A plausible benefit and practical use of NOS DACs presented in HifiNews

Orb

New Member
Sep 8, 2010
3,010
2
0
Again in the August HifiNews, this time from Paul Miller.
There has been a lot of discussion about the benefits and cons of going Non-Ovesampling, however one aspect all of us may not had considered is that of high quality downloads in conjunction with NOS DACs.
This is one area NOS could actually work well, in that playing native high res recorded music so there is no oversampling in the 1st place, this means if the NOS DAC supports the higher rates the issues of image duplication should not be an issue, and therefore the benefits outweigh any cons.

This came about due to PM having access to AMR's CD-777 that also is an outboard DAC supporting 24bit/192khz and also NOS, so any true high res recordings can be played with the NOS setting and the associated issues are pushed way beyond 30khz for the digital image and distortion no longer an issue.
PM does a comparison that showed using NOS for 44.1khz and 20khz tone generated both distortion in the audible band and image problems occuring under 30khz.
With NOS and 192khz using same tone, no distortion was measured while the image problem was pushed up to 55khz.
Of course this may raise the technical and theoritical discussion on whether this can have an affect on well designed preamps and amps that have a wide bandwidth, but initial considerations are that this could be the ideal way to listen to high rez recordings.

More info can be seen in PM August Opinions page in Hifinews, definitely thought provoking and one PM-Hifnews is going into more detail next month with more products.

Cheers
Orb
 

DonH50

Member Sponsor & WBF Technical Expert
Jun 22, 2010
3,947
306
1,670
Monument, CO
A 20 kHz tone generates an image at about 24 kHz for 44 kHz sampling clock. That is one of the reasons for the steep output filters on DACs.

For technical geeks, note that the oversampling issue audiophiles have is with the PCM noise shaping process (typically some flavor of delta-sigma (DS) modulator) in the DAC and pre-filters, not oversampling in general (which can be used on any DAC). A Nyquist DAC can also be used to oversampling, thouhg with only modest gain (0.5 bit for each doubling of sampling rate, vs. the multiple bits attained with DS modulation.

For that matter, a DAC running at 192 kHz presented with a RB (44.1 kHz) input stream _is_ oversampling by definition...

I also wonder how many NOS ADCs are used on the recording end vs. the (usual, I suspect) DS types.

Finally, I think a lot of the problems have to do with the conversion process to create the upsamples, rather than oversampling itself...
 

Orb

New Member
Sep 8, 2010
3,010
2
0
Also it can create distortion at 8khz as was seen in the measurement PM showed in the article.

Not sure if you are disagreeing or not Don, I am talking about a DAC running at 192khz being presented with native high rez recording without relying upon oversampling, one way to tell is to see whether the ringing is there, which again PM shows measurements for in comparison.
I agree another considerations may depend on the recording itself, that said it sounds like you are suggesting all studio recording ADC have to oversample to reach even say 96khz (as an example).
If that is the case how were native DSD studios done, not disagreeing as it is one area we do not see talked about a lot and not something I have seen all the info on, also I thought the digital workstations and adc did not need to oversample if operating at say 88.2/96khz and above.

That aside, it is a worthy discussion you raise on the merits of nos at the DAC stage when considering what happens before this.
However, this may have no impact to DAC ringing and the reason of the Meridian type filter (apodising) as it relates to my 1st post , or going NOS where this does not occur but has other issues, but as mentioned those problems are overcome if fed say 192khz downloaded music, this is something seperate to the recording studio adc discussion I feel you touch on but it is still of interest for sure and if someone has info please post (just wanted to clarify that the ringing and benefit of feeding NOS DAC a 192khz downloaded music is not the quite the same situation or context as to your post I feel).

Cheers
Orb
 
Last edited:

DonH50

Member Sponsor & WBF Technical Expert
Jun 22, 2010
3,947
306
1,670
Monument, CO
No disagreement at all Orb, just dragging out the technical weenie angle. It's my job. :)

A DAC converting at 192 kS/s from a 192 kS/s data stream is not oversampling, we agree. However, that DAC could still be an OS type, e.g a delta-sigma running at 1.536 MS/s. You can substitute "ADC" for "DAC" in most of these discussions. My ADC comment was because I do not know what the majority of studios use these days, and if they have an oversampling or Nyquist ADC in them. You can buy either, of course.

Filter ringing can happen in the digital world (before the actual DAC) or analog (at the DAC's outputs). Filter theory is complicated enough that I have shied away from it in my threads to date. I have a reasonable amount of experience and a number of grad classes on the subject, just enough to make me realize how little I know. ;) All life and filters are a compromise...

