Schiit Bifrost 2

Yes, it's the same.

 

No, I haven't updated the FW yet. I thought I'd see if I could replicate it by putting those sweeps on repeat, not so much for the audio content but because they cycle through 4 different sample rates in short succession.

Correction on the audio sweeps: they sweep from 1 Hz (not 1 kHz) to the Nyquist. Not that it makes much difference for our purposes.

@JustAnotherRando asked me if the artifacts are "click" noises. I think there might be some clicks and pops, but they're probably artifacts of the recording process caused by lag in my setup (this laptop my work gave me sucks). The aliasing sounds like a (quiet) siren sweeping up and down. Here is a quick and dirty visualiztion using foobar2k's spectrum display. (I don't have any fancy screen video capture software installed so I just used my phone. And please turn down the volume or ignore the background noises. There's kitchen rattling and burping after take out from McDs, cbf'd to figure out how to mute the audio in YT.)

This is the file 04. 88.2 kHz sweep 1 kHz to 44.1 kHz being played (1 sweep) with no aliasing.


This is the file 01. 88.2 kHz sweep 1 kHz to 44.1 kHz.flac being played (2 sweeps) with aliasing that you can see sweeping back and forth in frequency.



Edit 1:

Ambient noise levels in my place are fairly high and I'm listening with open phones. Add to that the typhoon and that's probably why I don't hear much post power cycle. The noise floor was low to mid -50 dB according to Audacity's sound meter.

Edit 2:
I trimmed the 176 kHz post reboot file, leaving the tail end of the audible portion of the actual tone sweep, then amplified it by 30 dB. Yes, you can still hear some aliasing.
https://drive.google.com/open?id=1G-EJH5zKVtxky5ChtpwNqZWtQN5ZMlZd

Edit 3:
It still sounds great, w/ or w/o aliasing ...

 

I spent some time again this weekend with the Bifrost 2, and was able to replicate the issue @gixxerwimp identified.

To respond to @atomicbob's question, this is a separate and different issue than the audible IMD spike (which @atomicbob measured in the other thread -- thanks for posting your measurements, BTW).

Here's a summary based on my experience:

Issue #1: The aliasing noted by @gixxerwimp

Just for ease of reference, I'll call this "degraded performance" mode. When the Bifrost 2 enters this mode, there are a variety of test signals that highlight the problem. A single 19kHz or 20kHz sine actually highlights the issue very audibly -- you don't need a sweep to hear it. I can't figure out exactly what causes the Bifrost 2 to enter degraded performance mode, but switching sample rates seems to cause it sometimes. It's not directly related to Unison/USB issues; I was able to replicate this by feeding the Bifrost 2 a Toslink signal as well. The Bifrost 2 returns to normal after power cycling.

The thing about degraded performance mode is that listening to music, it's not immediately obvious that something is wrong. Test signals highlight the issue pretty clearly, but for actual music, depending on what music you're playing, you might miss it. Classical flute does highlight the issue pretty clearly though. When in degraded performance mode, flute concertos have fairly obvious odd overtones that make the presentation hazy. Hopefully this is fixable with a firmware update.

Issue #2: The audible intermodulation distortion

This is a more subtle issue, but you can hear it clearly (although at a much lower level) with relatively simple test signals. I think this is largely the cause of the euphonic presentation I commented on earlier in this thread. I'm not sure if this is fixable via firmware; it might just be how the Bifrost 2 is designed. I think it does contribute to the greyer background than the Modi Multibit that I noted in earlier comments, and it's probably something to consider if you're primarily a classical or jazz listener and trying to decide between this and the Modi Multibit.

 

@AudioNut just to confirm are you using the new Bifrost 2 C0088, D0105 firmware released 9/26/19? Thanks.

 

I seem to be able to reliable replicate this "mode" by repeating a playlist of all my test sweeps which have different sample rates (44.1 kHz, 88.2 kHz, 96 kHz, 176.4 kHz, 192 kHz). I wasn't listening the whole time (for obvious reasons), but I came back half an hour later and it was doing it again.

I'm going up upgrade the firmware now and see if that changes anything.

 

No -- I'm using the original firmware. I'd definitely like to run tests with the new firmware, but I'm still just within the RMA window (I ordered it on the day it was announced, but it was held up in customs in Montreal for more than a week, so I received it later than most people) and I took it out of my system this morning to get it ready for a return, so I won't have a chance to try the new firmware.

 

As has my Bifrost 2, which has been on my desk since the day of the launch, and I use it every day. Also, according to our records, as has about 93% of all Bifrost 2s.

