1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.

    Dismiss Notice

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

Discussion in 'Dedicated Source Components' started by purrin, Dec 5, 2013.
418 419 420 421 422 423 424 425 426 427
  1. Sonic Defender Contributor
    I do take your point, and agree, we can't possibly know all of the laws of physics. And I will certainly look into Robert Rosen's work. One thing I guard against is the tendency to only look at information that is consistent with what we believe. I do try, I don't always succeed, to keep a truly open mind. Otherwise I would be an arrogant pr**k assuming I have all the answers and nothing more to learn. Just shoot me if I ever get like that.
    Maxx134 and Sapientiam like this.
  2. Sonic Defender Contributor
    I'll come back and reply properly when I return from a few errands I need to do.
  3. KeithEmo
    I absolutely agree with you - at one level.
    However - at another level - I disagree.

    My background is engineering (let's say "engineering first; audiophilia second").

    Here's the way I see it. You are correct. A lot of gear, especially what many of us would term "boutique gear", does in fact experience significant interactions with other equipment, and is sensitive to a little bit of noise on the power lines, and even to the length and sort of interconnects you use, or even to things like vibration. However, from an engineering perspective, those sensitivities are design flaws. A properly designed preamp should be able to deliver the performance it was designed to, when connected to the power grid in a typical home, using ordinary commercial grade interconnects, and should do so when connected to any other piece of equipment that has also been designed equally well.

    I don't doubt that many DACs sound best when connected to power that's been filtered by a $2500 laboratory grade power conditioner. However, the reason is not that "they're that good" or "incredibly revealing"; the reason is that the person who designed them was apparently good at designing DACs, but not so good at designing power supplies. The purpose of a power supply is in fact to take whatever commercial grade power you feed it and convert it into power of sufficient quality to get optimum performance from the rest of the system. If that particular circuit requires ultra-clean power, then that should have been considered when the device was designed, and the power supply it comes with should be up to the task. You should not be expected to purchase extra equipment to make up for the shortcomings of his design. Likewise, if a particular product is sensitive to vibration, then that should be taken into account, either by altering the design to eliminate the problem, or including some sort of sufficiently good vibration dampers in the product. And, if it's super-sensitive to electromagnetic interference, then the cabinet should be made out of solid copper, and it should ship with a ten foot copper bar, a sledgehammer to drive it into the ground, and ten feet of triple-zero battery cable to connect it with. Or, at the very least, they should list top-grade EMF isolation as an installation requirement. (I would suggest that, if a certain piece of consumer equipment won't work as designed when plugged into "an ordinary outlet in an ordinary home", then it's specific requirements for proper performance should be included on the specifications page. If that DAC only sounds really good when connected to a high-quality power conditioner, then that should be listed under "requirements".)

    I absolutely agree that the room itself is an important part of the system, and room treatment is an often-overlooked and critical piece of the puzzle, but it's also true that room treatment is "generic". Excluding certain interactions with speakers, a properly treated room should sound good with any DAC, or any amplifier or preamp, because room treatments are targeted at removing unusual sound characteristics. So, properly treated rooms should sound quite similar, and individual pieces of electronic equipment should not have specific requirements in that regard. As far as loudspeakers, I would agree that 'the speakers and the room make up the parts of a system", and always need to be considered together.

    Therefore, I would leave it this way.....

    If you are willing to accommodate a specific piece of equipment by making special provisions, or adding specific other equipment to go with it, then you should look for reviews by reviewers who are willing to expend a similar amount of effort to optimize the installation of that equipment when they test it. And, if you are unwilling to do so, then you should probably seek reviews by reviewers with similar priorities. And, yes, it is well worth noting that, for at least some equipment, the conditions under which it is reviewed are going to make a big difference in the results. This is why, at the very least, EVERY review should include a detailed description of the room and associated equipment used for the review.

    Maxx134 and richard51 like this.
  4. KeithEmo
    Let's just say it this way.....

    There is nothing that can be heard that cannot be measured; if it's real then it can be measured.
    However, that's not to say that we always measure everything that's important, or that we necessarily know how to interpret all the measurements we make.

