Morpheus technical measurements

Discussion in 'Source Measurements' started by atomicbob, Aug 2, 2020.

  1. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
  2. 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:
    Quick tests suggest my original Pavane does not exhibit this behavior. Or, if it does, it would be closer to -70dB.
     
  3. 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
    The discrete ladder appears to be about a 90db ~15bit part. The part gets slightly distorted when the output is hot hence SINAD doesn't keep increasing as the signal gets above -10dbFS. Two ladders are then glued together at the -55dbFS point. One ladder configured to run at a higher output than the other. The bitword is chopped into two and distributed accordingly. The ladder that covers the LSB half has about 1 or 2 bits left below a -120dFS signal.

    The parts could have been glued higher up with an advantage in SINAD at the point where the second ladder takes over, but with the disadvantage of lower SINAD at lower levels.
     
    Last edited: Aug 3, 2020
  4. 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:
    That would make sense, but also would be peculiar if this measured trait is exacerbated on the Morpheus relative to the older Metrum stuff. I'll have to check my Pavane more thoroughly, of course. With Morpheus, it could be as simple as different DAC parts, whether for cost savings, easier manufacturing, better performance in other areas, greater stability over time, or maybe for legal reasons had to be different than the parts he made for (as) Metrum. Maybe it's changes to the analog stage. I dunno.

    Historically, I know Cees hasn't focused on how many zeros show up after a decimal point. More about noise, speed, bandwidth, etc. So, I suppose it wouldn't be surprising if the new approach, whatever that might be, is better in those regards without too much change or sacrifice elsewhere.

    Not that I'm worried, "Oh, this must be why the DAC sounds like X, Y, or Z." Just curious from a technical perspective, especially given the evolution of his designs over time.
     
  5. 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
    Well, he may have had to start from scratch and been unable to use the prior IP he developed during the much public Metrum meltdown where he left to do his own thing.
     
  6. 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:
    Right. Like looking over a classmate's code in a CS course to figure out how they did it, then "doing it yourself" after. Or, perhaps, you're the one that gave insights, and now need to rework some of your code out of fear the classmate will copy too closely and set off copying/cheating warnings in the automated submission system.
     
  7. 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:
    Here's a rough look at how distortion vs signal amplitude compare for the Morpheus and an original Pavane (aka Level 1, with DAC ONE modules).

    I say "rough," because I used the STEPS program (from ARTA). I calibrated levels in both scenarios. DACs were within a few mV of each other for RMS output, and the same ADC was used in an identical manner.

    Nonetheless, this is meant more for speaker measurements and the like, as I understand it. The results use voltage for the amplitude unit, so I had to manipulate and fudge the voltage numbers to represent dBFS. In this sense, STEPs seems to limit to just under 0dBFS and just above -100dBFS.

    I think it's close enough to compare their behaviors. The original Pavane seems to have the ladders glued together somewhere around 68dB rather than around 55dB for the Morpheus.

    Do note that the Pavane uses input transformers to convert balanced to SE out. SE outs on Pavane measure with noticeably higher distortion than balanced outs with higher signal levels, especially at lower frequencies.

    For this test, I measured both DACs via SE outs at 1KHz, 24-bit, 48KHz sampling rate.

    Even still, the Pavane's THD is very close to the Morpheus with signals above -10dB. It noticeably bests the Morpheus from -10dB to about -50dB, though it's debatable if Morpheus's increased distortion in that signal range really matters.

    The Pavane caps out near 2% THD before the next ladder, or part of the ladder, takes over, whereas Morpheus caps out around 1% THD. It's expected that lower level signals will increase in distortion, where something like 1-2% THD isn't completely out of the norm. Given that, I think I'm inclined to say I prefer the approach the Pavane takes here, even if it might cap out at a higher distortion rate before switching over.

    The ladder that covers the lowest level signals seems to perform very similarly on both DACs.


    Morpheus vs Pavane Lvl1 THD vs Amplitude.png


    Another note, THD on the Morpheus seems more 3rd harmonic dominant across the spectrum. Pavane, on SE outs, is slightly dominant on 2nd harmonic, though it's usually very close to the 3rd harmonic.

    The 2nd harmonic drops on Pavane considerably with balanced outs, effectively lowering the THD a fair amount. Thus, with balanced outs, the Pavane is also more 3rd order dominant, but will show reduced THD overall compared to the results above.

    It looks like the Morpheus performs more or less the same regardless of SE or balanced outs.

    Overall, I think this shows design choices have certainly carried over into what I'll call the 4th generation of Cees-designed DACs, but some variables have changed for reasons we can't yet know for sure.

    I'd be very curious to see how the DAC TWO modules compare in something like the Pavane...
     
  8. bilboda

    bilboda Florida boomer

    Pyrate Banned
    Joined:
    Nov 20, 2016
    Likes Received:
    834
    Trophy Points:
    93
    Location:
    Miama
    I'm trying to translate the graph to sound. Say you have a recording with 17db dynamic range. Doesn't that relate to the 17 db at the higher end on the right, leaving the rest of the graph academic?
     
  9. 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 assume it would depend where that range lies.

    Modern, compressed rock/metal, for example, averages to somewhere around -20dBFS in DAC/ADC loopback test in ARTA, last I checked. But I don't know max or min values off the top of my head.
     
  10. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    Sound levels and dynamic range are calculated many ways. One popular way is to average levels over a period of 50 mS. However a digital waveform has samples that vary by the sample period. It is important to reproduce every sample with the precise location in time (jitter) and level (amplitude distortion).

    Consider the following waveform for The Canadian Brass Bach Air Pour Les Trompettes:
    21 wave level -6 dBFS - overview.png
    Depending where the level is calculated a 50mS RMS may span -6dBFS to -36 dBFS during the music content.

    01 wave level -5 dBFS.png
    Here is one point where the level is achieving approximately -6 dBFS in the left channel.

    22 wave level -6 dBFS - zoom 1.png
    Increasing the zoom provides greater detail but we are still a long way from individual samples. Further zooming around 0:27.166 will follow.

    23 wave level -6 dBFS - zoom 2.png
    4 mS earlier we observe left channel with a peak to -6 dBFS.

    24 wave level -6 dBFS - zoom 3.png
    Back to 0:27.166 we find a -21 dBFS peak in the Left channel.

    25 wave level -6 dBFS - zoom 4.png
    Note there are Left channel samples between -21 dBFS and -infinity (the zero crossing) One sample is at approximately -46 dBFS. The sample right of the vertical cursor has a DAC count of 2195.

    26 wave level -6 dBFS - zoom 5.png
    Two samples to the right of the previous note sample has a DAC count of 2145. This is 50 counts lower in a 16 bit system of 65536 levels. That nuanced level change is on par with a level change of -62.35 dB at an absolute level of approximately -21 dBFS.

    Level meters are convenient for watching peak, average, RMS levels etc. during recording, mixing, mastering, reproduction. Reproducing every sample accurately in time and amplitude are the goals for a Linear, Continuous, Time Invariant, Sampled System. Those words will jog memories for many that have taken math classes oriented toward digital systems.
     

Share This Page