Metrum Acoustics Amethyst DAC Measurements

Discussion in 'Source Measurements' started by Hands, Jun 20, 2017.

  1. 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:
    Equipment Used: Soundaware D100 digital transport BNC out -> Metrum Acoustics Amethyst Coax SPDIF in -> Mixture of Creative SB1240 and Focusrite 2i2 (first gen) -> Surface Pro USB

    Note on new, experimental methodology: I wanted to experiment with a new measurement technique, so to speak. I noticed the SB1240 and 2i2 excel in different areas relative to each other, which is to be expected. One will have noise in a lower spot than the other, and vice versa. They have their own inherent harmonic distortion characteristics. So on and so on. Sometimes measurement results wouldn't be totally representative of the DAC alone, but rather have a few ADC characteristics thrown in. That's the downside of cheap (but surprisingly good for what they are!) ADCs.

    So, for many of these tests, I did some Excel work and more or less combined the results in such a way to minimize the individual characteristics of the ADC. However, in the case of the jitter tests, the results were very similar between the two ADCs, but just dissimilar enough that combining them basically cancelled everything out and made the results look clearly unrealistically good. In that case, I defaulted to the SB1240 rather than try to combine them. Everything else, though, I think is fairly indicative of the Amethyst's performance. (Sine wave output in ARTA handled by SB1240.)

    1KHz, -1dB, 16/48 - My rough calculations put THD at around 0.0077% up through 20KHz. This is similar to the Menuet. I briefly squeezed in a quick look at the Pavane during a mini-meet and saw it closer to or slightly below 0.005%. Believe it or not, I never properly measured the Adagio!

    Metrum Amethyst 1K -1dB.png

    Here's a look at the same results as above, but focused on 0Hz-1KHz. Very clean. This was about as low as my ADCs would go. (The SB1240 doesn't have the lowest low-end noise floor, and the 2i2 had a higher noise floor since I had to raise the gain due to single ended output on the Amethyst. With a balanced DAC into 2i2, I can get a lower noise floor.)

    Metrum Amethyst 1K -1dB Low End Noise.png

    1KHz, -3dB, 16/48 - From here on out, I thought it might be interesting to see how the Amethyst handled this test as I started dropping the level of the 1KHz sine. Very minor changes. THD roughly calculated at 0.0081%.

    Metrum Amethyst 1K -3dB.png


    1KHz, -10dB, 16/48 - THD roughly calculated at 0.013%.

    Metrum Amethyst 1K -10dB.png


    1KHz, -60dB, 16/48 - Less interesting than the sine output below.

    Metrum Amethyst 1K -60dB.png


    1KHz, -90.31dB, 16/48 - Refer to sine output below.

    Metrum Amethyst 1K -90dB.png


    Here's a look at the -60dB sine wave. You can some minor "wiggling" at the top and bottom of the waves. I have not checked this on many other DACs before to see if this is a consistent behavior, but I was mostly curious to see if this repeated itself with a 24-bit signal. The Bifrost Multibit actually had a similar and slightly more noticeable behavior on its -60dB sine wave, but I never published those results. I don't know enough about this, but perhaps it's quantization stuff like with the -90dB test?

    Metrum Amethyst 1K -60dB 16Bit Sine.png

    The 16-bit, -90.31dB, 1KHz signal shows the 3-step pattern we would hope to see in this case. This is a good looking result given the limitations of my ADC.

    Metrum Amethyst 1K -90dB 16Bit Sine.png
     
    Last edited: Jun 20, 2017
  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:
    1KHz, -1dB, 24/48 - I was curious if a 24-bit 1KHz sine at -1dB would change these results. Maybe a few tiny changes, but very steady.

    Metrum Amethyst 1K -1dB 24Bit.png

    1KHz, -60dB, 24/48 - The Amethyst definitely seems to be responding to a 24-bit input with the lower level 1KHz sines.

