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.
378 379 380 381 382 383 384 385 386 387
389 390 391 392 393 394 395 396 397 398
  1. Currawong Contributor
    I've moved discussion about listening testing to Sound Science here: http://www.head-fi.org/t/771254/testing-claims-about-the-sound-of-different-dacs
  2. Maxx134
    It is indistinguishable from the Matrix-X dac.
    It is Lesabre at it best(or worst lol)
    Suffice to say I had no problem with planars,
    and it has a wide beautiful stage,
    yet if you pair it with an HD800,
    you will realize what this thread is all about.
    You will hear that top end fatigue.

    While I do very much admire your efforts in explanation, which is actually interesting...

    There is no way in hell I am ever going back to a Delta Sigma dac of any kind, after owning a yggy..
    Just no way..
    Reality bites when you can hear the difference.
    Also not saying DS is bad, as I do like other DS dacs, (LIO, Gungnir)
    but we talking another level..
    blasjw likes this.
  3. Poimandres
    Thanks, with planars where would you rate the oppo DAC? Curious if it is at the Gungnir level.
  4. Maxx134
    I preferred my oppo over the Gungnir because of oppo soundstage and details while the Gungnir excelled at dynamics and musicality.
    In other words, more of a preference choice as both similar level (to me).
  5. Poimandres
    Thanks. I have heard that dac is the weakest link in the oppo, so assuming that that is true and the Gungnir is 849 the Oppo is a good deal at its current price. Thanks for the feedback.
  6. judmarc
    Thanks for the kind words.
    OK, you've cut the frequency response sufficiently to get rid of aliasing/images to an extent satisfactory to you, and now we're all done, right?  :)  Well, lots of folks would say no.  So far we've talked about aliasing, a problem with frequency distortions.  But what about timing and phase?
    Timing: In the math that goes into filter design, "frequency domain" stuff (like the aliasing distortion we were talking about) and "time domain" stuff are what are called "conjugate variables."  As one is optimized the other gets worse.  It's not bad design, it's just sheer mathematics.  (This type of math applies to lots of different stuff besides filters.  The quantum uncertainty principle arises from the fact that a particle's position and its momentum are conjugate variables.  Nothing anyone can do about it, it's just how the math works.)
    What is "time domain stuff"?  Primarily it's what's called "ringing."  The sharper the cut in your filter, the more it rings.  If you want to see a graphical demonstration, go over to http://src.infinitewave.ca and check out the relationship between the Transition (frequency response) and Impulse (ringing) tests for various filters.  There is lively discussion over the audibility of ringing.  People who look at things pretty exclusively from a frequency response standpoint say the ringing is at ultrasonic frequencies, which is true (though I don't know to what extent this can cause intermodulation - "beating" - with other frequencies to create audible products).  People who focus more on the time domain say ringing causes "smearing" of transient response - drum and piano attacks, string plucks, etc.
    Many folks are especially concerned about "pre-ringing."  In what's called a "linear phase" filter, half the ringing energy occurs before the impulse.  Nothing in nature works this way, so that's why folks are more concerned about pre-ringing - they think a steep linear phase filter will sound particularly unnatural.  A way this has often been resolved is to use minimum or intermediate phase filters.  This will not stop ringing.  What it will do is push the ringing energy more (intermediate phase) or completely (minimum phase) to after the impulse, so any time smear is heard as a (supposedly) more natural reverb or tail to the transient.  But this comes with its own problems.  Intermediate or minimum phase filters are "dispersive" (another way of saying it is they have "group delay"), meaning the time it takes to get through the filter is frequency-dependent.  This can give a recording a feeling of depth.  But Keith Johnson (designer of that DAC purrin likes) has said in an interview he feels group delay is the worst problem with RedBook and the typical filtering used with it.  And I've written at CA about why I think I'm sensitive to it in my system with my particular speakers, which are designed to be "time-aligned" (i.e., speakers are designed so all frequencies arrive at the listening position simultaneously).
    Now what does this all have to do with 44.1 resolution and 8x oversampling?  'Cause you can get the response cut you need to avoid audible aliasing with a relatively gently sloped (less ringing) filter if you start the cut at 352.8 or 384KHz rather than 44.1 or 48KHz.  So that's why 8x oversampling became a standard so quickly: it allows use of filters that don't ring as much for a given amount of anti-aliasing effect.
