Volume 5 No. 3
Listening To Storage
Listening tests reveal significant sound quality differences between various digital music storage technologies.
Article By Andrew Harrison and Stephen N. Harris
E-mail HIFICRITIC By Clicking Here.

  Press 'play' on the remote control, and the chosen music selection sounds out into the room, sweet and clear. However, instead of a revolving record or compact disc, that music is now preserved as a digital snapshot, stored around the home on computer hard disks, perhaps a specific music track among tens of thousands that are randomly accessible in an instant. A decade or two ago, it would have been science fiction — but now sofa-punishing levels of convenience allow us to hear any music in our collection, without so much as opening a jewel case.

However, computer network audio is about rather more than convenience. It has as much to do with consistency, arguably more so than the 'perfect sound forever' CD ever did. Once favorite CDs have been carefully ripped into a stateless digital form — hopefully using software respected for its fastidious bit-accurate transcriptions — we just need somewhere safe to store all that data. The same goes for higher resolution material we may have archived from recordings or downloaded from online 'HD music' sites. For playback, the initial sound quality will presumably be determined by high quality D-A conversion and subsequent analogue replay chain.

Problem is, the sound of 'bit-identical' computer audio may well be just as inexplicably inconsistent as analogue. High-end audio has long been affected by seemingly insignificant environmental factors; and as we move further towards hard-disk — and, ultimately, solid-state — storage, we're discovering that another variable can unhinge the final musical experience. Only this time, there are even fewer answers why.

 

Background 
Anecdotal murmurings and some limited first-hand experience suggested that digital music files can sound different when played from different computer media sources. Take the simple playback of a stereo audio file, such as FLAC, Apple Lossless or uncompressed WAV, for example. Such a music file is typically played from either a computer's internal hard-disk drive, the network-attached storage (NAS) on the local home network (LAN), or maybe a USB thumbdrive. Is it really possible that the sound quality of bit-identical audio files' is influenced by their storage medium before being delivered to the hi-fi system's DAC?

Fellow computer audio enthusiast (and Naim PR person) Stephen Harris and I launched into some preliminary listening tests, to establish under reasonably controlled conditions if audible, repeatable differences could really be heard. We readily confirmed that the final sound quality is influenced not only by the choice of network player, DAC, digital cables, or indeed many other long-recognized factors, but additionally — and quite markedly — by the manner in which we now store large quantities of our music at home.

Initial tests compared two NAS drives from the same manufacturer, played through a reference two-channel system. The two drives were solid, reliable units and essentially equivalent — yet they also had certain intrinsic differences which we hoped might reveal any audible differences if they were present. While both had the same overall chassis and software operating system, their four-bay storage had a different array of 3.5in hard disks. Additionally, the two units also had different processor architectures, which might also affect perceived audible differences.

Having established early on that these two NAS units did indeed sound quite different, we went on to isolate differences between the individual drives, this time mounting all types concurrently in the same NAS chassis. Now we were in for an even bigger surprise.

 

Listening to Two QNAPs 
With these two quite similar NAS units to hand, a good starting point was to compare them to see just how obvious was the sound quality difference. On the face of it, there was little to differentiate the QNAP TS-439 Pro and QNAP TS-419P+ (dubbed QNAP1 and QNAP2 respectively). Both are four-bay boxes for users who need terabytes of storage capacity, with various RAID redundancy options to protect data in the event of disk failures.

QNAP1 is able to read and write files slightly faster. It uses an Intel Atom 1.6GHz single-core processor, and was configured with four 2TB Seagate LP so-called 'green' type hard disks that spin at around 5900rpm (rather than the more common 7200rpm). Though this makes it marginally quieter, a multi-bay NAS drive is rarely suited for sitting in a listening room unless inside a thoroughly sound-deadening cupboard.

QNAP2 has a very different Marvell ARM 1.6GHz processor, optimized for lower power consumption, and was fitted with four 2TB Hitachi Deskstar 7K3000 disks.

QNAP1 was found to serve up music with a similar level of rhythmic drive and image soundstaging as a good CD transport playing directly into our system's DAC. If anything, there was perceptibly more 'drive', in the sense of bass euphony and articulation, but this came with increased level which made the sound a tad bass heavy.

