more than capable computing platform should sound the same?


Hi Barry,

Welcome to the community.

Personally I do not hear any difference between local content and content streamed from Qobuz or Tidal. If the latter sounds inferior to you, it might have to do with your internet setup or service.

To add: I do not use Mosaic for playback, but Roon.

1 Like


I have just spent some time doing exactly what you say that you did : comparing purchased downloads from Qobuz with their streamed equivalents. The result is that I am unable to detect any difference.

This was using Mosaic. I am using Vivaldi Upsampler ( for network connection only, no DSP)/ Vivaldi DAC however this should be equivalent to Bartok as the comparison is still like with like.

Thanks for giving me the stimulus to do this as I have not carried out such a comparison for a long time.

So I do not think that there is any inherent audible difference between Qobuz “live” and Qobuz “stored” ( as it were) and agree with Erno that it has to be some factor related to your internet service or setup.

Pete and Ermo - Thanks for experimenting.

Were the files that you played from your local server FLAC or WAV / AIFF? I wonder if this affects the outcome?

I note when the file is streamed from Qobuz some form of file size reduction algorithm is used which may be different from the .wav / AIFF files I downloaded?

Previously (when listening on a different USB DAC) I have detected similar sound differences after ripping the same CD’s to FLAC (non compression) and .wav. Also, Hifi Critic have noted differences between file formats following their own experiments?

I will explore the internet connectivity to see if data is being ‘delayed’ or repeated for example?


Yes I am aware of reported audible differences between .wav files and decompressed FLACS. Indeed I have heard it myself in my system when using a Win 10 source and JRiver even when the decompression was selected before the file is sent to memory for replay. However the exact same test files played from my Melco NAS/Vivaldi setup exhibited no audible differences. I would not care to speculate why this may be.

I’m in the “ripping process” of all of my CDs (…sigh) with Melco D100 in flac on Melco NAS. Qobuz has no chance against this SQ wise, even some redbooks ripped sound better than highres streaming with Qobuz. Very sceptical at the beginning, but now it’s no question…

I’m afraid that you can’t draw too many conclusions from that comparison at least in the sense of one being fundamentally better (or worse) then the other. You have no way of telling if the streamed hi-res file and the CD are from the same master nor how the hi-res has been derived. You heard what you heard but that doesn’t necessarily make it universally applicable.

Even two CDs of the same repertoire made at different pressing plants may not sound identical. Nor may files supplied to Qobuz by, say, the French rights owner be identical to what superficially appears to be a CD of the same repertoire from the USA rights owner (or vice versa).

NB: All of my rips are stored on Melco too. However using dbpower amp/PerfectTunes to carry out the rip does save quite a lot of pain when it comes to metadata if that is causing you frustration.

Compared many tracks from qobuz vs ripped from different albums, f.ex Michael Franks, David Sylvian, Brendel, Haebler, Geneaux, Volodos…My ripped CDs sound better, more dynamic and great colours. Of course I don’t know which master they used, I can only say what I like more.
Regarding Metadata: this is such a s…t, I have to tag almost every album again, horror are box sets f.ex Riccardo Primo G.F.Handel…2 CDs are recognized , others not… a fun to tag Track1, Track2…

Fair enough.

You have my sympathy born from experience. I have ripped around 1.5K albums most of which are classical music. This latter fact causes all sorts of problems simply because the historic cultural structures of the genre are not recognised in the album/artist/track format of just about all players ( excepting Roon?) and ripping software.

When I started to rip 5 or more years ago the online databases that some ripping software referred to for metadata were hopeless. This is because they had been built by users with a lack of knowledge ( I came across a complete discography of the Beatles where the guy who created it apparently believed that John Lennon wrote everything) or use individual work arounds by people trying to find a solution to the standard format issue if applied to classical works.Most common was the use of the composer’s name instead of artist. So one ended up with a recording that was apparently of Handel himself as conductor or soloist playing with the LSO in 1975!

Why I ( and so many others) recommend dbpower amp , aside from PerfectRiP, is that it offers up to 4 different databases for the metadata. So you can choose to ignore the one auto selected and pick another if it looks more correct. Details can be edited as you wish. You can also reject the chosen artwork and chose a substitute from the internet or from your own files.

Basically things have improved over the years and most metatada available via it for albums created over the past few years is pretty accurate. Most importantly the part requiring the most effort for classical music is over 90% correct. That is the track level information; name of work/key/tempo etc.

You may still need to play with artist name ( my convention is conductor before orchestra. ) and it may be necessary to add the composer name at the front of track details when more than one composer is represented on the disc so as to distinguish Elgar Symphony 1 , 1st movement from ,say, Bax Symphony 1, 1st movement . However this is as nothing compared to having to enter every detail and it only relates to a minority of discs.

If you have a new Melco with access to SongKong then I must admit that haven’t tried it. However it still relies on one database which I think is MusicBrainz, sometimes reliable, sometimes not ( it is also one of the databases offered in PerfectTunes). However having another three options to pick from saves so much effort.

I have only encountered 2 or 3 albums that dbpower amp has not recognised. All of them admittedly pretty obscure and from the very earliest days of CD issues. So I had to manually complete every field on the metadata “form”.

In Windows version at least, DBpower amp also comes with a format converter. So if you want you can take a whole batch of files ripped from CDs in bulk and turn them into MP3s for playback on your phone. It also provides an ID3 reader so that you can amend the preset metadata from downloads or other rips without having to re-rip.

1 Like