Thoughts on a bunch of DACs (and why delta-sigma kinda sucks, just to get you to think about stuff)

[prot]

This thread is getting so serious .. Next time you know someone mentions DBTs and we all end up in the dark science section.

Heeey! That was my beer!

Were was I? Ah ya, life is dangerous...

Wait, you did not get that? ... wait ...

 

[jcx]

perhaps not too far afield - feedback theory is central to the internals of Delta Sigma converters and of course the analog buffers, I/V converters and filters after (almost) any DAC will employ feedback in some form or another
 
virtually all useful active circuits use negative feedback - the audiophile marketing "no feedback" really at best means no Global loop feedback and ignores local degeneration that feedback relations are needed to describe fully
 
but it is too big a subject to do justice to in any depth here
 
I did find the Pass audio distortion and feedback article unfortunate - I did press him over on diyAudio and he did admit my points of criticism - I consider his "exaggerating for effect" as pandering to the expected biases of the audiophile magazine audience
 
the "trick" for correct analysis that Pass skated past is that the signal level decreases by the gain of the intermediate stages as you follow the signal backwards from output to the input of a high loop gain global feedback amp
the reduced signal levels hugely reduces the distortion generation in each stage - at the input even undegenerated bjt diff pairs can add less than -120 dB distortion to the difference signal under ordinary operating conditions
 
there is no "tsunami" of IMD products in well designed high feedback amps operating within their limits - until you clip or try to "correct" a deadband
 
I do link, recommend just about everything else Pass has published
 
 
Putzeys feedback paper is a much more reliable popularization for audio enthusiasts with some EE knowledge - still not enough to learn the subject from http://linearaudionet.solide-ict.nl/sites/linearaudio.net/files/volume1bp.pdf (only a little bit of Laplace transfer function math...)

 

[evillamer]

Here's another article on Sigma Delta Conveters:
https://www.eecs.qmul.ac.uk/~josh/documents/Reiss-JAES-UnderstandingSigmaDeltaModulation-SolvedandUnsolvedIssues.pdf
 
In the article, it mentions several issues with sigma delta:
 

 We’ve identified several key issues in sigma–delta modulation: limit cycles, idle tones, dead zones, harmonic distortion, noise modulation, and stability.
 
To some extent limit cycles may be considered a mostly solved problem, whereas for each of the other problems the issues are understood for low-order design but the theory is not yet established for the high-order designs.
 
Dead zones have been effectively characterized for loworder designs, but they have not been reported to be problematic in highorder or commercial designs. Noise modulation is well understood for PCM, yet there is no well established theory for even low-order sigma–delta modulators.
 
Idle tones and harmonic distortion, though not well understood, are clearly related phenomena. With idle tones in particular, well-defined and simple relationships between the input signal and the frequencies of the tones have been observed that do not yet have a theoretical basis.
 
Noise modulation, idle tones, harmonic distortion, and limit cycles may be dealt with, at least in part, through the application of dither.
 
However, dither is less effective when used with a low bit quantizer, which is also when these issues are most serious.
 
Furthermore, dither is not helpful in dealing with stability issues, and will actually decrease the stable range of a sigma–delta modulator.
 
A better understanding of stability (and of other issues) is needed so that robust, high performance implementations may be developed.
 
There have been many promising recent results that may lead toward a theory of stability in sigma–delta modulators; this remains an active area of investigation for the author.

 

[Ableza]

... or someone will soon state the old audiophoole lie, "We really don't know how to measure everything in audio."  It always happens.

 

[Sapientiam]

OK I'll bite - 'we currently have no measurement for how well an amplifier conveys a recording's soundstage'.

 

[Ableza]

  OK I'll bite - 'we currently have no measurement for how well an amplifier conveys a recording's soundstage'.

HA HA HA!  Define the electrical signal properties of "soundstage."  Or even the acoustic field properties of this thing. 

 

[Sapientiam]

If I knew what the electrical signal properties of a recording's soundstage were I'd come up with a measurement.

 

[evillamer]

I think soundstage can be defined by the accuracy/resolution of the dac in reproducing the tiny plankton details accurately at the decay & release of the sound envelope. From my (rudimentary) understanding is Sigma Delta averages/smoothes things out using a feedback & noise shaping or some sort clever engineering tricks(which I understand that R2R DAC doesn't do this). Not sure if that averaging/noiseshaping/filtering affects the ability to reproduce very tiny(micro-rapid) changes in the soundwave accurately. 
 
of course there are many other dac performance aspects like(e.g.) stereo channel separation and IMD that also affects soundstaging.
 
 

 
 
 
 

Measuring reverberation time

 

 

 

[DreamKing]

The problem is that you're using a perception-based term that has nothing to do with science. You will never be able to effectively measure it. It needs to remain the same, under the same conditions no matter what.
 

