A Taxonomy of DACs?

[davidbeinct]

Is there an article anywhere that explains the main categories of audio DACs out there, and the (purported) advantages and disadvantages of each? I think, but am probably wrong, that we have two main types currently in favor, with each having a sub-species that is almost its own group:

 

R2R, or multi-bit, of which I gather there aren't too many left, with Schiit being the example with the largest following here.

 

NOS, which for the longest time I took to be "new old stock" but I now realize is non-oversampling. Many (most?) of these, I gather, can only do Redbook, but I have read of at least one that can take HD. I gather perhaps there might be some advantage to using something like HQPlayer to upsample everything if using an HD capable NOS DAC?

 

Delta-Sigma, which seems to be the way most DACs are these days. I gather these use single "switches" (although now some use multiples of switches?) run very fast.

 

I believe that this method is a pseudo-DSD process in its own right, and that therefore DSD capable DACs are a subset of these, but am fully prepared to be wrong. If I am correct, what, assuming one has mostly Redbook with a few hi-rez downloads thrown in, would be the advantage of a DSD capable DAC? I know that many people are preferring using HQPlayer to upsample everything to DSD and use a DSD capable DAC.

 

Please feel free to tell me how little I understand, I promise I won't take offense.

[rodrigaj]

The Well Tempered Computer is a very good reference.

[esldude]

I will help get this going better. What I am writing is my opinion, only occasionally informed by the facts, and as much informed by audiophile wife tales and from having lived through it all. This makes me at least an authority in my own mind. Anyone wishing to argue the facts is obviously wasting their time no matter how wrong I am. I therefore expect you will get lots of replies where posters will do just that.

 

I'll start with CD though digital audio existed before that.

 

In the beginning was the R2R ladder DAC. Its strengths were that is existed and sort of worked. It featured highly fatiguing, etched two dimensional sound, with super solid lows, no wow and flutter, and no noise. WOW! Sold like hotcakes. It was like magic compared to LP, and it would never wear out a disc.

 

Sony was first featuring a single ladder DAC running at double speed. It would switch between right and left channel samples which caused a channel to channel timing shift of 11 microseconds. Why the heck would anyone do that? Beats me, but it sure sounds messed up. Their DAC also had low level linearity issues. Meaning it had 12 good bits and the rest were rather variable. Sometimes a lower bit had a higher output.

 

Philips was there too with a ladder DAC. It was dual 14 bit running at 4 times oversampling. Why? Because Philips had planned on CD being 14 bit until Sony forced the issue. They didn't have any 16 bit DAC chips to use. They ran their 14 bit chips at 4 times oversampling and it was the equivalent of letting the serpent feed you an apple from the forbidden tree. For every 4 times you bump the sample rate, with some digital filtering and noise shaping you can gain one bit of dynamic range. So Philips DACs were really 15 bit for a 16 bit medium. Which made them better though not perfect in low level linearity, and better than Sony chips.

 

In time people made 20 bit chips just to get the first 16 bits to be close to linear and employed digital filtering with oversampling. Digital was still hard, fatiguing and 2D while being the most successful audio medium in history.

 

Then we were saved by one bit MASH or PWM designs. One bit being super oversampled giving super low level linearity, soft analog like sound, and all was well. It did have some odd idles tones, and some said it lacked real resolution and a soft low end.

 

Finally in the fullness of time the lords of audio gifted us with the final solution. Genuine multi-bit delta-sigma DACs featuring low level linearity, capable of real 24 bit performance (other than thermal noise limiting us to 20 bit real world), great noise shaping, and it could be turned up to 96 khz, 192 khz and more and more and more. It had all the benefits of R2R ladder DACs and the pleasing sound of one bit, and comes closest to the theoretical perfection of how 24 (and lately 32) bit DACs are supposed to work. They have become excellent, and cheap beyond belief.

