• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Measurement and Review of Schiit BiFrost Multibit DAC

Thomas savage

Liberation through innovation
Moderator
The Watchman
Patreon Donor
Joined
Feb 24, 2016
Messages
5,952
Likes
1,916
Location
uk, taunton
I'm nwavguy and so is my wife.

Eric
I don’t know anyone involved but my gut is telling me the toilet attendant at Harrods department store London who I saw a few weeks back is NwAvGuy..

Amir can’t be NwAvGuy , as NwAvGuy is funny , charismatic, clever and knows what he’s doing...
 

Soniclife

Addicted to Fun and Learning
Joined
Apr 13, 2017
Messages
586
Likes
199
Location
UK
I was going to get a DAC, but not if I can't surf the web and listen to music at the same time. Does everyone have a dedicated audio PC these days?
No, I use my aging I7 machine for everything. If something else is hammering the CPUs (rare) getting roon to do a simple upsample (about 3% max usage) can occasionally stutter, but usually doesn't, that's fair enough by me, computing resources aren't infinite, and it only happens when I've kicked something off that uses a lot of CPU. If I don't upsample CPU usage is very low. The PC is a non-issue.
 

Candlesticks

Active Member
Joined
Jan 19, 2018
Messages
108
Likes
125
@amirm

https://superbestaudiofriends.org/i...lies-and-usb-dac-technical-measurements.5770/

I think that post is directed at you.

From my reading of the post by AtomicBomb ;)

1. Differing settings used in the analyzer causes differences. One using WDM and the other using ASIO will cause differences. The same should be used for both. WDM should be avoided.

2. Full documentation is important.

I agree on point 2 because it's useful for when conflicting measurements arise and both people can look at each others settings beforehand. On point 1 it's just duh of course the settings have to be the exactly the same. I don't know if they were for yours amirm but you would know.
 
Last edited:

drconopoima

Member
Patreon Donor
Joined
Jan 14, 2018
Messages
49
Likes
27
@amirm

https://superbestaudiofriends.org/i...lies-and-usb-dac-technical-measurements.5770/

I think that post is directed at you.

From my reading of the post by AtomicBomb ;)

1. Differing settings used in the analyzer causes differences. One using WDM and the other using ASIO will cause differences. The same should be used for both. WDM should be avoided.

2. Full documentation is important.

I agree on point 2 because it's useful for when conflicting measurements arise and both people can look at each others settings beforehand. On point 1 it's just duh of course the settings have to be the exactly the same. I don't know if they were for yours amirm but you would know.
Well, I would assume Atomicbob's testing methodology doesn't vary, so, his results would be comparable against others among his own measurements:

http://www.superbestaudiofriends.or...infinity-technical-measurements-dac-only.223/

Considering that, what do we have here? It's 18-bit linearity and near 0 distortion on a -90 db sine wave test. So, there is something to worry about the Bimby that atomicBob measured.

Comparing against his technical and methodological section in the Bimby review, he used the same methodology in both: http://www.superbestaudiofriends.org/index.php?threads/schiit-bifrost-mb-technical-measurements.235/

So, that the Bimby measures badly is not a particular reviewer's manipulation.

http://www.superbestaudiofriends.org/index.php?threads/denafrips-ares-technical-measurements.4774/

What do we have here? A DAC with a similar price range of the Bifrost has 16-bit linearity and the -90dB sine wave appears sinusoidal (although, the LH Labs sine wave was better).

This is the same person that STARTED the Bifrost review with the conclusion that the numbers we were about to see below were excellent performance. Wow.

Forgot to include the -90 test I'm referring to https://imgur.com/a/TBs76
 
Last edited:

Candlesticks

Active Member
Joined
Jan 19, 2018
Messages
108
Likes
125
I'll point out that nobody has accused Atomicbob's measurements of being wrong, please provide quotes if I'm wrong, whereas multiple have for amirm.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
14,080
Likes
6,469
Location
Seattle Area
@amirm

https://superbestaudiofriends.org/i...lies-and-usb-dac-technical-measurements.5770/

I think that post is directed at you.

From my reading of the post by AtomicBomb ;)

1. Differing settings used in the analyzer causes differences. One using WDM and the other using ASIO will cause differences. The same should be used for both. WDM should be avoided.

2. Full documentation is important.

I agree on point 2 because it's useful for when conflicting measurements arise and both people can look at each others settings beforehand. On point 1 it's just duh of course the settings have to be the exactly the same. I don't know if they were for yours amirm but you would know.
Thanks for bringing that to my attention. It is a very good effort on their part to show how same equipment can measure differently. We also have our own thread here where we are trying to standardize measurements and source files so that even when using different equipment, results can be compared.

