Schiit (MultiBit) Bifrost

you rock Tuco1965 and Im4der!

 

So Multifrost then?

 

That definitely has my vote!

 

My OCD is driving me crazy as I watch the tracking of my Uberfrost's return to Schiit for its brain transplant.

 

Ah! I am so happy for you! I cannot wait until I can afford my multi-bit upgrade to transform my Uberfrost into a Multifrost!

 

I'm down with Multifrost and Multibifrost.

 

One of the interesting things about Bifrost Multibit is its direct input mode, which would allow you to compare different interpolation filters on the same device. For example, you could test the Burrito 96 kHz filter against UB's fir1 or the Cheby sinc, with its corner frequency set at 96 kHz. Anyone up for the challenge? I'm don't know how one goes from a filter design in Matlab into one that you can run in realtime. I guess you can prefilter a file, too.

 

Yep, I was considering doing this. Don't have time, though.

I have the Gungnir Multibit, though. Don't remember at what FS it is NOS. Is it starting 176 or 352?

 

Cute naming convention rules state you use the first two letters of the product name, add mb for multibit and add a y or something else that makes it sound cute. Maybe an "o"

Gungnir = gumb+y or gumb+o
Bifrost = bimb+ y or bimb+o
Modi = momb +y or momb+o *** not trying to start a rumor, just thinking of the future

The cute naming convention rule really works

 

Thanks UB. Some quick searching turned up stuff like this: http://convolver.sourceforge.net/ So I guess it's possible, and not that hard. Hmm .... I'm traveling for most of this month, otherwise I'd be down to try this. I've been looking for an excuse to get more Schiit.

BTW, I don't think Gungnir Multibit has a direct NOS mode. Bifrost Multibit has it for 176.4 and 192. I imagine the file would be upsampled to one of those two sample rates so Bifrost Multibit uses NOS mode, but a <48kHz lowpass filter would be applied to the signal, so the signal would be directly comparable to Bifrost Multibit's 96kHz comboburrito mode.

Of course, that introduces a variable that may make things complicated: Comboburrito's low-pass filter appears to be closer to half the sample rate than a quarter, but it's a start.

 

Convolver says it can act as a plug-in in JRiver, and appears to accept impulse responses in the form of .wav files. So you could generate the impulse response of a filter, and convolver can then use that to filter the input. Anyway, it appears this is all possible in a relatively straightforward way, so now someone just has to put it together and try it out.

 

To play with oversampling filters I just use sox to create separate files (or use sox as the player).

# Create a scaled reference track in higher bits 24/32, if oversampling clips
sox -V4 track09.wav -b24 track09_24.wav gain -1.5 dither

# 8x oversample example: the upsample effect does zero insertion, and the fir effect does the filter.
# the input to fir is just a text file with the coefficients printed out in a column(1.0 as max) --> 3.324854911850877e-14 /// 0.061162684994738131
sox -V4 track09_24.wav -b24 -r352800 track09_24_gauss8x.wav upsample 8 fir gauss8x.txt dither

 

Thanks guys. No rush. I'm not sure anyone with a Bifrost Multibit has volunteered to do this yet.

Maybe when Atomicbob's measurements on Bifrost Multibit come out we can start to guess at how fast they run the DACs. I'd guess it's all standard rates.

 

I feel like we ought to break out the dac engineering discussion into its own thread partly because it's super interesting and partly to stop crapping in Bambi's forest.