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.
389 390 391 392 393 394 395 396 397 398
400 401 402 403 404 405 406 407 408 409
  1. prot

    Your answer makes sense ... but mr Watts seems to disagree completely (at least that's what I understood from the big quote posted here, more taps= much better sound). Don't have the time or enough curiosity to dive into that kind of math & acoustics so color me confused.

    I am pretty sure that it can be done with less that 4xSLI cards. Especially since they dont need a generic CUDA-core but one very specific chip that just implements a single algorithm. They already have good expertise with custom FPGA's and since they sell $5K+ DACs, the budget should be there too.
    Sounds like something for their next DAC ... the Chord Threesome DAC ... one Dave, two Chordettes and some 'special sauce' :D
  2. judmarc
    More taps can be good, because they allow for more sophisticated filters.  But it's not an unrestricted, "more taps equal better filter forever and ever amen" type of thing.  There's the potential disadvantage that more taps will tend to mean a sharper cut and thus more ringing, so that has to be accounted for in good filter design.
  3. prot

    Most probably, all PCs/laptops older than 1-2 years do sound quite bad (and I guess that's what >90% of people use or have heard). Also agreed about the creative cards.

    But that "extremely clear and detailed" chip you are mentioning is pretty much what I expect from a DAC. I want a DAC to deliver 100% neutral sound and maybe also a clear, 3D soundstage (although the soundstage is not exactly/entirely a DAC's responsibility). I want the exact same (neutrality) from my source and amp.
    If I want to add that musical/enjoyable 'thingie' I'll try with speakers/HPs or maybe a tube preamp. Preamps & transducers are the components who add the most 'color' to the sound anyway and have the worse THN/etc measurements in any stereo chain ... all other components could be almost 100% transparent nowadays, no need to have any coloration from them.

    I am wondering if that so called "treble harshness" of the Sigma-Delta chips isn't just simple neutrality. I remember I read somewhere that a perfectly linear-to-20kHz component would sound quite harsh.
    The very neutral/linear O2 fits that theory, it's highs aren't the most 'musical'. Also the most linear speaker I ever heard (and owned) did sometimes sound a bit too strong in the upper treble area ... e.g. the noisy-clocks intro of Pink Floyd's Time was not particularly enjoyable.

    guess it's more clear now ... but in audioland ppl dont even agree 100% with mathematically proven theorems like Shannon's so I'll be still graying around those taps for a while :)
  4. mikoss
    What I heard comparing the resonessence invicta to the yggdrasil was treble (hi hats) sounding like ch-ch-ch-ch on the invicta to being a resolved tst-tss-tss-tst on the yggdrasil. Going back to cheaper DACs, hi-hats are even less resolved sounding. I also notice playing records that the percussion has a different sound than playing FLACs through a cheap DAC. It can sound more resolved and "natural", whereas I suppose the DAC sound to me has something like a compression going on. Just my opinion and after hearing the yggdrasil, I can't I unhear the difference.
  5. wink
    Gotta love those tst tss tst tss's
    mikoss likes this.
  6. purrin
    Invicta was strange. Full bodied tone, but with a strident and sharp treble timbre.

    You read wrong. Lots of non-harsh sounding gear has bandwidth that fully extends past 20kHz. Also, you also can eliminate roll-off in some DACs by using hires or upsampled source material.
    O2 / nwavguy isn't the only person in the world with a 'scope and makes neutral/linear amps. The reason the O2 doesn't sound as good as it could have been is because he went chasing after measurements instead of actually listening.
  7. prot

    exactly .. and who needs a DAC when we have such beautiful tst, tss, tst-s
  8. prot
    Could very well be wrong, I was just thinkin out loud. But there are quite a few empiricals that seem to fit my "linear is quite sharp" hypothesis. Most studio speakers will fit it too and those are known for being very neutral/linear. (and of course nwavdude is not the only builder of neutral components, it was just a sample.)

    But I still wonder, what exactly makes a well built and neutral/linear component sound harsh !? The only theoretical explanation I know is about the odd harmonics sounding harsh (as opposed to the pleasing even ones) ... and apparently a lot of negative-feedback may produce lots of those. That could very well be the O2's case. But modern DACs do not produce much of any harmonics ... at least not in the audible range (I'd put that at -100, -110dB and most DACs are below that already).
    Any other ideas?
  9. ultrabike

    I'm not sure what "pulse array" means, and never heard of "enharmonic distortion" and "noise floor modulation". Maybe he means quantization noise modulation? and I guess enharmonic means some sort of pitch error? Who knows. Also, in almost all cases, as signal gets smaller, non-linear distortion gets smaller. This is the case with most equipment, not just Chord stuff. If what Rob means is noise floor due to perhaps quantization, that is more a function of signal statistics and topology.