Also, QNAP1 did not sound as clean as CD in the higher registers. Some edgy grain exaggerated the sampled horns that sets the scene in the opening of Primal Scream's Loaded, adding to the color but nudging it off neutrality. Splash cymbals lived up to their name.

QNAP2 rendered the same song more tunefully. It was more organic and made more sense, the lines of melody and rhythm cooperating better. As well as showing better individual instrument distinction, the whole piece sounded tidier, tonally less messy without the roughened HF, and perhaps better integrated in musical intent.

So that's rocky electronica, sounding a bit more wiry and rougher through one NAS drive compared to another.

Penguin Café Orchestra's Union Café has an altogether more natural recorded acoustic. On Scherzo and Trio QNAP1 promoted the leading edge piano transients, following through with a lighter, brighter instrument tone – possibly Steinway-like? The same piano had more lower mid body on QNAP2 and slightly softer hammer impact, perhaps more like a Bosendörfer.

That hint of glaze on QNAP1 also showed an impaired subjective noise floor elsewhere. In hi-fi parlance, QNAP2 had the blacker silences and deeper spaces between notes. If anything, this track highlighted a fundamental shift in timbre between the storage sources. This wasn't the gentle tweak of a DAC's digital filter option; we felt it was more akin to changing loudspeakers. System sound was improved as if the DAC itself had been upgraded, say from a £500 to a £2000 model.

Madonna's Swim also proved illuminating. William Orbit's judicious dollops of bass synth showed how the QNAP1 had been exaggerating some of that low energy (albeit at the higher end of the range, so leaving 'infrasonic' impact somewhat weakened). QNAP2 was undoubtedly tighter here, and more disciplined.

So, having lined up two NAS drives from the same manufacturer, mixing up their disks and processors, it took just a few tracks to render the sonic differences obvious. Now it was time to try and isolate parts of the storage chains. For example, we had no idea if we were hearing differences in the processor architecture or between the hard-disk (or solid state) drives.

 

Different Drives, Same NAS 
While most NAS drives fit the standard 3.5in Serial ATA hard disk, it's not uncommon now to find support for the laptop-size 2.5in units too. The Synology DS411slim is built expressly to take these smaller hard drives.

The DS411slim is an ARM-powered NAS. Like QNAP it uses a versatile Linux operating system, but Synology's firmware allows easy setup of individual, separately addressable volumes. We selected two traditional hard-disk drives (HDDs), and two solid-state drives (SSDs): a 500GB Hitachi Travelstar 7K500,a 500GB Seagate Momentus 7200.4, a 128GB Kingston SSDNow and a 120GB Corsair F120.

In our initial listening tests, I couldn't discern any tangible difference in sound between the two hard drives. Harris thought the Hitachi sounded very ethereal, almost out of phase, and rated it lowest; the Seagate was sharper with a more thumpy bass, slightly brighter with a slight tendency to sibilance. Both lacked much drive in presenting the Madonna track, and were certainly 'mushy' compared with the best sound quality we'd heard from the QNAP stable.

Drive three (a solid state type) gave a far from subtle shift in tone and soundstaging. I thought that here this Kingston SSD spread the stage wider, could really pull apart the multi-track layers, and certainly led in blackness too, sounding agreeably quieter than it had any right to. Yet there was also a dull flatness to its presentation, even a graying of timbre.

If the Kingston SSD stood apart from the disk drives for its mostly good yet quite alien character, drive four made itself known for entirely the wrong reasons. This Corsair drive (another SSD) conspicuously highlighted vocal sibilants, and had a hard, relentless quality that was impossible to miss. Strangely, it also robbed the music of pace; it was the least engaging on any emotional level thanks to an enveloping tunelessness that appeared to carve up a song like an MP3 rip.

(It may be quite significant that the Corsair Force Series F120 takes a rather unusual approach to handling its data I/O. The Sandforce SF1200 controller chip uses what Sandforce calls DuraClass technology, which suggests this SSD's microcontroller is exerting itself more than usual in order to maintain a high throughput of data.)

In a quick personal ranking to date, covering all these tests, my notes put the Kingston SSD's sound quality first — with hindsight more a cerebral than an emotional choice — then the QNAP2, followed by either of the 2.5in drives in the Synology DS411slim. QNAP2 still had perhaps the most engaging bass performance of the crop, but its involving musicality came at the expense of a somewhat narrower soundstage.

