DSD handling in NDX 2

Streaming Audio
1

In my understanding NDX 2 uses Burr Brown PCM 1792 which can handle both DSD and PCM.

So does Naim still first converts DSD to PCM in case of NDX 2 before sending it to the chip ?

I know that is the case with ND555 as the chip used in ND555 (PCM1704) cannot handle DSD. So Naim first converts DSD to PCM (which is not lossless I shall say).

But is that the case with NDX 2 also ?

Regards,
Sourav

2

Yes Naim currently use several Texas Instruments DAC converters branded under the Burr Brown label. They all offer PCM or delta sigma (DSD) conversion other than the PCM1704K which is purely PCM. However the up stream components in the Naim products, specifically the oversampling and reconstruction filter DSP using the Analog Devices SHARC processor are configured as PCM only.
Consequently any DSD is converted to PCM and then processed as PCM prior to physical PCM conversion within the Naim streamer/DAC products.

1 Like
3

Hi @smazumder

Simon has summarised the situation in his post. Just to fill in the full picture of the how and why:

  • On the Burr Brown DAC’s they can be configured into an external oversampling filter mode, which means we feed either 705.6 or 768kHz dual channel I2S data at the DAC, which directly drives the sigma delta modulator.
  • When doing PCM we run a 21489 SHARC to do a custom over sampling filter that has far superior capabilities then the oversampling engine built into the DAC’s. It uses about 40% CPU time on a 2GFlop DSP, so it’s doing some serious number crunching.
  • When doing DSD mode we do a single step DSD bit stream to 705.6kHz conversion so the data is direct to feed to the modulator. Hence avoid multi stage conversion, where every digital filter will not be totally transparent.

So why not just feed the DSD straight into the DAC? In practice the internals of the Burr Brown DAC (and all dacs) are built at a cost and the performance of the DAC’s DSD support (+ external analogue components to filter off the HF noise inherent in DSD) is not as good as using the DSP. It also compromises PCM mode.

Overall the power of a SHARC running Naim custom DSP code, a great clocking circuit, the Burr Brown DAC running in external over sampling mode, plus living in a very nice electrical environment and a top notch analogue output gives very competitive results.

Best wishes

Steve Harris
Software Director
Naim Audio Ltd.

7 Likes
4

Thanks @Stevesky and @Simon-in-Suffolk for the detailed answer.

I am not technically sound enough in this space. But I think I got what you guys are saying. With that few followup Qs from

  1. So if I take the digital out from NDX 2 - the oversampling to 705.6 KHz does not happen at all. Right ? So the downstream DAC which is getting the DOP over SPDIF would get the DSD stratight. Right ?

  2. From your explanation it sounds like the DSD handling capability of ND555 would be slightly poorer than NDX2 as 1704K in ND555 does not give the configuration option for 705.6. So Naim has to convert DSD to PCM (which is surely lossy) to make it 76kkHz at DSP level before sending it to the DAC chip. Right ? Or is that little over simplification of the reality ?

Regards,
Sourav

5

I like that… living in a very nice electrical environment :ok_hand:

6

With my Naim DAC, although I’ve not done extensive testing, I’ve found that using Roon to convert to PCM seems to sound better.
I do use some DSP though so even when using DoP Roon converts to PCM to do its DSP and then back again.

7

@Guinnless If u mean using nDac then ur observation makes sense to me. Because nDac also uses 1704K chip which can only handle PCM. I also had nDac (got it in 2010) but sold in 2014 because of lack of proper DSD support

So then comes whether Roon uses better algorithm and processor vs Naim and also the h/w capacity.

With NDX2 I have found I prefer NDX2’s conversion than Roon’s. But my Roon core is running on a modest MacmIni 2015.

Regards,
Sourav

8

Yes I mean nDAC. My Roon core is on an i5 7th gen is about 10x for PCM conversion and 4x for DoP.

9

Hi @smazumder

  1. on S/PDIF out we supply native audio to 24/192kHz and DSD64 DOP. That is a limitation of the S/PDIF format. Beyond 192kHz the system can do compatibility conversion to PCM to nearest integer sample rate so it will play. The idea being that the external DAC will take the native stream and then process accordingly - you don’t want the streamer limiting the external DAC’s capabilities data and clocking wise.