    I agree with you about most Sabre DACs... to me they all, to some degree, tend to sound "bright and over-detailed - and as if they emphasize the upper midrange".
    (And, no, that perceived difference in sound is not visible in their frequency response measurements.)

    However, I would disagree that they deliver "immaculate measurements".
    Most modern DACs, including Sabre DACs, deliver exceptionally flat frequency response, exceptionally low noise, and very low THD and IM distortion.
    So, in terms of those three or four particular measurements, their response is indeed "exemplary".
    However, they vary widely in terms of other performance metrics, including things like the impulse response of their oversampling and reconstruction filters.

    Therefore, the reality is that Sabre DACs measure very well, and very much like other high quality DACs, on certain commonly used measurements.
    They also measure very differently on other measurements.
    But, since those other measurements are more difficult to perform, and their correlation with how a DAC sounds is less well understood, those differences tend to be overlooked.

    I have absolutely no doubt whatsoever that, if you sum and subtract the outputs of two DACs that sound different, you will find that they ARE in fact different.
    And, that being the case, obviously it is possible to measure those differences (probably in a variety of different ways).
    However, that doesn't mean that I can tell you which specific measurements will show those differences the best, or enable us to interpret them in a meaningful way.

    GrussGott likes this.
  5. KeithEmo
    All Delta-Sigma DACs operate by a rather complex mathematical process that involves altering the sample rate and bit depth.
    Part of the process involves the mathematical equivalent of converting the signal to one with fewer bits of resolution at a higher sample rate (which is oversampling).
    Therefore, while a particular D-S DAC may choose to bypass oversampling outside of the D-S process itself.... it's really kind of silly to describe a D-S DAC as "a non-oversmapling DAC".

  6. artur9
    We need someone to take your Emotiva DAC as the benchmark and do the above with another DAC. Publish the recording, I guess of the output. Maybe use the Jil as the benchmark ADC.

    Then we can create a library of all the differences from a wide variety of DACs and maybe have a prayer of finding the measurement that actually accounts for all the excitement in the DAC world.
  7. KeithEmo
    I already answered this in more detail - but I would like tom provide a more direct and simple answer.

    I agree with you that "the WORK of an audio DAC is to deliver a signal which listeners are able to interpret as music".
    However, at its most basic, before being converted into a pressure wave by the transducer, that signal is simply an electrical signal that varies with time.
    Therefore, if the signal itself is "perfect", and the DAC performs its job "perfectly", then the result MUST BE perfect.
    So, if you start with a specific digital audio signal (a list of numbers), and do a "perfect" job of converting it into an analog signal, then there is only one possible output.
    And, from that, if two DACs both do a perfect job, and you feed them the same input signal, then they will both deliver the exact same output signal, and it will sound exactly the same.
    (And, if two DACs are even slightly different, then one or both of them must be delivering an imperfect output signal - because its performance is less than perfect.)

    In reality, nothing is perfect, and we are always choosing between different flaws, based on which we find most problematic.
    The simple reality is that no DAC is perfect... so "immaculate measurements" is a relative term.
    (And, if two DACs "measure perfect enough", but sound different, the you simply aren't measuring the right things carefully enough to detect and quantify the differences that must be there.)

    Humans, and especially audiophiles, have a habit of picking out a few measurements that are easy to make, and that they're familiar with... and then proudly declaring that "nothing else matters".
    Sadly, this oversimplification of reality often leads to errors.

  8. KeithEmo
    One of the major ways in which DACs vary is in their impulse response.

    Note that an actual single impulse is an invalid digital audio signal, which can never occur in a valid band-limited digital audio signal.
    However, it works very well as a test signal to visualize the response of the filters used in various DACs.
    And, if you look at the impulse responses of various DACs, and even various filter choices in the same DAC, you will see that they are often quite different.