However, the difference between 93% and 99.5% (a typical launch) is night and day. We’ve had a capacitor issue (now fixed for all shipping units as of this week) and we’ve had firmware issues (also now fixed with all shipping units and available for download on the site, and going out to all owners this week.)

These problems are despite an extensive, one-year-long beta program (some people here have had plastic Bifrost 2s, they can vouch for that). The beta program didn’t identify any issue with the firmware. And no beta program will find bad components.

So, if you’re having problems, install the new firmware, or wait for the SD cards. I have confirmed that the new firmware fixes the remote issue, USB lockout, and the switching glitch (fuzziness or distortion).

So, again: install the new firmware, or wait for cards, and then let us know how it goes.

 

Confirmed it will fix it. See above.

 

@gixxerwimp, gotta let you know... as soon as it happened, the very first thing I thought of was your avatar...

 

So this is the only "issue" that remains, right? If only audible with test signals it's no biggie?

 

I'm not going to wade into a debate about what's audible or not for an average person for music -- I agree with the "don't be a measurement fascist" post on this forum. However, I think there's a consensus from the impressions posted that this is a fairly euphonic sounding DAC, and to my ears (don't crucify me for posting my impressions) even when it's not misbehaving it's not as convincing with acoustic instruments as the Modi Multibit. Whether that's related to the relatively high levels of IMD that @atomicbob measured (-64dB in one channel; -62dB in the other channel in the PDF he posted for the 19+20kHz FFT balanced, relative to the stimulus signals not FS), who knows.

 

@schiit What about your distributors? Do they still carry potentially defective Bifrost 2 units, or is it safe to order from them now? I am especially hinting at the malfunctioning capacitors...

 

Speculating here, but it seems like Bitfrost 2 has a slow, "leaky" filtering scheme, and when music has energy at the corner frequency (e.g. you flute concerto) you notice it.

edit: if so, this may or may not have much to do with the general "euphonic" character of the DAC

 

Nothing got into distribution, but everything will need the firmware updated.

 

How can one know if their unit is one with bad caps?

 

It dies and does not work at all. It's pretty obvious.

 

Just a bit of correction. That was -64 dBu and -62 dBu in th other channel. The IMD values reported in my measurements were dBu relative to +11.29 dBu signal generator stimulus which translates to approximately -75 and -73 dBFS.

In retrospect I probably shouldn't have changed y-axis formats in my new measurement format as I see it may be confusing. Traditionally these are reported dBFS on y-axis. I thought having dBu would assist those wishing to relate dBu to SPL using sensitivity of headphones via dB/mW. I am reconsidering returning to traditional dBFS instead of dBu for distortion y-axis graphs.

 

There's nothing to correct. I made this clear in my post... you even quoted it: "relative to the stimulus signals not FS".

If you're not testing with full scale signals, it doesn't make sense to do what you're suggesting and revise the distortion numbers down to be relative to full scale. You might as well re-run your test with a -80dBFS stimulus then revise the distortion numbers down to -144 and -142 dBFS. That's what you're saying. (Numbers chosen for rhetorical purposes only -- obviously the distortion components would be below the noise floor.)

 

Ok, I guess I didn't state that clearly enough.

Bifrost 2 balanced 0 dBFS (FULL SCALE OUTPUT) = +14.35 dBu.
Twin tones must be at -6dB relative to the full scale to avoid clipping when they are summed. The RMS value for a 19KHz + 20 KHz tone pair, each at -6 dBFS peak from +14.35 dBu will be +11.29 dBu. This is the ABSOLUTE MAXIMUM before clilpping. The signal generator is running FULL SCALE OUTPUT in the IMD tests reported.

To reiterate, for a DAC with 0dBFS = +14.35 dBu
+11.29 dBu = maximum output for two tones summed before clipping

The tests published were running with maximum output before clipping but had the y-axis report in absolute dBu and the difference between the dBu reported and signal generator stimulus which is adjusted for maximum before DAC clipping must be considered to achieve dBFS.

 

Here is the IMD test run with y-axis dBFS instead of dBu relative to signal generator stimulus at maximum output:

Bifrost-2 IMD CCIF dBFS 19+20KHz spdif input balanced output:



This graph has been annotated for clarity. Note:
Signal generator set two two tones, each at -6.03 dBFS to avoid clipping during summation peaks.
Analyzer level reported at 11.29 dBu RMS approximately 2 dBu below 14.35 maximum output for a single sine at 0dBFS
IMD CCIF today reads -77 dBFS in left channel.

11.29 dBu - 77 dB = -65.71 dBu