Audio-GD Singularity 19 DAC - Measurements

Discussion in 'Source Measurements' started by Hands, Mar 23, 2017.

  1. brencho

    brencho Friend

    Pyrate
    Joined:
    Dec 18, 2015
    Likes Received:
    7,978
    Trophy Points:
    113
    Location:
    LA
    f**k that i agree with @Hrodulf let's get our american commodity units back for this hot heaping pile of garbage. or just sell it on head-fi where people may still test chi-fi metal by biting down on it.
     
  2. SSL

    SSL Friend

    Pyrate
    Joined:
    Nov 12, 2015
    Likes Received:
    1,239
    Trophy Points:
    93
    A-GD claims 0.001% THD on their website. Are they known for making wild claims off by several orders of magnitude?
     
  3. TheIceman93

    TheIceman93 El pato-zorro

    Pyrate
    Joined:
    Dec 17, 2016
    Likes Received:
    1,300
    Trophy Points:
    93
    Location:
    SoCal
    I wonder if this unit is defective in some way. Not defending AGD but its hard to imagine that someone with this much experience building DACs has managed to build something that measures horribly and sounds just as bad. I'd be curious to hear Kingwa's thoughts on these results but it sounds like he bit off way more than he could chew with a fully in house design.
     
  4. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    My first thought reading this is Audio-GD thinking "we know our discrete DAC module sounds and measures like arse" so we better get on with choosing a new champion as we've shagged the PCM1704 supply to death.
     
  5. TheIceman93

    TheIceman93 El pato-zorro

    Pyrate
    Joined:
    Dec 17, 2016
    Likes Received:
    1,300
    Trophy Points:
    93
    Location:
    SoCal
    I have to believe that Kingwa thinks this thing sounds good. Otherwise, why wouldn't he just play it safe with the 9038 chips? Was anyone really asking for a discrete DAC from AGD? Its a pretty big gamble since one really bad product could ruin your entire reputation. I have a feeling the S19 is going to disappear pretty quickly once SBAF finishes sending through the measurement and impression wood chipper. Any bets on what the excuse will be? My guess is that they will say that its too pricy and time consuming to build.
     
  6. landroni

    landroni Friend

    Pyrate
    Joined:
    Aug 12, 2016
    Likes Received:
    1,164
    Trophy Points:
    93
    @baldr has alerted in the past that discrete designs tend to come with unspecified INL / DNL errors. This is actually one of the selling points of the R2R chips used in their MB line, as they come with tight INL / DNL specs.


    This BTW seems to be the reason for the needed warm-up time with the Schiit MB DACs:


    @Hands Maybe you need to burn the thing before estimating measurements or listening to it? Remember to put it in a brown bag...

    More seriously, thanks so much for the light comments around the graphs, going through the things you see. Not all of us are hazed... err... initiated in these things, so it helps to no end.
     
  7. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    Then, absent a problem with the sample unit SBAF got (which I'd really surprised at), it could be down to even him thinking "different" is "better", falling to expectation bias (none of us are immune), or just wanting to have a product in this space (see below) - but understanding that he can't compete with the performance of R2R ICs with the kind of discrete board that can be built, and sold, for the price he's doing it at.

    More likely, to me at least, is that his various "marketing" stunts have painted him into a corner. Regarding the PCM1704 as the second coming (maybe the first coming), waffling around it's pending unavailability as a tool to drive sales, focusing on the number of ICs in his units, and R2R ones at that, didn't leave him many options without a major departure in product architecture.

    Depending on your purchase volume, these aren't particularly inexpensive chips. It's one thing if you're using a single chip for all channels (and it's other functions), and another entirely if you start doubling up on them. In small volumes that issue gets exacerbated. And I have to assume Audio-GD's volumes are relatively tiny based on how long he made the dwindling supply of 1704s last ... (per his claims, not mine).

    Asking for one from Augio-GD specifically? Maybe not. But if you follow the major DAC threads in various places, over time, there's a definite fashion for discrete R2R DACs right now. Discrete R2R NOS DACs in fact. It's hard to stand out when your product is just another overpriced, badly BoM biased, reference sheet design with upgraded clocks and a fancy transformer.

    So, at least from an overall market trend, the desire is strong for R2R stuff, especially discrete, and at the lower end of the price spectrum it's underserved (and for a reason).

