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.
408 409 410 411 412 413 414 415 416 417
419 420 421 422 423 424 425 426 427 428
  1. hans030390
    Maybe the Audio-GD DAC-19? Unless you're interested in Chinese DACs based off the old Philips TDA chips...but I would not expect particularly good performance out of anything like that, especially TDA1543-based. You might like the tone? A lot of those are also NOS (non-oversampled), which can be pretty hit and miss with some people. Seems it's a minority that prefers NOS. Granted, I have seen some people try those cheapy 1543 NOS DACs that still end up liking them enough to move on to and primarily use nicer R2R DACs, vintage or otherwise.
  2. KeithEmo
    I would like to interject one or two "philosophical points" here. (These are sort of in direct response to things mentioned a few weeks ago; I didn't quote them here because my responses are more "general purpose", and so apply to the entire subject in general.)
    When we talk about DACs, we often seem to fall into a simple philosophical trap of deciding: "Is a DAC supposed to reproduce THE SAMPLES IT IS GIVEN, or is it supposed to reproduce THE ORIGINAL ANALOG CONTENT?". Unfortunately, any answer to that question is a sort of paradox. The job of a DAC is in fact to reproduce the original source material. However, THE ONLY INFORMATION IT HAS TO USE WHEN DOING THIS IS THE DATA WE FEED TO IT.
    One posting quoted a very informative article that suggested that (I'm paraphrasing the actual quote) "a DAC can't accurately reproduce the original signal because the original signal may have information at frequencies above half the sample frequency". While entirely true, this is simply a restatement of sampling theory.... If you play a digital audio signal through a DAC, it cannot reproduce any information in the original content that was above half the sampling frequency. However, this is NOT a limitation of the DAC. The simple fact is that any information in the original sample that was at higher than half the sampling frequency ISN'T IN THE DIGITAL FILE TO BEGIN WITH, and the DAC cannot reproduce what isn't there. And, since this is a limitation of basic information theory, rather than a limitation of ANY particular DAC or type of DAC, it is absolute. NOBODY DOES OR CAN EVER DESIGN A DAC THAT CAN REPRODUCE INFORMATION THAT ISN'T THERE IN THE RECORDING. (A DAC could be designed that used "fake" information to "fill in" information that it "guessed" was missing, and such synthetic information might sound better in some situations, but it would simply be "an artist's rendition of reality", and not a magical resurrection of the information that was lost when the recording was made. It's sort of like colorizing a black and white movie.)
    Perhaps the "safest" way to spell this out would be thus: The goal of a DAC is to reproduce the original analog signal as accurately as possible. Since the original analog signal is no longer available, the best way to do this is to reproduce the information that is contained in the digital audio signal as accurately as possible. (Since we don't know what was lost, we can't add information to replace it without the risk that our added information actually makes the result less accurate rather than more accurate, so the best we can do is to ensure that we don't lose or alter any information that we still have.)
    The paradox, or perhaps dilemma, is that we are starting with what are often limited or flawed recordings. A Red Book CD can't contain any legitimate information above 22 kHz - because of the Nyquist limit. Therefore, if there is anything there above 22 kHz, it is either noise or distortion, present because of faulty processing when the CD was recorded. Likewise, if poor quality or incorrect processing has resulted in a digital audio file that sounds "harsh" or "fatiguing" when you play it back accurately, you can't "fix" it - your only choices are to play it accurately, or risk making it even less accurate by altering it in some way to make it "more pleasant" or "less annoying". (And what if the original artist actually intended it to sound harsh and annoying? You can't know for sure, and saying that "you know what a real piano sounds like" in no way proves that the piano on this particular recording is supposed to, or originally did, sound like you believe it did.)
