Naim DAC - Is there ANY word?

quote:

Originally posted by jon honeyball:

quote:

Originally posted by DHT:
Kuma I can't quite understand how this sits, when some posters have sold their CDs3's in preference for the considerably cheaper Lavry dac.

I heard it recently at the UHES event. And not just for one track, we listened to it on and off across a whole day. And tried all the various settings from the mac.

No, Jon, that is not quite true. We stopped using the Mac/Lavry when we migrated above the two lowest priced Naim CDP's (whichever they were). Thereafter we only compared Naim options. Also, perhaps more significantly, due to my deafness, the Lavry settings were in their worst settings, i.e. Wide bandwidth (instead of Crystal) and absolute phase reversed. It was also being used straight out of the box, so to speak, having only been switched on immediately prior to the demo.

In another forum, someone has wondered why the Lavry always appear to do badly at a dealer demo but performs superbly on a home demo. Interesting, if true.

Richard

quote:

Originally posted by Richard Lord:
In another forum, someone has wondered why the Lavry always appear to do badly at a dealer demo but performs superbly on a home demo. Interesting, if true.

Not true in my home.

My experience wasn't a dealer demo. A friend of mine just so happen to have one and not using it.

He gave up on it but wanted me to try one to see my thoughts of it.

So apparently both of us weren't too enthused about it but the third owner seems to be very happy and now he found the matching DA to go with his Lavry AD for a song.

quote:

Originally posted by Richard Lord:
Also, perhaps more significantly, due to my deafness, the Lavry settings were in their worst settings, i.e. Wide bandwidth (instead of Crystal) and absolute phase reversed.

Interesting - according to the DA10 manual, the Wide PLL setting should only be used "for non-standard frequencies between 30KHz and 200KHz". And I (and most other users, as far as I am aware), use it in either Crystal or Narrow modes (I have settled on Crystal). I have briefly tried Wide mode and in my opinion it is by far the worst sounding.

quote:

Originally posted by Richard Lord:
the Lavry settings were in their worst settings, i.e. Wide bandwidth (instead of Crystal) and absolute phase reversed.[QUOTE]

[QUOTE]In another forum, someone has wondered why the Lavry always appear to do badly at a dealer demo but performs superbly on a home demo. Interesting, if true.

Richard

Not so in our demo. Factory set-up for 2 pin hot. We also experimented with all of the different settings for "mode" and "phase" and while these had an effect on the DA-10's performance with narrow being the best, it did not change the overall impression.


Hard to believe since others have posted that Naim dealers in the UK are now selling the DA-10so no issue of "warm up", etc... However, as you said "Interesting,IF TRUE."

It should also be noted BTW that the Wide PLL mode is the one that uses asynchronous sample rate conversion (upsampling) which is the feature that was briefly discussed on another thread some weeks ago as an explanation for what some viewed as the DA10's unsatisfactory performance.

quote:

Originally posted by Michael Trotz:
It should also be noted BTW that the Wide PLL mode is the one that uses asynchronous sample rate conversion (upsampling) which is the feature that was briefly discussed on another thread some weeks ago as an explanation for what some viewed as the DA10's unsatisfactory performance.

I'll respond as both these points involve me.
Please read Dan's white paper and visit his forum for answers as to the DA10 upsampling eveything below 88k in every mode. Crystal and narrow are synchronos upsampling as opposed to asynchronos because 88.2 and 96 are multiples of narrow input clock freq. I brought it up and never said that upsampling had to be bad, just that I thought I heard that characteristic but noted that it could be something else and that everthing has a character. Never said upsampling couldn't work well and that lay people are not in a position to argue circuitry with guys that do this for a living. We can only evaluate the resuts and I'll not revisit my conclusions.
For the record, this dealer warmed the unit for hours, tried all modes, different known sources with coax besides MAC tos and inverted at the speakers to hear all possibilities. We also tried some 96K material but weren't aware that it would be a non-upsampled source at the time.

quote:

Originally posted by js:
I'll respond as both these points involve me.....

.....Never said upsampling couldn't work well and that lay people are not in a position to argue circuitry with guys that do this for a living. We can only evaluate the resuts and I'll not revisit my conclusions.

Thanks for your response. I'm not sure why you felt that these points concerned you. I do recall that you had said that upsampling was not necessarily a bad thing, but my comment above was made with one or two other contributors in mind.

quote:

Originally posted by Michael Trotz:

quote:

Originally posted by js:
I never said that upsampling had to be bad, just that I thought I heard that characteristic burbut noted that it could be something else and that everthing has a character. Never said upsampling couldn't work well. I'll not revisit my conclusions.

Thanks for your contribution. I'm not sure why you felt that these points concerned you. I do recall that you had said that upsampling was not necessarily a bad thing, but my comment above was made with one or two other contributors in mind.

I am a dealer that demo/commented (per R. Lord) and was first to mention upsampling so I thought the shoe fit. Hope the upsampling bit helped.

