State-of-the-Art Digital

[microstrip]

I was never confused but perhaps DCS is? They talk out of both sides of the mouth at the same time with their ad copy. It is obvious if you read it carefully.

Well, what they write in these articles is technically consistent with their products and independent measurements and agrees with the private talks I had with their people and many interviews and articles all over the net and magazines along twenty years. But surely you are free to consider it just marketing.

And you are just telling us your preference for DCS...plenty of alternative opinions about them here as well.

Yes, and I am explaining the why's of my preference with proper references. At less I know why I prefer this type of sound. I have a similar attitude towards the very different sounding Metronome's or the Kondo DSL DAC, that I also consider SOTA.

Surely we can't expect that everyone likes the sound of the DCS, but taking a broad overview in this forum it seems a lot of people prefer it.

 

[acousticsguru]

All great and some love the result and others...less so. Just because a company conquers RADAR doesn’t mean it can do audio better than others. It means they will likely have a complex solution because RADAR of that type probably required a complex solution rather than a simple solution perfectly executed.

I cannot comment on the Vivaldi because I have only heard it briefly at shows, but earlier DCS was not SOTA sound IMO. SOTA technical for sure...sound quality? Nope not buying it.

Radars are all about low level resolution and signal-to-noise ratio, needless to say, in order to get great sound, one still needs to build a DAC around any converter board or chip set. Why do you think 99% of all audiophile DAC use off-the-shelf chip sets. The number-crunching part takes engineering chops the audio engineers simply don't have.

Earlier dCS were SOTA in the context of their era - 20 years have gone by since then. Even the Vivaldi 1.0 didn't take full advantage of the revised RingDAC, but relied on the bit-mapping algorithms of the older models. I'm guessing that is why Vivaldi owners on this forum and elsewhere make mention of the fact they own the updated 2.0 version.

As it being a 5-Bit converter, that's partly historic as it is what Sony's Sonoma recording system uses:


Technically speaking, the standard could have been anything from 5 to 8 bits at the DSD64 sampling rate of 2.8 MHz, 5 bit being the minimum (more bits, more headroom for noise shaping). In case anyone is wondering, yes, Sony DID lie to us…

As a matter of fact, I don't know of any single-bit modulator other than the Grimm AD1 - at least that's what Grimm claim it is (in turn, I know of only one recording label that uses a Grimm AD-1, at least occasionally).

Greetings from Switzerland, David.

 

[Al M.]

I don't have any subscription, so don't get to read many reviews, but I do remember reading Harley's description of the sound of the Vivaldi as (paraphrasing) making a piano sound like a piano etc., saying to myself, now look here, a critic who expects from a component what I want in an audio system (in contrast, ironically, to so many audiophiles who expect an audio system to produce something other if not "better" than reality).

Greetings from Switzerland, David.

That was the main reason why pretty quickly, after some break in, I preferred the sound of the Yggdrasil DAC to my Berkeley Alpha DAC. A violin sounded more like a violin, a flute more like a flute.

As for "better" than reality, yielding an opposite experience: initially I was smitten by a perceived greater transparency of the Denafrips Terminator DAC over the Yggdrasil, but eventually I discovered that this "transparency" was bought at the expense of a falsification of timbre, and thus not real, causing me to return it.

 

[microstrip]

Quoting dCS

"Everything presented to a Ring DAC is oversampled, noise-shaped and converted to 5 bits at either 2.822 or 3.07MS/s. This arrangement gives 24-bit performance in the audio band, low noise and a frequency response that extends beyond 100kHz."

After the 2.0 upgrade the Vivaldi DAC uses 5.644 or 6.14 MS/s conversion - twice the rate stated in their article. Each 5 bit DAC is implemented using 32 switched resistors of the same value - 2^5 =32 - the opposite of a ladder DAC. In order to avoid systematic errors or correlations the resistors for each conversion are picked by a random algorithm. There are four DACs in each Vivaldi DAC - L/R and +/- phase.

 

[acousticsguru]

Stereophile flat out calls the Ring DAC in th Elgar a 5-bit unitary weighted resistor ladder DAC (High speed latched switching metal film resistors) running at 64fs oversampling. It had a true measured resolution of around 20 bits.
Vivaldi I read mixed info. Stereophile says it is an improved ladder type but Tone magazine says it is using a FPGA...perhaps they refer to the digital filters?

It needs the computing power for the bit mapping process. I believe the measured record resolution is 21 bits thereabouts? The Vivaldi having lowered the noise floor of earlier RingDAC incarnations by another 3 dB would suggest it's in the ballpark. Logic would have it that the total system noise floor must make it (near-)impossible to reproduce anything close. It never ceases to amaze me that audiophiles, of all people, tend to be obsessed with numbers, much like car owners with horsepower? Having said that, ADCs and DACs do need computing overhead.

