DAC

Simon-in-Suffolk posted:

Interesting that the 272 does indeed use the ADSP 21489, there were several posts - and I think even from Naim (?) saying a different ADSP was used around the launch time  - perhaps that was for a prototype or or purely incorrect. If anyone can whip the top of the 272 they should be able to confirm

Curiously CreativeAudio have a close up photo of a ADSP-21369 on their NAC-N 272 images page.  But the processor shown in that image looks different to the processor shown in situ on the NAC-N 272 board.   They look to be different package types (if I'm interpreting the pictures right the 21369 shown is a SBGA package type while the board shows a DSP with LQFP package) - that may not help much as the 21369 is available in both package types (if I'm reading the AD website right).

And no ... not going to "whip the top" off my 272.

Eloise - interesting - yes perhaps its a prototype or there is mixup somewhere - but this is an image from 'their' 272 apparently

Simon-in-Suffolk posted:

Eloise - interesting can you infer the link please.. the ADSP chip should be just visible  - its on the lower PCB below a daughterboard

I don't think I'm allowed to post the link ... so if you goto Creative Audio's website, search for NAC-N 272 and then "more info" it has a page with "Images and Video" the two pictures I'm referring to are there.  A gold coloured DSP chip then a close up of the NAC-N 272 board.

Simon-in-Suffolk posted:

Eloise - interesting - yes perhaps its a prototype or there is mixup somewhere - but this is an image from 'their' 272 apparently

Looking closer, that image doesn't appear to match the other image showing the DSP chip more in situ as the capacitor shown lower right in the image above doesn't match up on ...

The DSP is the chip seen (mid picture) "under" the white connector with 5 wires (red at the right hand end) and "above" the f-connector.

Yep - perhaps Naim can confirm directly - (Phil or Steve for example) - but I wonder if the 21369 image is actually from the 172 ????

The image resolution just doesn't allow me to read the writing..

Not that it is that important - just curious....

That image looks like it could be from an ND5xs.

Maybe Naim need to speak to Creative Audio about the need to present things properly. If the picture is wrong of course. 

Thanks Richard - there is also this discussion from the forum from some while back  - which I based my earlier reply on - but its not clear to me checking back where the poster confirmed their  information from other than with some apparent discussion with Naim.

https://forums.naimaudio.com/to...21#70272142551529321

Simon-in-Suffolk posted:

Yep - perhaps Naim can confirm directly - (Phil or Steve for example) - but I wonder if the 21369 image is actually from the 172 ????

Does the 172 even have a SHARC DSP?  The nDAC uses the ADSP-21369...

as does the ND5 XS which has the DSP and has the DSP chip and capacitor in the correct place to (perhaps) be the picture Creative Audio has as being the NAC-N 272.

Just a thought - but do "we" have this the wrong way around and the NAC-N 272 has the ADSP-21489 and the earlier ND5 XS (and maybe NDX and definitely NDS) streamers along with the nDAC have the ADSP-21369.

Simon-in-Suffolk posted:

Thanks Richard - there is also this discussion from the forum from some while back  - which I based my earlier reply on - but its not clear to me checking back where the poster confirmed their  information from other than with some apparent discussion with Naim.

https://forums.naimaudio.com/to...21#70272142551529321

It appears to me that the statement...

So there it is. Slighty slower DSP (400 MHz 21369 over 450Mhz 21489) but same round-off error (both use 40 bit floating-point arithmetic).

...is (or has been interpreted) about face.  Its the ND5XS / NDX / NDS which have the "slightly slower" (21369) DSP; while the NAC-N 272 actually has a faster (21489) DSP.

As the 272 is the newest of the boxes it makes sense that it has the more recent chip. Whether it matters is something else entirely. 

Could be - the reply from Naim (apparently - this is taken from a post on this forum) about the differences between ADSP in devices and a customer's concern about the ADSP differences between the Nova and N-272

"We use the same 40 bit floating point algorithm in all our SHARC based products. The algorithm oversamples incoming digital audio to either 705.6kHz or 768kHz and buffers the audio for subsequent jitter removal.

Both the ADSP 21489 and 21369 DSPs can run in either 32 fixed point or 40 bit floating point; we only use 40 bit floating point.

The 489 is a little newer than the 369 and runs a bit cooler for the same rate of math otherwise the digital audio is absolutely identical. We differentiate the sound quality using the DACs, PSUs, analogue components and mechanical treatment such as suspension."

So it could all be irrelevant in Naim's eyes.... although running cooler has got to be a better thing for drawing less current and therefore less potential for ground plane modulation.