    For those less technically inclined....
    These are differences that occur and are easily seen on dynamic signals like pulses - and music.
    They are basically a way of characterizing how the DAC responds to rapidly changing signals rather than to continuous sine waves.
    Many devices, but DACs in particular, respond very differently to rapidly changing signals than the do to continuous sine waves.
    However, they do NOT affect static measurements - like THD, frequency response, and S/N ratio.

    Unfortunately, while the differences are easy to see with the proper tests, how they relate to sound quality is much less well understood.
    (And different people seem to vary widely in terms of how they respond to certain differences.)

    Maxx134 likes this.
  9. KeithEmo
    That's an excellent idea.

    The only catch is that you'll have to figure out a way to do it without the sound being affected by the ADC you use to record it or the DAC you use to listen to it. :ksc75smile:

    However, in terms of gathering data, it should be possible to have a bunch of different people listen to a bunch of different DACs...
    Collect all the measurements...
    And all their listening impressions...
    And search for correlations between the two.

    However, doing so is a lot of work, and doing it well enough to actually produce usable information even more-so.

  10. artur9
    Yeah, :) throughout. I wasn't thinking people would listen to the "difference recording", just run FFTs on it or something.

    As to the "affected by the ADC" that's why I specified the Jil as the benchmark ADC. Known quantity is the touchstone here, no?

    Full process something like:
    (1) New DAC on the market.
    (2) Create "difference recording" with benchmark DAC and benchmark ADC.
    (3) People listen to new DAC.
    (4) We find some capable person to correlate those listening impressions to the "difference recording"'s meaurements, somehow.
  11. Sonic Defender Contributor
    That is the issue, we don't really understand how the sound is affected, how much it is affected, and who is sensitive enough to these differences. I would also suggest that the number of people actually sensitive to these effects is not as high as it may seem. Expectation bias is scientific fact, it can't be denied, denying the impact of expectation bias is akin to denying that ice can melt in the heat. Once people started theorizing and reporting about these DS sound problems, it spread like wildfire all over the Internet and that is all it took to create a contagion effect. Many people would start hearing what they were told to expect to hear and then it builds from there.

    I have no doubt that there would be some people more sensitive to such sound anomalies, that is very reasonable to predict. And equally likely some early DS implementations were probably poorly done leading to the notion that DS conversion processing was inherently flawed. This is all likely to result in at least some over-reporting of how audible these sound anomalies are and how-widespread they are. My personal belief is that in blind listening tests, the vast majority of people evaluating well implemented DACs would have a hard time differentiating one from the other. That assumes a general ethos of sound reproduction whereby the circuit designers elected to not deliberately colour the sound.

    I'm not representing my ideas as fact, I'm not that arrogant, and I would be interested in your feedback as I always learn from your unsensational, to the point posts.
  12. KeithEmo
    I agree that the differences are often very small, and, to be quite honest, not all that significant.

    However, because the differences are so small, while I am normally a major supporter of double-blind testing, in this case I suspect that it is not going to serve the purpose. From my experience, many of the differences I hear between various DACs are barely audible when switching back and forth, with no delay at all, and become very difficult to notice with even a few seconds delay between samples. And, in many cases, while I can hear a difference between them, I wouldn't necessarily claim to be able to identify one or the other specifically... simply that there is a difference when switching between them.

    As I mentioned, when it comes to major differences, I am a big supporter of double-blind tests. However, when it comes to tiny differences, they often fail to do the job - especially if they aren't carefully designed to detect a particular difference. I'm tempted to make the comparison to our ability to detect differences in color. Let's test how different the colors of foot-square tiles have to be before we can tell that they are in fact different. Hold two red tiles up, one after the other, and you will find that our ability to detect small differences in tint is not at all precise. However, hold them up at the same time, a foot apart, and we can detect much smaller differences. And you will find that, as you move them closer together, we are able to detect smaller and smaller differences. Move them so they touch edge to edge, and most humans can distinguish amazingly tiny differences in tint or hue. (And, when you match swatches, you typically hold one up over the other for the best possible ability to detect the smallest differences.)