    Therefore, when it comes to DACs, as with a lot of other audio equipment, you may well be faced with the decision of whether you want a device which is accurate, or one which sounds good. I suggest that choosing a reasonable answer to that question should and does depend on what you listen to. (If you listen to mostly early, and badly recorded, CDs - then maybe it makes sense to sacrifice accuracy for a little euphonic "smoothing over" of the flaws common in that type of recording. However, if you listen to a significant amount of modern, well recorded, content - then maybe it makes sense to choose a DAC that renders it as accurately as possible.) I have personally never understood how the term "analytical" can be used in a negative sense... but that's probably because my choice is to have the DAC render whatever it is given as accurately as possible.
    Also, unlike with analog content, digital content gives me an opportunity to have "the best of both". If I have a digital recording that is actually badly flawed, and I do consider it reasonable to try and correct those flaws, I can do so using one of the many excellent digital editors out there, and a virtually unlimited choice of plugins and modules designed to fix specific flaws, and to add an amazing variety of "euphonic alterations". If I have a neutral "analytical" DAC, then I can always use software to make any euphonic alterations I like, but I can also listen to the recording without altering it. If I have a DAC that adds euphonic coloration to everything I play through it, then I have no way to avoid that coloration, even if I have good quality content that sounds better without it. (And, if I have a pair of headphones that are a bit "sharp", and so sound better when EQed to -3 dB at 20 kHz, I'd rather use the EQ setting on my music player - which I can turn on and off - to do so, rather than own a DAC or headphone amplifier that has a "fixed EQ" that alters everything that way, but which might not sound very good with the next pair of headphones I buy.)
    abartels, wink, dan.gheorghe and 3 others like this.
  3. KeithEmo
    You might check out these guys.... they make those 1543 NOS DACs - and you can usually find them on eBay.
    (and they usually are in fact quite cheap)
  4. abartels
    I completely agree.
    Nice to hear someone talking about "analytical" sounding equipment the way I think about it, just like there are no classical loudspeakers and popular loudspeakers.
    A loudspeaker has to reproduce in the best way, and not adding something, or leave something out,,,,,
    Maybe a little offtopic, but did someone heard of the new AKM AK4497? Any news on that front?
  5. BassDigger
    What's happened to this thread? [​IMG]
  6. abartels
    Sorry for asking [​IMG]
  7. BassDigger
    Chill; I wasn't really getting at anyone in particular. [​IMG]
    (Although....did you really have to re-post Keith's entire comment....?)
    kugino likes this.
  8. jimvibe
    A loudspeaker or a headphone driver is much less perfect than Amps and even less so than DACs so there is always some coloration present. Since there is no way to get rid of this coloration anufacturers have to chose the coloration they like most.
  9. wink
    Anything that adds to the actual recorded material is spurious information.
    I'm awaiting the MAGIC DAC which is not dependent on the recorded material, but accesses the original sound when it was recorded. It comes with a gallon can of snake oil to lubricate the metaphysical machinery manufactured by  Morgul minions securely ensconced within the hermetically sealed Faraday cage casing.
    Pixie dust and unicorn tears are not required on this model, but are reserved for the future TOTL model which is currently under intensive ensorcelling to meet the stringent parameters required by this proof-of-concept delving into the ultra-accurate reproduction of sonic reproduction.
  10. lukeap69