Morning gents,

To clarify this:

272 - 21489 dsp

ndx, nds, nd5xs, superuniti - 21369

nDac - 21369.  This was the original product that introduced the dsp tech.

dacV1 - 21489

Atom, Star, Nova, ND555, NDX2, ND5XS2 - 21489 with new programmable clocking system.  

Best regards

Steve Harris

Software Director

Cheers Steve - that does indeed clear things up and indeed corrects various posts over the years on this forum. I have updated my knowledge as appropriate 

(But it does suggest a particular customer might have been mis led or confused Naim Sales support over apparent differences between the ADSP in the Nova and N-272 ) as they both now we are informed use the same ADSP. The relevant link is above - and this was a little time back.

And also thanks for confirming the ADSP in the forthcoming ND555 - which sparked this off 

[I will delete my post that stated contrary info based on previous posts and elsewhere - so it does not get repeated by Google]

Patu posted:

Got to love your posts Simon, always full of first class information. Thumbs up! 

Unfortunately in this case I was incorrect - please see Stevesky's post above

Hi,

No probs.  Also for clarity if a given product is not listed in my post above, it's because it doesn't have a DSP in it.    Instead we typically use a dedicated SRC IC:

Qute (all varants) - TI SRC4392

NAC172 - TI SRC4392

UnitiLite - TI SRC4392

NaimUniti 1 & 2  - TI SRC8421 with CS8416 SPDIF Rx

Regards

Steve

Since this thread has morphed into dsp centric, the next question is why does the 13000 pound streamer use the same dsp as a 2000?

RE: Same DSP part used in all

The 21489 is a very powerful DSP and it's not the limiting factor.  It takes a lot of time to master making the DSP code sound good, so it's better to build on knowledge rather than move around different solutions and take the pain each time on discovering on what needs doing to make them really sing.      The DSP code is tuned on a per product basis though and often the smallest of changes can have some interesting results on the musical presentation. 

In all Naim products its a sum of the parts that make up the higher end products, it's not just a case of throw a better DAC chip in and go ship it!   So clocking, DAC, output stage design and components, PSU structure, PCB layout/materials, internal wiring, vibration management,... on and on it goes.    I wouldn't get hung up on a given piece of tech too much - the devil in the detail in the overall implementation will have far larger implications on the musicality of the product. 

I've worked in the Hifi industry for a long time and I'm a fairly no nonsense engineering minded kinda chap, but I do find the interaction of digital electronics+software vs the musical presentation that comes out is a fascinating puzzle.   You really have to think outside of conventional wisdom to move to the next level of the game.    For sure it's not voodoo, but to unravel what is going on when equipment like a top end Audio Precision analyser can no longer guide you is really heavy stuff.     No wonder I'm going bald.

Regards

Steve

Thanks Steve for the reply. I wonder how naim engineers tune equipment for the naim sound. Is there a checklist to go through?

Pace? checked

Rhythm? checked

Timing? checked?

Final check. Does my heart rate on the apple watch goes up after a listening session? 

Engineers work with objective data. no?

I don’t expect any new technical stuff in there

its all down to implementation and those 13 statement DR regulators

maybe the low voltage differential  signaling is new but that’s it

kaydee6 posted:

Since this thread has morphed into dsp centric, the next question is why does the 13000 pound streamer use the same dsp as a 2000?

Hi - an answer was presented a few posts earlier

"We use the same 40 bit floating point algorithm in all our SHARC based products. The algorithm oversamples incoming digital audio to either 705.6kHz or 768kHz and buffers the audio for subsequent jitter removal.

Both the ADSP 21489 and 21369 DSPs can run in either 32 fixed point or 40 bit floating point; we only use 40 bit floating point.

The 489 is a little newer than the 369 and runs a bit cooler for the same rate of math otherwise the digital audio is absolutely identical. We differentiate the sound quality using the DACs, PSUs, analogue components and mechanical treatment such as suspension."

Hardly any day and night differences between the 2 dsp chips

No - but less power for a given level of maths is good - and that was behind some of Mr Watts thinking as well with the Chord devices with his use of next generation Xilinix DSP FPGAs with very low power consumptions s.. and subsequent significant reduction of ground plane modulation. In fact with DAVE the level was so low he told me he couldn't measure it.

agreed simon, but all in all I think it's hardly a game changer.

Well, that was fun, and toward the end, quite insightful.  It's not very often we see a thread that attracts R&D gurus from naim, and gain an understanding of the thought process.  The OP will no doubt think we're all nuts, and wouldn't be far wrong.