NDX vs Linn DS?

So the NDX is not a bit-accurate streamer?

I had chance to compare the NDX with my CDX2/XPS and i would keep the CD player. The NDX is not ridiculous at all, it is even very impressive considering the price difference , simply amazing. But the CD is more alive and true. Surpisingly, this difference stays even whie changing the power supply...? But as said, the value for money is outstanding !
On my side, i would say the NDX is the ideal machine to complete a modern system and integrating non material music with a lot of convenience and ease of use.

Hi, Simon. It's absolutely clear that all DACs do DSP before running the bitstream into a D/A chip. It's absolutely clear that NDX is doing the same with its built-in DAC. I think likemusic is asking if NDX does any DSP before it streams the bits to the S/PDIF out. That's a fair question. It's similar to what software players do: like dithering and upsampling (if the corresponding checkbox is checked). I'm sure NDX does not do upsampling on S/PDIF out. AFAIK it doesn't do upsampling for built-in DAC either -- unlike Linn KDS which permanently upsamples all incoming bitstreams to 24/192. But I'm sure all of them (NDX, nDAC and Linn DS) do various types of smoothing DSP (like dithering) before sending data to D/A.

So the question is -- once the TCP data pack is received and resolved and stored in a memory buffer and then transcoded to PCM format -- does NDX process the actual bits before sending them to S/PDIF out? 

Thanks AMA, that was what I was trying to get at. If an NDX does process the bits in some way as I thought Simon was suggesting, then an NDX will feed an nDAC a different set of bits to a UnitiServe feeding the same rip to an nDAC, which would be kinda worrying.

> Apparently, Naim say that the SPDIF output on the Unitiserve "is not audiophile quality"

One distinguished member on this forum - suggests that UnitiServe + Rega DAC sounds better than wonderful - hard to fathom if Naim says it is "not audiophile quality". No wonder folk get confused. 

Lots of interesting posts here - seems there are lots of ways of manipulating what comes in to a box like the NDX to give the best possible output. My Sonus W4S upsamples, which is slightly different from the NDX approach, but nonetheless produces a fine sound through a Naim DAC.  This goes a long way to helping me make sense of why a rendering engine is important. 

Doesn't seem like there is a right way, but it does seems as there will be sonic differences that may make one method preferable to another for replay. 

So am I right in saying that if I could examine the bits out put from the S/PDIF on various rendering devices these would indeed be different. It would then be down to my ears to tell me which I liked better. 'Twas easier when I thought they'd all the same, but they don't 

I would rather hope the NDX (with the arguable exception of user selectable upsampling) left the bitstream alone and just passed it through to the s/pdif output. True, that makes it a very very expensive streamer when used like that, but I suppose it's not just 1s and 0s, but Naim 1s and 0s.

Hi Simon -

First of all, I will not pretend to understand how this stuff actually works, but since that has never stopped me from expressing an uniformed opinion in the past, I'll do so again now....

My interpretation of the NDX white paper is that Naim has decided to use oversampling and digital filtering (inside of the NDX) in order to be able to produce the cleanest (lowest noise) S/PDIF output signal possible.  And apparently they are not alone, as designers of other network players (as well as software players like Pure Music) are always advertising their efforts to improve their algorithms (with the goal being less effort).  The NDX white paper makes the point that:

"...With every increase in number crunching the DSP consumes more power, and the more power the DSP consumes the greater the power supply noise. By optimising the DSP algorithms and controlling the way the data is buffered, power supply noise is kept to a minimum – to the benefit of audio performance."

My assumption is that "native" setting, the NDX's default, means that the digital output would be the same sample size and bit rate as its input, and that the "96 kHz 24 bit" setting is only to be used in the case where the NDX is connected to a non-Naim DAC that can not reliably handle higher resolutions.  So in addition to the oversampling inside of the DSP (which is apparently a prerequisite for using a delta-sigma converter to move noise up to higher frequencies), there does also seem to be the potential for downsampling of higher bit rates prior to output.

It would be great to get additional clarification from Naim, but given their history of not commenting on product design issues in this forum, I doubt we'll ever see that!

Hook

IMHO, digital audio seems too complicated, and afflicted by non-digital factors, to be imagined as just a row of soldiers of 1s and 0s in a parade.

Hi Aysil -

I totally agree.  These are electrical signals, and the goal is transmit them with the minimal amount of accompanying noise. 

I'm no engineer, but I have read the NDX white paper, and have tried to use wikipedia to get a basic idea of what's going on.  I am sure I have plenty of misconceptions, but hopefully I have some basic concepts correct. 

The NDX, like many other streamers, oversamples its input by stuffing 15 zeros for each bit of audio data.   Apparently, this makes it easier for the DSP to run an algorithm that moves noise in the audio band up into the higher frequencies (i.e., "noise shaping"), and outside the range of human hearing.  In this way, it can deliver a "cleaner" S/PDIF signal.

Apparently there are some other well-respected designers do not do this (e.g., Empirical Audio), so I guess it will always come down to using our ears to determine whether or not any one streamer sounds better than another.

Hook

Hi Simon -

How did you come to that conclusion?  

