Musician Draco R2R DAC

Discussion in 'Digital: DACs, USB converters, decrapifiers' started by purr1n, Apr 16, 2024.

  1. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    90,543
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Subjective impressions here: https://www.superbestaudiofriends.o...to-make-it-work-dna-weiss-musician-zmf.14367/ The unit is hefty and has a really nice thick metal faceplate. The LEDs are dim and do not blind you.

    musd.jpg

    In a nutshell (subjective and measurements commentary): https://www.superbestaudiofriends.o...sician-draco-r2r-dac.14378/page-2#post-428699

    First set of measurements are SE output. Input is 48kHz 16-bits via coaxial. NOS on.

    Musician Draco
    1kHz 0dbFS
    upload_2024-4-16_19-47-38.png

    1kHz -30dbFS
    upload_2024-4-16_19-47-58.png

    1kHz -60dbFS
    upload_2024-4-16_19-48-17.png

    1kHz -90dbFS
    upload_2024-4-16_19-48-56.png
     
    Last edited: Apr 22, 2024
  2. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    90,543
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Below are using 1Mkhz bandwidth

    No analog filter, as usual for today's "NOS" designs.
    upload_2024-4-16_19-51-40.png

    48kHz beat frequency (sampling rate), with another extraneous 47kHz (48-1k) that is IMD???
    upload_2024-4-16_19-54-26.png

    But wait, looks like I am mistaken. Same thing with 2kHz stimulus. But we get 46kHz and 50kHz tones?
    upload_2024-4-16_19-56-14.png
     
  3. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    90,543
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    15kHz stimulus tone.
    upload_2024-4-16_20-0-36.png

    Same as above, but expanded out to 90kHz
    upload_2024-4-16_20-4-5.png

    Here is the scope below. Don't look like a 15kHz to me. This is what happens without an analog LPF / reconstruction filter when we are too close to nyquist.
    upload_2024-4-16_20-0-41.png

    NOTE:

    I don't have a problem with NOS DACs doing this. Down the component chain, the amp and definitely the transducer / headphone will serve their part in being the LPF / reconstruction filter. However, I think it may be fair to say that YMMV to a very large extent because of this.

    However, we need to put these claims that NOS is the one method that yields the truth. Oftentimes it doesn't. I suspect the aliasing is the cause of the NOS tone.
     
    • Like Like x 4
    • Epic Epic x 1
    • List
  4. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    90,543
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Other frequencies, 37kHz and 8.2kHz at full scale output

    0dbFS 37Hz
    upload_2024-4-16_20-9-55.png

    0dbFS 8.2kHz
    upload_2024-4-16_20-16-38.png
     
    Last edited: Apr 16, 2024
  5. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    90,543
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    500Hz Square Wave

    Pulsing ringing behavior. Changes constantly between no overshoot with ringing to none.

    upload_2024-4-16_20-19-56.png

    upload_2024-4-16_20-20-12.png
     
  6. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    90,543
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I turned off DC high-pass on the APx555 to allow DC. Looks like some DC offset.

    -20dbFS
    upload_2024-4-16_20-27-12.png

    -40dbFS
    upload_2024-4-16_20-27-28.png

    Starting below here, all signals are negative volts.

    -50dbFS
    upload_2024-4-16_20-28-20.png

    -70dbFS
    upload_2024-4-16_20-29-51.png
    NOTE: huge channel imbalance
     
    • Epic Epic x 2
    • Like Like x 1
    • List
    Last edited: Apr 16, 2024
  7. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    90,543
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I'm done, I know enough already, unless you guys want to see more, then please request.

    One last one. Balanced outputs do not have the DC offset:

    upload_2024-4-16_20-32-33.png

    Probably explains why balanced sounds better, clearer, and not muddled shit.
     
    • Like Like x 4
    • Epic Epic x 1
    • List
  8. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    90,543
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Oh, maybe not. Still some DC offset, just more noticeable at lower levels. Note blockiness because 16-bit and NOS with APx555 at 90kHz bandwidth.

    -80dbFS
    upload_2024-4-16_20-43-17.png

    One good thing, very clear and distinct blocks with no glitchyness or spikes.

    -90dbFS (balanced output)
    upload_2024-4-16_20-44-39.png
    Note: the waveform is actually slightly off - too extended at the top and bottom. We should be toggling one bit back and forth with each cycle.
     
    • Like Like x 5
    • Epic Epic x 2
    • List
  9. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    11,320
    Trophy Points:
    113
    Location:
    Philippines, The
    Right, my high frequency hearing doesn't even hit 20kHz at this point (more like 17~18 in the one ear and definitely 17 in the other) but I'm more than happy to accept that needlessly high levels of ultrasonic and infrasonic content is gonna mess with a person's hearing/perception somehow based on evidence and anecdotal experience.