    Edit: Another way to say that last bit is that 8x oversampling allows filters that have both acceptably low aliasing and acceptably low ringing to be designed more simply and cheaply.
    yfei, arnaud, murrays and 1 other person like this.
  7. gournard
    Hi guys and gals,  a friend owns an Oppo 105 which when the music gets busy sounds congested and noisy and has me diving for the remote to turn it down. On simpler fare like a piano trio etc. it sounds good, but big band stuff makes me cringe. I am after a dac that I can relax with and hopefully be drawn into the music, not necessarily looking for the maximum detail.
    I have found a Wadia 15 for sale, which uses 4 X PCM63P-K. Has anyone heard or compared this.
    I also have access to a Bryston BD-1/BDA-1 combo which I heard and thought it sounded great. It uses a CS4398 which is delta-sigma but the sound I heard was way beyond what the Oppo was doing and I could happily live with it.
    Although impressed with the Bryston the Wadia is much cheaper and would be my choice if it performs well, but I have never heard any Wadia product.  Perhaps some of you may be able help me choose.
  8. prot

    You keep talking about that "superior R2R approach" and you do not give a single freaking *argument* for it ... neither technical, nor sound tests/comparisons/etc ... nothing, nada, zilch.
    Sorry man, but for all I know you could be just as well talking unicorns. Could you please add *something* to support those R2R claims !? I'm not picky, I'll take *anything* ...
  9. KeithEmo
    As far as I can see, the thread is actually about differences, audible or not. (I've seen quite a few claims that Delta-Sigma is "technically inferior" based on how it performs the conversion rather than on claiming that it actually sounds bad - or different - although I've seen plenty of those as well.) I've owned quite a few "reasonable middle of the market" DACs, and I've also listened to quite a few belonging to other people, including some very expensive ones, and some very cheap ones, and some very good-sounding ones, and some really poor sounding ones. However, I can't say that I've personally noticed a "type difference" between the few that weren't Delta-Sigma and the rest that were. (Many of them sound different from each other, but I can't say that I heard something that all the NON-Delta-Sigma DACs had in common that was different than all the Delta-Sigma ones.) For one thing, most of the non Delta-Sigma DACs I've heard are also non-oversampling (but there ARE R2R DACs with oversampling - like Yggdrasil). Unfortunately, unlike oversampling, there's no way you can really compare "a Delta-Sigma and non Delta-Sigma version of the same DAC" - because a given DAC uses a chip that is either one or the other, and you can never simply "drop an equivalent R2R chip into an existing design and see if it sounds different".
    My comment on the noise floor was very general. On most of the reasonably good sounding DACs I've heard, I found the noise floor to be totally inaudible - and, as someone once said, you can't hear differences between different things if they're all inaudible. I simply don't buy into the idea that "something wrong with the noise floor could have some sort of intangible effect on the way something sounds". Sure, noise that's inaudible could cause excessive modulation of audible frequencies, or perhaps might cause additional intermodulation distortion, or some other type of anomaly - but it would be measurable. (I can imagine a situation where, for example, a major noise spike at 21 kHz might be inaudible yet might give you a headache, but I haven't actually run into that situation personally, and such an anomaly would certainly show up on the overall S/N measurement. The noise spectra I've seen on most "reasonable" DACs lately not only have a very low noise floor, but even the occasional spikes that are there tend to also remain below relatively low levels.)
    Yes, I was basically talking about "current production chips". The TDA1387 is a 16 bit DAC chip, which would disqualify it for most "current designs" - since these days a DAC that can't handle 24 bit signals would be a niche market item at best. The TDA1387 also doesn't appear to support oversampling internally, and this leads me two conclusions- depending on how you implement it. If you don't use oversampling, then it's going to be very difficult to design a reconstruction filter that is sharp enough to remove what it has to, yet avoids significantly altering the audio frequency response. (In that situation, I would expect the reconstruction filter to have such a profound effect on the overall sound quality that it would completely overshadow and differences in the DAC topology itself). If you DO use oversampling, it would have to be external to the DAC, and you would have to design and program a custom oversampling filter, in which case I would expect the sound of THAT to overshadow the inherent sound of the DAC topology.
    (Since the Delta-Sigma topology itself is basically an oversampling topology, you can only really compare Delta-Sigma DACs directly to other oversampling DACs - like Yggdrasil. Otherwise you're really just comparing NOS DACs to oversampling DACs.)   