The very first thing the NDX white paper says under "Digital filtering using Naim algorithms" is:

"...If we insert 15 zeros between each pair of input samples as the first stage of 16x oversampling, the frequency spectrum of the zero-stuffed signal will contain lots of unwanted components, as shown previously."

If this does not apply to S/PDIF output, then it sure is misleading!

Hook

Hi Simon -

My last comment here, and then I think it best for me to withdraw from the conversation, as I am certain my ignorance is only contributing to the confusion here.  And besides, it's giving me a headache.  

I stumbled across an engineering doc entitled "Programming S/PDIF on ADSP-2136x and ADSP-21371 SHARC® Processors":

http://www.google.com/#sclient...;biw=945&bih=413

It talks about oversampling on p. 9, so perhaps it is relevant to this discussion.

Anyway, good luck with figuring this out.  I'm going to go take a couple of aspirin and lie down. 

Hook

Simon, I think so because Linn ADS definitely doesn't do this and claims for a bit-transparent transmission. Thus, theoretically, ADS can translate the bit-perfect copy of original file to the S/PDIF output. And afaik this is explicitly stated by Linn. The same we can say about Weiss INT202 and Weiss DAC202 which both provide a built-in transparency check (!!!). Obviously it does not guarantee the further path of a bitstream will be clean as it's a function of S/PDIF link and S/PDIF receiver "smartness" on the other side of a line.

I think that a properly built streamer (like UnityServe, NDX, HDX, Linn DS, Logitech Transporter) never do something to bits on the S/PDIF output and the only difference in the end quality of bitstream may fall down to the S/PDIF generator implementation (massively dependent on clock quality and power quality). In fact, the output bitstream may stay constantly bitperfect but in case the clock timing is not good and the noise distorts the S/PDIF waveform the output bitstream may look very complicated for the receiver (external DAC) and that's where the bits are get screwed up resulting in sonic degradation. This explains why HDX/NDX may sound better than UnityServe when run into re-clocking DAC.

We had an interesting discussion before on how exactly the waveform distortion may cheat on receiver. I had an idea that timing errors between the bits are the main source of the bit flips and it was objected by Andy (a very good engineer who was charmed with nDAC and didn't believe that various sources may sound different through nDAC). I found his objections reasonable but it didn't mean that a concept of getting the bits flipped on the receiver input was wrong and my personal audio tests supported this. Later on I got a consultancy from our corporate digital engineer and he agreed with Andy that timing errors today are not that big to confuse the turn of the incoming bit for the majority of commercial on-board clocks but he noted that S/PDIF waveform (attack and decay) may be distorted by noise in such a bad way that even if a clock is recovered properly the receiver may fail to identify the next bit properly. Though it does depend on the receiver's S/PDIF abilities.

Hook,

The reason I doubt that the oversampled signal is sent via the SPDIF is that Naim perform 16x oversmapling in the DSP, therefore that would be a sample bandwidth 768kHz which is way higher than 192kHz which is what I see refer to as a top SPDIF frequency - even though the standard I don't beleive has a max frequency. Also if you look at the block diagram this higher bandwidth clock uses the I2R which is a synchronous transmission method to communicate between the DSP and the internal DAC.  Therefore  I can't in hindsight how the NDX DSP processing can likely be conveyed through the SPDIF.

AMA - on reflection it looks like the NDX might not do anything to process the the data on the SPDIF after all. But I guess earlier comments by members around using the NDX as just a SPDIF generator as being extravagent might have merit ....

Right now listening to  24bit 96kHz Eagles; Hotel California using the NDX/555PS and it sounds very very enjoyable and rocky. The recording of the drum kit is fantastic

It's all very confusing.

There's an earlier comment that, in some Naim circles, a UnitiServe is not considered to be of 'audiophile' quality. That makes it a very expensive NAS.

Presumably an NDX does deliver 'audiophile quality' bits to an NDX, albeit at a very high price, and with considerable redundancy in the system. Or do you need a power supply too, to get really really audiophile bits?

Ah well, it's only money

Keith, I agree with you UServe is absolutely more "audiophile quality" than a lot of things on the audiophile market. It is just not up to the level of its more expensive brothers, which is just normal. On the other hand, yes, everything is too expensive!

reading all this, i am confused. i have UnitiServe + nDAC. I can already play my rips via the DC1 cable from the UnitiServe to the nDAC. Why would i then need an NDX on top of all this -- what would the NDX add?

enjoy

ken

I don't think an NDX would add anything, and have yet to hear a rational explanation why it would. Others on this forum say the "magic beans" in its streamer function improve SQ.  I'm setting up an audition to see if I agree.

Presumably it would add:

- Internet Radio

- A better power supply

- USB input

- DACs/Analog Out in case you ditched your Serve and used another UPNP Server

- A source for a 2nd system with local output on nServe

- Front panel screen

- A different flavor DAC.

- And of course, another opportunity for an XPS/555ps

Whether or not those matter is another issue.

I think people have reported here that the Qute's digital output exceeded that of the nServe.

But if you want to go completely computerless it may be the best there is next to the HDX.

-Patrick