Non oversampling DACS: snake oil?

[davetroy]

I've been seeing a lot about non oversampling DACs at varoius websites. They seem to be big with the DIY crowd, although there are numerous small companies putting out versions, too. So, I have some questions:

1. What is the main difference between a NON OS DAC and a regular DAC?
2. What advantages could a NON OS DAC possibly have?
3. What disadvantages?
4. I'm guessing--and I could be wrong here--that the sampling features of DACs were created because non-oversampling didn't do the trick. So, is there a new breed of NON OS DACs that overcome these old negatives?
5. Are NON OS DACs nothing but snake oil?

I'm suspicious because a lot of the NON OS DAC supporters seem to also love the snake oil stuff: $200 wood platforms, expensive ICs, expensive power cords, etc. But, then again, a lot of them also like TVCs, which I like. So I'm not willing to dismiss them outright.

Thanks,
Dave

 

[tbewick]

I've been seeing a lot about non oversampling DACs at varoius websites. They seem to be big with the DIY crowd, although there are numerous small companies putting out versions, too. So, I have some questions:

1. What is the main difference between a NON OS DAC and a regular DAC?
2. What advantages could a NON OS DAC possibly have?
3. What disadvantages?
4. I'm guessing--and I could be wrong here--that the sampling features of DACs were created because non-oversampling didn't do the trick. So, is there a new breed of NON OS DACs that overcome these old negatives?
5. Are NON OS DACs nothing but snake oil?

I'm suspicious because a lot of the NON OS DAC supporters seem to also love the snake oil stuff: $200 wood platforms, expensive ICs, expensive power cords, etc. But, then again, a lot of them also like TVCs, which I like. So I'm not willing to dismiss them outright.

Thanks,
Dave

I should say firstly that I have read about oversampling ADC's and DAC's and still do have any understanding how they work, but I do have some comments. Obviously, digital converter design is rather complex, and there is more than one way of going about designing a converter. Rather than me incorrectly describing how converter designs differ, I think it's better if I just refer you to some sources of information:

http://www.scalatech.co.uk/papers/jaes496.pdf
Dunn and Sandler. J. Audio Eng. Soc., Vol. 44, No. 4, pp. 227 - 244 (1996 April). 1. 'A Comparison of Dithered and Chaotic Sigma-Delta Modulators'.

In its introduction, this paper briefly describes the basics of digital converters.

http://www.essex.ac.uk/ese/research/audio_lab/malcolmspubdocs/C27 AES lecture Introduction to digital audio.pdf
C27 INTRODUCTION TO DIGITAL AUDIO, (tutorial paper), Hawksford, M.O.J., Images of Audio, Proceedings of the 10th International AES Conference, London September,1991

General info on digital audio.

http://www.era.lib.ed.ac.uk/bitstream/1842/297/1/gu96.pdf
'Sigma Delta Modulation Of A Chaotic Signal', PhD thesis by Gary Ushaw, University of Edinburgh, Oct 1996.

Mathematical background on chaos and oversampling.

It's probably best when talking about the performance of digital converters to consider specific design implementations. One oversampling converter may be better or worse than another non-oversampling converter design. This excellent paper by Julian Dunn describes measurements which can be used to assess digital converter performance (registration at Audio Precision is free):

http://ap.com/library/technotes.htm
Dunn, J. (2003). "Measurement Techniques for Digital Audio", Audio Precision Application Note #5, Audio Precision.

 

[WmAx]

I should say firstly that I have read about oversampling ADC's and DAC's and still do have any understanding how they work,

The non-over sampling part is not what I am concerned about. Even with a non-over sample DAC, and using old-fashioned steep non-linear phase low pass filter (and assuming the filter is precisely implemented) to remove the aliasing, the phase distortion should remain inaudible according to prior perceptual research.

However, the particular non-oversampling DACs that seem to be becoming popular in the DIY community do not appear to have proper anti-alias filters! Refer to Audio Note brand DACs. They use a shallow low pass filter that allows for substantial aliasing to remain. These units seem to be broken by design. Audibly different? I suspect that if any DAC has a chance of being audibly different when used for real-world music playback purposes, these are the prime candidates!

-Chris

 

[davetroy]

They use a shallow low pass filter that allows for substantial aliasing to remain. These units seem to be broken by design. Audibly different? I suspect that if any DAC has a chance of being audibly different when used for real-world music playback purposes, these are the prime candidates!

-Chris

Well, that seems to fit in with what I've been reading: supporters say they're broken by design but work anyway. How is that possible? But (and I might be speaking ignorantly here) if an oversampling DAC is filling in information, they're not exactly 100% correct, either, are they?

 

[WmAx]

Well, that seems to fit in with what I've been reading: supporters say they're broken by design but work anyway. How is that possible? But (and I might be speaking ignorantly here) if an oversampling DAC is filling in information, they're not exactly 100% correct, either, are they?