    I suspect that many of the audible differences people claim to hear are in fact simply expectation bias, while others are tiny differences caused by other things, and totally unrelated to the architecture of the DAC. And, from what I read in many forums, the majority of audiophiles aren't quite clear on the details of the different technologies anyway. For example, many people seem unclear on the fact that, while the Schiit Yggdrasil is an R2R DAC, it is also an OVERSAMPLING DAC... (The original R2R DACs were developed before oversampling, and so don't employ it, but many modern ones do - because oversampling has major benefits - while avoiding oversampling causes major issues that need to be overcome or simply lived with. So, how much of the difference between a modern AD1955 and a vintage Phillips 1543 chip is because one is only 16 bits, how much is because one is R2R, and how much is because one uses oversampling? Your guess is as good as mine.)

    There are also major differences in how different technologies react in different situations. For example, most D-S DACs are more significantly affected by jitter than most R2R DACs. That means that, all else being equal, an R2R DAC will in fact probably perform better if your source has a very high level of jitter. However, because of other limitations, all else is rarely anywhere near equal. And, since there are many modern methods for eliminating virtually all jitter, is this really an issue at all?

    As you noted, different people are simply more or less sensitive to specific types of errors.... Some people are actually better able to hear certain types of errors. Others, while able to hear them, simply find some more annoying than others. (I personally hate vinyl, because the presence of two or three loud pops entirely ruins any enjoyment I get from listening to a song. Others claim not to notice them at all.)

    And, yes, many DAC vendors do go out of the way to add various sorts of coloration to the sound. Some do so in the honest belief that the final result is closer to "right". Others do it simply to differentiate their product. It would be difficult to sell a $2500 DAC after admitting that you used a $10 DAC chip, followed the circuit design provided for free in the App Notes, and it worked really great... when someone else is selling the same thing for $250. It's much more feasible to deliberately make it sound a bit different, then convince your customers that your version is "better". (If you read any of the original Sabre DAC white papers, they actually stated that, rather than choose the filter that provided the most accurate output, they held focus groups... and chose the one that "the most people liked the best". I can't fault them for doing it that way, but neither am I convinced that their choice ended up being the most accurate one. I've always personally found Sabre DACs to exaggerate detail to some degree - which I find mildly annoying - but many other people seem to find pleasing.)

    GrussGott, castleofargh and arnaud like this.
  13. intlsubband
    Thank you for the detailed explanation! For the record, this is how the sampling was described in the 6moons review: "the Gold always processes incoming data at its native resolution without internal upsampling. What goes in is what comes out." However, as you note, there is upsampling in the DAC chip itself - in this case, a dual 1792A chips, which according to their spec pages, indeed upsample... so I'm not sure why they are making this claim. http://www.6moons.com/audioreviews/antelope2/1.html
    Last edited: Jul 5, 2018
  14. Sapientiam
    Right (IM here being two-tone stimulus I take it).

    This is an interesting avenue of study. The filters in S-D DACs are in general built down to the lowest silicon area possible, which means adopting a half-band filter for the first stage upsampling (1X->2X). Half band filters are a compromise to save on MACs - half the coefficients are zero. The downside is the FR violates Nyquist because instead of the Nyquist frequency being firmly in the stop band, its in the transition band and only single digit dB down (I forget if that digit is 3 or 6)

    Summing up so far then we look to be in broad agreement that today's measurements aren't telling the whole story about a DAC's performance as we have the 'Sabre brightness' not being shown up in any measurements so far. I would add to the 'brightness' issue the biggest thing lacking for me in an ES9023 and that's subjective dynamics. I also haven't much clue what measurement is needed to quantify subjective dynamics but my first stab at it would be noise modulation.
    Maxx134 likes this.
  15. Sonic Defender Contributor
    @KeithEmo as always, thank you for your very succinct, well written, and well reasoned response. I wonder how much of the notion that DS chips are inherently "unnatural" sounding" or flawed derives from what seems like many people conflating the early egregious errors (depending on one's taste) of the first Sabre DAC designs with all DS architectures? I get the sense that many people with a strong bias against DS implementations erroneously think that they are all essentially Sabre like.
418 419 420 421 422 423 424 425 426 427

Share This Page