We'd both agree, I'm sure, that there is a whole world of difffernce between upsampling 44.1 to 88.2 versus resampling 44.1 to, say, 96 :-)

quote:

Originally posted by js:
Hope the upsampling bit helped.

Yes it did. The difference between the Crystal/Narrow and Wide modes has been quite clearly explained on their forum and is as you have described. As I mentioned above, I have briefly tried the Wide mode and found it unsatisfactory and I wonder whether the negative impression of upsampling that some (not you) appear to have might be related to the way it has been implemented in that mode.

I should also add, in response to another comment above, that the difference between the Crystal/Narrow and Wide modes is not just a matter of a different target frequency (88.2 vs. 96). The DA10 actually uses a different "crystal oscillator circuit" for each incoming frequency in the Crystal/Narrow modes and a fixed crystal for all incoming frequencies in the Wide mode (with conversion being done to the internal crystal rate). This involves a recomputation of the incoming data for the new rate - something that the designer himself admits involves a tradeoff in performance.

quote:

Originally posted by Michael Trotz:

quote:

Originally posted by js:
Hope the upsampling bit helped.

Yes it did. The difference between the Crystal/Narrow and Wide modes has been quite clearly explained on their forum and is as you have described. As I mentioned above, I have briefly tried the Wide mode and found it unsatisfactory and I wonder whether the negative impression of upsampling that some (not you) appear to have might be related to the way it has been implemented in that mode.

I should also add, in response to another comment above, that the difference between the Crystal/Narrow and Wide modes is not just a matter of a different target frequency (88.2 vs. 96). The DA10 actually uses a different "crystal oscillator circuit" for each incoming frequency in the Crystal/Narrow modes and a fixed crystal for all incoming frequencies in the Wide mode (with conversion being done to the internal crystal rate). This involves a recomputation of the incoming data for the new rate - something that the designer himself admits involves a tradeoff in performance.

That is in tune with what I posted. 44.1 goes to 88.2 and 48 goes to 96. The frequencies are referenced to crystal oscillators. that's true of both Narrow and crystal. It's a neat and commonly used solution for upsampling from most sources. They'll use one of those 2 for everything else in wide, interpolate up to it and continue normally downstream from that point on. I just don't know if 88.2 or 96 is used for wide but either would involve the same process. The crystal input involves more at the input than narrow as the original source clock is ignored and the stream is held in a buffer and reconstructed with a new clock before upsampling to correct for more jittered input.

quote:

Originally posted by Michael Trotz:

quote:

Originally posted by js:
Hope the upsampling bit helped.

Yes it did. The difference between the Crystal/Narrow and Wide modes has been quite clearly explained on their forum and is as you have described. As I mentioned above, I have briefly tried the Wide mode and found it unsatisfactory and I wonder whether the negative impression of upsampling that some (not you) appear to have might be related to the way it has been implemented in that mode.

I should also add, in response to another comment above, that the difference between the Crystal/Narrow and Wide modes is not just a matter of a different target frequency (88.2 vs. 96). The DA10 actually uses a different "crystal oscillator circuit" for each incoming frequency in the Crystal/Narrow modes and a fixed crystal for all incoming frequencies in the Wide mode (with conversion being done to the internal crystal rate). This involves a recomputation of the incoming data for the new rate - something that the designer himself admits involves a tradeoff in performance.

It seems this approach (resampling and reclocking) is also used by others brands. For example the new Bryston bda-1 external DAC. Why does this solution involve a tradeoff in performance?

Goldfinch - beyond the statement that asynchronous upsampling involves recomputing the data, which compromises performance, I have not found any additional details (maybe it is proprietary information?). However, there is quite a lot of information on their website and a more careful search may well unearth more information.

Another interesting claim that the designer makes on their forum ( which seems to confirm your observation) is that, while DACs did not have upsampling before around 1990 (as the concept did not exist then), all modern DACs (at least the ones of which he is aware) have at least some upsampling. In fact, he appears to believe that it is not possible to design a good DAC without using it. In his view (and as I believe JS has observed), upsampling in and of itself is not necessarily a problem - what is important is how it is implemented.

I have no way of knowing whether these claims are actually true, but he appears to be very knowledgeable about DAC design in general, and also quite willing to share information about his design philosophy.

quote:

Originally posted by goldfinch:
........

It gets trickier here. Quantisation noise that needs to be dealt with, interpolation at certain freq. or all depending on upsampling method chosen, how you upsample as 2 bits never quite behave as 1 and more processing in general. etc. You also get advantages like easier and possibly less audible filtering, consistently good results at moderate cost with currently available bits and simple common circuits after the upsample which is the tricky part. I'm sure there's more that I haven't a clue about on both sides of the ledger. There are always trade offs and the best engineers are the ones that recognize them and find the best workarounds. These are often not readily apparent, may only get discovered after the fact and sometimes still are disagreed on by even the top engineers. It's nice to know about some of these things but consumers should never choose based on anything but a listen. Everything works in theory or it wouldn't be used.