NAS, Streamer or Local/USB Drive


I’m a long term IT professional and longer music lover, I don’t like the term audiophile, but at the end is what I’m.

Since late '90s I’m playng around with computer reproducing music, first was macintosh then in 2005 I met squeezebox and i liked so much that most of the music I listen to today come from some sort of SB, while the best - of course - still come from my turntable.

Good old black disks apart, I really believe a good digital setup could let us enjoy our music in a very convenient way and the way squeezebox environment has grown to be is probably the best in functionality and reliability. More than this and thanks to volunteers from the large squeezebox users community it’s open to be used in hardware starting from 30 US$, for free.

Volumio is making thing lot easier now and you could have a very good, full funcional squeezebox in minutes on every hw supported by Volumio.


But, what about sound quality?

I think is impossible full answer to this question: too many hardware, operating systems, software, network, dac, cable,… combinations are in play, moreover sound quality is a matter of perception and taste, so is higly subjective.

What I could say is that in my experience, in systems I’ve tried and for my personal point of …listening I’ve never felt squeezelite (the software emulator player) sounding worse than other players in the same environment (both audio and pc/networking system). Not that all the sofware players sounds the same to me, but differences are minimal and really smaller than the ones I could feel switching the hardware (same audio system, same network, same files, same software player but different hardware to me could sound different even if all of them output a bit perfect stream).

Strange? I don’t know, but this moved me to try to understand.

Reading and asking a lot I’ve found that the majority just don’t care about, the rest of us, audio enthusiast, engineers or professionals has two ‘forma mentis’:

a. It’s impossible, bit are bit, it’s just placebo effect, come back after doing some ABX double blind test…
b. really? probably is just placebo, but let we see what’s happening…

Sure who don’t care are the one who better enjoy their music, maybe “a” people are righ - quite annoying, often they are - but I feel more confortable with “b” people, also becouse some things we today accept as normal were spotted out by b people and initially refused by a people, while majority simply did not care about, so I keep on reading, asking and listening.

Doing this I’ve encountered many different opinions, but a reliable and well established company as SonicStudio, the vendor of Amarra, suprised me a lot:

“Computers with a Solid State Hard Drive installed can sound significantly smoother than their spinning hard-disk counterparts. Same goes for a Network Attached Storage (NAS): it can sound grainy as compared to a locally-connected SSHD” (

Wow, this is 180° from where the world is going today! let me try…

I have a macbook air with an SSD so the trial was easy:

a. Itunes + Audirvana.
b. LMS + squeezelite.

Just for your information, I was not able to detect any difference in both case, but this mean nothing other than for me.

After sometime, few days ago, I’ve met Volumio and discovered Michelangeloz is the founder.

Michelangeloz is a very fair citizen of many forums and it happened to me to read is mind about bad influence the network could have on sound quality, quite the same as Sonic Studio, but while Amarra is a stand alone player, I suppose most of us will use Volumio to power a network player, so for sure is not talking ‘pro domo sua’ (nor I think SonicStudio do).

He was very kind to answer my questions and suggested to open a dedicated thread here to discuss this matter, so we are.

Next two posts are my questions and his answers, hope we could continue and involve other peoplle in the discussion.


Hi Michelangelo,

could you please clarify to me your belief that playing songs residing on NAS can affect the sound quality?

Here some questions:

a. It refers specifically to the use of NAS with SAMBA share and its mount by SO Client or more in general to a negative influence of network activity, and therefore also to the transmission and reception of a stream?

b. I’m I correct to infer that you consider preferable to work in an 'isolated ’ playback environment (HDD or SSD local library or USB)?

c. In your experience, is a generalized problem or specific to an hardware / software configurations?

d. Always in your experience / opinion, can be minimized?

tanks in advance,


Here the answers from Michelangeloz.

Please note that the original conversation was in Italian, the translation is mine, so any typo, mistake or other is up to me.


In general, a data stream is less reliable than a local-read. Although it is appropriately buffered, the possibility of packet loss is always possible.
In any case, I was referring to scenarios that use SAMBA or NFS.

Yes, the reason is this:

I am not suggesting a reproduction from the local is always better than a NAS. The problems arise from the fact that an architecture that use network streams brings with it all the limitations and problems of the system in which it operates. By system I mean routers, network interfaces, network speed itself, any bottlenecks and software issues.

To achieve a good result, or one comparable to a local reading, you need to optimize and study the specific system in which it operates. Thing that is not available to everyone, and it does not guarantee certain results.

An example: I tried for months RaspyFi with a proper combination of parameters rsize and wsize with which to mount the NAS. These parameters should be fine for most users, and mitigate the problems to the bus network of raspberry.
A reading from the local has simply less chance of failure, and does not need particular optimizations since it is much more independent of the hardware and software environment in which it operates.

