NAS vs. internal

August 4, 2021
 by Paul McGowan

A growing trend in media servers is to employ the use of a NAS rather than rely upon the built in HD inside of a music server.

It’s probably not a bad idea.

NAS is an acronym for Network Attached Storage—a hard drive accessible over a home network. For years, NAS were shunned by most of us (including me) because they were slow and a pain in the keester. But, like all things digital, over time they’ve gotten easy and fast to the point they are hardly distinguishable from an internal HD.

The advantage of a NAS over an internal hard drive is noise, or better said, the lack of noise. In most products, components built-in to a single chassis must share a power supply, ground, and close proximity. In digital audio, that can be a recipe for degraded sonics due to internal noise.

A NAS has its own power supply and is physically and electrically isolated from the server. Your NAS could easily be in the basement while serving digital audio to your living room system.

In the upcoming release of PS Audio’s PerfectWave Octave streamer we’ve all along been planning on an internal hard drive because, well, that’s what people expect. But more than simply expectations, the added simplicity and reduced number of components required to build an all-in-one media server do not contribute to sonic degradation in our system because of our galvanically isolated output stage—the AirGap Interface.

That said, we’re about the only folks designing such equipment.

Without the advantages offered by isolating the media server’s output stage, it might be a good idea instead to consider a NAS.

Subscribe to Paul's Posts

37 comments on “NAS vs. internal”

  1. With my first streamer (a Linn Akurate DS) which I bought in 2009 it only works with network storage, and that is still the case. I used an Iomega network drive, basically a hard drive with a bit of network hardware included. It was pretty much plug and play.

    I’ve had an Innuos server for just over two years, which is galvanically isolated from my music player as it sends data over an ethernet cable that carries no power.

    Power issues are dealt with by having multiple internal power supplies – the more you want, the more you pay.

    Power to the internal hard drive is turned off whenever possible.

    The biggest problem for those running Roon is the electrical noise from running Roon Core, and some find it better to run Roon Core on a separate server.

    Innuos designed their own streaming software optimised for their hardware and designed to run at very low power and noise levels. It does sound better and I’ve moved Roon to an external server.

    There are numerous solutions at a vast range of prices, and galvanic isolation is fairly widespread, it seems over the last 15+ years to boil down to a balance between convenience and the last 0.1% of sound quality.

      1. It’s a popular budget device, but it only does Roon and I chose to pay for a device that is n to tied to Roon and with a better power design.

        1. I am only concerned with the digital noise generated inside the box. My SSD drive is external. I never mounted a drive inside the Nucleus. I have no problem with being a ROON only user.

    1. I have a music server with a large cache memory and an internal hard drive. The server will stream, which I use for Radio. For my own music I store my 3TB music library on the servers internal hard disc. When music is selected on the server then it turns on the hard disc and loads the cache memory after which it shuts down the hard disc. This solution seems to make good sense as electrical or mechanical noise from the hard disc is eliminated as is communication hash associated with the music files themselves (there is still some communication for control purposes).

  2. Although I personally use a NAS on a separate from audio power cirquit myself and galvanically isolate the LAN connection from it, after a 2 hour call recently with a very knowledgable person of another leading streaming server manufacturer (to get a bit deeper into the general matter), I see it more differentiated.

    And although I’m keen on the PS Audio approach, PS Audio is certainly NOT the only company caring for a best possible isolation of the server internals…maybe different than others.

    Some interesting approaches, findings and opinions of different manufacturers here:

    – noise not only comes over the power cirquit, but also through LAN or other signal connections

    – not all the noise can be filtered/suppressed by galvanic isolation

    – some manufacturers say, the use and need of galvanic isolation somehow proves the missing avoidance of noise beforehand

    – normal SSD’s are common standard, but slow spinning HDD’s with as low as possible power consumption are sonically better. Reason is, that SSD’s contain a processor causing noise and HDD’s not. Their spinning drive is also suboptimal, but not as bad.

    – the proper solution for a superior SSD integration is to modify its firmware to be able to completely switch it off when tracks are playing (preferable for HDD’s, too). This firmware modification option seems very rare.