Greetings from Switzerland, David.

 

[morricab]

Quoting dCS

"Everything presented to a Ring DAC is oversampled, noise-shaped and converted to 5 bits at either 2.822 or 3.07MS/s. This arrangement gives 24-bit performance in the audio band, low noise and a frequency response that extends beyond 100kHz."

After the 2.0 upgrade the Vivaldi DAC uses 5.644 or 6.14 MS/s conversion - twice the rate stated in their article. Each 5 bit DAC is implemented using 32 switched resistors of the same value - 2^5 =32 - the opposite of a ladder DAC. In order to avoid systematic errors or correlations the resistors for each conversion are picked by a random algorithm. There are four DACs in each Vivaldi DAC - L/R and +/- phase.

A sigma/delta converter doesn't have metal film resistors at all...as I said it is something of a hybrid but has elements of both. Why this should or should not be inherently superior is not at all clear to me. It looks like just another way to slice and dice the digital signal.

 

[Al M.]

A sigma/delta converter doesn't have metal film resistors at all...as I said it is something of a hybrid but has elements of both. Why this should or should not be inherently superior is not at all clear to me. It looks like just another way to slice and dice the digital signal.

I suggest you read the attachment (from The Absolute Sound) that I posted yesterday in #935. The Ring DAC is a quite impressive concept that does solve certain problems.

I downloaded the article at a time when I had more interest in dCS sonics than I do now. Perhaps I might not mind having the full Vivaldi stack in my system (I loved some very important aspects of its sound but have not had the opportunity to evaluate all of them, such as rhythm & timing), but not at the expensive price to pay for it.

 

[microstrip]

A sigma/delta converter doesn't have metal film resistors at all...as I said it is something of a hybrid but has elements of both. Why this should or should not be inherently superior is not at all clear to me. It looks like just another way to slice and dice the digital signal.

Sorry, it is clear you are not interested in understanding how the ring DAC works. The technical advantages are clearly explained in the DCS pages and reviews that we are referring. They show in the measurements. The subjective aspect must be left to the listener.

Surely we can imagine that a DAC using 2^24 switches and equal resistors would measure and sound better ...

 

[morricab]

Sorry, it is clear you are not interested in understanding how the ring DAC works. The technical advantages are clearly explained in the DCS pages and reviews that we are referring. They show in the measurements. The subjective aspect must be left to the listener.

Surely we can imagine that a DAC using 2^24 switches and equal resistors would measure and sound better ...

Measure better? Sure that can be tested. Sound better? You have quite an imagination indeed if those technical details translate directly into better sound. How do you justify this?

What those self-serving documents from DCS don't discuss are the potential downsides from how the technical perfection is achieved. Now, maybe they are right and if so, then why can so many people still be critical of the sound? I think there are many here that might consider MSB (who make their own DACs) or CH Precision (which despite being a VERY technically competent company STILL use an off-the-shelf BB DAC chip (ok 4 per channel but still off-the-shelf) or even Lampizator, TotalDAC(who make their own as well) or Aries Cerat to be at least on the same level sonically. Why hasn't that technical superiority translated into total dominance in digital where all consider it to be the end all and be all (clearly DCS thinks their basic tech is as it has stayed constant now for a long time).

Would you claim that the DCS is simply the most true to the signal it is being fed and if we don't like the sound it is because the recording is poor or the system inadequate?

Is there another DAC out there that you consider a truly worthy challenger?

 

[morricab]

Radars are all about low level resolution and signal-to-noise ratio, needless to say, in order to get great sound, one still needs to build a DAC around any converter board or chip set. Why do you think 99% of all audiophile DAC use off-the-shelf chip sets. The number-crunching part takes engineering chops the audio engineers simply don't have.

Earlier dCS were SOTA in the context of their era - 20 years have gone by since then. Even the Vivaldi 1.0 didn't take full advantage of the revised RingDAC, but relied on the bit-mapping algorithms of the older models. I'm guessing that is why Vivaldi owners on this forum and elsewhere make mention of the fact they own the updated 2.0 version.

As it being a 5-Bit converter, that's partly historic as it is what Sony's Sonoma recording system uses:

View attachment 72464
Technically speaking, the standard could have been anything from 5 to 8 bits at the DSD64 sampling rate of 2.8 MHz, 5 bit being the minimum (more bits, more headroom for noise shaping). In case anyone is wondering, yes, Sony DID lie to us…

As a matter of fact, I don't know of any single-bit modulator other than the Grimm AD1 - at least that's what Grimm claim it is (in turn, I know of only one recording label that uses a Grimm AD-1, at least occasionally).

Greetings from Switzerland, David.