    Metrum Amethyst 1K -60dB 24Bit.png

    1KHz, -90.31dB, 24/48 - More curious to see how the wave output actually looks.

    Metrum Amethyst 1K -90dB 24Bit.png

    The -60dB wave output at 24-bits does look better and more consistent.

    Metrum Amethyst 1K -60dB 24Bit Sine.png

    The Amethyst resolves the 24-bit, -90.31dB, 1KHz sine wave fairly well. For any noisy portions, my ADC will have some limitations here. I can't rule out some of the "noise" being due to the Amethyst itself.

    Metrum Amethyst 1K -90dB 24Bit Sine.png

    50Hz, -3dB, 16/48 - Harmonics are all fairly low. Rough THD calculation puts it at around 0.0042% up until 1KHz, for what it's worth.

    Metrum Amethyst 50Hz -3dB.png

    19KHz, -3dB, 16/48 - Not many surprises here for a Metrum, non-oversampling DAC. (Note that most -3dB tests I ran were due to the -1dB or higher test clipping my ADC.)

    Metrum Amethyst 19K -3dB.png


    21KHz, -3dB, 16/48 - Again, about what you'd expect.

    Metrum Amethyst 21K -3dB.png
     
    Last edited: Jun 20, 2017
  3. 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:
    250Hz and 8KHz, 4:1 Ratio, -1dB, 16/48

    Metrum Amethyst 250Hz 8K -1dB.png

    13KHz and 14KHz, -3dB, 16/48

    Metrum Amethyst 13K 14K -3dB.png

    19KHz and 20KHz, -3dB, 16/48

    Metrum Amethyst 19K 20K -3dB.png

    Jitter (JTest) results for 16-bit and 24-bit. The spurs we'd expect to see are at a low level. However, there are other spurs that ideally would not be there. Thankfully, everything is low in level. I'm not sure what this would ultimately mean for jitter performance on the Amethyst, but I don't think it's cause for much alarm. (This is via coax SPDIF.)

    Metrum Amethyst 16Bit Jitter.png Metrum Amethyst 24Bit Jitter.png

    With USB, jitter results were...interesting. I tried a few different computers and setups. It always tended to look like this.

    Metrum Amethyst 24Bit Jitter with USB.png
     
    Last edited: Jun 20, 2017
  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:
    Overall, there's not much to be surprised by here given my earlier tests of the Menuet as well as other non-oversampling DACs I've tested. Not a whole lot to say, so let me poop out some obvious words to wrap this up. No, a non-oversampling, and especially Metrum DAC, will never match the squeaky clean look of a really good oversampling DAC, but these results are fairly good given what the Amethyst is. It's nice to see Metrum has solidified the performance gains they've made with their latest generation of DACs, albeit I was a bit more bored since this wasn't the measurement-shitstorm that was the S19.
     
    Last edited: Jun 20, 2017
  5. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Supplemental measurements

    1kHz -1 dbFS 16/48
    Amethyst 1kHz.png

    1kHz -60 dbFS 16/48
    Amethyst 1kHz -60db.png

    1kHz -90 dbFS 16/48
    Amethyst 1kHz -90db sine.png
     
    Last edited: Jul 25, 2017
  6. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    -90db 1kHz Sine 16/48 (don't trust voltage values - it's not calibrated)
    Amethyst 1kHz -90db sine.png

    -60db 1kHz Sine 16/48 (don't trust voltage values - it's not calibrated)
    Amethyst 1kHz -60db sine.png
     
    Last edited: Jul 25, 2017
  7. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    -90db 1kHz Sine 24/48 (don't trust voltage values - it's not calibrated)
    Amethyst 1kHz -90db sine 24.png
     
  8. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    19kHz and 20kHz two tone
    Amethyst 19 and 20kHz.png

    19.kHz
    Amethyst 19.1kHz.png
     
    Last edited: Jul 25, 2017
  9. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Amethyst white noise 44.1kHz
    Amethyst white noise 44.1kHz.png
     