  10. BassDigger
    Am I right(?) in thinking that the key difference, between Delta-Sigma and R-2R, is that r2r processes the bits (up to 20 of them) in unison; all at the same time. The entire process is real-time, from start to finish.
    Whereas DS (which started life as a way to add extra bits to a 14 bit r2r dac) processes them individually or separately, at high speed, and then has to put the bits back together, to reconstruct the real-time signal.
    This is why the timing (the clock and removing jitter) is so important for DS. And maybe it's why r2r sounds better; it isn't prone to the timing errors, in the same way that DS is.
    After all, timing is at the heart of all music; the beat, the rhythm, the notes themselves; they're all frequencies; timing is everything. [​IMG] 
  11. KeithEmo
    It is correct that an R2R DAC basically takes the full value of the sample as "input" and puts out a single voltage value "all at once" while a Delta-Sigma DAC essentially "slices it up in time, processes each piece in sequence, then sums the results". However, the conclusion that "that's why it sounds better" simply isn't logically valid. While the way a Delta-Sigma DAC works is certainly more complex, and intuitively seems "messier and less precise", the fact is that all that really counts is the result - and both deliver very accurate output signals. (The fact that the process used by an R2R DAC is simpler and easier to understand in no way suggests that it produces a "better" output.)
    Neither Delta-Sigma DACs nor R2R DACs are "prone to timing errors". What's happening is that, because of the high level of oversampling used in Delta-Sigma DACs, they are more sensitive to timing errors that are present in the signal you send to them. This same factor is present to a degree in any oversampling DAC - because, the higher the clock rate, the more of a percentage error a fixed amount of jitter is in relation to it. This affects Delta-Sigma DACs more than other DACs because they oversample at a higher rate. If you send a bad signal to both an R2R DAC and a Delta-Sigma DAC, odds are that the Delta-Sigma DAC will produce more distortion as a result. Note that this situation doesn't exist if you send a GOOD quality signal to both. It just means that you have to be more careful what you send to a Delta-Sigma DAC if you want good results.
    To put a bit of perspective on this.... Assuming a perfect input signal, with absolutely no jitter, and all else equal, a $5 Delta-Sigma chip will deliver performance equivalent to or better than that you get from a $50 R2R chip. However, since the Delta-Sigma chip is more sensitive to jitter, you're going to have to spend an extra $10 on the input circuitry to ensure that the Delta-Sigma chip gets a clean enough input signal to avoid having its performance degraded by jitter. However, assuming you deliver a clean signal to both, their outputs will be equivalent.
    (However, this can be a "deal breaker" if you aren't able to design your other circuit elements well enough to deliver the clean signal that the Delta-Sigma DAC requires to perform well. This might suggest that, if you're designing a DIY project, or are a small company without the design know how and expensive test equipment required to design and test for low levels of jitter, the less strict signal requirements of the R2R chip might be a distinct advantage to you.)
    Your final comment about "timing and music" also calls for additional comment.... (you are laboring under a common misconception there).
    When we refer to jitter as a "timing error", we are talking about nanoseconds or picoseconds - that's BILLIONTHS and TRILLIONTHS of a second. To put this in perspective, at the 44.1k sample rate used on a CD, the samples are about 20,000,000 picoseconds apart. There is no way a human (or any other living creature) is going to HEAR an error of even tens of thousands of picoseconds directly. (A "decent" input stage, by today's standards, should limit the jitter to several hundred picoseconds at worst). In order to be audible as a beat "out of place", you would need an error of several milliseconds, or a speed error of several hundredths of a percent.
    Producing a clean and correct output relies on converting samples that have the correct values at the correct times. If you have jitter, then the timing is slightly incorrect, so you're converting the right values at the WRONG times, which produces a result quite similar to what would happen if the timing was perfect but the sample values were wrong - you get distortion. As it turns out, the distortion you get is related to the frequency characteristics of the jitter, and is related to the content itself, but not in a "harmonic manner" (you get distortion that is related to the input signal, but doesn't consist of "simple harmonics" - which means that it doesn't sound exactly like "ordinary THD".)
    When you see those graphs, with a sharp peak surrounded by a bunch of smaller peaks and assorted junk, what you're being shown is the overall spectrum of "what's coming out". The theoretical perfect output would be a single sharp narrow vertical line, and those other peaks are signal that shouldn't be there but is (distortion). Since harmonics tend to be masked by the music signal itself, and a lot of music already contains harmonic content anyway, we can reasonably assume that this unrelated and non-harmonic distortion will quite possibly be more audible and more annoying when it is present. This is why, with a DAC, we would hope to find not only an overall noise floor that is on average inaudible, but we would also hope that no individual "spike" would extend high enough above the average noise floor to itself be audible. So you look for a low noise floor ("the grass") and for there to be no peaks that extend very far above it.