    – a proper isolated/noise optimized server solution with included hard drive should be better than a NAS integration running into a streamer, as there measures can be applied to the drive, which are not available for external components

    – an example for the previous mention is, that tracks can be buffered in RAM before playing, which seems not possible yet for external drives. If buffering or a temporarily switched off hard drive are not realized, there might be no real advantage against a NAS.

    – as long as there are not proper solutions for “audiophile” routers, switches and other network components and their power supplies (no matter if in or out of the audio power cirquit), the effect of later optimizations are limited

    – extensively optimized power supplies for servers/streamers (besides other isolation efforts and output stages) are a main factor for good sound

    These are just few approaches of one other manufacturer and I think it will be interesting, which of those PS Audio also addresses or not, or additionally. My wish would be a unit, that enables at least stream buffering also for external drives like a NAS, which the other manufacturer took into his perspective.

    What we see is, that there are multiple approaches and options and that streaming is a can of worms, hardly fully optimizable. It seems optimal digital is just possible (in terms of the source) when everything (physical spinner, hard disk etc.) is not only isolated but temporarily switched off during play (which can lead to some operating/convenience limitations).

    I would love to have a deeper than marketing level exchange on the PS Audio streaming offers here, too.

    1. That’s an excellent list and, whoever it is, it is not dissimilar to what Innuos do. One of the reasons for their new operating system is that running their device as a Roon Core does not permit buffering. Roon forces servers to operate in a way that is not optimal for sound quality, because Roon is more about functionality. It is not a bad thing, it is just a design choice. A chose an HDD drive rather than SSD.

      One thing you missed is that a decent server should have good vibration damping. Innuos include 3 asymmetrically placed feet and rubber inside the lid. The latter is a good idea and an easy $1 fix with a sheet of Dodomat.

      1. All this talk about noise and galvanic isolation, that is nice.
        Has anyone showed in a controlled experiment that al\ll those “noises” are detectable by sensitive instrumentation or not?
        Has anyone shown that the signal reaching the amplifiers is any different in those systems compared to others that don’t have it?
        Or is this more of the mythology based audiophilia nervosa?

        1. Possibly that’s the same challenge as measuring soundstaging improvements from the speaker’s sound wave output. Audible, existing but not measurable 😉

          1. If it is so easily heard, then you should be able to demonstrate that in a “controlled” experiment.

            Toole did it with speakers for decades. He decided that mythology did not help him design better speakers, so he developed the science for it.

            If we can measure the time since a black hole swallowed some kind of dead star, we should be able to figure out this.

            Paul, I know you don’t like it, but that doesn’t make it true.

            1. And yet you haven’t. The modern human central nervous has been a work in progress
              for well over a 100,000 years with literally billions of trial and error experiments in the great laboratory called Life on Earth. It is by far the most elaborated and sophisticated system in known space. Your spectrum analyzers et al. are glorified Tinker Toys by comparison. But keep the work up, you may discover something that you can measure and quantify in electro-acoustic performance that correlates to what messy, biochemical people actually listening to and enjoying music experience. Blind (deaf) allegiance to orthodoxy has not come up with anything beyond gross approximations so far, and is not likely to do so.

              p.s. I know about LIGO and the confirmation of the existence of the long hypothesized gravitational waves in the 4-dimensional space-time continuum. Professor Janna Levin had the sense of humor and apparent love of music to name her book Black Hole Blues and Other Songs from Outer Space (highly recomended by old, hand waving, environmental geologist emeritus, by the way).

          2. Paul and CtA, I can see both of your viewpoints and agree with both of them! I am a retired physicist, so I love measurements and hard data and I am an audiophile who understands that there are things that we can hear that cannot be measured using today’s technology. Maybe someday, but not today.

            I was a beta tester for the new PST ( Perfectwave SACD Transport ) which replaced my old DMP ( Directstream Memory Player ). When I got the PST I got my wife to do what I call controlled listening with me. We listened to 4 tracks from 3 SACD’s first on the DMP and then on the PST. It took me about one to two minutes to completely unplug the DMP and replace it with the PST using the same cables. I checked equal volume levels with my SPL meter. We took notes while listening and there was not talking until the listening was over. My wife knows nothing about galvanic isolation or any of the technology. To her it is old box versus new box. When we compared notes and talked we both had the same conclusion: The sound from the PST was clearer and cleaner than the sound from the DMP.