Filling in information? That is false if you mean this to apply to the end-result audio signal. The entire point to the over-sample process is to have as little distortion as possible added to the final generated analog signal. The non-oversampled units at discussion here actually do ADD information ( frequency components that were never meant to be in the final audio signal, as well as incomplete signal re-construction as a direct inter-related consequence ) to the audio signal by lacking an anti-alias ( anti-imaging filter to be more precise ) filter.

As for how they can work? One needs to define what they mean by work. SET ( single ended triode ) tube amplifiers that output 5 - 10 watts ( and with single digit percent THD at any given time ) and alter the frequency response by several dB in some cases work, if you define this as amplifying a line level signal to a point where it can drive a loudspeaker. Just the same, a non-oversampled DAC as discussed here generates an analog audio signal that can be used as a source to feed a sound system. In both cases, if one prefers the sound of considerable added distortion(s) as compared to transparency, it is up to that individual.

-Chris

 

[Rob Babcock]

I suspect that the 'additional distortion' will not be audible. Perhaps you should arrange a DBT to verify that. I suspect you'll find no statistically meaningful difference, WmAx.

 

[WmAx]

I suspect that the 'additional distortion' will not be audible. Perhaps you should arrange a DBT to verify that. I suspect you'll find no statistically meaningful difference, WmAx.

I would gladly do so if someone sends me a unit on loan.

However, it is interesting, as one group has already done exactly that...

http://www.matrixhifi.com/contenedor_discmandac3.htm

Under blind conditions, the Audio Note DAC ( a unit constructed with no proper anti-alias/image filter ) was identified 90 percent of the time. All listeners, under blind conditions, identified the device under test (when it was the Audio Note) as sounding slightly worse as compared to a Sony Discman ( a Discman that was DBTed by this same group against a high end proper DAC, and no audible differences found in that comparison ).

It is interesting in that a DAC that was incredibly poorly designed is only slightly worse sounding as compared to a standard (transparent) DAC.

I should point out that the exact cause of the sound difference was not concluded. Was it the lack of proper anti-alias/image filtering in the Audio Note DAC? Excessive tube based distortion designed in to the pre-amp output stage of the Audio Note DAC? Other design blunders of the Audio Note DAC?

-Chris

 

[davetroy]

Unfortunately, the story that links to is not in English, so I can understand it about as well as I can understand non-oversampling vs. oversampling.

And I'm not sure I'd call it a design blunder when they did it on purpose. An intentional design blunder? I'd like to know what they considered "worse."

 

[Nick250]

The question I am curious about is, how much audible influence does the DAC make in the big picture, if any. Let's compare a $2500 receiver to a $500 receiver for the purposes of the discussion.

Nick

 

[WmAx]

Unfortunately, the story that links to is not in English, so I can understand it about as well as I can understand non-oversampling vs. oversampling.

Use an online translation engine.

-Chris

 

[WmAx]

The question I am curious about is, how much audible influence does the DAC make in the big picture, if any. Let's compare a $2500 receiver to a $500 receiver for the purposes of the discussion.

Nick

How does the cost of a receiver apply?

Any properly designed DAC should produce no audible distortion(s) or noise, regardless if it is the one included in a $39.99 Sony Discman or a $5,000.000 stand alone audiophile DAC.

-Chris

 

[tbewick]

The question I am curious about is, how much audible influence does the DAC make in the big picture, if any. Let's compare a $2500 receiver to a $500 receiver for the purposes of the discussion.

Nick

I really think that it's difficult to make generalized statements, because there are so make different DAC's available. DAC's can introduce audible errors, some may be obvious, like a portable CD player having more background noise than a separate CD player. Other errors may be fairly subtle, and require careful subjective and objective testing to establish -

'Fielder has presented an objective technique suitable for evaluating the audibility of nonlinearity produced by digital converters. In [1] it is shown that noise and adjacent signal masking effects can often ameliorate nonlinearities, but that high-level tests show that seemingly small discontinuities can be quite audible under the right circumstances". In the absence of an exact requirement for inaudibility of distortion, a tight overall specification must be sought.'

- http://www.ee.usyd.edu.au/~jimr/pubs/papers/aes97.pdf
Rathmell, J. et al. (1997) "TDFD-based Measurement of Analog-to-Digital Converter Nonlinearity", Journal of the Audio Engineering Society, Volume 45, Number 10, pp. 832-840; October 1997.

One thing that has recently been discussed a lot are non-linearities introduced by overload of a DAC's digital anti-imaging filters. The earlier Audio Precision paper I mentioned describes tests that examine this and concludes:

'Most signals driving into a DAC are not likely to cause any overshoot, and so the amount of headroom beyond 0 dB FS is not relevant to the faithful reproduction of those signals. However, some signals may cause overloads in DACs. The MLS signal is not meant to be a representative signal. It is being used as a (nearly) worst-case signal in order to measure other effects. For example, if—in another device—the kind of signal inversion that occurs in the trace of DAC “A” were to occur only just above full scale (rather than at +5.3 dB FS) it may then produce audible artifacts in the presence of some high-level material.'

page 102.