  2. ND555 - not really. In the ND555 we drive the resistor ladder in the DAC at bypassed oversampling rates. However in a ND555 it chews off issues like mechanical isolation and damping, better domain management of cleanliness (generic computing processing/DSP clean domain processing/analogue output), plus removes PSU noise out of chassis. It costs a lot more as it does a lot more :slight_smile:

Best

Steve Harris
Software Director
Naim Audio Ltd.

1 Like
10

Thanks again Steve for the detailed input.

  1. I am ok with DSD 64 for the time being in SPDIF out. My external DAC (P S Sudio Direct Stream) neither supports more than DSD 64 because of spdif limitation. But I always wondered why Naim Streamers (especially the high end ones) do not support USB output which typically can support till even 256 DSD. Eventually when there would be more downloads of DSD available in 128 format (or higher) this would become a big limitation.

  2. ND555 - I understand (and appreciate) the supreme level of isolation of every aspect in ND555 buy its mechanical design and mandatory use of external PSU. But once information is lost in DSD to PCM conversion no other thing can make up for that. So are you saying driving the r2r ladder in the dac chip of 1704k at bypassed sampling rate (I think that means oversample the data to the max so that the DAC does not oversample at all) can stop the information loss ?

Regards,
Sourav

11

Hi Sourav,

On data loss conversion its primarily a battle of who does the best filter with DSD. The big curse of DSD is that there is a huge amount of high frequency noise that needs to be filtered off. In a native DSD system that is done in the analogue domain with a low pass filter - typically rolling off at about 50kHz, but it’s not uncommon for manufacturers to aim for about 25-30kHz.

In the Naim solution we do it in the digital domain inside the DSP as that is more accurate than the analogue domain equivalent and the 705.6kHz ‘raw data’ output going to the DAC is pre-cooked ready to go to the modulator or resistor ladder driver. We also didn’t want to compromise the analogue output stage by switching in analogue domain DSD correction filtering and gain stages.

There is a reasonable article on the subject here on PCM vs DSD:

Best wishes

Steve Harris
Software Director
Naim Audio Ltd.

2 Likes
12

HI Steve,

Steve thank you again for these details explanation. I can understand how busy you are given your position. But you r still taking time in forum to explain these things does really show Naim’s confidence on the solution. I have seen in case of P S Audio’s dac, Ted Smith (the chief designer/architect/engineer of the dac) always taking time in forum to explain these things to the novices like me.

However, to be honest I am not fully convinced (or understood - may be) with your point that doing the filtering at digital domain is more accurate. But this is a comment based on my limited understanding. You may have strong points to prove that.

However, I surely accept the fact that finally it is about who does Math better or implement the Algorithm better given hundreds of other constraints. And surely with bigger product like ND555 you have less compromises on these and you have less constraints as well as less unwanted behaviour of the overall system at runtime. Also shifting the switching the filtering to digital domain than analog domain surely makes sense to me.

At least these explanations now make me aspire for ND555 when I have money. But surely given the price point that I wud compare it with other contenders like Rossini, MSB, Lampizator, TSS at that time.

BTW, (from my last 3 months experience with NDX 2) I do like its DSD representation very much. May not as good as P S Audio’s dac. Or I should rather say different but equally enjoyable on many specific recordings. What NDX 2 does better is a little more representation at the low frequency level. That makes few Jazz and Rock records sound more lively than P S Audio’s representation in my ear and setup.

This is in comparison to what I experienced in nDac back in 2014 when the DSD support was introduced to it. I sold it after that. Uniti Core’s DSD conversion is also pretty bad.

A last point. I surely think it may be a good idea to have a white paper on what you described in last 2-3 posts to me. That would demystify many of wrong myths around how DSD is wrongly/not properly handled in these new streamers of Naim. Naim is surely taking a different path but not necessarily better or worse than others. It is up to end users to compare the sound and decide.

Thanks again.

Regards,
Sourav

13

I think Steve was simplifying many points with regard to filtering… one key benefit of filtering in the digital domain is that it can reduce ultra sonic noise energy prior to the down stream conversion components… this will reduce intermodulation artefacts and therefore a better reconstructed signal compared to filtering in the analogue domain after conversion… intermods are not harmonic based at source, and therefore are best avoided when you can, as they are difficult or impossible to filter later… as they can create artefacts in the pass band.

1 Like
14

I believe for the similar reason naim does not recommend use of flexible volume control in NDX 2

15

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.