Sonnet Morpheus DAC Measurements

Discussion in 'Source Measurements' started by purr1n, May 6, 2020.

  1. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    57,327
    Dislikes Received:
    98
    Trophy Points:
    113
    Location:
    Antarctica
    Conclusions thus far:
    • Measurement behavior hard to predict or extrapolate from one measurement to another - some odd behavior and interactions.
    • However, steady state distortion measurements too low to make a difference or tell us anything that we can correlate to human subjective perception.
    • Roll off on the high end starts gradually at 10kHz and is down -4db at 20kHz. Hard to say if people can hear this. I think I can.
    • "Preamp" should still be treated as digital attenuation where we still want to keep the attenuation to a minimum.
    Subjective Impressions Here:
    https://www.superbestaudiofriends.o...gital-audios-morpheus.9093/page-2#post-298790
    https://www.superbestaudiofriends.o...gital-audios-morpheus.9093/page-3#post-299204

    I am having some issues with balanced output measurements at lower levels from combination of specific gear, verified with other measurement gear I have on hand, so just showing SE output measurements for now. AES input used below. (EDIT: I figured out it was a grounding issue - bad AES cable. BAL measurements here: https://www.superbestaudiofriends.o...heus-dac-measurements.9275/page-2#post-299929).

    0dbFS - full scale 1001Hz SE out
    upload_2020-5-6_21-26-49.png

    -30dbFS 1001Hz SE out
    upload_2020-5-6_21-24-34.png

    -60dbFS 1001Hz SE out
    upload_2020-5-6_21-25-24.png

    -90dbFS (toggle of smallest bit for CD quality content) 1001Hz SE out
    upload_2020-5-6_21-26-2.png

    What's interesting is the behavior of the lower level test tones, much less distortion and AC noise. Also the excess grass (at 1001kHz) is discussed here: https://www.superbestaudiofriends.o...t-morpheus-dac-measurements.9275/#post-299776.
     
    Last edited: May 11, 2020
    exocer, lagadu, Firegon and 2 others like this.
  2. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    57,327
    Dislikes Received:
    98
    Trophy Points:
    113
    Location:
    Antarctica
    BAL outputs

    0dbFS - full scale 1kHz BAL out
    upload_2020-5-6_21-37-21.png

    -90dbFS (toggle of smallest bit for CD quality content) 1kHz BAL out
    upload_2020-5-6_21-35-51.png

    Lower noise floor (and higher output), otherwise same characteristics as SE.
     
  3. fp627

    fp627 Friend

    Friend
    Joined:
    Sep 26, 2018
    Likes Received:
    777
    Dislikes Received:
    8
    Trophy Points:
    93
    Location:
    SouthernCA
    Any idea why this is? Also thought this was pretty peculiar (+ no 60Hz noise/bump, this is the first time I've seen this as far as I can recall on SBAF).
     
  4. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    57,327
    Dislikes Received:
    98
    Trophy Points:
    113
    Location:
    Antarctica
    0dbFS THD+N sweep 20-20kHz BAL out
    upload_2020-5-6_22-2-0.png

    -30dbFS THD+N sweep 20-20kHz BAL out
    upload_2020-5-6_22-4-0.png
     
  5. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    57,327
    Dislikes Received:
    98
    Trophy Points:
    113
    Location:
    Antarctica
    48kHz white noise

    upload_2020-5-6_22-43-10.png
     
  6. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    57,327
    Dislikes Received:
    98
    Trophy Points:
    113
    Location:
    Antarctica
    Yes. Seems a bit strange. I'm getting odd results from the AverLAB with this one. Here are the measurements from the QA400 (SE):

    -6dbFS
    upload_2020-5-6_22-52-20.png

    -60dbFS
    upload_2020-5-6_22-54-19.png
     
    Last edited: May 7, 2020
  7. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Friend MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    10,322
    Dislikes Received:
    38
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    @purr1n I'm very curious to see if you will be able to determine what causes the noise floor discrepancies between the two analyzers. If this then has any bearing on how the DAC will "react" with different amps and the like, perhaps that might somewhat explain variable subjective impressions. I'm not banking on it, though, but am curious nonetheless even just from a pure technical perspective.

    Time permitting, can you test measurements with the pre-amp functionality totally disabled and how results change when using the pre-amp?

    On the Adagio, I found adjusting the DAC's pre-amp volume to have noticeable, and sometimes interesting, results on measurements. Interestingly, it performed best when volume was set a decent amount lower than the 2V RMS SE out standard. When matched to that standard and compared against an original Pavane, the Pavane measured a fair amount better (and perhaps better overall, no matter what volume you set the Adagio at). Nonetheless, anyone that listened to the Adagio found it an improvement over the original Pavane.

    Another note: Metrum, and now Sonnet, DACs have always seemed to suffer from increasing distortion as you hit higher frequencies (with high output level signals), or at least to a much higher degree than many other DACs.

    I'll have to see how the original Pavane compares in this regard. It's one of the better measuring products they've put out, at least of those I've tested.

    I'm not sure how much this has to do with their choice of DAC converters themselves or their generally, and purposely, simpler output stages. I've measured enough NOS DACs to know some don't exhibit this behavior, though it is a more common problem.
     
    Aklegal and Baten like this.
  8. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    57,327
    Dislikes Received:
    98
    Trophy Points:
    113
    Location:
    Antarctica
    Here is another set using a proto AD Schiit AD converter (not noise floor is higher)

    -60dbFS
    upload_2020-5-7_12-17-55.png

    -90dbFS
    upload_2020-5-7_12-18-33.png


    And yet another using the Focusrite (note dbFS is not according to the scale - I couldn't normalize it to the graph)

    -60dbFS
    upload_2020-5-7_12-25-46.png

    upload_2020-5-7_12-27-26.png

    I'd say the AverLAB results for these lower level measurements are probably legit.
     
  9. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    57,327
    Dislikes Received:
    98
    Trophy Points:
    113
    Location:
    Antarctica
    Let's see what's actually happening. (48kHz)

    -15dbFS 1kHz
    upload_2020-5-7_13-10-0.png

    -15dbFS 5kHz
    upload_2020-5-7_13-10-19.png

    -15dbFS 8kHz
    upload_2020-5-7_13-11-30.png

    -15dbFS 11kHz
    upload_2020-5-7_13-15-46.png [/user]
     
    Last edited: May 8, 2020
  10. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    57,327
    Dislikes Received:
    98
    Trophy Points:
    113
    Location:
    Antarctica
    Now this is interesting, when I use a frequency that isn't a factor of 48000 (the sampling frequency out) at higher levels, I get the a lot of grass, similar to the AverLAB output at higher levels. Note in the plots below, 960Hz and 2000Hz, when I increase the frequency just a bit so the number wouldn't integer multiply cleanly into 48000, I get a ton of grass.

    I suspect IMD at play.

    -15dbFS 960Hz
    upload_2020-5-7_13-23-5.png

    -15dbFS 961Hz
    upload_2020-5-7_13-22-54.png

    -15dbFS 2000Hz
    upload_2020-5-7_13-23-32.png

    -15dbFS 2004Hz
    upload_2020-5-7_13-24-12.png
     
    Last edited: May 7, 2020
  11. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    57,327
    Dislikes Received:
    98
    Trophy Points:
    113
    Location:
    Antarctica
    -15dbFS 2kHz, PREAMP OFF
    upload_2020-5-7_13-30-20.png


    -15dbFS 2kHz, PREAMP ON, 0db attenutation
    upload_2020-5-7_13-29-47.png


    -15dbFS 2kHz, PREAMP ON, -50db attenutation
    upload_2020-5-7_13-31-31.png

    No changes with preamp on and off. Well, if anything, we know that the "preamp" really isn't a preamp, it's simply digital attenuation.
     
    Hands likes this.
  12. Clemmaster

    Clemmaster Friend

    Friend
    Joined:
    Sep 28, 2015
    Likes Received:
    1,517
    Dislikes Received:
    3
    Trophy Points:
    113
    It's not digital attenuation, it's a change in ref voltage to the DAC chips, which, being resistors, tracks nicely.

    @purr1n : do you have time to measure my Jade, too? I could drop it off (with wipes :D) tomorrow or this weekend?

    I noticed a lot of distortion when setting the pot to a low position, up to ~11AM position. As in, the windows tone (that plays when moving the slider) would cause audible modulation (heard through both speakers and headphones).
     
  13. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    57,327
    Dislikes Received:
    98
    Trophy Points:
    113
    Location:
    Antarctica
    Kind of the same difference because if we had a true preamp, would have brought the spurs down along with the signal. But using the preamp function to attenuate, we just increased THD from 0.10% to 1.54%. The manufacturer knows this, and thus recommends in the manual to not use the "preamp" function on super high gain power amps (and by inference, not on super efficient speakers). The idea is that you still want to treat this like digital attenuation and keep the attenuation within a good sweet spot.

    --

    Here is one of the audio ayatollah's favorite measurements:
    (-10dbFS)
    upload_2020-5-7_13-42-46.png

    A little grass, but very good.

    A little tidbit: contrary to what someone else may have said, this signal is not reflective of actual music. Most music will have very little under 50Hz (even higher if small classical ensembles) and roll off in the highs. See example below for a Mozart winds recording and a Daft Punk (bassy) track. Yellow is peak. Pink is geometric mean.

    upload_2020-5-7_13-48-53.png

    upload_2020-5-7_13-50-4.png
     
    Last edited: May 7, 2020
    crenca, Clemmaster and Psalmanazar like this.
  14. Clemmaster

    Clemmaster Friend

    Friend
    Joined:
    Sep 28, 2015
    Likes Received:
    1,517
    Dislikes Received:
    3
    Trophy Points:
    113
    Ha, I couldn't tell appart the marker line to the actual fundamental (I'm color blind, deficient in green, mostly)... It just looked the same in all cases.

    Now I get why the Jade sucks at high attenuation in general, and in particular driving my Focal Solo 6 directly (these have way too much gain :rolleyes:)
     
  15. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    57,327
    Dislikes Received:
    98
    Trophy Points:
    113
    Location:
    Antarctica
    600Hz and 1700Hz dual tone. -10dbFS
    upload_2020-5-7_13-57-53.png

    50Hz and 1700Hz dual tone. -10dbFS
    upload_2020-5-7_13-59-7.png
     
    Psalmanazar likes this.
  16. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    57,327
    Dislikes Received:
    98
    Trophy Points:
    113
    Location:
    Antarctica
    Conclusions thus far:
    • Measurement behavior hard to predict or extrapolate from one measurement to another - some odd behavior and interactions.
    • However, distortion measurements too low to make a difference or tell us anything that we can correlate to human subjective perception.
    • Roll off on the high end starts gradually at 10kHz and is down -4db at 20kHz. Hard to say if people can hear this. I think I can.
    • "Preamp" should still be treated as digital attenuation where we still want to keep the attenuation to a minimum.
     
    Psalmanazar likes this.
  17. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    57,327
    Dislikes Received:
    98
    Trophy Points:
    113
    Location:
    Antarctica
    It's a novel approach, and likely to yield results where a little bit of attenuation isn't noticeable. Some digital attenuation solutions seem to suck immediately upon -1db applied. I figured it's the implementation, math, filters, etc. Same goes for switching in an active preamp from bypass - that can also immediately suck.
     
  18. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    57,327
    Dislikes Received:
    98
    Trophy Points:
    113
    Location:
    Antarctica
    Measurements with 95% of DACs is mostly stupid although interesting on an academic level. The measurement errors are too small relative to amplifiers and especially transducers (which will be the primary source of non-linearities).

    Much easier just using your ears. Or if you don't believe, do a blind test like what I've done in the past. Double blind even. The blind test will ensure that you are not subjective to confirmation bias from measurements.
     
    Last edited: May 7, 2020
    exocer, Senorx12562, Grattle and 2 others like this.
  19. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Friend MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    10,322
    Dislikes Received:
    38
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    @purr1n Thanks for checking the "pre-amp" results! It does not seem to exhibit the behavior I measured on the Adagio.

    Do other DACs show odd measurement behavior when testing a frequency that's not a multiple of the sampling rate? I haven't really checked for that myself to know if that would be the case.
     
  20. Psalmanazar

    Psalmanazar was: dingleberry; now: chill + interesting

    Slaytanic Cliff Clavin
    Joined:
    Sep 27, 2015
    Likes Received:
    3,762
    Dislikes Received:
    121
    Trophy Points:
    113
    Hands, this is very common with gear. Most gear collapses with higher and lower frequencies at hotter levels. Good gear gets slightly worse until clipped. Bad internet popular gear (RME, SMSL), turns to absolute dogshit. Rare is stuff like Lavry that gets better by saturating in a cool way. And probably. Nobody tests this shit. The audio gear industry relies mostly on selling crap to deaf people or people who don't care. Semi-respectable publications hand wave a lot of absolute dog shit pieces. So much shit sounds weird and nobody will talk about it. Most masters peak at what? -1 db if we're lucky? Modern pop and rap is always in the red. Old masters will often have clipped snare and cymbal hits. None of it is hitting most converters with -10 dbfs peaks.

    Purrin, THD isn't that important either if the ear can't hear it. Often slight or even massive THD (for some genres and recordings) can make things sound better. Even with classical music. 5% THD gear can often sound cleaner to the human ear than .1% piece in many circumstances.

    The mullah's measurements are haphazard, usually bad in the first place, he doesn't measure most important things, he doesn't install drivers, and most importantly he is deaf and shilling what he can make money off of. Many technically minded people will hear something wrong with something, measure it, and if they don't find what measures up with their ears, they will open up the gear and start measuring and swapping out parts. This is how a lot of popular gear that is ass to people with good ears gets destroyed and the internet objectivists larping as audio spinoza cry their eyes out and are revealed to just be apologists for shitty gear when they argue you can't hear butthole chinkese knockoff parts and the shittiest, cheapest parts in the power supplies imaginable. The PRC manufacturers are basically just throwing shit at the wall for people who listen to spotify as background music, watch youtube and youporn videos, and play videogames.

    It turns out that the plain jane parts done well in good designs almost always beat meme expensive part with bunghole parts that will fuck anything with them in the audio path or power path. Often they're salvageable and just defective by design while some cheap old power amp or DBX shit from a pawnshop often does 1 cool thing well and a bunch of things bad and was just cost reduced by a major corporation. You can crack those open and see the spots for the Jensens, replace some caps, and bam mediocrity is good. That's not the case with Far East White Van shit.
     
    Last edited: May 7, 2020
    Senorx12562, Grattle and mokobigbro like this.

Share This Page