    Quick query, why 37Hz and 8.2kHz specifically for the stimulus tones? Also, wait, you're inputting a steady 500Hz into the DAC and its ringing behaviour changes on the fly? That's... curious.

    Never thought about using amp and transducer to serve as a hard LPF to dampen DAC noise up high, makes me wonder about how much discrepancy between DAC impressions is attributable to that. Cool. Also makes me wonder what the folks who developed this DAC used to test stuff with.
     
    Last edited: Apr 17, 2024
  10. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    90,543
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Somewhat arbitrary, but 37Hz is the lowest frequency of the third bass drop in Daft Punk's Do It Now. Really chosen because I feel it's a realistic low end that we can hear - and also to emphasize that in 99% of cases, music rarely ventures lower than that. I'd be OK with a speaker/headphone spec that goes down to 37Hz -3db. However, the objectivists measurebators of today (and Stereophile, because a speaker must go down to 20Hz for obtain the A+ Recommend Components rating) would probably have a cow.

    8.2Hz because the second harmonic lies well under 20kHz and also because that tends to be a common stridency point in headphones and IEMs.

    The discrepancy in experience depending upon downstream gear could be significant. The use of transformers in the chain could limit bandwidth more, thus serving as a LPF / analog reconstruction filter of sorts (FWIW, I believe the Metrum Pavane DAC used transformers at the output) Headphones like STAX will reach up higher in frequencies, thus having less of a LPF.

    NOTE: I realized the DC offset on the balanced is OK. 500uV is nothing. However, let's take a look at things relatively for a reality check. Both -50dbFS below with DC high pass turned off at the APx555

    upload_2024-4-17_9-14-22.png

    upload_2024-4-17_9-14-5.png
     
  11. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    11,320
    Trophy Points:
    113
    Location:
    Philippines, The
    HAHA I knew 37Hz was strangely familiar for some reason! And yep, seconding ~8kHz and above as being a pain point, though lots of folk seem to have more sensitivities lower down circa 5~6kHz. But that's too much data for most to bother actually reading I think :p

    Aye, DC offset makes for an alarming change in recording perception which is why even rudimentary recording software like Audacity automatically compensates therefore automatically when making gain up. Think I mentioned I was spending some time learning more about the other end of the recorded music "hobby", and it's just all sorts of wild seeing how problems with things like absolute polarity are even more annoyingly salient when working with dry recordings, especially for complex things like voice (not that anything recorded in real space isn't by defaults all sorts of amazingly complicated in waveform, but voice is what I'm most familiar with). That said, yeah balanced out seems to be none too bad in terms of absolutes.

    This really does remind me of someone on here (memory fails me) who opined that NOS actually sounded more fatiguing to them than OS, which they surmised was on account of the very higher frequencies being left unfiltered. My only experience so far with OS/NOS has been the MMB2 so I'm bowing out of opining too strongly there other than to say NOS actually is less fatiguing to me on that DAC on account of the more sedate treble presentation and less aggressive attack throughout the frequency range.
     
  12. crenca

    crenca Friend

    Pyrate
    Joined:
    May 26, 2017
    Likes Received:
    4,227
    Trophy Points:
    113
    Location:
    Southern New Mexico
    Holo's products (May, Spring, Cyan 2) have an analog filter before the SE & balanced outputs (although I don't know its design), so NOS does not = unfiltered necessarily. Apparently, the Draco has none based on @purr1n measurements...
     
  13. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    11,320
    Trophy Points:
    113
    Location:
    Philippines, The
    My ESL and word salad is likely showing again, whoops. Just to reiterate I was sharing someone else's surmisings about why NOS appeared to be more fatiguing to them in the long term (whereas I actually find OS on the MMB2 to be more aggressive wholesale, never mind that it does better suit my preferences). Still trying to look for that conversation, it was somewhere on here and the person shared a very pixellated looking YouTube video featuring a piece from a video game soundtrack featuring a woodwind (IIRC) instrument that they found well exhibited that fatigue.

    EDIT: Found it, Pillars of Eternity OST thing shared by @/RedFuneral:

    https://www.superbestaudiofriends.o...-piety-magni-pi-etc.12827/page-18#post-424871

     
    Last edited: Apr 17, 2024
  14. crenca

    crenca Friend

    Pyrate
    Joined:
    May 26, 2017
    Likes Received:
    4,227
    Trophy Points:
    113
    Location:
    Southern New Mexico
    For fun I just now streamed this track (Roon via Qobuz, 44.1/16) through my Cyan 2 at 1x NOS (> Freya +, SE in passive > Freya + in SS mode > Focal Utopia). Despite having this DAC for a week I have not actually bothered to feed it 1x content as I have not like NOS "mush" in the past. I then listened to the track upsampled in Roon to 16x (352.8, linear phase filter). Can't say it made much of a difference. That woodwind sounds very close miced either way - like it is (or about to) overload the mic. In fact, everything on this track sounds harsh (despite it being a slow, laid back vibe). I am willing to blame the recording ;)

    edit: Using Roon to resample this 44.1/16 track to DSD x256 helps a bit, but then Roon's DSD conversion always seems to take an "edge" off just a tad... After the track ended Isabelle's Faust w/ the Freiburger Barorchester playing Mendelssohn's 5th came on - now that's music :)

