HDX vs. UServe - via nDAC and via NDX; are we failing by the source components?

Posted by: aysil on 09 April 2011

NDX is out, and there is a lot of excitement around its sound and different comparisons with similar devices, with nDAC, different PS, etc. Are we thereby overlooking another important question – the importance of the source component?


Part one: different source components via nDAC


When I had, in an earlier post, reported big sound quality differences between HDX and UServe through my earlier DAC, some members asked if I hear the same differences through my nDAC. Some suggested the differences would diminish through nDAC. I was also very curious.


One by one, I connected HDX and UServe to the same input of nDAC with the same digital cable and listened to different kinds of music from the rips I had made to the on-board hard drives of both devices. (Both are non-ssd versions) The results were identical to my previous findings using my previous dac: “UnitiServe sounds more nervous, a bit more rough, and the micro-dynamics more flat (in comparison to HDX). The HDX has a more relaxed presentation, more resolved, better flow, and more precise rhythm.” I can add further: HDX has deeper soundstage and more correct tone colors.


I don’t know why some members on this forum did not hear the difference in their own system. (It is not a negligible one. My friend accompanying me during the auditioning did not even believe we were playing the same track!) Of course, in a more revealing system the differences would be augmented. My Kondo Souga amp is definitely an extremely revealing amp; annoyingly revealing at times! I hear the different positioning angles of microphones, and thereby sometimes “see” a distorted soundstage or a bent forward piano or things like that. However, any system from the different Naim ranges would also have, to varying degrees, the resolution to convey this kind of differences. (Edit: it turned out later in this thread that those members who have compared these two devices had used a NAS and they were playing the same file on the NAS, contrary to my case. See later on this thread)


I am not writing these to discredit UServe. Its weaknesses are only in comparison to HDX. UServe is probably one of the best sources in its own price category. I wish I could compare it to CD5XS, which is almost the same price. I asked our distributor, but it was not in stock at that time. The only thing I could find at hand for comparison was a less than £1k CD player from a neighbor. It was definitely much worse. Compared to UServe, it had much less detail, and a blurred and muddy presentation.


CONCLUSION 1:


Naim DAC’s “buffering/re-clocking architecture” is Naim’s ingenious solution for the task of GETTING THE BEST OUT OF A DIGITAL SOURCE. It is by no means there to disguise the quality differences of source components, nor does it make up for the weaknesses of inferior sources! Unfortunately, we just cannot escape the issue of the quality of the source component.

Part two: different source components via NDX


What happens when NDX comes into the picture and we use UPnP streaming instead of the S/PDIF connection? Does the characteristic of the source component play as important a role? (With source component, I mean the component which comes before the NDX, for reasons I will explain below) There are already some test reports online, which compare different source components connected to Linn DS devices and describe the differences. In fact, in one of them, HDX scores much better than RipNAS, for example. I wanted to reach my own conclusions. Taking HDX and UServe as two examples of different computer audio source components, I connected both to the modem with the same length, brand of CAT6 cable and auditioned them through the UPnP input of NDX.


First, they both sound quite different through UPnP. UPnP transfer and S/PDIF transfer have very different sound characteristics. I don’t have a clear winner yet; they have different strengths and weaknesses. (This should be the topic of another post.) However, the difference between HDX and UServe remains! In fact, the difference is much bigger because UPnP is the more detailed (and revealing) method of digital transfer. The characteristics of the two source components are exactly the same as above in the previous comparison. It can be quite disappointing to switch from HDX to UServe. The beautiful silky textures of string instruments from HDX give way to a somewhat scratchy presentation from UServe, and the nicely swinging body of the clarinetist from HDX gives way to a somewhat more stiff musician body from UServe.


I must say again: these remarks should not discredit UServe for those who are enjoying this device in their Uniti systems. The description is only relative to HDX, and in regard to HDD versions. The only conclusion to draw is that different computer sources may sound significantly different. (HDX and UServe are, in principle, dedicated computers.)


One interesting note: the difference was consistent whichever server I used to stream the data, HDX server, UServe server, or Asset server on my laptop. That means, for example: when I used the UServe as server, the sound was better streaming data from HDX’ hard disk drive than streaming data from its own hard disk drive. (Edit: this actually coincides with the experience of some members who have compared the devices when playing the same file from the same location and did not find a major difference. See later on this thread)


I cannot know to what extent the above-described differences btw the two devices stem from the actual ripping and to what extent they occur during the process of reading the data from the drives. Many would assume that if the ripping engine and algorithms of the HDX and UnitiServe are the same, the data on the drives should be the same. However, I don’t want to rule out the effect of the different cd-drives of these devices on the supposedly same ripping engines. One interesting test would be to change the save location of the rips and save the rips of both devices to a common third location on the network. (I don’t have a NAS, so this could be a folder in my laptop computer.) I will try to do this test during the next weeks and report to you. It will be interesting to see if the rips from these two devices will sound different when saved to the same third location, and the sonic differences between these rips and the rips saved directly onto the on-board drives.


