R2R vs Delta Sigma DACs - same results?

[xophere]

I have tested some DACs in blind testing of vastly different prices and designs. I had trouble hearing a clear, better sound above, say, around $130. Sometimes I could hear a difference but not be able to describe or have a preference for it. We did find at one point that the preamp we used swayed us much more than the DACs behind it. There have been DAC's over the years that allow you to swap the op amps on the outputs to change the sound of the DAC. I haven't had two with just the op amps changed to be able to test, but I assume that the amplification change could change the interaction with the downstream amps and have an effect. Of course you would expect to see a change in level or of frequency to show it.

I wonder if we end up in this crazy space because we are constantly trying to use metaphors to describe some aspects of our systems and then mistakenly take that to areas where there should be negligible effects. Any valid DAC should be good enough. And then just features and style remain. I could not hear a clear difference between a high end R2R and a $200 desktop DAC. But I certainly can hear the difference between the DACs in computers and crapy TVs and my $100 DACs or even my Qudelix. Again, I am probably hearing the difference in the analog amp circuits. I should redo my test between the DAC in my receiver and my DAC in my headphone AMP. I do tend to prefer the external DAC's but maybe it is just me being silly.

 

[gwing]

Right. So the record company produces a CD. 16 bit. They do the mastering in 32 bit. Then when they create the CD master they dither that down to 16 bit.


I buy it, take it home, put it in my CD player connected digitally to my external DAC. The dac is now converting 16 bit dithered samples. The dynamic range produced can be as high as 120dB. I was pointing out that the R2R DAC does NOT get better dynamic range compared to a Delta Sigma Dac with 16 bit data - so the discussion included both architectures.

I've become interested in this and would like a better understanding.

From what I can gather dithering from a higher resolution master can indeed increase the *perceived* dynamic range of a 16bit signal from the theoretical max of 96db to 120db or so which indeed seems great in providing headroom we can truncate in our DAC. However this dither is I think only applied to the least significant bit of the signal so when you apply a bit of volume reduction does that 'scale' SNR as you would expect or does it rapidly degrade the dithering benefit?

Also, is that perceived dynamic range benefit across the entire audio spectrum or just for specific frequencies? Can we actually measure this (which I guess would be tricky as we would also have to have access to the original undithered master for comparisons) or it just a human perception effect evaluated by listening tests?

 

[antcollinet]

I've become interested in this and would like a better understanding.

The best dither primer I know is from our old friend Monty - he covers it in detail from 11:36 in his digital show and tell here. The whole video is well worth the 25 minutes or so if you've not seen it - or even if you have.

 

[aagstn]

That video is a favorite. He really makes the process understandable.

 

[gwing]

The best dither primer I know is from our old friend Monty - he covers it in detail from 11:36 in his digital show and tell here. The whole video is well worth the 25 minutes or so if you've not seen it - or even if you have.

Thank you, that was a really interesting video. Watched it all

 

[earlevel]

I've become interested in this and would like a better understanding.

From what I can gather dithering from a higher resolution master can indeed increase the *perceived* dynamic range of a 16bit signal from the theoretical max of 96db to 120db or so which indeed seems great in providing headroom we can truncate in our DAC. However this dither is I think only applied to the least significant bit of the signal so when you apply a bit of volume reduction does that 'scale' SNR as you would expect or does it rapidly degrade the dithering benefit?

Also, is that perceived dynamic range benefit across the entire audio spectrum or just for specific frequencies? Can we actually measure this (which I guess would be tricky as we would also have to have access to the original undithered master for comparisons) or it just a human perception effect evaluated by listening tests?

Dither (TPDF usually) itself doesn't decrease noise, it's slightly increases it. but during the process of truncation the error is still known and can be used in a filter to modify its spectrum. Typically, it's pushed up to the high end where our ears are relatively insensitive, but some can also be pushed lower. So it's an exaggeration to say we get 120 dB of dynamic range, we just get lower noise in our most sensitive area of hearing—and very little chance of being in a musical situation where you can actually tell the difference, as opposed to a lab and test tones.

It can really work wonders for small sample word sizes. but we're almost always talking about 16-bit (when we're talking about it at all, considering the easy access to 24-bit audio which obviates the need for dither, much less shaped dither—cue the rocks and garbage). At 16-bit properly flat dithered, the noise floor of typical audio gear will be higher than the dither level.

Someone will argue, "But we need to do it for those with the gear and ears that do support 100dB+ of dynamic range!" (typed with tongue in cheek). My question would be, "Why are such people listening to 16-bit audio?"