  10. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Amethyst Triangle Wave Analysis 6 LSBs up and down. 100 cycles laid on top of each other.
    Amethyst 65LSBs Triangle Analysis.png

    Gungnir Multibit for comparison
    Gumby 65LSBs Triangle Analysis.png

    And Audio-GD S19
    AGD 65LSBs Triangle Analysis.png
     
  11. Luckbad

    Luckbad Traded in a unicorn for a Corolla

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    3,408
    Trophy Points:
    113
    Location:
    Holly Springs, NC
    As a former Audio-GD fanboy, may I just say... :(

    I still really liked the Master-11 until I started hearing good tubes amps and proper NOS DACs, but the Singularity is singularly disappointing.
     
  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:
    The S19 results make me laugh every time.

    I take it the Amethyst has a glitch at the tip of the triangle? I want to say those results look fairly decent otherwise? Certainly not the mess than is the S19.
     
  13. Priidik

    Priidik MOT: Estelon

    Pyrate BWC
    Joined:
    Sep 27, 2015
    Likes Received:
    2,153
    Trophy Points:
    113
    Location:
    Estonia
    Doesn't look like sinc function to me. Analog filtering?
     
  14. Grahad2

    Grahad2 Red eyes from too much anime

    Pyrate
    Joined:
    Jan 26, 2017
    Likes Received:
    1,162
    Trophy Points:
    113
    Location:
    Singapore
    I suspect it's the NOS at play, since Marv uses 96kHz files. Don't have enough frequency up top to sharpen the triangle. (I believe supermegaburrito upsamples to 384kHz for 96kHz input).
     
  15. 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:
    So, Metrum's approach to 24-bit processing, as mentioned, is to split a 24-bit signal in half (and maybe 16-bit signals too? I'm not sure), process those halves separately on the more accurate part of the DACs, since each module has two DACs under the hood effectively, and then attenuate + combine the lower level signals with the other half. Something like that. Is that just a wordy way of describing already existing DAC architectures, or is that a more unique approach?

    Either way, I am curious if there are tests that might be able to uncover how well this actually works on the new Metrum stuff, if it's implemented well, if it has downsides, and how close they actually get to resolving 16 and 24-bit signals respectively.

    I mean, it's at least doing something with a 24-bit signal. The old Metrums I don't recall being able to do that with a -90dB 24-bit signal.
     
  16. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Tip at the triangle could be from aliasing or filter. If you think about it, it's like an impulse response at the tips. I wouldn't worry about it too much.

    I'll run more routines to better differentiate the results - but the differences are definitely noticeable here and do correlate what we hear.

    It looks like it's getting 18 bits.
     
    Last edited: Jul 28, 2017
  17. 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:
    18 bits would be a respectable result. Big step up from last gen!

    Shouldn't the white noise look more like this, being NOS and all? I thought Metrum only used a very gentle analog filter.

    [​IMG]
     
  18. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    More geekery. 6LSB triangle wave crawl. Waveforms successively overlaid each other using a different random color for each wave.

    AGD S19
    AGD S19.png
    LOL, I don't think the S19 ever plays successive waves the same way.

    Amethyst
    Amy.png
    I love how repeatable the R2R ladder is.

    Gungnir Multibit (balanced out)
    Gumby.png
     
    Last edited: Aug 9, 2017
  19. Luckbad

    Luckbad Traded in a unicorn for a Corolla

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    3,408
    Trophy Points:
    113
    Location:
    Holly Springs, NC
    The S19 waveform is an artistic impression of euphony. Unfortunately, they meant to choose Joseph Karl Stieler to paint Ludwig van Beethoven, but actually chose Picasso.
     
  20. 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:
    Damn, Gungnir Multibit is looking fine! That S19 though...

    Amethyst seems OK. Glad their new DACs aren't total shit. Curious to see how the Pavane or Adagio do with all their paralleled modules.
     

Share This Page