CONCLUSION 2:


During the online chat on NDX, Gary Crocker called the NDX a “source component”. I doubt if this is the correct term. The digital audio data is processed in the inner topology of a computer source before being transferred to NDX. Therefore, we should call NDX, like other “renderers,” an intermediary component, where the computer IS the source component.


If two computer sources from the same manufacturer can sound so different, I guess different computers, different NAS, even different usb sticks sound considerably different. In fact, I start to believe that, with regard to computer audio, we are in a similar phase to the early 1980’s with regard to cd-playback, when most people believed cd-transports have no influence on sound quality, as “they were merely transferring the ones and zeros to the dac”!


We will soon find out, how the power supply of our computer, or the material of the outer case, or the brand of the hard drive, or the quality of the inner cabling, or any element of the topology of our computer, or even the CAT cable we use to connect to the network influences the sound. When I search back on this forum, I actually find many posts that point at such influences. They have not been the focus of attention of the majority. As the quality of the “renderer” components rise, such aspects will become more audible.


There are some members on this forum, like Tog, who regularly advise that any worry about the influence of a factor on the sound would make sense only if it makes a considerable impact in the current level of the whole set-up, and that we should be able to sit back and simply enjoy the music! I find such warnings very valuable, because audiophiles do tend to forget this basic fact and get lost in worries and details. However, we cannot escape such worries altogether as it is this quest for better sound that brings us together in this forum. Total piece of the mind exists only in Heaven!

Posted on: 13 April 2011 by Tog

Now let me repeat and be more specific on my claim: if two uPnP servers stream data from the same NAS to the same Client streamer and the sound differs this means that one or both uPnP servers are FAULTY (I mean they FLIP the bits occasionally).

 

I would dispute your logic here. If the streams are of identical files served over the same UPnP it is far more likely that any percieved differences will be generated by the renderer or the perception of the listener. You find a fault by reducing the number of variables to a level where you can identify with reasonable probability where the issue lies.

 

Two identical servers = issue either with media (unlikely with identical rips)  or wifi/ethernet (unlikely if same network infastructure is used. THEREFORE > problem lies with renderer / dac / amplification / speakers or power supply. Keep reducing the differences until the issue is resolved.

 

IMHO - the biggest varaible will be the human being, so reduce that variable by staying sober.

 

Tog

 

Posted on: 13 April 2011 by DavidDever
Originally Posted by AMA:
Originally Posted by aysil:
It's obvious that in digital audio, reading the bits correctly is not enough for sound quality, unlike Goldman Sachs banking operations.

Asyl,

 

Once the bit is safely delivered to the buffer of re-clocking streamer/DAC it will always produce the same sonic signature, no matter of bit origin (providing the power for output gain stage does not change in time).

 

Reading the bits correctly from FLAC is also NOT an issue.

 

The actual challenge of network audio streaming is to transmit the bits from NAS to the streamer/DAC memory buffer. It's quite a long chain: transcoding the bits in uPnP, encoding them into TCP packages, sending the packs over ethernet/Wi-Fi, receiving the packs with streamer's ethernet receiver, decoding the packs, storing the bits into a memory buffer for further re-clocking.

The drivers of ethernet generators and receiver provide multiple check ups to secure the packs are transmitted bit-perfectly.That's it -- since that moment the sonic signature will be pre-defined.

 

Theoretically the ethernet transmission chain can be flawless as the existing throughput standards exceeds the audio bitrates by a big margin (even for hi-res).

 

Now let me repeat and be more specific on my claim: if two uPnP servers stream data from the same NAS to the same Client streamer and the sound differs this means that one or both uPnP servers are FAULTY (I mean they FLIP the bits occasionally).

Faulty is a strong word–but under-powered, under-specified, over-taxed all come to mind–and very frequent, I might add, in the NAS world.

Posted on: 13 April 2011 by likesmusic
Originally Posted by DavidDever:

Faulty is a strong word–but under-powered, under-specified, over-taxed all come to mind–and very frequent, I might add, in the NAS world.

So, in the context of the original post, which was under-powered, under-specified or over-taxed - the UnitiServe or the HDX?

Posted on: 13 April 2011 by aysil
Originally Posted by AMA:
...

 

Now let me repeat and be more specific on my claim: if two uPnP servers stream data from the same NAS to the same Client streamer ...

 

Originally Posted by Tog:

...

 

 If the streams are of identical files served over the same UPnP ..

 

Dear all,

We are NOT discussing this case here; in my case, the streams are of two distinct files (although they are the ribs of the same CD made by two different Naim devices) located on two different locations. And using different UPnP servers do not make a difference!

 

 

Posted on: 13 April 2011 by aysil
Originally Posted by AllenB:

Well said Tog  That's what I have been trying to get across to Aysil in my probably rather long-winded posts, but you have wrapped it up quite succinctly!

Allen, could you please clarify what you have been trying to get across to me, and what I have not been able to understand?

Posted on: 13 April 2011 by AMA

asyl, don't you have an itch to run NDX/nDAC with XPS or 555PS to see how the top Naim streaming sounds like? I'm sure it will be very different from Kondo 

Posted on: 13 April 2011 by likesmusic

Well, maybe we're all deaf AMA.