    {crenca's being a snob again}
     
    Last edited: Apr 17, 2024
  15. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    90,543
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    MMB2 still has the analog reconstruction filter in the output stage, albeit one designed for x4 or x8 OS, not x1 (NOS). I suspect this is the reason why Schiit DACs in NOS mode aren't so typically "NOS" sounding. I feel the change to NOS mode on Schiit DACs is more subtle.

    Quick measurement of MMB2 in NOS. There's still the spray. 48kHz is high up, but then we see the spray much lower throughout.

    upload_2024-4-17_10-37-7.png

    Also still weird (aliased) 15kHz signal, but not as weird without any analog filter.
    upload_2024-4-17_10-47-30.png
     
    • Epic Epic x 2
    • Like Like x 1
    • List
    Last edited: Apr 17, 2024
  16. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    11,320
    Trophy Points:
    113
    Location:
    Philippines, The
    I was just chatting with @ChaChaRealSmooth about how it was weird that I couldn't find any comprehensive comparative measurement suites of the MMB2 in both OS and NOS mode (the AP report on Schiit's page only features one set of measurements, presumably OS, unless my reading comprehension is pooping itself again). Yeesh, if this is supposed to be a more subtle NOS sound than typical I'm afraid to try a proper NOS DAC out! The NOS mode on this is already enough to sway my preferences on things a fair bit (I like OS more-- yes perhaps it's less microdynamically nuanced with less plankton but it suits me better, haha).

    Honestly it wasn't a bother to me either, but evidently the NOS mode I have access to isn't a "true" NOS implementation so I'm going to qualify my impressions as being less pertinent here. Rather curious what might have been causing the fatigue on RF's end.
     
  17. Clemmaster

    Clemmaster Friend

    Pyrate Contributor
    Joined:
    Sep 28, 2015
    Likes Received:
    3,284
    Trophy Points:
    113
    Are you asking, or do you already know the answer? :p

    This is the normal behavior of sample and hold converters. They realize what we call a "nearest interpolation", which exhibits this behavior: every tone gets replicated on either side of each multiple of the sampling frequency, unless attenuated by a filter (which always occurs, either intentionally via the use of a band-limiting filter, or simply because no practical system has infinite bandwidth).
    If your sampling frequency is 48kHz and your input signal is 1kHz, you will get replicas at 48-1 (47kHz), 48+1 (49kHz), 96-1 (95kHz), 96+1 (97khz), ...
    There is some attenuation, as dictated by the sinc envelope (= Fourier transform of a rectangular pulse, which is what sample and hold realizes in the temporal domain). For this example, the first 2 replicas are expected to be ~32dB down from the source tone and then it goes down from there at a much slower rate, just like this plot shows:

    [​IMG]
     
    • Like Like x 3
    • Epic Epic x 1
    • List
  18. RedFuneral

    RedFuneral Facebook Friend

    Joined:
    Jun 8, 2016
    Likes Received:
    215
    Trophy Points:
    43
    Location:
    CT, USA
    Hello! For added context I find NOS less fatiguing 99% of the time. The fatigue was in a specific system(pretty sure DAC, amp, and headphones are all 'hot' in same frequency.)
     
  19. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    11,320
    Trophy Points:
    113
    Location:
    Philippines, The
    You have my sincere apologies, I actually did reread that bit but forgot to amend my statement here. Got sidetracked! Makes perfect sense
     
  20. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    90,543
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Now, this is super weird.

    Musician Draco
    SE outputs NOS
    0dbFS
    THD (no N) vs Frequency
    upload_2024-4-22_9-55-18.png
    (50 point crawl to get finer granularity)
    upload_2024-4-22_9-50-31.png

    Let's get some granularity of the interesting area:

    Musician Draco
    SE outputs NOS
    0dbFS
    THD (no N) vs Frequency
    90kHz bandwidth
    (50 point crawl from 1 to 10kHz)
    upload_2024-4-22_9-53-44.png

    Let's take a look at harmonics broken down (will only show ones interesting)

    5th harmonic
    upload_2024-4-22_9-55-55.png

    7th harmonic
    upload_2024-4-22_9-56-45.png

    Let's look at the spectrum for 3.5kHz and 6kHz next...
     
    Last edited: Apr 22, 2024

Share This Page