    That'll probably be the claim. Or some twatwaffle about it needing four million hours of burn-in. But if they go down the price-as-a-problem route, it's going to be relatively trivial to call that out based on the cheap parts, pure SMT build, and the fact that the R&D is already done. If they're selling it (the DAC module) at a profit, for between $30 and $45 (depending on which resistors you want), it's not going to be justifiably more expensive than his old 1704 based models, and about on par with 9038s in low volume.
     
  8. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Pyrate MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    12,285
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    It's definitely the 100-120V version. Or so the sticker says. I can't see anything out of order by looking at the unit. Doesn't look like anything was damaged in shipping.

    I suppose I can't rule out that their new SPDIF implementation is bad or plain broken. I don't know why, seeing as every other DAC likes my source, but maybe the S19 doesn't play nice with it.

    Seeing as I was going to ship it out tonight but won't be able to, due to blizzard conditions, maybe I'll be able to test out USB real quick just to be sure.

    I don't know all of their THD test conditions. It looks like they might be doing a -5dB or -6dB signal? Maybe the S19 really hates the higher level stuff, like -3dB or greater. Or maybe there's some other fuckery going on with how they do their tests to make the results look good. I tested at 48KHz. Maybe the DAC prefers 44.1KHz? If so, that would make it pretty broken in my mind. And the overall floor looks kind of dirty and noisy, not unlike what I measured.

    Maybe their measurements were for the non-upgraded unit? i.e. Half the DAC modules, non-upgraded TCXOs. Remember that not all hardware upgrades actually make things better.

    Let's not forget the DAC modules themselves are using 0.1% tolerance resistors. That's pretty bad. I don't know what sort of architecture they use or other methods to help mitigate resistor tolerances and matching issues, but I wouldn't be surprised if that's lacking in this case.

    If the unit is defective, that calls into question all the burn-in shit AGD performs on each unit. For some reason, I thought they tested each unit before shipment. Am I making that up?

    Plus, any other time I've tried to ask a DAC manufacturer about measurements that didn't match their spec sheet, I get the Jedi hand waving trick and an excuse that DACs are hard to measure.
     
  9. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    Looking at the DAM boards (from admittedly way-less-than-ideal pictures), if they're doing any kind of mitigation then it's being done in DSP and that means it's almost certainly guesswork rather than anything based on the actual (and current) performance of the resistor network.

    The results are probably what they look like ... a poorly toleranced ladder yielding lack of resolution as values blur into each other. Hell, even with the 0.01% parts that TotalDAC uses, I think they only manage 14-bit resolution from a 16-bit source, and there's a lot more cleverness going on in Vincent's design than I can see potential for in Kingwa's.

    And that's probably combined with exacerbated switching noise from driving the shift registers faster than is necessary (and perhaps there's some latch/dwell/release timing issues as a result as well).

    ...

    Is the version you received the "neutral" one? I'm assuming so from the wording on the Audio-GD website, but I figured I'd ask.
     
  10. Rockin_Zombie

    Rockin_Zombie Facebook Friend

    Joined:
    Jan 2, 2017
    Likes Received:
    117
    Trophy Points:
    43
    Location:
    Toronto
    May be you can e-mail Kingwa? Seeing this potentially hurts his business, I think he will take it seriously. In general Kingwa is pretty responsive, and things don't get lost in translation as much ( as opposed to say contacting AGD support).
     
  11. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I honestly do not think it's broken.

    These results are NOT unexpected for 0.1% resistors. I wrote some code a while back that generated random INL plots given a R2R DAC using X tolerance resistors. I can't seem to find it. But the results for 0.1% resistors were all over the place (with some good and some extremely bad), especially if there is no calibration and adjustment circuitry or software wizardry (it would take time to measure each module to obtain the calibration, thus add more to the costs).

    Look at the -60db sine wave. It's pretty clear the word codes are ultimately being translated into voltages which all over the place, with curves in the sine wave not even being straight. It's not just the chunks, but if we examine closely, some parts of the curves are higher slope than should be, others less. It's pretty ridiculous. This behavior is also reflected in the distortion plots.
     
  12. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Pyrate MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    12,285
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    I think so. I ordered it prior to them introducing a warm version. I don't even want to know what their "warm" sounds like if the one we got is "neutral."

    I was just about to say it would be up to @Marvey, but he posted just before I did. The S19 is a low-end DAC using low-end DAC modules with resistors at a tolerance level, 0.1%, an order of magnitude worse than what you get from the best Soekris board or TotalDAC, which even then 0.01% resistors theoretically can't do more than 14-bits without some sort of calibration and other architecture wizardry that I don't understand fully.

