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.
372 373 374 375 376 377 378 379 380 381
383 384 385 386 387 388 389 390 391 392
  1. azteca x
    Right. I hope to see standard transient measurements sometime soon. The discussion around the Yggdrasil and it's filter reminds me of similar discussions over tons of feedback to get super-low distortion vs. minimal or no feedback and less perfect measurements. For the uninitiated: https://passlabs.com/articles/audio-distortion-and-feedback
    Time will tell as more people get their hands on the Yggdrasil, the eventual trickle down into the Gungnir and Bifrost, blah blah...
  2. frenchbat

    Sorry about that. Hope it didn't take too long to clean :wink:
  3. goobicii
    that is exactly whats wrong with oversampling and sigma delta chips,that these dacs spit out signal vastly different to what really is on pcm
  4. prot

    Could be. Still, food bloggers have all the rights to ask Qs and criticize as they please and havent seen many ppl in lynch mood yelling at them.

    Anyway, my bad, apparently I missed a lot of that discussion. Btw thx for the update AustinValentine.
  5. Defiant00
    Audacity isn't showing you "what's really on pcm" it's showing you a simplified and incorrect view for the sake of performance.
    Please watch the video linked first before responding to my post. What you're viewing in Audacity is not the actual waveform, it's straight lines connecting discrete values. It is likely done this way for performance reasons (lines are much quicker to draw and are likely close enough for a high-level view), but just because the Audacity team chose a convenient/fast way to display the data doesn't show anything about how it's actually converted to sound (which is not a linear interpolation between the steps as displayed).
    Again, this is all addressed in the video I linked above, including an entire portion on why stairsteps and other linear views are incorrect, including examining the output of a DAC directly using an entirely analog scope. Perhaps you should watch it?
    Sapientiam likes this.
  6. goobicii
    only difference is that PCM is series of straistepsand not straight lines,that doesnt disprove anything I posted,the aplitude jumping is there
  7. Defiant00
    That's exactly one of the main misunderstandings this video is about. PCM is not and does not produce stairsteps, and anyone representing it as such is wrong.
    Since you apparently missed it the first time, here, again, is a good rundown on why this is wrong: http://xiph.org/video/vid2.shtml
    Skip to 3:36 if you're in a hurry.
  8. goobicii
    I watched it before
    PCM is stairsteps,only reason it isnt becose the equipment that runs it is not fast and pure enough,if you read my original post you will notice I explained this,I adressed this
    its actualy good thing for low samplerates that our dacs are so slow and imperfect that they to give curves to the staircase,it makes the ****ty low samplerate not so ****ty,far from truth but atleast it have organic curves
  9. azteca x
    ....No. Watch it three more times and really listen.
  10. Defiant00
    What? No, producing sine waves that match the samples (or at least get very, very close to them in the case of delta sigma) is exactly how they're supposed to work.
    That's kind of the fundamental point of the sampling theorem ( http://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem ), that given a specific band-limited set of data (aka, nothing over 20khz), you can reproduce the original waveform given a discrete set of samples. So no, no stairsteps; the samples are at discrete points in time, but that doesn't mean that's the value in between points.
    And that's it for me; if you bother to watch the previous video or actually read about sampling theorem then great, perhaps we can have an actual conversation about it.
  11. KeithEmo
    That's sort of a trick question.
    My point is that most "good DACs" these days have specs that are good enough that - based on the numbers - I wouldn't expect the differences between them to be audible. There are so many DAcs today that have "exceptionally good specs" that I find it difficult to see how any could "stand out" in a meaningful way based on specs. To pick a few at random, check out the specs on the latest Benchmark model, and on the Emotiva DC-1, and the Wyred4Sound DAC2.... All of the "common specs" like THD, S/N, IMD, and frequency response are so good on all of them that I wouldn't expect the differences to be audible - and I wouldn't expect it to be possible for any of them to be improved on in a meaningful way. Therefore, even if the Yggdrasil's specs on any of those measurements is better, I wouldn't expect that difference to be audible. I don't look at its specs and think: "Aha, xyz is a lot better than everyone else; I'll bet that's why it sounds so good."
    To put that another way, IF YOU'RE BASING YOUR EXPECTATIONS ON THE COMMONLY QUOTED SPECIFICATIONS, then the Yggdrasil and a lot of other current DACs out there are all so close to perfect that I can't see how ANY of them can be claimed to be significantly better than the others. Therefore, the differences most likely lie somewhere else. Personally, I suspect that the digital filter is probably the single biggest factor there, and perhaps the one in Yggdrasil really does sound better than most of the others. Unfortunately, besides saying whether the ringing is symmetrical, and how much ringing there is overall, we don't really have measurements that describe those differences in any way that I know how to relate meaningfully to how the device actually sounds. (You can see that the ringing on different DACs, all of which may have symmetrical ringing, looks slightly different in other ways - but I've never succeeded in figuring out which visible differences there correlate with which audible differences. I'm also personally convinced that jitter is more audible than many people seem to think, so perhaps the Yggdrasil's excellent jitter immunity is audible - although, again, many other DACs out there also have excellent jitter immunity.)