    As far as filters, I have no idea what WTA does but it's seems pretty clear it's not a half-band filter. It is interesting that a 1M tap 16-bit coefficient sinc filter is seen in a possitive light, because when sampled at Nyquist, and with the appropriate window, that is a half-band filter. It is also pretty clear the WTA filter is not a linear phase filter, which IMO makes the WTA filter suboptimal in some ways. So it seems we know what the WTA filter is not, but that leaves us with a large set of random posibilities.

    I would not be concerned about the pre ringing of a filter as that only affects transcient response (like a few ms before the song starts playing) and delay. But I would be concerned of a very large filter if the number of bits is not sufficient to carry the arithmetic with out incurring into quantization issues.

    Also, I don't get this "timing error" issue Rob keeps alluding to. If he is talking about jitter then some of Chord's products didn't seem to do so well on the bench:


    If he is talking about square waves perfectly straight and fast transitions, then more than timing errors we maybe talking about ultrasonic bandwidth limitations. Which are ultrasonic.


    So all in all, from what I can tell, all that write up amounts to saying is that the Chord uses a very large, possibly numerically challenged, non-linear phase FIR filter (which may or may not be minimal phase). And the topology solves "enharmonic distortion" and "noise floor modulation", whatever that means (for "small" signals), cuz that's how it sounds to Rob.

    Would love to see a more straight explanation of what WTA does, and less of this "pulse array" "cascade of half band filters" "Kaiser" preceived verbal wanking business.
  10. Wildcatsare1

    Just a thought, Prot how often do to get out to hear un amplified music?
  11. prot

    last time a few weeks ago ... small venue and some acoustic, jazzy music. But not as often as I'd wish ... and nowadays it is quite hard to find any un-amplified performance. Also, the amplified ones are so bad sometimes that it makes me wish I stayed home. E.g. a few months ago a dumb-deaf audio-engineer did set up a Gregorian-Chants concert same as a Metallica one: huge bass and generally very loud instrumentation-track ... it just drowned the beautiful choir-voices which should have been the star of the show. It was much worse than listening to a CD or even the lousiest mp3s.

    Anyway, now that I answered ... what was your point?
  12. Wildcatsare1
    Just that live, non-amplified music can sound "rolled off" compared to the harsh, glaring treble you were speaking about in your posts. Real music usually doesn't have a hot treble. Some recordings of course do have that artifact, but a great DAC, amp, speaker, headphone should convey the tone, dimensionality, naturalness of un-amplified music, no?
  13. purrin
    You need to be more specific on the empiricals. What studio speakers? How where they measured? At the listening position? At the "standard" 1 meter with mic leveled with the tweeter in an anechoic chamber? What is the response off-angle? How are the speakers setup / where they setup properly? How were the monitors intended to be used? How were the monitors loaded, i.e. whole-space, half-space, quarter space, and those with adjustable baffle step correction, where they set to correct boundary loading.

    For DACs, crap in ultrasonic region before the analog filter. For amps, poor slew rate and peaks in ultrasonic region. No rule, just tendencies.
  14. evillamer

    It seems that the chord pulse array is a 2048FS 5th order multi bit pulse density modulation with dither and noise shaping,
  15. prot

    You're asking a bit much, not a pro-studio engineer. Saw some of those though (nothing of say abbey road size & prestige) and my general impression was: very detailed and clear sound but not a relaxing listening space. Some studio speaker samples: k+h o300-400, focal (biggest 3way ones), barefoot, pmc (dont remember models, pretty big & shinny new ones). I did not measure anything but the sound was quite linear, no obviously accentuated freqs and all had some sort of room treats.
    Their components did sound quite linear/neutral but also kinda shrill for my ears. Same for many ultra highend (and suposedly very neutral) hifi setups in various shops.

    As about the slewrate, I think most modern amps (at least the ss kind) are linear to 50khz or more .. even "lousy" receivers are. Doubt that is a big issue nowadays. A poor slew rate would prolly also defeat the linear to 20khz rule .. same about the peaks you mentioned.

    So, I think I still have a Q. Other than odd order harmonics why would a linear-to-20k component sound shrill?
389 390 391 392 393 394 395 396 397 398
400 401 402 403 404 405 406 407 408 409

Share This Page