            Paul, You are right, it is easily heard.

            Paul, As a scientist I wonder if this has already been measured. When the DMP was being developed were there routine measurements made on its performance? If so, were those same kind of measurements made on the PST when it was being developed? If these measurements exist can they be compared to see if there is a lowering of the noise floor?

        2. Having listened mostly from streaming devices for ever 10 years, the improvement in sound quality has been quite dramatic. The signal does not change, but electrical noise on the signal path is very easy to hear.

          Paul is correct, no mythology.

          It is not very expensive, but it help[s to consider the entire signal path from the consumer unit onwards.

  3. Today’s post raises a few questions for me.
    “…PS Audio’s PerfectWave Octave streamer…” and a little further “an all-in-one media server”
    1. So the upcoming Octave is a COMBINATION of a STREAMER (a.k.a. networkplayer) and a SERVER (storage). (like the ones from Innuos). Can it also be used as a stand alone server or stand alone streamer ?
    2. Does the Octave make the built-in streamer in the Direct Stream DAC/DSJ (Bridge2) obsolete ?
    3. Is the internal storage HDD or SSD. Or both in the same device, with HDD storage and SSD caching for playback. (like the ones from Aurender). Does the customer have a choice ?
    4. Will the Octave have its own DAC (I think not).

    1. I could answer most questions from Paul’s previous info in the forum, but leave it to Paul here.

      What I wanted to tell to your topic 3. is, that as I described further above…a caching of a HDD by SSD would make no sense, as it would rather make things worse than better. The caching should then be done in a memory block within the streamer, not inside an add. unit coming with one more processor and its own memory block. So far my understanding. Reading/playing from a connected SSD is not better than from a spinning disc, may be even worse. That’s what a friend I trust practically verified once and what a manufacturer of very sophisticated streaming servers confirmed to me.

      1. Over and over again I am “accused” of things I never wrote/meant.
        Do people not read carefully enough, is it the foreign language (nuances), or what…
        Anyway, in my comment today I did NOT write about a CONNECTED SSD, but about an INTERNAL HDD and INTERNAL SSD in the same device.
        Like, as I wrote (!), in Aurender devices.
        And since their streamer/servers are the measure of things this can’t be bad (understament).

        Off topic, but addition to yesterday : mr. “Blowin’ in the wind” accused me of complaining.
        Well, the only one complaining (about my comment) and almost ordering (haha) me to stop was he.
        In my comment I only wrote what the situation in the real world is. Not my opinion, by no means a complaint, just the facts.
        And yes, call it as you like, but we are seduced by advertisements of manufacturers/magazines.
        Do I blame them ? NO, they have to make a profit, I get that.
        But it is a fact, audiophiles are notorious for “upgrading” (more than in any other hobby !), and susceptible to the beautiful words of PmcG and other manufacturers.
        If that wasn’t the case, we’d all have the same setup for 30 years or so.
        So, no steady (solid) bedrock, but an everchanging bedrock (what I called a swamp).
        And of course nobody is obliged to buy new stuff, but that’s not the point.

        1. Then you were not misunderstood, I meant the same. An internal SSD is anyway just another device with its own processor, no matter if connected inside the housing of the server or external.

          Aside of that, whatever the one or other manufacturer does, it’s always the best in his eyes, but not all equally good in an absolute sense (for whoever can judge this 😉 )

    2. 1. So the upcoming Octave is a COMBINATION of a STREAMER (a.k.a. networkplayer) and a SERVER (storage). (like the ones from Innuos). Can it also be used as a stand alone server or stand alone streamer ?

      The upcoming Octave streamer will not have any built in storage. Instead, it will have several USB ports available to connect external HDs for that purpose.

      2. Does the Octave make the built-in streamer in the Direct Stream DAC/DSJ (Bridge2) obsolete ?

      Yes, indeed it does. The upcoming Mark II DS will not have a Bridge option. This upcoming standalone device will be the new means by which our products connect to the outside world.