The points they make don't apply to my testing though. First, they should know that it was my team and initiative I took to implement the entire new audio stack in Windows when I was at Microsoft. The Windows XP kmixer had to be shot in the head and I made sure that happened in Windows Vista development :). Brand new pipeline was put in place and that is what is in use in Windows 10.

Importantly to the points they make, we implemented a new API called WASAPI for direct access to the sound driver bypassing the kernel processing. We wanted to support ASIO but that required licensing and it was cheaper and easier to build WASAPI instead.

What is wonderful in Creator Edition of Windows 10 is that full USB class drivers are now available. Windows not only recognizes all the DACs I am testing but automatically creates WASAPI interfaces for them. This eliminates the hassle of finding and installing ASIO drivers. Reliability has been superb although sometimes event driven does not work.

My testing is all done with Roon in exclusive mode using Wasapi. This is documented in every review I write although sounds like I should expand more on it for newer readers. This is what is in this review for example:

upload_2018-2-11_12-36-47.png


Roon only shows the above when WASAPI exclusive mode is used. So the issues they raise with WDM do not apply to my testing. Everything is bit-exact. No testing I do is with Windows processing.

There seems to be a tendency over there to keep assuming incompetence and lack of knowledge because they have not heard of me. All they have to do is click on my signature, do some googling, etc. and not make assumptions about me not knowing the difference between bit exact and not when my team implemented all of this.

On FFT points, my analyzers maxes out at 32K which is a blessing. Folks like Atomicbomb who use Prism Sound Dsound have access to far longer FFT kernels, resulting in a whole bunch of spikes that can be distracting. It also artificially lowers the noise floor that is seen in the graph, giving a highly misleading impression of how low the noise floor of the DAC really is. Even in my measurements there is a 32 dB "FFT gain" (artificial reduction in noise floor). Many people look at those graphs and assume the noise floor is what is shown when it is not at all the case. So if anyone needs to be cautious, is those folks with their large FFT kernels.

It is on my TODO list to write an article on my testing and expand significantly on the above.

Finally, I have done testing using Prism Sound that atomicbomb is using. Here is the article: https://audiosciencereview.com/forum/index.php?threads/a-deep-dive-into-hdmi-audio-performance.56/. As you see there, there is much more documentation on the test setup, mostly because that is an article that was going in print as opposed to me trying to get a review out.

So point taken but there is no smoke here let alone fire as my as errors in my measurements. :)

Importantly, AtomicBomb's linearity measurements match mine. As do the glitching. What differs is our write-ups is that he closed his eyes and praised the product as being excellent. Whereas I told it as it is.

If they want to telegraph messages back and forth to me, I like them to make a full disclosure as to origins of the products they are testing. Are they free loans or are they purchasing them with their own money?
 
Last edited:

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
14,080
Likes
6,469
Location
Seattle Area
Does AudioScienceReview even have sponsors?
No. The only funding we get are through donations and that amount so far is a tiny fraction of the money I have spent purchasing products.
 

Candlesticks

Active Member
Joined
Jan 19, 2018
Messages
108
Likes
125
Thanks for bringing that to my attention. It is a very good effort on their part to show how same equipment can measure differently. We also have our own thread here where we are trying to standardize measurements and source files so that even when using different equipment, results can be compared.

The points they make don't apply to my testing though. First, they should know that it was my team and initiative I took to implement the entire new audio stack in Windows. The Windows XP kmixer had to be shot in the head and I made sure that happened in Windows Vista development :). Brand new pipeline was put in place and that is what is in use in Windows 10.

Importantly to the points they make, we implemented a new API called WASAPI for direct access to the sound driver bypassing the kernel processing. We wanted to support ASIO but that required licensing and it was cheaper and easier to build WASAPI instead.

What is wonderful in Creator Edition of Windows 10 is that full USB class drivers are now available. Windows not only recognizes all the DACs I am testing but automatically creates WASAPI interfaces for them. This eliminates the hassle of finding and installing ASIO drivers. Reliability has been superb although sometimes event driven does not work.

My testing is all done with Roon in exclusive mode using Wasapi. This is documented in every review I write although sounds like I should expand more on it for newer readers. This is what is in this review for example:

View attachment 10571

Roon only shows the above when WASAPI exclusive mode is used. So the issues they raise with WDM do not apply to my testing. Everything is bit-exact. No testing I do is with Windows processing.

There seems to be a tendency over there to keep assuming incompetence and lack of knowledge because they have not heard of me. All they have to do is click on my signature, do some googling, etc. and not make assumptions about me not knowing the difference between bit exact and not when my team implemented all of this.

On FFT points, my analyzers maxes out at 32K which is a blessing. Folks like Atomicbomb who use Prism Sound Dsound have access to far longer FFT kernels, resulting in a whole bunch of spikes that can be distracting. It also artificially lowers the noise floor that is seen in the graph, giving a highly misleading impression of how low the noise floor of the DAC really is. Even in my measurements there is a 32 dB "FFT gain" (artificial reduction in noise floor). Many people look at those graphs and assume the noise floor is what is shown when it is not at all the case. So if anyone needs to be cautious, is those folks with their large FFT kernels.