 

[mtrycrafts]

Use an online translation engine.

-Chris

Interestingly the Google translator is not translating that page. Went to the main website and other pages are translating fine, not this one, even from the main page.

 

[davetroy]

Interestingly the Google translator is not translating that page. Went to the main website and other pages are translating fine, not this one, even from the main page.

That's because Google likes Non OS DACS.

Seriously, though, the people in the group don't seem to say what they liked or didn't like about the Discman against the Non OS. Still, it was pretty overwhelming. I'm still curious, though; I would like to hear a Non OS DAC, although I'm not sure if I'm willing to pay $350 for the privilege. Mhdt Labs does, however, have a seven-day refund policy, so the only risk is shipping.

 

[Nick250]

I really think that it's difficult to make generalized statements, because there are so make different DAC's available. DAC's can introduce audible errors, some may be obvious, like a portable CD player having more background noise than a separate CD player. Other errors may be fairly subtle, and require careful subjective and objective testing to establish -

'Fielder has presented an objective technique suitable for evaluating the audibility of nonlinearity produced by digital converters. In [1] it is shown that noise and adjacent signal masking effects can often ameliorate nonlinearities, but that high-level tests show that seemingly small discontinuities can be quite audible under the right circumstances". In the absence of an exact requirement for inaudibility of distortion, a tight overall specification must be sought.'

- http://www.ee.usyd.edu.au/~jimr/pubs/papers/aes97.pdf
Rathmell, J. et al. (1997) "TDFD-based Measurement of Analog-to-Digital Converter Nonlinearity", Journal of the Audio Engineering Society, Volume 45, Number 10, pp. 832-840; October 1997.

One thing that has recently been discussed a lot are non-linearities introduced by overload of a DAC's digital anti-imaging filters. The earlier Audio Precision paper I mentioned describes tests that examine this and concludes:

'Most signals driving into a DAC are not likely to cause any overshoot, and so the amount of headroom beyond 0 dB FS is not relevant to the faithful reproduction of those signals. However, some signals may cause overloads in DACs. The MLS signal is not meant to be a representative signal. It is being used as a (nearly) worst-case signal in order to measure other effects. For example, if—in another device—the kind of signal inversion that occurs in the trace of DAC “A” were to occur only just above full scale (rather than at +5.3 dB FS) it may then produce audible artifacts in the presence of some high-level material.'

page 102.

Ok, the tech part of this way over my head. Thanks for trying however.

Nick

 

[krabapple]

The non-over sampling part is not what I am concerned about. Even with a non-over sample DAC, and using old-fashioned steep non-linear phase low pass filter (and assuming the filter is precisely implemented) to remove the aliasing, the phase distortion should remain inaudible according to prior perceptual research.

Exactly. All players used to be 'non oversampling'. The key there as you say was high-quality, precision filter implementation. The introduction of oversampling made life easier for CD player makers.

However, the particular non-oversampling DACs that seem to be becoming popular in the DIY community do not appear to have proper anti-alias filters!

Yup. Typical audiophile silliness. Combine that with a belt drive and tube electronics and you have the perfect audiophile CD player -- a high-cost unit that really will be likely to have a 'sound' of its own.

 

[mtrycrafts]

Exactly. All players used to be 'non oversampling'. .

That may have been the very first ones. Soon after that I started seeing 2X on up, 8X oversampling. this goes back to the late 1980s.

 

[krabapple]

That may have been the very first ones. Soon after that I started seeing 2X on up, 8X oversampling. this goes back to the late 1980s.

Yes, and CD players appeared circa 1984 ..I'd already owned two by 1990.

 

[westcott]

Filling in information? That is false if you mean this to apply to the end-result audio signal. The entire point to the over-sample process is to have as little distortion as possible added to the final generated analog signal. The non-oversampled units at discussion here actually do ADD information ( frequency components that were never meant to be in the final audio signal, as well as incomplete signal re-construction as a direct inter-related consequence ) to the audio signal by lacking an anti-alias ( anti-imaging filter to be more precise ) filter.

As for how they can work? One needs to define what they mean by work. SET ( single ended triode ) tube amplifiers that output 5 - 10 watts ( and with single digit percent THD at any given time ) and alter the frequency response by several dB in some cases work, if you define this as amplifying a line level signal to a point where it can drive a loudspeaker. Just the same, a non-oversampled DAC as discussed here generates an analog audio signal that can be used as a source to feed a sound system. In both cases, if one prefers the sound of considerable added distortion(s) as compared to transparency, it is up to that individual.

-Chris

Danger Will Robinson!!!!!

You have more guts than I do. You are speaking heresy, as some would lead you to believe.

I will say that I have run tubes and solid state and over the years, all my equipment is SS. I too prefer the cleaner signal.