The fact it can spot in certain configurations makes it inherently a generalized problem.

Of course, I personally believe these steps as excellent starting points

  • Use the wired network (with galvanic isolator, for others and not so obvious reasons)
  • Optimize parameters rsize and wsize in relation to hardware and network
  • Use the latest versions of samba \ cifs
  • Make sure the processes that “read” the network streams always have higher priority than playback.

I hope to be exhaustive. I was pleased to receive this question, is not usual.


@Michelangelo thanks again for your answers.

As you suggested I opened this thread, I hope the section is the right one and hope I have not done wrong with the translation!

Resuming where we left off, I understand that you feel the local solution is preferable for these reasons:

a. Intrinsic ability to Ethernet ‘lose’ packages.
b. greater fragility / complexity and less control in a network infrastructure.

I understand that, but - I think, correct me if I’m wrong - that the problems introduced in these situations are ‘obvious’ and occur as holes or clicks during playback.

Not to mean they are easily removed, or that do not represent a major obstacle to the enjoyment, but in my opinion are more to be compared to a ‘failure’ to delete than to a negative ‘characterization’ of the reproduced sound.

In other words, a system with the above problems will not sound even acceptable and at the end or you remedied or is abandoned (except for intent of experimentation and development).

What “worries” me is that even in well designed systems, free of this ‘macro’ problems there could be a systematic (in the proper sense) sound characterization (Amarra speaks of ‘grainy’ sound ).

Thanks again,


Wow Marco,

As you say there are many variables in this discussion. I will give you my brief point of view. I sell and configure bespoke digital music hardware systems (on a very small scale by the way), so I react to what my potential client base require.

  1. Most people that have smart devices are streaming regularly from the Internet, you tube, spotify, inet radio, etc. etc.

  2. A large number of people like the idea of a central library and to play music in multiple rooms, so we arrive at the need to stream around the house too.

  3. The bandwidth involved in streaming music is not significant (particularly if we stick to CD Quality 16bit 44.1 kHz = 1411kbps).

  4. I have achieved acceptable sound quality in many different hardware and software configurations using streaming from a NAS drive.

  5. In a comparison SSD vs HDD in a NAS a colleague of mine concluded that the SSD was better but I couldn’t detect this difference. It wasn’t a blind test and he reads a lot of the audiophile stuff on the Internet !

  6. I have built Music Server PC’s with SSD drive installed. The main advantage is size, operational speed and noise IMHO. Sound quality was not an Issue.

  7. As Michaelangelo points out the Network Performance is CRITICAL. I always start with this. I will propose Home Network improvements as part of the system. I don’t agree to put in streaming equipment in a house where the home networks is “mierda” !

  8. If someone said they wanted one sound system and had a USB drive with all their music on it I would consider it best practice to use the USB drive locally mounted. I would always apply the principal of least number of failure points will be more reliable.

  9. Finally regarding sound quality, nothing I have used or tried or heard so far compares to a really good CD player connected to really nice DAC.

Hope this is a positive contribution to the discussion.


P.S. We can have a discussion about “lost packets” another day :smiley:

Hi Kevin,

pleased to meet you here, tanks for your valuable opinion, think is higly condivisible, at least from me.

If I can, I would like to kindly ask you and everibody to help me keep the discussion about file reproduction and (if and) how networks could degrade sound quality.



  1. Finally regarding sound quality, nothing I have used or tried or heard so far compares to a really good CD player connected to really nice DAC.

I agree - but I was hoping that the wolfson dac on the rpi would at least be equal since they are both using the DAC in my reciever.


It’s a complete minefield of comparisons :unamused:

I wasn’t explicit in my point 9 but I was referring to CD & DAC combinations in the £5k-£10k (for the pair) :open_mouth:

The cost/quality issue enters into the mucky world of audiophileland.

I think the Wolfson DAC + RPI is extraordinary value for money. As are many other DACs on the RPI platform :stuck_out_tongue:

So what is the receiver you are comparing with ?

I’m always up for testing and investigating. However I am always very conscious that we are in a very subjective domain with HiFi/Audio. I firmly believe if one is satisfied with audio quality/cost then then to some extent that is the end of any personal debate. However if somebody feels ripped off then that is worth questioning.