Why do you denigrate off-the-shelf chips? Do you think the engineers at Texas Instruments (BB) or Analog Devices are not at least as skilled in what they do as DCS engineers? Shrinking ladder DACs to small chips is really something and interestingly, some SOTA companies still think ladder DAC tech is still the best sounding, regardless of the technical merits of RADAR chips are. It is clearly not all about noise floor and low level resolution. That is not to say they are not important, they are, but there are clearly other factors that somehow come through that leads others to go another road sonically.

 

[microstrip]

Measure better? Sure that can be tested. Sound better? You have quite an imagination indeed if those technical details translate directly into better sound. How do you justify this?

We are discussing a source element - a DAC that IMHO must recreate a signal as close to the originally encoded as possible. I think that good measurements are essential in very top units, although they do not assure us per se of good sound. Again IMHO excellent linearity is essential to avoid digital artifacts - it is not imagination, it is mathematics and signal processing.

What those self-serving documents from DCS don't discuss are the potential downsides from how the technical perfection is achieved. Now, maybe they are right and if so, then why can so many people still be critical of the sound? I think there are many here that might consider MSB (who make their own DACs) or CH Precision (which despite being a VERY technically competent company STILL use an off-the-shelf BB DAC chip (ok 4 per channel but still off-the-shelf) or even Lampizator, TotalDAC(who make their own as well) or Aries Cerat to be at least on the same level sonically. Why hasn't that technical superiority translated into total dominance in digital where all consider it to be the end all and be all (clearly DCS thinks their basic tech is as it has stayed constant now for a long time).

Would you claim that the DCS is simply the most true to the signal it is being fed and if we don't like the sound it is because the recording is poor or the system inadequate?

Is there another DAC out there that you consider a truly worthy challenger?

I am not interested in DAC preferences and market shares, as I have now told several times - an interesting subject for another thread. Being a curious audiophile I would not mind listening to the MSB Sellect II or the Atlantis Reference in my system.

 

[Al M.]

Again IMHO excellent linearity is essential to avoid digital artifacts - it is not imagination, it is mathematics and signal processing.

Excellent linearity precisely is a main reason why Mike Moffat chose the Analog Devices AD5791 precision DAC chip for the Yggdrasil DAC and was determined to make it work. This despite the zero-crossing distortion, which precluded it from recommendation by the DAC chip designer for audio applications and which he was able to suppress.

It offers excellent INL (Integral nonlinearity) and DNL (Differential nonlinearity) specifications that apparently significantly exceed that of typical audio ladder DAC chips.

 

[ack]

Measure better? Sure that can be tested. Sound better? You have quite an imagination indeed if those technical details translate directly into better sound. How do you justify this?

What those self-serving documents from DCS don't discuss are the potential downsides from how the technical perfection is achieved.

FWIW, I've never been fully convinced of this alleged ring-DAC superiority. Its main feature is that the ladder resistors are of the SAME value, or as close as possible, but in the end, you end up with more such resistors and latches. The linked article claims 96 resistors, but over here https://www.hifiplus.com/articles/m...what-changes-were-made-to-the-dcs-vivaldi-20/ their Director says there are 48. Nonetheless, that's more complexity than the typical 24 latches-and-resistors you will find elsewhere, albeit with different-value resistors. And on top of that, you have this randomization of the signal path in the so-called ring-like fashion, which creates a tad of "random white noise" as the linked article says.

Different? Yes. Superior? Not so sure. It has always felt to me that their main goal is to randomize noise; not sure, but if it were to be true, not sure why this actually helps more than it hurts.

 

[acousticsguru]

Why do you denigrate off-the-shelf chips? Do you think the engineers at Texas Instruments (BB) or Analog Devices are not at least as skilled in what they do as DCS engineers? Shrinking ladder DACs to small chips is really something and interestingly, some SOTA companies still think ladder DAC tech is still the best sounding, regardless of the technical merits of RADAR chips are. It is clearly not all about noise floor and low level resolution. That is not to say they are not important, they are, but there are clearly other factors that somehow come through that leads others to go another road sonically.

The reason is the precision required of the resistors in an R2R ladder, not merely of the individual resistor, but of each one of the resistors in relation, a fundamental problem that has little to do with engineering chops (other than e.g. improving the laser etching in the chip design etc.). Just do the math, it's really quite simple, or read Page 10 of the review Al uploaded. I know you love the sound of R2R, as do I, I'm merely aware of the technical limitations (among other, in resolution and noise floor, non-linearity etc.). The cost for a discrete R2R ladder goes up exponentially with the quality and precision of the resistors, whereas there are technical limits to what can be done in a chip design (this is, after all, a thread about SOTA DACs).

In short, I'm merely stating a fact, not "denigrating" anything or anyone, least of all chip designers. One of my best friends from Highschool is an electronics engineer (and audiophile) who spent half of his life in Silicon Valley doing just that, chip design. It's an art in itself, no doubt about it.