      3. Is the internal storage HDD or SSD. Or both in the same device, with HDD storage and SSD caching for playback. (like the ones from Aurender). Does the customer have a choice ?
      4. Will the Octave have its own DAC (I think not).

      No. We’re still on track for a separate new version of the DirectStream due out late this year.

      1. Regardless of storage type, a failsafe architecture is needed for music collections. You said multiple external drives through USB (I assume 3) will be present. If so, the entire collection needs to be managed under a single logical tree structure. The USB needs to be the interface to a RAID architecture for purposes of never loosing a drive and thus the data. NAS will tend to support both USB and ethernet interfaces, the latter allowing for a home network and the NAS server thus isolated. Sure, there are backups. I run about 30 TB into S3, but that is not sufficiently satisfactory. Local storage needs to be redundant in any product.

        1. Exactly. All digital data should be backed up, preferably with more than one system and even by diverse techniques and off site, i.e. RAID and/or Cloud storage. The reliability of HDDs has hopefully improved, although increasing data density frequently also increases fragility. SSDs are probably better, but again ever increasing data densities on the chips also contributes to reduced ruggedness. And your connection to the Cloud is subjective to less than ideal internet connections and not infrequently malevolent human actors. Your data can all go “Ker-poof!” with the greatest of ease.

          I also realize that my CDs, SACDs, LPs, et al. are also subject to loss by catastrophic fire or sudden flooding. Some of the aforementioned bad actors could pull up with a bogus moving van and clear the place out in my absence. It happens.

          Again it’s all a balancing act and an ultimately futile striving against entropy. “Things fall apart, it’s scientific.” — Talking heads.

      2. I’d love to (in the forum) know more about PSA’s strategy and experience with the effects and impact of the grade of possible isolation from noisy sources vs. their deactivation/avoidance vs. importance of power supply vs. importance of output stage…and all that in relation to other upgrade options like we know from firmware, transformer and ext. PSU upgrades of the DS DAC.

        I suspect the impact for the buck goes down the further we get into the rabbit hole for an ideal realization of the streaming source infrastructure.

  4. Paul, I agree with your observation that NAS technology has greatly improved since the concept was first introduced. With optimized redundant disk schemes to protect against disc failures, very fast storage access times and gigabit network speeds, today’s NAS servers have comparable performance to that of a connected device’s internal storage. The acronym has even gone through some evolution, now also interpreted as Network Aware Storage, in order to transcend the idea that such devices need to be physically ‘Attached’ (tethered by cables) to a network. This lends itself to the idea of improved galvanic isolation schemes for use with music servers.

    Like you I also think that modularization or componentization has the useful properties, at least intuitively, of better isolation potential and replacement flexibility as functional and performance improvements are developed.

    1. My problem and why I use a NAS is, that there’s still no SSD in a server with more than 8TB. And the backup flexibility is probably bigger than the one of a complete server. Ok, and it’s much cheaper than a high storage capacity inside a server.

      1. It makes for an interesting architectural question, if I interpret your comment correctly. My view is that the best configuration is to keep the DAC, music server and storage server components separate. I believe that the AirGap interface Paul will use to isolate the output stage will effectively prevent any upstream distortions from polluting the analog signal, whether they be from the processing activities of the music server, moving data from HDD, SSD, LAN, WLAN or the internet. This, of course, assumes that the bitstream that ultimately enters the DAC is already bit-perfect.

        1. DAC and streamer separate: definitely yes

          What Paul says about streamer/server and storage separated is, that, if no care is taken for isolation inside a server, a separation (e.g. NAS) is probably better. I agree with that.

          What Paul also says is, that PS Audio cares for proper isolation between hard disc, server/streamer within the upcoming server unit, so there should be no disadvantage to a separate NAS architecture use. Nothing to doubt here, too.

          What I say after my research is, that a server with proper isolation AND caching/switching off SSD during playback should even have advantages against an external NAS integration.

          In case folks have the NAS and router/switches not separated from the audio power circuit, I’d personally say, that a well designed, internally cross-isolated integrated server should be better than a NAS infrastructure generally, even without caching/switching off SSD during playback.

          Finally we all have to choose the level of noise avoidance we want to apply for the money and find out the impact of those influences in relation to the expense.

