Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[kstuart]

This thread is mainly for future JRiver MediaCenter users with Schiit Bifrost USB DACs.  The settings probably work for Gungnir as well.

I just purchased a Schiit Bifrost DAC to upgrade from my HRT DAC.  This was due to very positive reviews of the two upgrades made to the Bifrost about six months ago - the new Uber Analog card and the new 2nd Generation USB card.  Okay back to settings:

Go to schiit.com/drivers and download the 1.03 drivers.  These are dated May 03, 2013 for XP, Win7 and Win8 and August 20, 2013 for Vista ( BTW, if you have drivers from earlier, you should probably update your drivers ).  Note that the Vista drivers are a separate zip and are only found on the /drivers web page.  Install as directed on that page.

In JRiver MC19, go to Tools->Options->Audio and the first line, Audio Device, select:

ASIO for C-Media USB Device

On the second line, click Device Settings and:

Make sure that "Use Large Hardware Buffers..." is not checked.
Make sure that "Device uses only most..." is not checked.

Under Buffering, in the drop-down, select:

Minimum Hardware Size

Click on Open Driver Panel, and there are two settings:

Bit-depth should be   32/32 bits
Latency should be            4 ms   (if you hear any pops or noise, increase this setting to the next level and try again, but 4 ms  seems sufficient so far)

On the third line,  DSP & Output Format:

Sample Rate ->  every line should be "No Change" except for the last line "Greater than 192..." should be set to 176,400 hz .   Do not up-sample any sample rate - the DAC is specifically designed to provide optimum sound quality by not changing the sample rate.  Schiit says:

Q: I heard about this upsampling stuff, where they take 16/44.1 and magically make it into 24/192. Does your DAC do that?
A: Not just no but hell no. None of our DACs will ever do sample rate conversion. Our goal is to perfectly reproduce the original music samples, not to throw them away and turn everything into a mystery-meat soufflé. Sample rate conversion destroys all the original samples. What goes in isn’t what comes out. We worked hard on a microprocessor-controlled, bit-perfect clock management system to ensure that all the original music samples going into Bifrost are delivered to the D/A converter, whether you’re running 16/44.1 or 24/192, rather than cheaping out and throwing in a sample-rate converter so we could use a single clock.

[kstuart]

Verified to work fine with MC20.

[Dil]

Thank you so much! I'm good to go now.

[Arindelle]

Hi just recommended the Uber version of the Bifrost or the Gungnir to a friend of mind.

couple of questions/remarks for you Kstuart

Sample Rate ->  every line should be "No Change" except for the last line "Greater than 192..." should be set to 176,400 hz .

why not over 192 set to 192khz? Their USB speak goes to the full 192,000hz spec - is there a reason for this.

Do not up-sample any sample rate - the DAC is specifically designed to provide optimum sound quality by not changing the sample rate.  Schiit says:

Yes I understand the point of view of the people at Schitt as this is essentially a NOS dac. They are pretty clear about it. But I have also read their comments about bit rates (just can't find it now or I'd link to it) and upsampling. Paraphrasing to the best of my recollection, it mentioned JRiver .. something to the effect that if you want to upsample, better to do it in JRiver as it in 64bit where the noise floor is so low as to not screw things up .. that way if you send the 192k upsample by JRiver (64bits) it would be processed bu their 32 bit dac as 192K. They admit no 24bit DAC is really 24bits (or the noise floor would be -144db), and theirs is missing a couple of 0s so its not really 32bit either. Wouldn't it be ok to just suggest to my friend to use his ears on this one, rather than it being categorically no upsampling at all?

Now, the guys at Schiit basically say that they put the USB in to be "buzzword compliant" and satisfy the demand. They seem to prefer SPDIF ... have you tried this? If so would a straight WASAPI set up be OK or do you have any recommendations for a special config?

Lastly, hate to ask questions about SQ, but do you find the sound stage pretty large? Are you happy with the DAC in general, any cons to be aware of ?

Thanks in advance

[kstuart]

Hi just recommended the Uber version of the Bifrost or the Gungnir to a friend of mind.

couple of questions/remarks for you Kstuart

why not over 192 set to 192khz? Their USB speak goes to the full 192,000hz spec - is there a reason for this.

