Gungnir Multibit wants to kill Kattefjaes in his sleep

That is caused by some form of monitoring being active, the mixing of the signal and the delayed copy creates that characteristic comb filtering. Make sure 'Listen to this device' is not selected in the Windows Audio Properties.

 

Hey there, nice of you to chime in! Thanks for REW, it's handy software

I considered that, but it happens under Linux on another machine too, annoyingly.

I ran through a series of tests (all at 48kHz unless specified otherwise) -

• 2i2 into UA-25 - fine
  
• UA-25 into 2i2 - fine
• Gungnir Multibit into UA-25 - wiggly
• Fulla 2 into UA-25 - wiggly
• Fulla 2 into UA-25 at 44.1 (ignoring that the input cal was at 48kHz, just for a crude confirmation) - wiggly

This happened reliably. The "wiggliness" came and went predictably depending on the sound device. I dd also try other combos, including into the 2i2 with controls in various positions.



The REW data:

https://www.dropbox.com/s/sg45av2z06bko57/dac test.mdat.gz?dl=0 for the file

So, it's periodic, like comb filtering or beating - but I'm really not sure that there's any software monitoring going on, with either O/S. That said, the layers of cruft on top of ALSA on your average Ubuntu install (which I have to use for Reasons) makes that far from clear. Monitoring certainly was disabled on the hardware controls of each soundcard.

Also, everything fine using the UA-25 or the 2i2 as a source, too. Now they do have the capability of being an input device too, maybe that get treated differently. However, that's too convenient a hypothesis says sod's law.

At this stage, it boils down to "it's really not a problem at all, just a weird puzzle". I can always use either/both soundcards for input and/or output (though I will probably want to make/buy some TRS to XLR leads for one amp). However, it'd be cool to know what's causing this, so I could be lazy and use my nice DAC when measuring headphones- plus puzzles are there for unraveling.

(I would try under a somewhat ancient MacOS too, with a less cluttered sound control panel, but last I checked, the obvious MacOS JVM options didn't do so well with sound inputs.. was that ever improved?)

 

1. Yeah, I'm not worrying about the the bass much- getting a seal with Lambdas is always a total chore. At peace with that for now- can always do something ugly later if I need to get sane behavior.
   
2. The sample rates have been consistently 48kHz across the board for the sake of simplicity- except for when I deliberately changed them to see if the behavior changed (e.g. if something wasn't managing to set the sample rates and failing silently).
3. Ah, that's a really good point. Would the humble Fulla 2 have that issue too- or is the filtering something off the sheet (and lower latency) for the 4490?
4. Good point. One day I should try ARTA. Comfort zones and all that. Do we have a good crash guide to it around here, for the lazy?

If it is latency due to filters, that would make sense. Too bad I don't have any Chord DACs to try any more, they were latency monsters, from what I recall- would have been interesting.

As above, it's really not a problem, no-one should lose sleep over it. This is just me trying to work out what's going on. When I need sane results, either or both of my USB sound cards will work just fine. However, I do appreciate the clues.

 

You are making measurements with 2 sweeps. Multiple sweeps requires precise sample clock alignment, you can't usually achieve that when the input and output device are different. Use 1 sweep.

 

Today I Learned(tm). Thanks! I shall heroically make all the dumbass mistakes so that no-one else has to. Mystery solved, I suspect.

Edit:

Yes, you nailed it, thanks a lot. Round-trip between DAC and soundcard is flat as a pancake:



That's really cool, means I can start poking around answering my original question.

 

Is this where having the external clocks and the squid's worth of tentacle cables to every single component becomes relevant?

 

That was an interesting and educational thread. I'm glad things turned out to be alright. I did like the original thread title better:
My Gungnir Multibit Wants to Kill Your Mama
(Pardon the particularly horrible YouTube sound.)