Oversampling is another of those definitions in the audio world that differ from my world. "Oversampling" to most audiophiles implies a PCM modulator (demodulator for DACs, I guess) and related noise shaping. However, if I use a DS DAC clocked at 1.536 MHz (8x OSR) to provide 192 kS/s effective sampling rate, is that NOS or "native"? Not to me... I translate "native" and NOS in the audio context to a Nyquist DAC (no noise shaping, no modulator) running at the specified conversion rate (e.g. 192 kS/s). While several Nyquist chips will facilitate oversampling, that does not seem to be the norm in the audio world (some high-end units use them that way, IIRC).

A long-winded way of saying I agree with you, I think. - Don

p.s. Would you have a link to the article, or is it paper-only?
 

Orb

New Member
Sep 8, 2010
3,010
2
0
AH yes agree and makes sense, and yeah totally agree - native to me is that 192khz is not derived from a lower rate or modified in terms of internally OS as you say to like 1.53 MS/s or 500khz (which I think ESS Tech DAC chips used to be but these could be bypassed).
I should also mention in my example of the 8khz distortion this was relating to a 48khz sampling rate with the 20khz tone.

Don, if the theory is that the image only has to be suppressed to say 50khz or more, what is the additional benefit in audio (emphasis audio context as appreciate it does matter in other subjects) to OS to the 1.53 MS/s or even 500khz as some manufacturers do?
I appreciate some products are going for insane numbers these days, but I thought the oversampling was only to assist with the filter and push the image outside of the audio band.
I assume one benefit may assist widebandwidth preamps and amps, but I do not think this has been proved with well designed-engineered ones.
Cheers
Orb
 
Last edited:

Phelonious Ponk

New Member
Jun 30, 2010
8,677
23
0
I'm in over my head here, but the only NOS DACs I've heard just didn't sound very good to me. They just came off as slightly muted, or soft, or....something.

Expectation bias?

Tim
 

Orb

New Member
Sep 8, 2010
3,010
2
0
That is possible Tim and not to do with expectation bias I would say.
In theory NOS should work ideally only if the sampling rate of the music is high enough to push the duplicate image beyond the audible frequency.
So if NOS is used for standard CD we have potential problems for the following reasons; A mirror image will be realised from 22khz upwards.
So in this example a 20khz tone will create a mirror image at 24khz (44.1khz minus 20khz), and in terms of energy this will be very high (nearly as high as the 20khz tone and this is also repeated further up the FR still at extremely high energy - meaning loud in relation to other signals).
Also on top of this is the creation of IM or beats that generates distortion, in this example at 4khz.
Now the problem is that music is far more complex than just a single tone and this will mean that the in-band distortion will be spread throughout the whole of the audible frequency of CD music 20hz to 20khz.

However, if NOS is used with say 192khz music track then the issue of the mirror image and distortion (to some extent) are resolved, because the mirror image is pushed up to above 50khz while the distortion is low even in terms of measurement - PM showed various graphs so this is easier to understand with them in front of one.
Now with the test results PM seems to have become interested in the concept of using NOS IF one is using true high rez music tracks, so in the case of the audio product that fits is the AMR CD-777 as it has I think 6 to 8 filters that also include NOS.
The benefit of NOS is that it does away with digital filter affects; consider those who prefer apodising over the usual FIR.

Possibly one other consideration is that going this way with the NOS would enable implementation of gentle analogue filter, without the ringing/apodising/or distortion, depending upon what was used.
I appreciate for some the discussion in comparing analogue filter to digital filter is possibily contentious and maybe should be in another thread as I am not sure if there is a conclusion either way.

Cheers
Orb
 

DonH50

Member Sponsor & WBF Technical Expert
Jun 22, 2010
3,947
306
1,670
Monument, CO
Orb: I went over delta-sigma DACs a little in this thread. The benefits of higher rates are potentially greater dynamic range and filters that are easier to design and with lower slopes. In-band SNR is a function of the oversampling ratio (OSR); you gain more SNR as the OSR increases. And, you can use output filters that are much lower order, with lower ringing, phase shift etc. in the audio band. Also, in-band roll-off is much lower so no peaking of the response is required. A NOS DAC has a 3.54 dB roll-off at Nyquist (22 kHz for a 44 kS/s DAC), one-half the sampling frequency, from sinx/x effects due to sampling and before any other bandlimiting in the system. Delta-sigma's I have worked with use low-order loops (since they have to operate at GHz rates) and relatively high OSRs (32 - 128). So, an OSR of only 8 is not actually very high for a delta-sigma design.