    Ignoring the pictures, most people who claim to notice low but significant amounts of jitter usually describe it as "blurring the sound stage" or "making things sound blurry"....  I would personally describe the effects as "making a well recorded wire brush cymbal sound more like a leaky steam valve" - the frequencies are all present, but you lose the "sense" of individual wires hitting metal and it sounds more like a generic burst of noise at the proper frequencies. I also tend to notice a difference on sibilants - to me they seem more exaggerated but less natural when a high level of jitter is present.
    (Note that I'm talking about "jitter being present at the DAC" - which is all that counts. If the DAC has some sort of jitter reduction mechanism, which many do, then all that matters is how much jitter remains when the signal arrives at the actual DAC chip to be converted. As it turns out, it requires VERY careful circuit design to be able to remove or reduce jitter to a very low level, and to avoid introducing new jitter to the signal on its way to the DAC itself. Simply using a good clock is not enough to ensure low jitter on the audio signal - although using a bad clock can be enough to ensure a bad jitter spec.
    arnaud, gevorg, Mortalcoil and 8 others like this.
  12. Sonic Defender Contributor
    Great information, thank you. I have confirmation that our July meet will indeed have a well broken in Yggy there along with the owners Rag amp so a great stack of Schiit. What I hope to see as I've said is whether or not there is this overwhelming preference for the R2R presentation of the Yggy over the terribly flawed, horrible, error-riddled, only built because it is cheap and could be designed by idiots M51 (sorry, couldn't help the vent). I will not be a subject in the testing of course (but I most certainly will be enjoying listening to the Yggy).
  13. KeithEmo
    The one bit of advice I would give (sort of) is that Yggdrasil is well designed in many ways, including having a specially designed oversampling filter, and a very nice analog section. Therefore, I wouldn't necessarily assume that the way it sounds is entirely, or even mostly, due to the fact that it uses an R2R DAC (although that's almost certainly part of it). I also wouldn't necessarily expect similar performance from other R2R DACs.
    I would also note that many R2R DACs are ALSO non-oversampling (because the same "philosophy of simplicity" often seems to favor both design choices), and NOS DACs, because of the requirements and limitations of their reconstruction filters, tend to sound rather different from other DACs. Yggdrasil, while being R2R, is an oversampling DAC - so I definitely wouldn't expect other NOS R2R DACs to sound like Yggy.
  14. Ableza
    Hello and glad to see you posting here on Head-Fi. 
    I'm a long-time Emotiva fan and have been supporting Dan Laufman since the early days.  But I will trust the design decisions made by Mike Moffat over yours or Lonnie's or mine or, well, over pretty much anyone's when it comes to DACs.  As a long-time Theta-luster I was very excited when I first read about the Schiit Yggdrasil and as a first-build owner I am happy with both the performance and the price point they obtained.  The two Emotiva DACs I have owned both left me wanting more (although I have not heard the Stealth.)  Whether this was about topology or execution or digital interpolation schemes and sampling algorithms I don't know nor really much care.  I just know what stayed in my rack and what was re-sold on Fleabay.
    As in all things audio, to each his own, but when it comes to design I'll stick with the word of the engineer who invented the market sector.
  15. Sonic Defender Contributor

    Fair, but come on, for several years Schiit was quite happy to sell D-S DACs, so it is somewhat disingenuous to suddenly start saying well, that was a crap technology, we have moved on. Nobody is questioning the skills or knowledge of MM, nor the prowess of the Yggy, quite the opposite actually. What some of us are saying is that regardless of the technical level of discussion how things sound being a very subjective field means that there are actually people who like the D-S signature, and again, I have yet to read of any blind listening tests that demonstrate that any R2R DAC is consistently preferred by listeners. I will say again, I have no opinion yet on the matter beyond having an open mind. I think the M51 for instance sounds quite good, I actually like it very much. Now it is very possible that after hearing the Yggy I will be saying something very different, but perhaps not.
378 379 380 381 382 383 384 385 386 387
389 390 391 392 393 394 395 396 397 398

Share This Page