Because every file available over 192 is a multiple of 88.2, i.e. some DSD-related audio file.  So, you want to use the 176.4khz setting in order to keep the calculations at integer multiples.

[blgentry]

Because every file available over 192 is a multiple of 88.2, i.e. some DSD-related audio file.  So, you want to use the 176.4khz setting in order to keep the calculations at integer multiples.

A.  You're responding to a comment made in 2014.

B.  Apparently, in digital audio, integer multiples just don't matter.  Because EVERY SAMPLE is recalculated when doing sample rate conversion.  At least that's what Hendrick schooled me on when I made a similar comment a few months ago.  I was aghast!  But apparently he's right.

This is actually one reason that Schiit's Mega-Combo-Burrito filter is so special:  Because it can preserve original samples while upsampling digital data!  That's why it took so many years and so much math and research to make it happen.  Because preserving samples is not a normal part of any other sample rate conversion algorithms.

Do I remember correctly that you own a Bifrost Multi-bit?  I'm more than a little jealous. 

Brian.

[kstuart]

I just wanted to add that I am still using the above with MC21, and with the upgrade to the outstanding Multibit version.

(While doing so, I saw the 2014 question for me, and answered it for future readers.)

(I can't take this further, because I have never studied DSP math, but pages like https://en.wikipedia.org/wiki/Decimation_%28signal_processing%29 state that downsampling by an integer multiple is just a matter of throwing out the extra samples, whereas by a non-integer factor means interpolation first.  As you mentioned in regards to Schiit's filter, SRC produces different results depending on the algorithm, however, simply throwing out every other sample should come out the same every time.)

In fact, I am now using ZoneSwitch to use the Schiit Bifrost Multibit DAC whenever Video playback has an audio track with 2 channels.

In order to do that, I found that I had to increase the Cmedia ASIO control panel setting from 4ms to 10ms to prevent dropouts or stuttering when using the Bifrost with Video files.  This means that it also needs to be 10ms for audio (since Cmedia only has one setting, it is not a JRiver internal setting that can be different per zone).

So, I tried a couple of my audio test tracks that I am very familiar with, and could not detect any audible difference between 4ms and 10ms latency.  So, I can recommend 10ms if you have any issues at all with any kind of dropouts or stuttering with the Cmedia ASIO driver for Windows that Schiit uses.

[blgentry]

(I can't take this further, because I have never studied DSP math, but pages like https://en.wikipedia.org/wiki/Decimation_%28signal_processing%29 state that downsampling by an integer multiple is just a matter of throwing out the extra samples, whereas by a non-integer factor means interpolation first.

That's what I thought too.  Hendrick corrected me here:

https://yabb.jriver.com/interact/index.php?topic=100716.msg699113#msg699113

I'm assuming he is correct based on what I've since read.

Brian.

[kstuart]

I found that I still have some dropouts and stuttering with some Video files when using 10ms latency.  That prompted me to do another quicker, simpler test with one part of one audio file, first using my original 4ms setting and then using 20ms, and this time I can definitely hear a difference.  (20ms loses some of the decay detail on acoustic guitar, as if you were resting a finger lightly on the string.)

So, I am going back to 4ms latency and abandoning using the Bifrost for Video (since I do not want to lower the audio quality - even slightly - to improve the playback of the compressed audio that is usually found in Video that I am playing.)

BTW, I found this comment by mastering engineer Barry Diament:

In the "old" days, sample rate conversion algorithms did less damage (injected fewer artifacts, such as the brightening and hardening many do today) when performing integer conversion (i.e., even multiples - halving or doubling the sample rate). This may still be true of the lesser algorithms today. However, the better algorithms are better at doing the math and to my ears, will deliver more transparent results even when performing non-integer conversion than lesser algorithms will when performing the easier, integer conversion. I find this particularly true of my long-time favorite, Alexey Lukin's 64-bit SRC from iZotope. The algorithm is smart and just doesn't care whether the conversion is integer or not. It can handle the math.