Does the transport to an nDAC make a difference? Oh - and other musings...

I like the water analogie good to see you thinking outside the box not staring down that tunnel, pity it isn't quite correct.

Think of jitter as being manifestation of noise, jitter is a measurable indication of noise. Correcting the effect of noise doesn't eliminate noise. It is still there.

It's in the White paper, it's what Naim claim.

You won't get Naim commenting. Tried that 2 years ago.

BTW, I have a mail from Naim support that appears to come from the designer of the nDAC commenting on the design of the nDAC. I had asked them a couple of questions and if they could comment on some of the issues raised by the extremely long thread of a few years ago. They answered some of the questions and posed some other thoughts and I asked if I could post it but was asked not to publish it.

Without going into the details, they essentially say theory doesn't hold 100% because complex interactions, no matter how small, occur. I'm perfectly happy with that as a response and can totally see why it would allow minor differences to be heard - I work with this sort of hardware on a daily basis so understand that sort of response. 

If I can summarise from my own viewpoint:

The nDAC is an extremely well thought out design. It all but eliminates the transport from the equation and delivers consistently high performance from a variety of sources. It does this via a very clever and innovative architecture which is theoretically perfect, but due to real life isn't quite perfect in practice. The fact that it isn't perfect allows different transports to be perceived as slightly different as they will each have different noise characteristics which will present itself as slightly different renditions of the same data. Hence why transports sound different.

In context, whilst these differences can be heard by some, they are not heard by all and the level of differences are relatively minor - in the grand scheme of things. The nDAC essentially minimises the differences between a good and bad transport - far more than a traditional DAC (which will sync with the SPDIF clock) does. The nDAC levels the playing field. It doesn't quite make it perfect, but it gets close. Closer than a large number of other DACs.

Fundamentally, if you have a bit-perfect source and you have £3k to spend, unless the rest of the system is maxed out, it is unlikely that spending the £3k on a source for the nDAC is the best value for money for your system. It may make A difference, but it won't make the best bang for buck difference. 

Are they padding the bottom line? That's a completely different question. There is another thing at work here. I'm a confirmed geek - I build computers and run my home network as a hobby. I know my solution is 98+% as good as the best streamer Naim can provide because I understand - at the deep technical level - what is going on. Would I like to make my home solution into a product... NO WAY! the technical support would be a nightmare. There are FAR too many variables that would mean I'd be on the 'phone 24/7 with customers supporting - for example - why their windows share didn't work like it did yesterday. You have to understand Naim is a business - just like any other. It defines products and the environment those products work, but occasionally, they let the enthusiast in (think alternative power supplies for example). They won't endorse them, but that doesn't make them any less of an alternative.

Bottom line: if you're techy enough to understand what you are doing, you can get almost the same performance out of the nDAC as you can with a "high end" (i.e. expensive) streamer from Naim - or any other manufacturer of streamers. The solution isn't for everyone, but it isn't to be dismissed as being totally inferior as some do here. That was proven to me when James came over. There is no magic in digits. The magic is getting them back into analog(ue) - something Naim do extremely well with the nDAC.

Fatcat... Please carry on pulling my arguments to pieces. I'm sure those who actually know what they are on about are finding it funny. I know I am....

Don't let anybody pull the wool over your eyes. . Trust your ears.

I don't think Naim have ever claimed the nDAC is capable of being front end neutral. If it was I'm sure they would say so, it would be a HUGE selling point. But they don't for good reason. It isn't.

More ramblings than arguments. People can take obsessions to far. LEAVE IT. Find something more interesting to occupy your time. Find a hobby.

<cough> Pot - here's kettle on the phone for you..... says something about being black.....

Ooh.. You are awful ???

:-)

You guys are funny.

Andy says:

"Fundamentally, if you have a bit-perfect source and you have £3k to spend, unless the rest of the system is maxed out, it is unlikely that spending the £3k on a source for the nDAC is the best value for money for your system. It may make A difference, but it won't make the best bang for buck difference. "

I have to agree on that one. I suspect into my level of system, a apollo-R ($995) would already be more than sufficient. That's if I wanted to do CDs, which I don't. It just doesn't make sense to me to spend something at the level of the CD5XS or CDX2.2. Does it sound better? I don't care either way. The money is better spent elsewhere.

I'm not saying I'll go get the nDac. I'm happy enough with the Squeezebox into the Rega Dac. Even the rega dac has that buffering thing going on. And I've been able to get the SBT where it's more than good enough for the rega dac.

To put it a different way, this whole argument is about $$$ isn't it? If money wasn't a concern, why waste time debating it. Yeah, do whatever NAIM recommends. A couple $10K up or down doesn't make any difference.

Really speaking, the argument shouldn't be about whether you hear a difference, whether the transport makes a difference or not. You'd rather say that it doesn't make sense to invest as much in a transport anymore since it doesn't make sense in a $$ sense.

Unless you're really extremely rich. At which point, you'll just ask you're dealer to put together a top of the line system, and he would. $10K up or down doesn't make a difference that point.