Tim and Orb: IME the analog output filter is a key component leading to audible distortion (distortion in the sense of altering the sound; the filters are linear from a systems point of view). A NOS DAC requires a very high-order (steep) output filter to suppress the images decribed by Orb (and many others). High-order filters tend to ring badly and introduce phase shift well below their corner frequency. A 22 kHz "brick-wall" filter is affecting the phase signficantly at 10 kHz and below, and these filters also tend to present fairly ugly loads on the output buffers that can cause other frequency (and thus time) aberrations in the response. By using a Nyquist ("NOS") DAC at high rate images are less a concern and high-order filters are not needed. In terms of the audio bandwidth, the DAC is not a delta-sigma or other "oversampling" architecture, but is in fact is operating at a rate well above the audio band. Thus, a "nice" filter can be used to roll off out-of-band audio signals and still suppress the (now very high) images generated.

HTH - Don
 

Orb

New Member
Sep 8, 2010
3,010
2
0
Thanks for expanding Don.
Not sure its clear Don in your example but ringing is applicable to the usual oversampling DACs unless utilising and apodising filter, and there are quite a few who it seems do not like this, or alternatively prefer it to FIR type filters we see in most DACs.

The one thing I am not sure of in the PM example is whether this is with or without an analogue filter, or even if he was using the AMR.
That said as he used the same setup for 44.1 and 192khz, it is interesting that the distortion was high with the mirror image at 44.1, while the distortion is pretty damn tough to identify in the 192khz one.
With what you mention in the 1st sentence Don, if the NOS works as shown in PM measurements it actually behaves as well and better in some ways to a typical audio delta sigma DAC.
I notice you calculate for 44.1 but PMs article is specifically about using NOS for high rez only, he is in agreement (and it would be hard to disagree) as it is pretty clear of its difficulties with traditional CD 44.1 music :)

Cheers
Orb
 

DonH50

Member Sponsor & WBF Technical Expert
Jun 22, 2010
3,947
306
1,670
Monument, CO
Where can I find the article?

Apodizing filters trade purely linear-phase design for less ringing by combining two or more filter types, yes? I did not think they completly eliminated the ringing, just reduced it to a few short and small (low-amplitude) cycles. I have not researched them in any detail...

Theoretically a DAC is a DAC, NOS or not... NOS (Nyquist) DACs are much more straightforward and do not have stability issues with the modulators and filters that delta-sigmas do. Their biggest issue is achieving the element matching required for very high resolution.

"High rez" means more bits to me; I keep forgetting audio "hi res" means higher rate as well as more bits. Using a NOS DAC at a high rate, easing the filter designs, makes sense to me.
 

Orb

New Member
Sep 8, 2010
3,010
2
0
It is in the actual August publication and not online unfortunately.
This was from Paul Miller's opinions page, but what makes it interesting is that he backs it up with measurements, none of this is disagreeing between us just potential perspective, as you point out "hi res" to some others including me also means higher rate and not just more bits.
You may be right about apodising filters but they are still evolving and do well with ringing when considering their very latest model; the closest modern comparison I guess would be the review by Stereophile of the Ayre QB-9, which shows traditional reconstruction filter, their own apodising type, and then a more advanced listen filter.
http://www.stereophile.com/content/ayre-acoustics-qb-9-usb-dac-measurements
JA touches nicely on the problem that you can have roll off (which PM does as well), but considering "hirez" and with his measurement it looks potentially acceptable - as I mentioned this is going to be an ongoing process (this was the 1st look and investigation) with PM with considering NOS DAC products combined with "hires" music, which seems an ideal solution.

Unfortunately I would feel bad quoting the article as I tend do that often recently (due to so many good articles) in Hifinews, but technically what it shows is possibly something all of us did not think of; NOS DAC potentially being the ideal solution when used specifically with hires music.
A food for thought, which is backed up to a point with measurements.
Although shame it is not mentioned whether those measurements were taken using the AMR that has 6 filter settings and what if any analogue filter (possibly not due to the high distortion showing on the 48khz graph and with mirror image, while the 192khz still has mirror image but over 55khz and the distortion not noticeable)

Cheers
Orb
 

About us

  • What’s Best Forum is THE forum for high end audio, product reviews, advice and sharing experiences on the best of everything else. This is THE place where audiophiles and audio companies discuss vintage, contemporary and new audio products, music servers, music streamers, computer audio, digital-to-analog converters, turntables, phono stages, cartridges, reel-to-reel tape machines, speakers, headphones and tube and solid-state amplification. Founded in 2010 What’s Best Forum invites intelligent and courteous people of all interests and backgrounds to describe and discuss the best of everything. From beginners to life-long hobbyists to industry professionals, we enjoy learning about new things and meeting new people, and participating in spirited debates.

Quick Navigation

User Menu

Steve Williams
Site Founder | Site Owner | Administrator
Ron Resnick
Site Co-Owner | Administrator
Julian (The Fixer)
Website Build | Marketing Managersing