 

[antcollinet]

Someone will argue, "But we need to do it for those with the gear and ears that do support 100dB+ of dynamic range!" (typed with tongue in cheek). My question would be, "Why are such people listening to 16-bit audio?"

My question would be "how have they not deafened themselves"

Or, to put it another way, even if they started out with the ability to hear noise 100db Down from the music - then they are not going to retain that ability for very long if they have the volume turned up enough to allow them to do so.

 

[earlevel]

My question would be "how have they not deafened themselves"

Or, to put it another way, even if they started out with the ability to hear noise 100db Down from the music - then they are not going to retain that ability for very long if they have the volume turned up enough to allow them to do so.

Yes—I've mentioned the same in discussions with audio recording folks.

I've been in discussions with luminary audio engineers in their late senior years who maintain they can hear the last bit level or 24-bit. In one discussion, one of them said that "masking does not exist". This is someone who worked on many Motown hits that certainly had tape bleed-through that's more than a thousand times the level of 24-bit audio effects, I guess that was OK.

But at the same time, these people think there is no hard limit below which they cannot hear something. I've had people get furious with me for saying even that at some point the pressure wave won't be able to move your eardrum, or that the sound of blood coursing through their body would drown out the levels they think they can hear. They think I'm demeaning their expertise in their craft. LOL (yes, I laughed out loud when I finished typing that.)

 

[fkam]

Because they all have a common goal: output a near perfect reconstruction of the sound waves (as voltage) that are represented by the digital data.

From dac chip specs, seems they all market themselves to produce more analog sound, more musical, and less clinical nowadays. I take it as they are not just faithfully reproducing digital to analog, but shaping it too

 

[voodooless]

I take it as they are not just faithfully reproducing digital to analog, but shaping it too

No. It’s just talk. The measurements of hundreds of DACs here in ASR show that the overwhelming majority have a faithful reproduction.

 

[Strumbringer]

I had been interested in this topic. Someone recently posted elsewhere that my beloved RME ADI-2 DAC FS was mid-fi at best. I didn't understand this, as it's an incredible piece of gear and does exactly what it should (and more with excellent features). The suggestion was an expensive R2R like Yggdrasil or Holo is what I'm needing. I've never heard either, but honestly don't know that I'd detect a noteworthy difference, based on my research thus far.

 

[oleg87]

Someone recently posted elsewhere that my beloved RME ADI-2 DAC FS was mid-fi at best.

An imbecile on the internet made a noise.

 

[Ropeburn]

From dac chip specs, seems they all market themselves to produce more analog sound, more musical, and less clinical nowadays. I take it as they are not just faithfully reproducing digital to analog, but shaping it too

They don't do "shaping". If they were, it would completely go against the goal and whole purpose of a DAC: conversion to analog as perfectly as possibly achievable.

They know that too. It's a fierce market and you need your products to stick out in a way. So you need a selling point. Why not cater to widespread audiophile nonsense terms? Find the right words, and they'll buy it. With extra nonsense tax too, if your words are better than others'.

 

[Jimster480]

No. It’s just talk. The measurements of hundreds of DACs here in ASR show that the overwhelming majority have a faithful reproduction.

Define "faithful"

 

[NTK]

Define "faithful"

The digital-to-analog conversion process is mathematically deterministic. It is a true encoder/decoder process. Given the digitized samples, we know exactly what the outcome should be. Therefore, we know what is "faithful" or not.

 

[antcollinet]

Define "faithful"

Come on - you know this.

Output is an audibly perfect analogue representation of the digital data from the source.

 

[Jimster480]

Come on - you know this.

Output is an audibly perfect analogue representation of the digital data from the source.

Except that we have found out that many R2R DAC's do not indeed accomplish this task without a high level of distortion.

 

[Hindsider85]

Burr Brown made a nice PDF, where they go into detail why Delta Sigma DACs are simply worse.
Also nice to see, how Delta Sigma DACs pretty much cheat to achieve their supposed superior SNR values.

http://www.jitter.de/pdfextern/DesignSem5.pdf

 

[Doodski]

I simply do not want to watch that video nor do I have 36 minutes and 38 seconds to see if a random video has anything of interest or relevancy to me in any way, shape or form. You should condense the video in text for us and if people want to watch the video that's their choice.

 

[antcollinet]

Except that we have found out that many R2R DAC's do not indeed accomplish this task without a high level of distortion.

Exactly - they fail to meet the definition of faithful**. They are not included in the "vast majority" that do meet the definition.


**though to be fair, some of the more recent releases don't do so audibly.