 

The System 
The reference system comprised a dCS Purcell upsampler and Delius DAC, feeding a Music First Audio system controller and Chord Electronics SPM1200C power amp, and Bowers & Wilkins 802D loudspeakers. Cabling was predominantly Nordost, with an Isotek Titan mains conditioner.

The link between the NAS units and the traditional hi-fi comprised two Naim units, a UnitiServe and an NDX. The UnitiServe was primarily used as a CD ripping tool and a network media server. The Naim NDX is a network music streamer with onboard DACs, bypassed here to use the dCS two-box DAC.

The various storage drives were sited in another room, connected to a Cisco Linksys E4200 wireless gigabit router in the listening room via 25m of Belkin Cat 6 Ethernet cable. Another gigabit switch in the remote room (a Netgear ProSafe GS108), enabled several NAS units to be online at the same time, each connected to the router/switch with high-quality Cat 5e patch cable.

 

Tentative Conclusions
This initial trial was not intended to be an exhaustive study into all the factors that can affect the sound quality of network and computer audio, only to confirm or deny the suspicion that digital bitstream coming from hard disks are not all equal. Which has to be somewhat surprising, to say the least.

By now we should know better, and acknowledge that digital audio is very far from immutable. Most of the troubling inconsistencies in CD playback, for example, have been at the point of domain conversion, typically the digital-to-analogue conversion in the playback chain. But the movement of digital bitstreams around the system, such as over a 75-Ohm cable from disc transport to DAC, have been prone to known and definable transmission problems — more so for the venerable S/PDIF standard that rolls up the clock signal within the data channel.

Carriage of digital data over Ethernet, for example, ought to be less troubled by the vagaries of the cable and interfaces. And what of the data being unraveled at the source, from the hard disk or solid-state drive? Why should the type or brand of the disk have such an audible impact on the final sound quality? Clearly, there's scope for more hypotheses to be set and tests undertaken, with lab measurement as well as subjective listening. Maybe we can solicit logical explanations from engineers who understand the low-level mechanics and operation of computer file and storage technologies, and can suggest specific avenues to explore.

This first test was also not intended as a buyers' guide to decide which disk or NAS is likely to work best in high-end audio applications. But our findings suggest there may certainly be some mileage in such reviews, once the several variables have been reduced sufficiently to be sure we're hearing the fingerprint of a specific NAS unit, or hard disk, or perhaps the way the disks are combined in a particular array.

 

Another Wildcard: The Striped Synology 
here are many variables in network audio, as many (quite possibly more) than in an analogue-only hi-fi system. Take RAID arrays. A RAID ('redundant array of independent disks) is a neat way to combine the capacity of several smaller disks to make one larger addressable volume.

A number of disks may be combined in three main ways, with benefits like increasing read/write/latency performance, or building in succinct redundancy by sharing the same data between disks, so that whole disks can be pulled out without any overall data loss. Some RAID setups — arguably the best, given at least four to play with — improve both speed and safety. These are RAID 5 (with enough redundancy to allow one disk's complete loss/ removal); RAID 6 (two disks); and RAID 10 (one or two, depending on which of the four actually expire).

The RAIDing of volumes could be the subject of a further investigation, to see if compounding disks may even exaggerate their audible signatures. For this initial study, we pulled out another Synology NAS unit, a DS211 two-bay box with two 2TB Western Digital RE4-GP 'green' disks set up in RAID 0. In other words, the two disks have data striped across both to augment performance. In the best case something like the sum of each individual drive's data throughput can be exploited, but at the risk of total data loss in the event of one disk expiring.

While this additional test was in no way scientific, we thought it would be worth finding out if this alternative NAS box/disk/RAID configuration offered anything different in our initial delve into the sonic differences of storage systems.

As it turned out, it was possibly the best sounding source yet. It could sustain pace and drive, and gave body and richness to music where the Kingston SSD, for example, had been heard as limpid and lightweight. Maybe higher frequencies still weren't as insightful as direct CD playback at its best, but the sound had a relaxed quality that this listener has found quite enticing enough to plan a migration of all music onto it — pending a test of other NAS combinations!

 

E-mail HIFICRITIC By Clicking Here.

 

Subscribe!

Click here to subscribe to HIFICRITIC