 

Malcolm Steward claims to hear a difference between the HDD cables in his NAS when using a Klimax DS. See http://www.malcolmsteward.co.uk/?page_id=2805 

Posted on: 13 April 2011 by totemphile
Originally Posted by AMA:

asyl, don't you have an itch to run NDX/nDAC with XPS or 555PS to see how the top Naim streaming sounds like? I'm sure it will be very different from Kondo 


Have I kicked off some bad habbits? A lot of name confusing going on in this thread here...

Posted on: 13 April 2011 by Guido Fawkes

An interesting, if at times confusing thread. Please remember that you're listening to entire systems; it is not possible to listen just to individual components. I have no doubt two systems sound different, but it is hard to isolate what causes these differences. I know of an early HDX that could make a system sound different, just by being part of it because, I surmise, its power supply was not helping the performance of other components; I'm sure this was subsequently fixed. 

 

I subscribe to theory that identical PCM streams delivered to the same DAC should sound the same and whenever I have heard this then my ears tell me it is a reasonable theory.  

 

What I often wonder is can some folk hear the difference between two CDs of the same songs on the same system. I find it just as likely that two CDs would sound different as two cables connecting the HDDs in a NAS or two UPnP servers sounding different. 

 

Surely, a US/nDAC is a very good source component and would be very hard to beat, but I'm also impressed with other source components. My most interesting discovery was the inexpensive MF V-Link for connecting my computer to my UnitiQute, it also works a treat with the nDAC. 

 

Lastly suppose the different methods of ripping data did result in slightly different PCM streams arriving at the DAC - in terms of SQ does it really matter if a few bits in a million are not quite the same. 

 

All the best Guy. 

Posted on: 13 April 2011 by AMA
Originally Posted by likesmusic:

Well, maybe we're all deaf AMA.

 

Malcolm Steward claims to hear a difference between the HDD cables in his NAS when using a Klimax DS. See http://www.malcolmsteward.co.uk/?page_id=2805 

likemusic, I don't argue that the problems in the network transmission may occur. Less with some data cables and more with the others. Less with one uPnP server and more with the other. Less with one NAS and more with the others (Naim has published a list of NAS which worked well with Naim streamers and referred that many other had failed).

 

So how did you come to the idea that I negate this?

Posted on: 13 April 2011 by AMA
Originally Posted by totemphile:
Originally Posted by AMA:

asyl, don't you have an itch to run NDX/nDAC with XPS or 555PS to see how the top Naim streaming sounds like? I'm sure it will be very different from Kondo 


Have I kicked off some bad habbits? A lot of name confusing going on in this thread here...

And what exactly confused you 

Posted on: 14 April 2011 by Tog
Mr Stewart makes lots of claims ....



My approach boils down to one principle - when trying to find a fault work systematically to exclude all those things that could be creating the problem in the first place. Having said that most of the problems I come across in life are generated by people rather than objects.



Just be logical and methodical is all.



Tog
Posted on: 14 April 2011 by aysil
Originally Posted by Guido Fawkes:

An interesting, if at times confusing thread.

Actually it should have not been so confusing. I caused it to be a bit more confusing by introducing more than one discussion in the same thread, thinking they are related.

 

In my opinion, there is just one clear conclusion we could draw from this thread: The local on-board storage version of UServe should not be recommended. It sounds way behind the similar version of HDX, whatever the reason! On the other hand, UServe works very well as a server when there is a NAS on the network. Many members are happy with the results in this second case. This is also my observation. My UServe performed perfectly well when it was streaming the files located in HDX' storage.

 

I said "whatever the reason" because it would be just speculation. However, I am pretty sure the reason(s) lie within the UServe and HDX, because all other factors were carefully controlled and constant in my system. I can also exclude the possibility that UServe was adversely affecting another component and that's why the system was not able to show its performance. I let the UServe on, and functioning (outputting tracks from on-board HDD to digout) also when I was auditioning the HDX. If there was adverse effects, they would be present when I am auditioning the HDX, too.

 

I also introduced another discussion on this thread, arguing if the components which come before the stream client contain elements or factors which influence the sound quality, they should be called the source components. This argument tossed into a lot of challenge, which I respect very much, and from which I learned a lot. However, I still think that it is at least a valid argument, given that, as likesmusic pointed out, some people hear difference even btw HDD cables in their NAS. Of course, this discussion is purely theoretical, and it really doesn't matter what we call them. It is a matter of term definition and convention. We should just keep an open mind to possible new discoveries in this field.

 

Btw, I had also reported that I heard the differences between the devices when connected to nDac as well. This observation was not disputed. Many other members have reported in previous threads differences between sound sources through nDAC before. In my opinion, it could be concluded, that "buffering and re-clocking architecture" does not wipe out all differences between sources. This is not its function anyway.

Posted on: 14 April 2011 by Hook
Originally Posted by AllenB:
...

I would dispute this as well in terms of the Naim units via a nDAC, but I am losing the will to live anymore on this particular thread. 

 

 

Hi Aysil -

 

Given your interest in this subject, you might want to read this old thread (and the others it spawned):

 

Why is the nDAC so cheap?

 

Hook