  If I knew what the electrical signal properties of a recording's soundstage were I'd come up with a measurement.

 

[Sapientiam]

  The problem is that you're using a perception-based term that has nothing to do with science. You will never be able to effectively measure it.
 

 
I agree its perception based but disagree that perception based things have 'nothing to do with science'. Some cognitive scientists are studying how perception works, in detail. To me it does sound rather defeatist to say I'll never be able to measure it effectively.

 

[DreamKing]

   
I agree its perception based but disagree that perception based things have 'nothing to do with science'. Some cognitive scientists are studying how perception works, in detail. To me it does sound rather defeatist to say I'll never be able to measure it effectively.

 
I'm saying soundstage isn't a scientific term (applied science). Too vague and varying to be measured. It's like saying you're gonna measure prat one day, even more confusing.

 

[jcx]

the Smyth SVS Realizer is a practical demonstration that typical electronics don't seem to have a big problem with presenting a convincing synthesized soundstage of a specific set of loudspeakers in a real room
 
they sidestep a lot of the modeling with real measurements, mics in your ears in the real room, in front of real speakers from mono up to 7.1 multichannel setups and calibrate as you turn your head through specified range of angles for your personalized hrtf
 
then they do the same for your circumaural headphones - they supply a Stax Lambda system as a default for those without their own preferred audiophile cans and amp
 
the result is very good - you can localize the virtual speakers as well as the real, compare in seconds by just lifting the cans off your head, tilting them down so the head tracker can't see the target
 
the reductionist position is that good electronics, a complete ADC/DSP/DAC chain inserted gives "out of the head" stable, "realistic" soundstage of "those speakers over there, in this room" - because the phase, frequency response conditions needed for the amps ADC and DAC are trivial
 
http://smyth-research.com/technology.html
 
after hearing a full personal calibration demo of the full Smyth system in a 5.1 setup it is ROTFLAMO funny reading most Head-fi "soundstage" discussions of even headphones, much less claims that specific amps have a serious role in "correcting" soundstage, "in synergy" with certain cans
 
at least crossfeed circuits or plugins, Dolby Headphone make real differences - but still are pale shadows of what the Smyth Realizer does with personal calibrations and active headtracking

 

[DreamKing]

Spoiler: Warning: Spoiler!

  the Smyth SVS Realizer is a practical demonstration that typical electronics don't seem to have a big problem with presenting a convincing synthesized soundstage of a specific set of loudspeakers in a real room
 
they sidestep a lot of the modeling with real measurements, mics in your ears in the real room, in front of real speakers from mono up to 7.1 multichannel setups and calibrate as you turn your head through specified range of angles for your personalized hrtf
 
then they do the same for your circumaural headphones - they supply a Stax Lambda system as a default for those without their own preferred audiophile cans and amp
 
the result is very good - you can localize the virtual speakers as well as the real, compare in seconds by just lifting the cans off your head, tilting them down so the head tracker can't see the target
 
the reductionist position is that good electronics, a complete ADC/DSP/DAC chain inserted gives "out of the head" stable, "realistic" soundstage of "those speakers over there, in this room" - because the phase, frequency response conditions needed for the amps ADC and DAC are trivial
 
http://smyth-research.com/technology.html
 
after hearing a full personal calibration demo of the full Smyth system in a 5.1 setup it is ROTFLAMO funny reading most Head-fi "soundstage" discussions of even headphones, much less claims that specific amps have a serious role in "correcting" soundstage, "in synergy" with certain cans
 
at least crossfeed circuits or plugins, Dolby Headphone make real differences - but still are pale shadows of what the Smyth Realizer does with personal calibrations and active headtracking

 

 
I think the booming VR development is going to introduce competition to the Smyth Realizer as well as more binaural content. It's the best thing that can happen. It won't be about replicating speakers for headphones soon but binaural content (much more of it and hopefully not just for video games) and hardware based on individual binaural HRTF. 

 

[evillamer]

On the topic of soundstage:
 
Although this is marketing video(at the very end of the video), It has quite valuable information with regards to how we perceive sound. see it for yourself.
 
 
 
longer video for those interested in speaker room accoustic:
https://www.youtube.com/watch?v=BroP_iFVJvo

 

[BassDigger]

 
... or someone will soon state the old audiophoole lie, "We really don't know how to measure everything in audio."  It always happens.

 
Bite No.2:
 
What difference does it make, for you? That sand must be really clogging up your ears! 

 

 
Saying that we can measure everything is surely like saying that we understand everything.
Who is ultimately the better scientist, one who thinks that he does or doesn't understand?