The samples that make up a piece of digital music should hit the D/A-converter w/o getting corrupted. How hard can it be?
A streamer with a built-in D/A-converter should have no problem with that. When the poor samples have to travel over S/PDIF, the receiving DAC must compensate for the shortcomings in this one-way, real-time protocol. Still Naim clearly think that they have overcome most of the problems with S/PDIF since both the NDX and the ND5 XS can be upgraded with the nDAC (but not the NDS, thank God!).
I'm interested in knowing (surprise!) if there's a difference when connecting the NDX and the ND5 XS to the nDAC with regards to missing or corrupt samples? (So, I think I've got a slightly different twist on this subject...)
If they sound differently, it must be because the samples that are fed to the D/A-converter (not the S/PDIF interface) differ, right? What else can influence the SQ? (RF noise?)
Can this be quantified? Imagine we could use a term like "error rate" which would be the ratio between the number of incorrect and the total number of samples.
How many samples per second would be incorrect per second (say)? 1? 100? I have absolutely no idea! It would be hard to measure though (opening up the box would be the easiest part)...
Apart from an interest in this subject, it also boils down to a question about whether I should keep my SB Touch and get an nDAC, or get an NDX (and use the SBT for Spotify and multiroom listening).
I've got a SuperNait and I've come to the conclusion that I need to replace the SN built-in DAC.
I've listened to both combinations at home, but not at the same time (loved the nDAC, and the NDX was also very good). So of course, I need to compare them side-by-side. If the SBT-nDAC doesn't sound as good as the ND[X5]-nDAC combination, what's wrong with those samples that finally hit the
D/A-converter and, if it can be quantified, what would be the "error-rate" (or whatever term to use)?
Posted on: 08 May 2012 by Guido Fawkes
I'm confused my Vortexbox and Mac feed a bit for bit identical PCM stream to a DAC (UQ) - I can capture it and compare it and they are 100% the same. There is very little difference in SQ, with possibly, the Vortexbox, a fraction better. The Vortexbox uses Ethernet so loads the FLAC into the UQ so the only jitter reaching the DAC is down to UQ rendering which is probably very low, but I have no way to measure it. The Mac streams raw PCM over S/PDIF with very little jitter or noise. I use an optical connection so electrical noise is banished. The W4S Sonos Connect sounds ever so slightly better on most tracks, but does not produce the same bitstream owing to some up-sampling it does. If you send the stream through a Naim DAC/555PS then there is no difference that I can hear.
I'm not sure what else to measure. The only differences are where the processing is done and the jitter.
I can measure differences is processing on a Mac between ALAC, AIFF and WAV - but the differences are very small (0.5% peak) with AIFF least cpu and ALAC the most - WAV being very close to AIFF. I can't hear any difference. I can't measure the processing within the UQ as I do not wish to take it apart and would know how to do it even if I did.
In my opinion the Naim DAC sounds better than the NDX and is only surpassed in the Naim armoury by the NDS (though I've only heard one demo of the NDS). I would use the SBT in to the Naim DAC given your choice.
Posted on: 08 May 2012 by Simon-in-Suffolk
Peter, the chances of any sort of data corruption in a regular SPDIF connection is very low, and in my expierience tends to cause a temporary mute in audio when it occurs.
However as Guy point out, RF electrical noise in all interconnects, whether they be audio or digital, can pollute albeit very subtly the downstream analogue and digital circuits. RF noise in analogue circuits can cause intermod artefacts (can sharpen or flatten the quality of the audio). Digital circuit noise can modulate clocks causing time domain distortion or jitter.
Now using using optical digital interconnects removes electrical noise from the connection, ( although the LED may have noise modulation) however there is another possible source of artefacts.
Naim in thier ndac and NDX electronics use an algorithm that tremoves transport jitter in SPDIF signals. This from all accounts provides a very accurate with respect to time data stream. However this de jitter-processing itself has the potential to cause electrical noise in its own right, hence we may 'hear' jitter in the signal, even though the NDAC effectively eliminates the jitter in the SPDIF.
Now in the NDS we see the decoupling and screening of the streamer boards is increased markedly, and so these artefacts should be less audible, I look forward to confirming whether this is indeed the case.
Simon
Posted on: 10 May 2012 by retepqt
Guy, how do you capture the data and where? Not a UQ feature, I'd guess.
I liked what I heard from the nDac, so your suggestion to use the SBT is in line with my original thoughts, however, I don't think Simon and some other members agree (but then again, I've seen proponents for the SBT-nDAC combo).
Simon, why (if it's true) would an NDX-nDAC sound better than a SBT-nDAC? I think you're saying that the probability for errors occuring during transmission is very low (neglectible) - even for a SBT hooked up to the nDac? (The clock in the SBT is good enough for the nDac, I assume.)
Is the conclusion that differences in SQ from different sources emanates from RF noise in the analog parts? Then there's nothing wrong with those samples... (And the nDac has taken care of the jitter associated with S/PDIF.)
So, would a good power supply for the SBT decrease the noise in the analog section of the nDac - both for optical and coaxial? Thought it wouldn't matter when using a digital connection.
Peter
Posted on: 10 May 2012 by Guido Fawkes
Guy, how do you capture the data and where? Not a UQ feature, I'd guess.
The UQ has a digital out so you can send the data to a CD recorder or to a digital input on a computer .... the PCM that comes out of the digital out is the same as in the input file - or at least it is on the files I've tried.
I'm not saying all transports sound the same, but I am saying I can't find a difference in the bitstreams that come out of them if they are set for bit perfect - the ones that up-sample are quite different and often sound a bit different/better to me.
I do not expect everybody to agree with the subjective things I write - as this is like music, some sounds we like and other we don't ..... I know Simon can hear differences between things on his system that I cannot on mine .... but both of us are happy.
Some things though are objective and that the bitstream is unchanged on the samples I tried is a fact ... so any difference in sound must be for another reason. It could be as simple as the power supply on a computer puts rubbish in to the mains, whereas the NDX doesn't ... but most likely it is more complex. I probably need a better system than I have to hear the difference though (and possibly better ears).
I wish we could measure the difference easily then it would be real;ly easy to decide which to buy.
One trick - that could be tried is to capture the analogue: I could do this on cassette and see if the resulting recordings sounded different (well I could if I had a NDX at home).
All the best, Guy