    Does the case need to be hermetically sealed?
  11. Articnoise

    Okay I will try to explain. If using the word analytical as meaning separating/dissect the sound into its elemental parts - I do think it can be seen as negative. Mind you I don’t think that analytical meaning that it is more or less resolving, more or less neutral or more or less accurately.


    It just means that is not coherent. The coherence factor is a very important quality for me and make the listening fun, logical and musical and I have found analytical to be the opposite if playing at home. If you are in a studio making music, yes then you want a more analytical gear to dissect the music in to pieces.   


    It is a bit complicated to explain exactly what makes one gear to sound more or less analytical or coherent, but to me they are the very opposite of each other. The analytical sound separates the music into individual parts that are then easy to detect and a coherent sound is more about the whole and less about the parts. I think of it as the analytical sound lack the low level combining glue which binds the instruments and music together. A good balance between the two is usually the best. 

  12. KeithEmo
    I think I do understand what you mean, but it also seems to me that different people use the term very differently, and to mean subtly different things... to me the term "over-analytical" seems more accurate to describe "the condition" you're talking about. I agree with you that something that over-emphasizes the separation between individual elements is not a good thing, however I don't see that as being part of the same "scale" between accurate and not accurate. In other words, to me, being analytical means "being totally uncolored and accurate and not covering anything up" - and exaggerating the differences goes past "being totally analytical" and into something else - being inaccurate in the other direction altogether.
    I prefer to use the analogy of a picture. When you take a digital photo, it can be sharp or blurry, and the color can be accurate or not, and the contrast and brightness can be accurate or not. These are all things that can be measured, and each can be more or less accurate, but I would say that "being analytical" or "being accurate" means that all of them would be "correct". Now, it may be that, from an artistic point of view, we may prefer a particular photo if it is less accurate (for example, it's common to deliberately blur photos of faces to hide blemishes, and to airbrush entire bodies to de-emphasize obvious flaws, and deliberately exaggerating colors looks very cool in certain cases). However, very few people I know would deliberately buy a blurry TV so old movies look less obviously bad, or deliberately buy a poor quality camera because it doesn't show up the flaws in their subjects. (Many photographers use a "gel" filter to "soften" certain pictures, or use equivalent post-processing in Photoshop, but very few would buy a camera that was incapable of delivering a sharp picture if called upon to do so. Likewise, you might consider a TV that includes a "soft picture" option for watching old movies, but probably only if it has an "off switch" for when you don't want it.)
    Now, Photoshop also has an option that allows you to "sharpen" a picture after it is taken. In reality, what this feature actually does is to boost the contrast ratio around edges. (Processing a picture to deliberately add slight halos around high-contrast edges - a dark halo on the dark on the dark edge and a light halo on in the light edge - makes it appear to be sharper.) However, it's really a sort of optical illusion. It doesn't actually add detail but, by making the details that are already there more apparent, it makes the picture seem more detailed. If properly applied, this trick can make a too-soft picture look very good, but, if over-applied, it produces an exaggerated effect that looks unnatural. To me, the way a picture that's been over-sharpened looks is exactly analogous to the way some equipment sounds - and like what you described - almost like someone has artificially outlined the edges of each instrument or note (and this seems to be what some people describe as sounding "etched" - which is a great description of how over-sharpened pictures look).
    My point of that somewhat long winded description was to demonstrate that the two "directions" are NOT really a continuum. Even though it may seem that way to people who've never taken a picture or used Photoshop, there isn't a single "control" that goes from "blurry" to "sharp". There are really two separate controls, one for "added blurriness" and another for "added sharpness". And, to me, if applied to a picture, the term "analytical" would mean - NO added blurriness and NO added sharpness. (No-one would describe the cartoon-like picture you would get if you turned the sharpness all the way up in Photoshop as "analytical" - they would describe it as having exaggerated sharpness - or even as looking like a caricature.) You can add "blur" to compensate for a certain picture being "annoyingly sharp", or to cover up details you'd prefer not to see, and you can add "sharpening" to compensate for a picture that's "too soft", but they are still two distinct colorations, and the the most accurate picture will have none of either (and, with images, there is a pretty obvious point where that is true).
    I believe that a lot of audiophiles don't understand this distinction... and to me it seems pretty important. If your headphones and amp blur the details, then the solution is to reduce the problem that causes the blurring, but that's not the same as taking something that's neutral to begin with and add something that artificially boosts the "audio contrast". You can't make the blur go away by adding sharpness - at best you can create an illusion that makes it look superficially better. And, likewise, if something alters the sound in the way you describe, which I would equate to over-sharpening a picture, then the solution is to reduce the flaw that's causing the error in that direction, but that's not the same as simply "adding blur". But you DON'T fix an over-sharpened picture by adding blur. (You might argue that adding sharpening to a blurred picture does in fact improve it, but most photographers would agree that doing so is a last resort, only used after you've done your best to eliminate the original error.)
    To me, from a technical perspective, it seem like an awful lot of discussions about headphones and amps come down to "this headphone is too sharp - I need to find a blurry amp to go with it", or the reverse, as if there were in fact a single control that went from one to the other. (To me, this seems rather like trying to correct nuances in frequency response with one of those old "Tone" controls, rather than using more accurate Bass and treble controls... and I can't imagine an audiophile using a Tone control to correct for his overly-bright speakers - because it's obvious that the solution almost certainly won't "line up" with the problem. (If we're talking about a simple aberration in frequency response, then boosting the treble in one device to compensate for a roll off in another may produce a good result, but that's usually NOT what we're talking about, which makes a proper and effective solution somewhat more complicated.) If you agree with my assessment, then it also becomes obvious that we must differentiate between correcting a problem where a device is adding unnatural and excessive detail, and starting with something that's simply totally accurate, then reducing the amount of detail because of a personal preference for a softer audio picture, or because we're listening to flawed source material which simply sounds better when we don't hear all the details. 
    Personally, I much prefer to do my best to start out with "perfectly neutral", removing as many colorations and imperfections as possible, then adding the precise colorations and alterations I like, rather than to try and find more or less random combinations of flaws and colorations that add and cancel in such a way that I like the result. (Both because it seems like "the more rational way to do things", and because that way I don't have to re-think my entire system every time I "throw off the delicate balance" by changing one component or another.)
    (That's why, to me, saying "over analytical" is sort of like saying that a picture is "too perfect"....  in that it sort of doesn't make sense .... To me, a "perfectly analytical" system would make it so each individual instrument, and even each note, was precisely distinct and not at all burred together but, at the same time, wouldn't exaggerate the separation between them either. Anything past that would have crossed from "analytical" to something else... )
    dan.gheorghe likes this.
  13. BassDigger
    You're comparing two dissimilar adjectives; they do not mean the same thing (no-matter how many analogies you use):