          I guess the PSA transport should be a good reference and Paul will know what’s worth to do to get close to that level. The reason for me not to use a transport is missing convenience of streaming and non-access to the huge availability of hires sources aside SACD and the reduction to redbook format for the main amount of digital albums.

    1. Thanks for the simplest and honest response.

      Perhaps the order of ascendance of my listening choices have been:

      Lowest: Streaming as a means of reviewing new music that I hear about and locating other musicians and their recordings that I discover through ROON. The sound is still quite reasonable.

      Second: Paying CDs on a quality player. An improvement of the digital signature and highly listenable leading to…

      Highest: Analog recordings because many records are most capable of producing the finest quality musical reproduction. Thanks for bringing me back to reality MF.

  5. I switched to a NAS for all of my files a few years ago and have not looked back. They can be accessed by any of my connected devices in the home for remote listening other than where the NAS is located. I upgraded to a Synology product a year ago and that has been even more rewarding. I use an AudioQuest Jitterbug in the USB port of the NAS (of dubious value, I know, but it was $50 so who cares ha ha) as well as other Jitterbugs in the router, the computer that runs Roon.

    I have debated a Noise Harvester or two for the system, does anyone here use them? All my computer related pieces and parts are plugged into battery backup units for when the power goes south, those APC units are probably very noisy in their own right.

    I did buy a HumZero from Scott’s promotion email the other day but have yet to employ it in my system.

  6. For nearly 20 years I’ve been using Synology NAS units for my network type clients.
    I keep one for testing on my bench and another in my office for our two company’s data and personal data as well as music.
    I’ve tested several streamers over the last couple years and a couple of them require some thing that doesn’t work well on the NAS.
    The distinction here is access to “open shares, or a DLNA server. I tried a few that will run *in the NAS” vs requiring it to be running on yet another PC – did not like the results, as it indexed my music “share” in a stunningly unusable manner.
    I looked at Roon, but found it unnecessarily complex… I just need to access music in the NAS.
    An older Node 2i works just fine and is not gross with an external DAC, but will not stream DSD.
    All Lumin units require DLNA.
    A few others sounded really good, but had the DLNA issue.

    I have a Technica SL-G700 SACD/ streamer on the way and it has rave reviews for sound/build quality, RCA/balanced analog outs as well as all the other accoutrements for in/out and excellent DACs in a dual parallel setup, excellent hardware build, extreme PS isolation and all that.

    I’m anxiously awaiting the PS Audio streamer, especially since I hear the price-point may be perfect, and may ship this year.

    So, if I have input at all, please make it simple and not require DLNA.

    PS – with any internal/external storage (USB), if one desires to access that across the network (like to copy media to it), it will require some sort of basic server software, and you will get into configuration and/or licensing issues issues.

    Thus it makes sense to keep this a pure streamer only.
    Cheers all.

    Thanks for these posts Mr. McGowan!

    John in Seattle, WA, USA

  7. After reading all of today’s comments so far , I’ve decided I’m pretty much a networking moron. Plus I don’t stream anywhere but to my main rig.

    So ease of set-up / detailed instructions will become paramount for me to consider expanding into networking.

    I keep 2 to 3 forms of back up. One ‘imaged’ hard drive of the main one in use. The main one in use is not internal but on a separate buss (currently thunderbolt). The internal drive only has the OS and playback program on it.

    The other 2 forms of back up are data dvd’s and 256 / 512 GB jump drives that live in a safety deposit box off site.

    Since I’m a neophyte with this streaming stuff, what’s the advantage of
    Something like an Octave server?

      1. Hey jazznut,

        So in its most basic form something like the Octave is a single purpose computer optimized for routing music playback of digital files? The special attention being given to noise isolation and / or handy features?

        1. Yes I’d say that catches it well when we speak of something like the Octave server.
          If we speak of the pure streamer, you’d have to care for an online or USB storage and its connection, setup and isolation yourself.

Leave a Reply

Stop by for a tour:
Mon-Fri, 8:30am-5pm MST

4865 Sterling Dr.
Boulder, CO 80301

Join the hi-fi family

Stop by for a tour:
4865 Sterling Dr.
Boulder, CO 80301

Join the hi-fi family

linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram