Sonnet Morpheus DAC Measurements

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

  1. Clemmaster

    Clemmaster Friend

    Friend
    Joined:
    Sep 28, 2015
    Likes Received:
    1,885
    Dislikes Received:
    4
    Trophy Points:
    113
    This seems a lot worse than digital attenuation, though :confused:
    At least with digital attenuation, the harmonics/spuriae go down in proportion, until they fade into the noise floor.

    The preamp just reduces the fundamental, while keeping the harmonics at the full level, which is bizarre!
     
  2. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    64,224
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    In the post first, I realized I ran 1000Hz to align with the FFT bins. Here is BAL out running at exactly 1000Hz. These are the cleanest I'm able to get it.

    0db 1000Hz BAL out
    upload_2020-5-8_16-49-52.png

    -30db 1000Hz BAL out
    upload_2020-5-8_16-49-20.png

    -60db 1000Hz BAL out
    upload_2020-5-8_16-47-51.png

    -90db 1000Hz BAL out
    upload_2020-5-8_16-48-25.png
     
    Baten and Psalmanazar like this.
  3. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    64,224
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    Now this is super funky

    1kHz THDN from -100dbFS to 0dbFS.
    upload_2020-5-8_17-13-39.png
     
    Priidik, Baten and Psalmanazar like this.
  4. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    64,224
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    Supplementary to above
    upload_2020-5-8_17-16-52.png

    upload_2020-5-8_17-17-45.png
     
    Senorx12562 and Psalmanazar like this.
  5. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    64,224
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    Seeing if I can replicate results elsewhere - yup

    -54dbFS
    upload_2020-5-8_17-21-32.png

    -55dbFS
    upload_2020-5-8_17-21-0.png
     
    Senorx12562, Grattle and Psalmanazar like this.
  6. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    64,224
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    Not a big deal really because even if the original line from 0dbFS continued on (without being spliced) to 100% distortion at 100dbFS, I wouldn't have cared since 99% of recordings, especially the privately remastered deals are 16/44.1.

    Could this sort of be the Honda VTEC (or similar variable valve-timing) approach for DACs? LOL!
     
    Psalmanazar and JK47 like this.
  7. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    64,224
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    Here are the sine waves:

    GRN = -54db (the one with the higher distortion spikes, slightly higher odd order than even)
    YEL = - 55db (the one with a little grass at higher orders)

    Looks like there is a little bit of squaring off of the -54db signal, which is expected based on the FFT pattern.
    upload_2020-5-8_18-39-43.png

    Does any of this matter? Probably not. I doubt people will be able to hear distortion -40 below a -50dbFS signal. You'd have to be listening extremely loudly in an anechoic chamber. And there is a huge advantage for lower level signals.
     
    Baten likes this.
  8. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Friend MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    11,051
    Dislikes Received:
    39
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    I know Cees made a big deal about his approach to the Pavane. Something about splitting a 16 or 24 bit signal in half, i.e. either two 8 or 12 bit signals, converting those on the most accurate part of the ladder, attenuating one of those two halves, and combining the signal in the end...

    Is that just a drawn out, convoluted, or weird way to describe an existing architecture?

    Either way, that this trait occurs where and how it does, I wonder if it's something related. I believe he's still using that general design approach even in his latest products.

    Just one guess. I could be dumb and way off.
     
    Baten and earnmyturns like this.
  9. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    64,224
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    54db is 9-bits. Looks like x2 9-bit ladders spliced together. Very cool. I'll do some more experiments to verify. As I said, Honda VTEC:

    [​IMG]
     
    Psalmanazar likes this.
  10. Clemmaster

    Clemmaster Friend

    Friend
    Joined:
    Sep 28, 2015
    Likes Received:
    1,885
    Dislikes Received:
    4
    Trophy Points:
    113
    Yes, the Metrum DACs usingthe transient modules and the Morpheus do what @Hands described: run both 12-bit halves on separate 16 bit DAC and merge them by adding both halves with the correct amplitude.
     
    Psalmanazar and earnmyturns like this.
  11. JK47

    JK47 The Beer Houdini

    Friend JK47
    Joined:
    Sep 26, 2015
    Likes Received:
    6,371
    Dislikes Received:
    102
    Trophy Points:
    93
    Location:
    "Surf City USA"
    I clearly remember just getting out of High School in 1994 and going with my buddy to pick up a 1993 GS-R at the Acura dealership. First time I felt VTEC, FFFFFUUUUUUCKKK. Bought a 1997 Prelude as soon as it came out, still one of my all time fave vehicles I've owned. Funny enough, a Merkur X4R4Ti was another all time fun ride, especially when re-routing a vaccum hose, having boost kick in at 2500rpm rather than 3500rpm, and rear wheel burnouts and drifting corners for days. All the while having a Pinto motor with a turbo, same motor as the turbo T-bird, and SVO Stang of that era, lol.
     
    exocer, Senorx12562, ufospls2 and 3 others like this.
  12. MisterRogers

    MisterRogers Ethernet Nervosa

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    1,520
    Dislikes Received:
    0
    Trophy Points:
    93
    Location:
    Aurora, CO
    Home Page:
    I've shared a link to this thread with Cees. He's wondering - what measurement equipment is being used?
     
  13. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    64,224
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    For the analyzers: AverLAB (BAL and SE), Focusrite 2i2 with ARTA software, and the QuantAsylum QA400 (SE).

    For signal generators: ARTA PC Eitr (coax) and AverLAB (AES).

    Would be curious to know if this unit is not representative; I am totally willing to re-examine. It's also possible I've made a mistake - something I am always worried about - we've seen how others have borked measurements in their haste.

    I'm not a measurement terrorist. I hope you can make that clear to him!
     
    Last edited: May 9, 2020
  14. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    64,224
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    XR4Ti, the American car for the enthusiast who didn't want to go sorority girl BMW. Too bad it didnt pan out in the long term, but that's Ford. I loved the styling.

    Prelude, packed with tech, classic and timeless design from Honda's glory days before they slumbered for 15 years making boring shit until recently... Civic Type R with all its tech is sort of the successor to the Prelude, but they kinda went overboard on the ricer aesthetic. Although I have to admit it's kind of cool now because ricer is passe.
     
    JK47 and Senorx12562 like this.
  15. Psalmanazar

    Psalmanazar Most improved member; A+

    Slaytanic Cliff Clavin
    Joined:
    Sep 27, 2015
    Likes Received:
    4,206
    Dislikes Received:
    132
    Trophy Points:
    113
    Purrin, how is the perceived sound at -55 with makeup gain an external pre vs -54 and above? I always thought the Metrum stuff was very clean for nos but way too soft.
     
  16. MisterRogers

    MisterRogers Ethernet Nervosa

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    1,520
    Dislikes Received:
    0
    Trophy Points:
    93
    Location:
    Aurora, CO
    Home Page:
    I don't think there's a risk of of him thinking that - I've known him awhile, and he's a bright, level-headed dude.
    I'll share any insights or questions he has - might even get him to join up for dialog.
     
    insidious meme, neogeosnk and JK47 like this.
  17. MisterRogers

    MisterRogers Ethernet Nervosa

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    1,520
    Dislikes Received:
    0
    Trophy Points:
    93
    Location:
    Aurora, CO
    Home Page:
    FWIW - Cee's believes that the above measurements are being impacted by a grounding problem within the test chain, as Morpheus is clean, having a noise floor of at least -140 dB (lower). He shared with me a graph he generated for a magazine writeup (he asked me not to publish it yet), showing a THD +N of 0.00344.

    I'll happily share the graph(s) and specifics as soon as I'm green lit to do so.
     
    msommers likes this.
  18. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    64,224
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    I'm actually not getting much different.

    Noise floor I am achieving below (this is with 8192 FFT window size), but I'm not getting 0.0034%THD (-90db), but -85db? instead, namely because the odd harmonic spikes are a bit higher. This is for the 0dbFS signal. See below.

    Ground contamination will result in 60Hz harmonics and associated IMD, which we are not seeing here. The AV mains filtering is excellent.

    1kHz BAL output into AverLAB
    upload_2020-5-10_9-42-53.png

    This is assuming we are talking apples to apples and he wasn't looking at the -54dbFS and -90dbFS results.
     
    Last edited: May 10, 2020
    Baten likes this.
  19. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    64,224
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    Here is 1kHz SE output into QA400
    upload_2020-5-10_10-14-27.png

    It's SE outputs here now, so seeing some AC mains slip through is normal. I had to back off from full scale output since the QA400 won't handle that voltage. However, the THD results here are in line with ~90db THD or 0.003%, except for the 4th harmonic bloop. Note that his analyzer tends to have more random spurs than the other.

    Note Y scale goes down to -150db.
     
  20. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    64,224
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    Here, I redid it with 32767 window size and set the Y-axis bottom -160db. It's the same plot essentially. Results are in agreement with Cee's statements. Highest harmonic is about -92db (close to it since it's not landing on the exact FFT bin). THD may be a little bit higher as we sum up all the harmonics.

    upload_2020-5-10_10-28-46.png
     

Share This Page