An argument against dsd

I counter with a present day argument for DSD as opposed to a 10 year old argument against dsd.

http://www.positive-feedback.com/Issue68/dsd_pcm.htm

Not so fast !

The latest issue of HiFi Critic has a fascinating piece on DSD. SACD is coming back, but in a discless form that can be played straight from a PC : ".... the development of the DoP protocol (DSD over PCM) repackages DSD into a PCM-like datastream that fits within the frames of a 16 bit/176.4 kHz pipeline that is easily conveyed over USB 2.0 "

More background from the BitPerfect people here

I must admit I still cant see the benefit of a format that is anchored to an effective 16 bit dynamic range  - but DoP is nothing more than packing a 16 bit stream into a 16 bit word within the 24 bit PCM sample word which can then be handled with regular framing protocols such as SPDIF etc - or even a private extension of the WAV format - and the 'extra' 8 bits because of the limited DSD word size are used for stream identification - which can be extended for multichannel etc. So DoP has advantages for multichannel over a PCM stream using 16 bit 'words'  - but why bother for stereo?

I must be missing something...

Simon

As I understand it, noise shaping extends the dynamic range "which can be greater than 150db in the audio band below 20kHz" :

http://www.positive-feedback.com/Issue59/dsd.htm 

There's a very interesting article by John Siau about the disadvantages of DSD here:

http://www.realhd-audio.com/?p=74

I am not really sure this is an update making the case for dsd better.

The first argument against dsd in modern recording is that you cannot edit it digital, so unless you favor analog recording and mixing dsd cannot be used.

The dsd backers seems to be against digital audio, wanting the good old analog tape sound back. Of course dsd may sound better than 24/96 highres recordings or better,  you can only know if you have heard both options.

I've been hearing about DXD that to me sounds very similar to what Jan-Erik was mentioning. That's the problem with emerging tech, too difficult to keep up.

Will the next stand-alone DAC incorporate any of these advancements? Will Naim be able to retrofit the streamers? GIves me a headache honestly so I'll just keep enjoying what I have for now. The DAC helps the SBT, no doubt, but what was surprising was how much it improved my TV and HT gear sound.

Arun

I have been told by a Naim insider that the 384 kHZ sampling rate on the DAC V1 will allow it to handle DSD, but further reading indicates that it’s to handle DXD in native format... Anyway, I have been reassured that the DAC-V1 is future proof...

Jan

DXD is 'just' a 384 kHz PCM sample rate and has nothing to do with the DSD format.

the article on Positive Feedback states that the quality of DSD is better than PCM,

The highest resolution 24-bit recordings speak well of PCM but in my system they don't quite equal the best DSD and analog recordings.

but DSD carries some editing issues that have to resolved, but DSD is best for archival purposes, even though they still prefer a fast analog tape though.

It could be great mastering and/or today's converters but the SACD sound like that found on the new ORG Music releases 'The Bridge' & "My Favorite Things' is totally convincing. This is what digital should have sounded like from the beginning -all IMO of course!

Jan, you will need to explan than... DSD sigma processing as used in SACD has, I read,  a theoretical maximum of about 120 dB, but in practice with current technology is about 105db. 105db approximates to about 17.5 bits of dynamic range.

24 bit PCM has a theoretical maximum of 144 dB, with real dynamic range of about 124 dB ( ie approx 21 bits)

As far as noise shaping, this is required because DSD introduces significant quantisation noise, so through DSP methods the noise is transformed up the frequency spectrum and consequently gets concentrated and becomes even more intense. This 'feature' of DSD ultimately limits the frequency spectrum compared to equivalent PCM, and usually heavy low pass filtering is applied above 20 kHz or 40kHz ( the next octave up) to reduce distortion and protect damage to tweeters and power amps.

Id be interested to see the explanation of how managing quantisation error noise this way adds to the dynamic range in the way you state.. The article you refer to has no explanation and carefully worded so perhaps only certain specific contrived conditions can yield such a dynamic range if the encoded signal follows a specific pattern.  As usually reducing quantisation errors reduces the difference between theoretical and realworld dynamic range. Or is this a new format of DSD with a higher bit rate  compared to SACD?

Simon

Simon

the article states

... typical dynamic range of a DSD signal sampled at 2.8224MHz which can be greater than 150db in the audio band below 20kHz.

and

... DSD at double the rate (5.6448MHz) has an extended audio range of 0-40kHz above where the noise floor then starts to rise gently.

The author of the article has been part of the original development of SACD and been involved/consulting in all kind of DSD-related issues, so I believe that gives some additional credibilitity to statements in this article.

Cheers

Aleg

Aleg, well also being somewhat familiar with encoding systems and work in the space where ICT and communications codecs and encodimg algorithms are used and described I am wary of such statements of 'which can be greater' with no qualification. The way delta sigma works there may be some extreme examples in a signal pattern such as an amplitude changing low frequency sine wave that lend themselves to a high dynamic rate, but has no or little consequence to the real world or music. Delta Sigma encoding for a given Fs has a lower dynamic range than multibit encoding. (though direct comparisons are a little loose)  To drive dynamic range up on single bit, you need to significantly over sample as otherwise the high frequency quantisation errors become too prominent in the audio band...

Delta Sigma preceded SACD and is a staple of anyone studying encoding systems and DSP, so I am up for advances, but they need to be explained to hold water as on their own to an engineer its just sales/marketing mumbo jumbo and should be treated cautiously.

IME the advantage of DSD was it was easier and cheaper to produce a DAC for a given encoding quality compared to mutibit, but DAC / DSP filtering technology has hugely moved on and 24 bit multibit has become more prevalent since delta sigma was first developed and therefore that advantage i suggest is somewhat diluted.

Simon

EDIT: Likesmusic interesting article posted above, with a more level headed view of audio recording engineering perspective of DSD, and interesting view that current single bit format is flawed and if doing it again he would suggest develop a 4 bit delta sigma algorithm which makes sense to me.... If feels like DAB doesn't it?

Simon

i came across an article from a former anti-dsd engineer who has been converted to dsd as best encoding format for music distribution,  albeit via pcm-recording to analog-mixing to dsd-distribution Route

read here in the section SACD

http://www.acousence.de/index....temid=60&lang=en

Any views on this one?

cheers

Aleg

 Hi Aleg, I must be tired, I have read and re read your link but am struggling.. The author complains about the HF noise in DSD obviscating some of the HF info and temporal info and so they have experimented by reducing the DSD dynamic range to reduce the effect of the HF noise above 20kHz.. But by doing so have produced pleasing results.. Specifically with his Wagner recording..

Hardly sounds like  a ringing endorsement but he seems quite happy with his new SACD master...

Having just read Adam Meredith's reference to Henry Reed's the 'Naming of Parts' on another thread   ... And its  parody of the drill sergeant with his comparison and contrast of the banal and beautiful and cold and emotionally warm... It seems quite fitting.

Simon

pure speculation - but i find it very interesting that audiophilleo now offers an audiophilleo 1 special edition which supports "384 Khz SPDIF output as well as DSD128 over the DoP protocol".  The usb implementation of the DAC V1 is from audiophilleo.

jerry 

Jan, I am sure the DAC V1 is no different but i seem to remember the NDAC can handle sample rates of 768kHz.  The limiting factor is the max clock frequency of the DAC chips  and the DSP.

Simon