It is on my TODO list to write an article on my testing and expand significantly on the above.

Finally, I have done testing using Prism Sound that atomicbomb is using. Here is the article: https://audiosciencereview.com/forum/index.php?threads/a-deep-dive-into-hdmi-audio-performance.56/. As you see there, there is much more documentation on the test setup, mostly because that is an article that was going in print as opposed to me trying to get a review out.

So point taken but there is no smoke here let alone fire as my as errors in my measurements. :)

Importantly, AtomicBomb's linearity measurements match mine. As do the glitching. What differs is our write-ups is that he closed his eyes and praised the product as being excellent. Whereas I told it as it is.

If they want to telegraph messages back and forth to me, I like them to make a full disclosure as to origins of the products they are testing. Are they free loans or are they purchasing them with their own money?
Cool response so everything is done via WASAPI exclusive?

It's worth having clarification in each review of what procedure was followed. I find other people just write a single post of all their settings and link to that at the top of the review each time stating that is the measurement procedure followed. Easy to copy/paste each time.

So in the jitter comparison between the Fiio E10K and Modi 2 the same settings were used for both?

No. The only funding we get are through donations and that amount so far is a tiny fraction of the money I have spent purchasing products.
So no manufacturer donations?
 

Thomas savage

Liberation through innovation
Moderator
The Watchman
Patreon Donor
Joined
Feb 24, 2016
Messages
5,952
Likes
1,916
Location
uk, taunton
(snip)


So no manufacturer donations?
Correct, the forum is funded by amir other than the recent initiative asking for member donations.
Asking for donations can actually make people feel more involved, more apart of what we are doing so it serves a purpose beyond just the financial one.

If this site ever takes on manufacturer sponsorship in any way shape or form you will know because I won’t be here.

 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
14,080
Likes
6,469
Location
Seattle Area
Cool response so everything is done via WASAPI exclusive?

It's worth having clarification in each review of what procedure was followed. I find other people just write a single post of all their settings and link to that at the top of the review each time stating that is the measurement procedure followed. Easy to copy/paste each time.

So in the jitter comparison between the Fiio E10K and Modi 2 the same settings were used for both?
Yes.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
14,080
Likes
6,469
Location
Seattle Area
So no manufacturer donations?
Definitely not. I am getting some loaner equipment for review and when so, I am and will be very clear as that occurring. Most manufacturers don't see what I do as good for them so I don't expect that to become common.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
14,080
Likes
6,469
Location
Seattle Area
And so it is clear, here are the faults clearly visible in atomicbomb's measurements:

This is his linearity graph. A perfect DAC would have a flat line. Here is what he measured of Schiit BiFrost Multibit back in 2015:



The red notations are mine. As we see linearity error starts to creep in right around the same -58 dB that I measured which corresponds to about 10 bits of resolution. Furthermore, the error becomes very significant at > 10 bits at CD's noise floor of -96 dB. So this DAC cannot reproduce 16 bit signals accurately let alone high-resolution. Other DACs I have measured are almost ruler flat to the end.

This is another test he ran which I also did:



This is supposed to be a sine wave but we see clear glitches near zero crossing. Again, my measurements show the same thing.

Note that both of these are time domain measurements so FFT is not involved. He does have a FFT spectrum in blue and red however which shows the effect of those glitches.

So instead of worrying about my measurements, he needs to explain how these two measurements of his show excellence.
 
Last edited:

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
14,080
Likes
6,469
Location
Seattle Area
Notice all the documentation I provide on the graphs on what they mean. He doesn't do any of that. Just posts graphs after graphs without anything for the reader to focus on. As a result, back in 2015 people saw poor measurements of Schiit BiFrost Multibit from them but it did not register with anyone. Had he noted the problems as I do, and shown comparisons in-situ to other DACs without the problem, we could be sitting here with the problem potentially being resolved.
 
Last edited:

drconopoima

Member
Patreon Donor
Joined
Jan 14, 2018
Messages
49
Likes
27
Notice all the documentation I provide on the graphs on what they mean. He doesn't do any of that. Just posts graphs after graphs without anything for the reader to focus on. As a result, back in 2015 people saw poor measurements of Schiit BiFrost Multibit from them but it did not register with anyone. Had he noted the problems as I do, and show comparisons in-situ to other DACs without the problem, we could be sitting here with the problem potentially being resolved.
He could have thought that this was excellent performance from a DAC, but that loses ground when you find other of his measurements clearly demonstrate that he has seen better performance from products on a similar price-range.
 
Top Bottom