 

Next we detour to SACD or what is called DSD. You run at 64 times the redbook rate, do the oversampling, noise shaping voodoo and have the true purity of the one true bit without the filtering (other than the filtering you have to have of course). Gives something like 50 khz bandwidth and 20 bit resolution. The idea of the purity being what makes it work. Another Sony invention. Of course when recording with DSD, the one true pure bit, you can't edit tracks, adjust levels, perform EQ or any other DSP. Nor can you on playback, unless you convert to PCM and then convert back. I have never decided if this is goofier than the first DAC of theirs which swapped between right and left channel every sample. I tend to think this is the craziest of the two ideas. DSD is crazy, impractical, inefficient and more expensive. So naturally audiophiles love it with a passion.

 

Somewhere around here were the Pacific Microsonics HDCD DACs, but no body cares about those.

 

There were hybrid DACs that use a few multi-bit, and some sigma-delta combos. Best of both worlds. And the Ring DAC which is sort of like having a few bits and varying the level they represent as needed to have the best of both worlds (sigma-delta and multi-bit) and benefits nothing else can provide. So crazy this didn't even fly.

 

So summing up most of the world is DSD or sigma-delta. Some say these are internally the same, but DSD is more pure and right(what poorly disciplined thinking that is). Others say the sigma delta is cheap and wonderful which will never gain credence in high end audio circles, I mean how could it? It's cheap, and wonderful. DSD is pure though sometimes now it uses a few bits, sigma-delta is usually based upon heavily oversampled and digitally filtered 5 or 6 bits.

 

Just don't forget modern NOS and multibit R2R DACs take us back to the real deal and save us from the harsh, flat, 2D digital fatiguing sound even though these are the same designs that earned digital sound that rep to begin with.

 

DSD fixes that by not having filters that ring (not counting the filters it has remember) while old sigma-delta is much maligned having those ringing filter artifacts that make digital sound harsh, 2D and fatiguing. It does this by having digital ringing at ultrasonic frequencies we can't hear. While DSD has ultrasonic noise we can't hear (which apparently is just fine if not better), and wouldn't be workable without this huge dollop of noise.

 

Every innovation is to fix the 2D, harsh, hard, and fatiguing digital sound quality so that we are just one more step closer to smooth analog LP like sound. But we just can't seem to get there from here.

 

Also there are now chipless DACs or what I like to call DACless DACs. They use the bitstream with clocking and filtering and multiplexing to be a DAC without being one. Sort of turns the purity of the one true bit into the personification of the none true bit.

 

It would have been good to include one of those tree like graphs like they use showing evolution of different animals. But I was too tired to do that today.

 

Again I remind you, some of the information in this post is actually true and correct.

[PopPop]

That's a great writeup, Dennis. There aren't a lot of people (well maybe a few on CA) who could summarize that the way you did. Thanks much.

[Jud]

That's a great writeup, Dennis. There aren't a lot of people (well maybe a few on CA) who could summarize that the way you did. Thanks much.

 

Yes, it was great fun and just enough of it was actually true.

[Paul R]

I will help get this going better. What I am writing is my opinion, only occasionally informed by the facts, and as much informed by audiophile wife tales and from having lived through it all. This makes me at least an authority in my own mind. Anyone wishing to argue the facts is obviously wasting their time no matter how wrong I am. I therefore expect you will get lots of replies where posters will do just that.

 

ROTFLMAO!!! Definitely takes the award for best writeup of the quarter.

 

-Paul

[semente]

This thread is very interesting:

 

http://www.computeraudiophile.com/f6-dac-digital-analog-conversion/multi-bit-dacs-vs-delta-sigma-dacs-2066/

 

I wonder why idiot_savant no longer posts here...

 

R

[master]

@esldude - Thanks for that write up.

[steven whitton]

Fantastic write up very interesting and wish i could have had things put so elegantly when researching for my last Dac!

[firedog]

Where's beerandmusic?

[master]

Where's beerandmusic?

 

Oh no... don't go ahead and ruin a good thing now.

[17629]

"Please feel free to tell me how little I understand, I promise I won't take offense."

 

Lets just say that you're 1 poll short re-igniting the cold war.

