Gungnir Multibit wants to kill Kattefjaes in his sleep

Discussion in 'Audio Science' started by Kattefjaes, Feb 24, 2018.

  1. Kattefjaes

    Kattefjaes Mostly Harmless

    Pyrate
    Joined:
    Sep 5, 2016
    Likes Received:
    4,521
    Trophy Points:
    113
    Location:
    London, UK
    This isn't really a thread that starts with informative measurements. It's more of a WTF thread- something weird is happening that I can't explain. People with more clue than me may be able to shed some light. Maybe.

    Some background.

    I wanted to characterise the very audible difference between two e-stat amps, driving a pair of SR-303. One of them sounded a lot less "thin" than the other, so I broke out my gonzo measuring rig* and did some measurements. The results were weird- very uneven.

    I swapped the other amp in, and the results were similar- still very uneven. Like seriously uneven..

    [​IMG]

    This is what I'd been getting all along. I tried controlling for various factors, changing headphones, amps and finally source. The common factor was that whenever I used the Gungnir Multibit as a source, whatever it was feeding, balanced or SE, I was getting weird voids in the FR. If I used a different source, the FR looked normal.

    Here's a set of measures that show what's going on:

    [​IMG]

    You can see that when I used the 2i2 (Focusrite Scarlett 2i2) as a source, the magenta and yellow traces, they don't have that pronounced set of voids (or possibly spikes). The yellow is the 2i2 driving Sennheiser HD-25 directly, and the magenta is the 2i2 into a Stax SRM-313, driving SR-303.

    The input device was also the 2i2, but the "direct monitor" was not enabled.

    When any other combo is used including

    Gungnir Multibit->Jot->HD25,
    Gungnir Multibit->SRM-313->SR-303
    Gungnir Multibit(bal)->Octave 2->SR-303

    (Measurements done in daytime, lots of domestic noise and it's hard to get a seal with Lamdas, so if it looks a bit ragged, just ignore that- it's the big trends we're looking at.)

    ...I get those weird results. The only factor that seems to need to be present to trigger them is using the Gungnir Multibit. It looks pretty bad in the measurements, but I can't really hear this behavior. I just generated a 1kHz to 2kHz sine sweep (which covers the most drastic void on the graph) and can't hear any real inconsistencies outside of the usual perceptual frequency/loudness stuff. I tried a sweeps from 30 seconds long down to 5, and couldn't pick anything much out in any of them.

    So, my question is- what's causing this artifact? I only see it when I use the Gungnir Multibit as a source. Something odd there.




    * Just a cheap home-made thing, built to OJNeg's old circuit design. I don't think specifics of its performance matter here, as it produces waveforms that look the way that you'd expect, normally, and doesn't seem to be busted. It's consistent when compared to itself.
     
  2. aamefford

    aamefford Nothing like chamberpot coffee

    Pyrate
    Joined:
    Nov 26, 2015
    Likes Received:
    1,092
    Trophy Points:
    93
    Location:
    Between SF and Sac
    Any chance you could try another Gungnir Multibit as another control? I know getting one’s hands on a second Gungnir Multibit may not be trivial though.
     
  3. Azimuth

    Azimuth FKA rtaylor76, Friend

    Pyrate
    Joined:
    Nov 9, 2015
    Likes Received:
    6,838
    Trophy Points:
    113
    Location:
    Nashville, TN
    Home Page:
    That looks like some comb filtering of some kind.

    Something certainly going on with the Gungnir Multibit.

    Have you tested just the raw output?
     
  4. Madra

    Madra Acquaintance

    Joined:
    Oct 18, 2015
    Likes Received:
    26
    Trophy Points:
    18
    Totally OT, @Kattefjaes , would you care sharing your thoughts on the Octave V2 on the Stax thread?
     
  5. Kattefjaes

    Kattefjaes Mostly Harmless

    Pyrate
    Joined:
    Sep 5, 2016
    Likes Received:
    4,521
    Trophy Points:
    113
    Location:
    London, UK
    I still haven't worked exactly what I think, actually, it usually takes me a while to get my head around an amp. You may have to bear with me while I get used to it- maybe trying a couple of different transducers with it too. I'd rather get my thoughts slightly coherent than produce horrifying giant bullet point listicles with the SNR of a detuned AM radio.

    I have a prototype unit in a bigger box, with uprated power caps, so it might be marginally different to the normal production model. However, it's probably not all that different.

    I can give you a spoiler though- it doesn't suck. This is important to know.
     
    Last edited: Feb 24, 2018
  6. numbercube

    numbercube Acquaintance

    Joined:
    Sep 29, 2015
    Likes Received:
    98
    Trophy Points:
    18
    Location:
    Germany
    Java or ASIO drivers?
     
  7. Madra

    Madra Acquaintance

    Joined:
    Oct 18, 2015
    Likes Received:
    26
    Trophy Points:
    18
    Trying to make your spoiler just that little bit more informative, do you consider that Stax amps (323, 353x, 717, srm-t1) suck?
    That said, will be looking forward for your Octave impressions, whenever you feel ready to share.
     
  8. Kattefjaes

    Kattefjaes Mostly Harmless

    Pyrate
    Joined:
    Sep 5, 2016
    Likes Received:
    4,521
    Trophy Points:
    113
    Location:
    London, UK
    The Gungnir Multibit does not have ASIO drivers.
     
  9. Cspirou

    Cspirou They call me Sparky

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    8,200
    Trophy Points:
    113
    Location:
    Northwest France
    You know what frequencies the nulls are occurring?