As to the "RADAR chip" you mention, not sure what you're referring to? There used to be a commercial RingDAC chip set, not sure if it's still in production (Meridian used it for while in some of their products), from what I've heard it's hardly SOTA.

Greetings from Switzerland, David.

 

[microstrip]

FWIW, I've never been fully convinced of this alleged ring-DAC superiority. Its main feature is that the ladder resistors are of the SAME value, or as close as possible, but in the end, you end up with more such resistors and latches. The linked article claims 96 resistors, but over here https://www.hifiplus.com/articles/m...what-changes-were-made-to-the-dcs-vivaldi-20/ their Director says there are 48. Nonetheless, that's more complexity than the typical 24 latches-and-resistors you will find elsewhere, albeit with different-value resistors. And on top of that, you have this randomization of the signal path in the so-called ring-like fashion, which creates a tad of "random white noise" as the linked article says.

Different? Yes. Superior? Not so sure. It has always felt to me that their main goal is to randomize noise; not sure, but if it were to be true, not sure why this actually helps more than it hurts.

As I said before, the good quality of the DAC shows in the measurements and concepts involved in how it is implemented. Do not expect for find the reasons in magazines and reviews. The reasons for the difference in the number of resistors are either different models starting with the Elgar or referring to the each DAC or the whole left/right DAC unit.

The concept of randomizing the use of the resistors is a key point to the excellent performance and ultra high linearity of the ring DAC - it is used in instrumentation since long to get top linearity. The objective is not creating white noise, but transforming the always existing low non linearity in extremely low harmless noise.

The subjective aspect of sound quality it a matter of individual preference - knowledge of the scientific and technical aspects can surely create a positive bias!

 

[microstrip]

Excellent linearity precisely is a main reason why Mike Moffat chose the Analog Devices AD5791 precision DAC chip for the Yggdrasil DAC and was determined to make it work. This despite the zero-crossing distortion, which precluded it from recommendation by the DAC chip designer for audio applications and which he was able to suppress.

It offers excellent INL (Integral nonlinearity) and DNL (Differential nonlinearity) specifications that apparently significantly exceed that of typical audio ladder DAC chips.

The Analog Devices AD5791 is a 20 bit DAC. One LSB error in this DAC is an error 16 times higher than a LSB in a 24 bit DAC. The excellent INL and DNL numbers as quoted in the datasheet per se are meaningless for audio and very hard to compare with the usual specifications of audio DACs. Curiously the Stereophile measurements show zero-crossing distortion.

Would we buy a streamer that would chop the four LSB's of all our hi-rez music files? I would not, but it is me.

 

[Al M.]

The excellent INL and DNL numbers as quoted in the datasheet per se are meaningless for audio and very hard to compare with the usual specifications of audio DACs.

Blanket assertion. Mike Moffat appears to think otherwise.

Curiously the Stereophile measurements show zero-crossing distortion.

Obviously you have not paid attention to my posts -- why would you, since you dismiss the technical performance of the Yggy anyway, whereas I appreciate the one of dCS DACs (how all the technical performance correlates with sonics is a different matter). The zero-crossing distortion of the first generation Yggy is already negligible (the graph in Stereophile is at -90 dB), and in version 2 it appears to be basically eliminated:

Schiit Yggdrasil V2 upgrade Technical Measurements

 

[acousticsguru]

FWIW, I've never been fully convinced of this alleged ring-DAC superiority. Its main feature is that the ladder resistors are of the SAME value, or as close as possible, but in the end, you end up with more such resistors and latches. The linked article claims 96 resistors, but over here https://www.hifiplus.com/articles/m...what-changes-were-made-to-the-dcs-vivaldi-20/ their Director says there are 48. Nonetheless, that's more complexity than the typical 24 latches-and-resistors you will find elsewhere, albeit with different-value resistors. And on top of that, you have this randomization of the signal path in the so-called ring-like fashion, which creates a tad of "random white noise" as the linked article says.

Different? Yes. Superior? Not so sure. It has always felt to me that their main goal is to randomize noise; not sure, but if it were to be true, not sure why this actually helps more than it hurts.

I'm counting 48 latches and 96 resistors per channel:



The converter board is a bit easier to see in this pic of the Vivaldi One:



Greetings from Switzerland, David.

 

[Al M.]

The Analog Devices AD5791 is a 20 bit DAC. One LSB error in this DAC is an error 16 times higher than a LSB in a 24 bit DAC. [...]

Would we buy a streamer that would chop the four LSB's of all our hi-rez music files? I would not, but it is me.

The relevance of 24 vs. 20 bit resolution at the DAC chip level vs. DAC analog output level has already been discussed.

 

[microstrip]

The relevance of 24 vs. 20 bit resolution at the DAC chip level vs. DAC analog output level has already been discussed.

And we all know the opinion of 20 bit DAC owners!