 

In all seriousness, I don't think this is the best way to classify dac's. I know that I've said this many times and have been ignored, but I'll say it again. In order to consider a dac in any meaningful way, you need to look at 3 things. The digital section, the analog section and the designer. The vast majority of audiophiles look at the digital section almost exclusively. The last thing a well designed dac is, is a generic item. The design of the analog section can easily have a larger effect on the dac's sound than the digital section. It all depends on who designs the dac and what type of parts they use, both digital and analog. Just for an example, if you take the same recording on Redbook, SACD and DVD-A (assume 2 channel), and play them all on my Ayre 5 universal player, you hear some differences, but regardless of format, it still sounds like an Ayre component more than anything. Take those same disc's and play them in my friends Esoteric Universal player and all 3 sound very different from the Ayre, and all they all take on the qualities of the Esoteric.

 

Knowing this should give you an advantage as a consumer, and make things easier for you. If, like me, you prefer Ayre equipment, you'll probably be happier with something from them regardless of what the specs are. If you like a different brand, that's just as good. If you know and like the sound of a particular brand/designer, that will probably be more of a factor than just getting a dac based on the chip or processing.

[master]

Lets just say that you're 1 poll short re-igniting the cold war.

 

In all seriousness, I don't think this is the best way to classify dac's. I know that I've said this many times and have been ignored, but I'll say it again. In order to consider a dac in any meaningful way, you need to look at 3 things. The digital section, the analog section and the designer. The vast majority of audiophiles look at the digital section almost exclusively. The last thing a well designed dac is, is a generic item. The design of the analog section can easily have a larger effect on the dac's sound than the digital section. It all depends on who designs the dac and what type of parts they use, both digital and analog. Just for an example, if you take the same recording on Redbook, SACD and DVD-A (assume 2 channel), and play them all on my Ayre 5 universal player, you hear some differences, but regardless of format, it still sounds like an Ayre component more than anything. Take those same disc's and play them in my friends Esoteric Universal player and all 3 sound very different from the Ayre, and all they all take on the qualities of the Esoteric.

 

Knowing this should give you an advantage as a consumer, and make things easier for you. If, like me, you prefer Ayre equipment, you'll probably be happier with something from them regardless of what the specs are. If you like a different brand, that's just as good. If you know and like the sound of a particular brand/designer, that will probably be more of a factor than just getting a dac based on the chip or processing.

 

I've heard this argument (countless times) and to me that's actually missing the mark (not the argument, but the DAC certainly is).

 

End of the day (it would seem) every DAC and DAC manufacturer is missing the mark. Maybe not in the digital section, but certainly in the analog section.

 

Your (or my) favorite singer or band would sound the same anywhere. Sure some of the room acoustics might come into play, as well as other components, but by how much?

 

DACs are supposed to do that, convert digital to analog, and represent everything accurately. They are not doing it right by coloring with their own sound signature... even if its in the analog section of the DAC.

 

That is the antithesis of the very definition of audiophile.

[17629]

I've heard this argument (countless times) and to me that's actually missing the mark (not the argument, but the DAC certainly is).

 

End of the day (it would seem) every DAC and DAC manufacturer is missing the mark. Maybe not in the digital section, but certainly in the analog section.

 

Your (or my) favorite singer or band would sound the same anywhere. Sure some of the room acoustics might come into play, as well as other components, but by how much?

 

DACs are supposed to do that, convert digital to analog, and represent everything accurately. They are not doing it right by coloring with their own sound signature... even if its in the analog section of the DAC.

 

That is the antithesis of the very definition of audiophile.

 

I've heard your argument too. It makes sense and if that's what works for you, then its the right way to do it. But for myself, I don't buy it.

 

"They are not doing it right by coloring with their own sound signature... even if its in the analog section of the DAC."

 

Take that statement specifically. You can't build something like that and not have it color the sound in some way. Its not like a designer can make a decision that he can build a product that had no color. So if coloring the sound isn't optional, every designer has to do what they think is the right thing. Then it all comes down to choices of how the sound is colored, and not if its colored.

 

In the end, however you look at it, we really choose components in one of two ways. The first is subjective. We listen, and if we like it enough and can afford it, we buy it. The other way is to try and look at the component objectively and make a decision based on factors other than how it sounds. I know some people here will strongly disagree with me for saying this, but I feel that those of us that go down the objective route, are really making choices that are subjective in nature anyway. They just don't see it. (That's not supposed to be an insult in any way, its just my honest opinion.)