    arnaud, gevorg, wahsmoh and 1 other person like this.
  12. goobicii
    PCM will be stairsteps even if it was  99999999999999999999999999999999999999999999999999999999999999999 billion gigahertz
    yes yes stairsteps,thats all pcm is,becose dac  cant start and stop instantly we get curves
  13. KeithEmo
    If you were to look at the output of a DAC WITHOUT FILTERING IT, then you would indeed see stair steps. However, the reason is that, if you do that, YOU'RE DOING IT WRONG. The reconstruction filter, which is a necessary part of "doing the process correctly", limits the bandwidth - which means that it removes the steps and leaves you with only a reasonable facsimile of the original signal. The reconstruction filter isn't "slow and imperfect" at all. In fact, it's doing just what it's supposed to do - filter out the extra out-of-band energy (which is what those steps are made up of). 
    The extra junk is supposed to be there, and it's also supposed to be removed, yet you insist on:
    1) complaining that removing it is a flaw
    2) then complaining that it's still there
    wahsmoh likes this.
  14. goobicii
    yes it would look more like stairsteps but it will still be not perfect since even fastest dac with no reconstruction filter cant start and stop instantly,that being said I umderstand that you misunderstood what I meant whitch is my fault and I pretty much agree with your post
    you also support my argument against Defiant00,just like you said,PCM is strairsteps,its only becose of reconstruction filter that the output isnt full of square edges
  15. Defiant00
    He said the opposite, but I guess if you can't be bothered to read or watch any of what others have posted then I wouldn't expect you to pay attention to that either.
    PCM itself isn't stairsteps, what he said is that the output of a DAC before filtering would look like a stairstep.
    PCM is just a set of discrete samples. Numbers, one after another, at a known rate.
    For example, is this a stairstep?
    No, it's a set of numbers that could be interpreted in any number of ways. If you interpret it as PCM data and feed it to any rationally-designed DAC you'll get an analog output very close to the set of sine waves that would pass through those values at the specified sampling rate.
    Alternately, if you decided to put those in a spreadsheet, then sure, you could make yourself a bar graph and say it's a stairstep (THIS WOULD NOT BE SOUND DATA THEN OBVIOUSLY, IT WOULD BE A GRAPH, PLEASE FOR THE LOVE OF EVERYTHING HOLY DON'T QUOTE THIS ONE SENTENCE AND CLAIM I SAID IT WAS A STAIRSTEP). But if used as PCM data then no, the 100% mathematically correct output would be a set of sine waves very close to the original input (which is what any reasonably designed DAC will get very, very close to).
    smitty1110 likes this.
372 373 374 375 376 377 378 379 380 381
383 384 385 386 387 388 389 390 391 392

Share This Page