    I am inclined to think the product we received is not broken. It's just a bad DAC.
     
  13. Clemmaster

    Clemmaster Friend

    Pyrate Contributor
    Joined:
    Sep 28, 2015
    Likes Received:
    3,268
    Trophy Points:
    113
    Ctrl+H -> Replace "ES9018" with "ES9038".
    That's how he did the documentation page. I hope he didn't do this for the actual DAC design, as well :rolleyes:
     
  14. skank

    skank Friend

    Pyrate
    Joined:
    Sep 30, 2015
    Likes Received:
    69
    Trophy Points:
    18
    Location:
    Florida
    I have an older Audio GD DAC and it is fussy about the SPDIF input. Mine actually responds well to adding a Coax Attenuator (aka John Kenny mod). It has something, I believe, to do with the fact that Kingwa's DAC's don't usually use an isolation transformer in the SPDIF circuit (or it's poorly designed). In any case it seems sensitive to reflections in the SPDIF circuit. Adding an attenuator is even noticeable using an I.A. OR-5 which I'm assuming has a decent output.
     
  15. Elnrik

    Elnrik Super Friendly

    Pyrate
    Joined:
    Jan 10, 2017
    Likes Received:
    8,973
    Trophy Points:
    113
    Location:
    Denver CO
    Home Page:
    Maybe there is a little Gundam build up in the toslink port.

    (Ok, ok... I'll quit with the anime jokes.)

    I am curious to see what the results from the USB port are. (Or any other input.) I'm not expecting an improvement, based on the resistors being crap, but... who knows.
     
  16. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    @Hands, just get rid of it and send it to me for the USB measurements. I'll take some measurements from both USB and coax with the QA400. I'd prefer to move this thing along sooner rather than later.

    As far as MOT provided measurements not jiving with our own. This happens all the time, going way back to Audeze's awesome looking CSDs for the LCD2, which was the impetus for me to start measuring shit and forming Changstar. Also, don't forget Sennheiser's 1/2 octave smoothed elongated frequency response measurements that came with the HD800 certificates.

    I have a general rule: never trust MOT provided measurements because they are essentially marketing materials.
     
    Last edited: Mar 24, 2017
  17. take

    take Friend

    Pyrate
    Joined:
    Dec 27, 2015
    Likes Received:
    185
    Trophy Points:
    43
    Location:
    Suffolk County, NY

    Nah, they definitely state that, on their page The babies in gestating:

    But of course, there is the classic listening test, too, so it's very strange that something anomalous would pass QC if it is as rigorous as they state.

    [​IMG]
    (Please turn up the temperature a few degrees and get this guy a better chair!)
     
  18. Dino

    Dino Friend

    Pyrate
    Joined:
    Jun 21, 2016
    Likes Received:
    3,332
    Trophy Points:
    113
    I was curious what people were saying on http://www.head-fi.org/t/831737/audio-gd-singularity-19 . I was skimming and skipping around (seemed like they went from "it needs more burn in" to "you just need the right SPDIF implementation going in").

    I got to the last page and a comment was made "There is some guy on the Super site doing measurements and listening impressions and they are pretty bad." Someone asked "Link?". He was told that there was a clue in the previous post. Then, someone posted "SBAF is always so negative. Is this really that bad?". I refreshed the page and poof, all of that had disappeared.
     
  19. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    This is normal for them. I got the warmer version (Master 7) and compared it with Amos' neutral version (forgot what model, I think it was SA2 or SA400, but those might be Russian missile systems; but it had half the PCM1704 chips, but in the same form factor chassis), and it turned out mine was the more neutral version. So go figure.
     
  20. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I'd love someone to do an analysis of the products we've had in our loaner program in the past 12 months and see how the results tally out in terms of positive, mixed, or negatively received.

    In this case, per the measurements, the reason we are negative is because none of us have ever seen such spectacular craptastic measurements for a DAC. And the characteristics of the measurements seem to point to the R2R modules as the culprit. Based on the measurements alone, this product deserves SBAF's very first Leaky Warmbutt Smear Award. If I used my own money to buy this (instead of collective funds here), I would be enraged. Actually I already am enraged, but I am holding back to a bit to see if I or someone else can replicate the results.
     
    Last edited: Mar 24, 2017

Share This Page