P.S. I think I’m saying the important thing is that people are happy with their investment (whether it be Hardware cost, Time spent installing stuff and experimenting). If you consider that you are getting back the enjoyment in listening to the money or time you have spent then it’s Happy Days :smiley:

Marco, the click and pops are something we’ve experienced a lot (and sometimes still experience) on the PI, but this is not strictly network related, but rather a bad implementation of USB bus (where pi’s ethernet resides).
The kind of sound degradation induced by a network, optimally configured or not, is quite intrinsic to tcp\ip protocol in general. As NWTStuff points out, the best way to tell it is that it is less optimal than USB and local storage.
There are some major issues that could affect this kind of connection:

  • Buffering affecting stochastic jitter
  • Packet Drops
  • Incorrect package size
  • Sound drops

Interesting discussion. Let me find some empirical data to keep it grounded to real matters…

Yes please, this will be a valuable piece of information.


Coming in late on this discussion, but one option I don’t see mentioned is a locally attached eSata device.

The Cubox-i for instance permits an eSata connection.

It is my understanding that USB connections for hard drives are not a good way to go for many reasons - polling delays, other traffic on the USB bus, etc.

So I am using the SATA protocol instead.

Also, reading Kevin’s notes with respect to advantages of network attached storage above, I would also like to suggest for consideration that several music computers with attached physical storage can use rsync to make sure they have copies of each others’ libraries, which seems to me to offer some extra security as well as eliminating concerns over live streaming of data files around the house.

I started out looking at hi-end dacs for my CD player in Nov '16 - the Prism Callia had caught my attention. Next step was thinking how I’d get HD audio streamed into that and whether I’d be able to hear the difference with my not-as-young-as-they-were ears, which led me to look at the Pi again which I’d played with when they first came out. I was pleasantly surprised by the choice of Dacs available and the Kali reclocker. I went for an RPi3, Piano and Kali, plugged it up and checked out Moode, Rune and, of course Volumio which is currently my favoured player.

One of the features of the platform that convinced me to go down this route was the fact that as time goes on better DACs will become available allowing you to experiment and update to the latest and greatest for a modest cost - I’d like to design a CS4398 based DAC ( beloved of Prism and Calrec ) but I’m a hardware engineer without the software skills to be able to do this other than by nailing the Dac to 44.1/16bit - any one interested in collaborating on the s/w with me?

Listening on my system ( Cyrus mono-blocks and pre-amp, Wilmslow Home Studio Monitors ) the Piano was on a par with the HHB CDR882 player I designed for HHB in '06 using an AKM combi A/D and DAC - previously I worked for AMS-Neve and Calrec audio on broadcast mixing consoles. After a few hours listening I added the Kali which took the audio to the next level of performance. For the cost the performance and musicality is excellent.
Next step was to set about ripping 1500 CD’s to Flac. A combination of Roxio easy audio + Gracenote, Bliss for first pass file tagging and covers and MP3tag made reasonably light work getting tags and cover for >95% of mainstream rock/pop/dance Cd’s and about an half of the classical ones.
I had hoped to be able to store / stream the library from a NAS drive - Synology DS116. That didn’t work out too well. Volumio reports 72Mbps but I’m still getting dropouts when streaming from the NAS - I’ve pushed up the pre-buffer to the 40% max which helped a bit. Also the RPi3 and NAS took >12 hours to scan the library, maxing out the NAS’s CPU too. It works pretty well with smaller selections presented via the NAS.
I’ve come to the conclusion that if you have a large library you will be far better off using local USB storage. 256GB USB sticks are <£50 - but watch out for very large low cost offerings on Ebay and Amazon - if it looks to good to be true it is! :blush: .
I’ve split the collection into genres. The largest being pop/rock with ~250 artists, 500 albums. Switching between views in the browser takes longer than ideal - around 10 seconds to switch in and out of pop/rock genre. The time taken is proportional to the amount of albums in a genre.

For large collections I don’t think it makes much sense to list all the Artists, Albums AND Tracks at the genre level. I think it would speed up access time a lot if the track listing was dropped at this level of the menu.

Also when rescanning the library the artist/album/track times should reset to zero at the start of the rescan - currently it doesn’t. Resetting the counters would help you know that the rescan is happening. At the moment these numbers only update once the totals exceed their previous values.

Lastly, a minor point, I think it would be good to have “Playing…By…From…” at the top of the pages in a larger point setting of the font.

Thanks for an excellent, open ended and sourced player.

I am experiencing “holes” or “dropouts” while I WiFi streaming music from NAS to my new ASUS tinkerboard. According to the hardware spec, it should not have resource conflict between USB and ethernet. I suspect these dropouts are due to WiFi. (No problem while playing music from a USB flash stick)

I am trying to adjust the router QoS setting to ensure the router gives enough priority while I stream music. But I am not sure which one should I assign the highest priority while I use Volumio. Can someone help?

Screen Shot 2017-07-24 at 3.41.23 PM.png

P.S.: Never mind the question. I just entered the MAC address of my tinkerboard to the route QoS list and assigned the highest priority to Volumio. It seems to wok so far. Sorry for the stupid question…