A better, more useful way to spin this argument  is:

Given a budget of say $5K/$10K/$20K what's the best system you could put together with used/new NAIM components. You can bring your own transport to the party if you like.

Maybe I'll make this a new topic.

And watch the fun.

Interesting thread... My observations so far :-). The comment about spdif and transport jitter, I think perhaps the point should have been about sample jitter and transport/spdif jitter being different. They are.. The extent they affect each other is largely down to the receiver implementation.

Also I don't think the argument is about $$$, it's about engineering design!! Better designs with fewer compromises tend to cost more, but not always.... In my expierience just because something expensive it doesn't mean it's better....

I agree. Isn't it also about getting the best bang for your buck though?

By sample jitter, I assume you mean the jitter that is due to the conversion between analogue and digital in the first place not being timed exactly 100% correctly. If so, no replay system can do anything about this with current data formats (you'd need timing information too).

No, but unless we take spectrum analysers to dems, we unfortunately have to let our ears do the measuring.

BTW: Hook mentioned you'd done some experiments with a spectrum analyser. Did you publish them - if so, got a link - I'd be very interested in what you found.

Andy

No, but unless we take spectrum analysers to dems, we unfortunately have to let our ears do the measuring.

and to me surely this is what its all about anyway?? ie how it sounds to us.

By sample jitter, I assume you mean the jitter that is due to the conversion between analogue and digital in the first place not being timed exactly 100% correctly. If so, no replay system can do anything about this with current data formats (you'd need timing information too).

Not quite, i would brush off your DSP maths. Source jitter in the recording encoding function  (causing a gausian distribution of noise over the time domain -> therefore the frequency domain) can be reduced in power density by oversampling the signal in the replay system transformation function and therefore the source distortion due to conversion jitter is reduced in power level. ie the effect of conversion jitter is reduced.

Do you not just love discrete and continuous signal/DSP mathematics - you can do so many cool things

Hi - no I have not published the audio spectrum analyzer, but lots of fiddling around here with basic analyzers - I would love to get my hands on an audio frequency Rohde and Schwarz - could you help?

Simon

Have to say I hadn't thought of it that way. But it's the DAC that does the oversampling in its DSP, not the transport - so the source jitter is unaffected by the transport. 

R&S: unlikely I have to say, if we do have one it will be on someones bench somewhere in France or India.

Indeed - if the jitter in the transport is fully elimated there will be no incremental noise. In practice though this is impossible to achieve, but is effectively made insignificant in effective designs.

This is so wrong, Gaussian is spelled with two esses. 

But seriously Simon, thank you for clarifying this. I couldn't have said it better myself. In fact I couldn't have said it at all.

But what does it mean ?...

Jan

What does it mean? It means I can use my squeezebox touch with the nDac.

Simon - can you run that one past us again - in baby steps?

Simon.

Are you saying a jittery transport equals incremental noise, a very jittery transport equals a lot of incremental noise.

Sorry for typo yes it's Gaussian of course :-)

I'll try... Jitter is essentially noise in the time domain. The distribution, or  chance of, of a distorted time value from a regular period is in line with a Gaussian distribution. Any value is possible but the further away from the ideal the less likely it will happen.

Now Gaussian distributions have some neat mathematical property. They retain thier same power or energy when transformed from the frequency domain to time domain and vice versa. When we do ADC and DAC we are essentially switching from frequency to time and time to frequency domains respectively.

Now in the DAC process there are many things you can do to the signal in the time domain. One is to oversample the samples. Now oversampling is interesting, by using zero value  added samples ( signal sample followed by a zero sample is 2x oversample for example) we increase the sample frequency but we are not adding any new values (zero xanything is zero). So we are distributing the original signal over thsa larger frequency spectrum, but the originalfrequencies are maintained. However that noise or jitter bears no direct relationship to the original sampled frequencies. Therefore this noise is distributed over the larger oversampled frequency range. Therefore when we filter and extract (transform) the oversampled signal back into an analogue one, the noise from the original conversion jitter is divided out by the oversampling ratio, and so the energy or power of the original conversion noise(jitter) is reduced in intensity in the recovered analogue signal.

Therefore oversampling reduces the effect of conversion noise or jitter in the recovsigns analogue signal.

Hopefully that makes some sense?

Simon

Good explanation, thanks.

Crikey Simon, I understand what you're on about and I have to read what you write multiple times.... 

All oversampling does is make it easier to design the filter that comes after the D->A conversion. What it does is move the artefacts you get from the digital processing up in frequency so the filter that gets rid of them can be less severe. A filter that is less severe messes with the frequency range you are interested in less -> a "better" DAC. 

All this is moot though - the oversampling in the DAC happens no matter what the transport is. What you are describing are the processes that are chosen by the DAC designer that will alter the way the DAC sounds. NONE of this is transport dependant - which is, after all, what the thread is about.

Thanks chaps, and  Andy of course you right, all noise is additive so any time domain or 'digital processing' noise in the DAC conversion process wI'll ultimately work against the noise reduction by oversampling. So even if you over sample you still need a well designed/implemented DAC.

I'm out for a few days, I'll catch up here later in the week.

Simon