    Full Definition of ACCURATE 

    :  free from error especially as the result of care <an accurate diagnosis>

    :  conforming exactly to truth or to a standard :  exact <providing accurate color>

    :  able to give an accurate result <an accurate gauge>


    Definition of ANALYTIC

    :  of or relating to analysis or analyticsespecially :  separating something into component parts or constituent elements

    • The definition of analytical is someone who studies and examines the elemental parts of something, or is something related to the study of small parts of a whole.
    I'd agree that for something to sound 'analytical' the sound character is somehow emphasising particular parts, or gives the impression that it is doing so, and is evidence of a sound signature or characteristic. This implies that it's unnatural, unrealistic and most probably an indication of an inaccuracy.
    Accurate sound reproduction should surely mean 'has total similarity to the original event'. The original event is a natural, real or synthesised sound, with only the characteristics that were included, intentionally or not, during the recording process.
  14. Stillhart
    I don't claim to know anything about audio engineering, I'm just a computer nerd who likes this hobby.  So please take this as it's meant:  just as a point of conversation, not contention.  
    It sounds like what you're talking about is the audio version of what gamers call anti-aliasing on graphics cards.  Why wouldn't something like that work well in audio if it works okay in video?  I can certainly see going overboard with it similar to how newer TV's will interpolate entire frames, giving the much-debated "soap opera effect".  But I'd think a little bit of anti-aliasing (where you average the differences between two given samples and stick and extra sample in between to smooth out the transition) could sound good if done right.
    Or am I misunderstanding your point and we're talking about two different things?
  15. Sonic Defender Contributor
    While I don't know this as a fact, I am not sure the visual system processing system is an analogue for the auditory processing system.
408 409 410 411 412 413 414 415 416 417
419 420 421 422 423 424 425 426 427 428

Share This Page