    To me it looks like 1500Hz, 4500Hz, 7500Hz.....
     
  10. Kattefjaes

    Kattefjaes Mostly Harmless

    Pyrate
    Joined:
    Sep 5, 2016
    Likes Received:
    4,521
    Trophy Points:
    113
    Location:
    London, UK
    More or less, yeah.

    1469
    4404
    7341
    (10282 is probably an outlier, as it exsts with the 2i2 as source)
    13205

    Almost like they're harmonically-related?
     
    Last edited: Feb 24, 2018
  11. TwoEars

    TwoEars Friend

    Pyrate
    Joined:
    Feb 2, 2016
    Likes Received:
    2,006
    Trophy Points:
    113
    We could try this:

    @schiit @baldr

    Maybe they've seen something similar in the lab at some point, could be a filter or some such.
     
  12. Cspirou

    Cspirou They call me Sparky

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    8,200
    Trophy Points:
    113
    Location:
    Northwest France
    Maybe a crystal went haywire or something.

    But yeah, would really need measurements of the Gungnir MB output to see what's going on. Can you use the Scarlett 2i2 input to measure it?
     
  13. Kattefjaes

    Kattefjaes Mostly Harmless

    Pyrate
    Joined:
    Sep 5, 2016
    Likes Received:
    4,521
    Trophy Points:
    113
    Location:
    London, UK
    Yeah, you're right, it is the logical next step.

    I'll only be able to do SE, though, as I don't have any XLR to TRS cables handy - the 2i2 engages the preamps when you plug XLRs into the combo sockets on the front. However, that's no bother, as the weird voids were visible via the SE out of the Gungnir Multibit too.

    I'll have a go when I get a moment. First order of business is to go to the supermarket, however!
     
    Last edited: Feb 24, 2018
  14. gepardcv

    gepardcv Almost "Made"

    Contributor
    Joined:
    Nov 20, 2015
    Likes Received:
    408
    Trophy Points:
    63
    Location:
    Terra, Sol System
    Not sure if this is relevant, but is your Gungnir Multibit upgraded to gen5 USB and what input are you using?
     
  15. Kattefjaes

    Kattefjaes Mostly Harmless

    Pyrate
    Joined:
    Sep 5, 2016
    Likes Received:
    4,521
    Trophy Points:
    113
    Location:
    London, UK
    OK. here's the Gungnir Multibit straight into the 2i2, single ended, with a Modi Multibit via USB added for comparison:

    [​IMG]

    It's weird stuff, wonder what I'm doing wrong?

    @gepardcv - For the record, using Gen 5 USB on the Gungnir Multibit. May see if I can find a DDC kicking about to try it over S/PDIF.
     
    Last edited: Feb 24, 2018
  16. Cspirou

    Cspirou They call me Sparky

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    8,200
    Trophy Points:
    113
    Location:
    Northwest France
    Modi Multibit too! I assume you are using USB with both. Do you get the same issue when changing computers?
     
  17. Kattefjaes

    Kattefjaes Mostly Harmless

    Pyrate
    Joined:
    Sep 5, 2016
    Likes Received:
    4,521
    Trophy Points:
    113
    Location:
    London, UK
    I dunno, charging my ancient laptop, but if we open the other bottle of red, I will become an unreliable narrator.

    However, I'm starting to think it's something in the 2i2 - maybe some weird mismatch causing beating or something? I grabbed the USB sound module for a gaming headset (tiny, cheap, simple) and it looked like this into the 2i2:

    [​IMG]

    That does sort of imply that it may not be at all the fault of Them Schiitheads(tm). I will try another computer at some point. Also, I have a DDC somewhere, which I can try with the Gungnir Multibit/Modi Multibit, with a TOSlink cable of reasonable quality.

    Edit:

    Just remembered this.

    One other data point is that I think I saw similar results with my old UA-25 (USB sound card), which is getting too cranky to use. I was getting the strange voids with it, but it was becoming wonky with USB- hence picking up the 2i2.

    If it is some bullshit processing in the Windows audio subsystem, the laptop will be a useful test. We'll see what I get on there under Linux at some point- probably not tonight, though.
     
    Last edited: Feb 24, 2018
  18. Azimuth

    Azimuth FKA rtaylor76, Friend

    Pyrate
    Joined:
    Nov 9, 2015
    Likes Received:
    6,838
    Trophy Points:
    113
    Location:
    Nashville, TN
    Home Page:
    Hey, you don't have the Windows sound coming back in do you? This looks like comb filtering which if a slightly delayed signal is mixed in, it would cause this kind of filtering. I think hardware latency on those audio recording interfaces is about 10-12ms on USB.

    Either that, or some kind of mix back. Don't k,ow if you are using ASIO or WDM, but I would check all of that.
     
  19. Kattefjaes

    Kattefjaes Mostly Harmless

    Pyrate
    Joined:
    Sep 5, 2016
    Likes Received:
    4,521
    Trophy Points:
    113
    Location:
    London, UK
    I don't believe so, no. It does look a lot like comb filtering or beating, though- as it's highly periodic in nature. Will do more investigation when I am more awake.
     
  20. Biodegraded

    Biodegraded Friend

    Pyrate Contributor
    Joined:
    May 28, 2017
    Likes Received:
    7,985
    Trophy Points:
    113
    Location:
    Vancouver BC
    What sample rate is feeding the DAC? REW samples at either 44.1 or 48k, and if the input signal is higher the measurements start to look like this - see the REW help for the SB Live (last item on the help topics list).

    @JohnM might be able to shed some light.
     

Share This Page