Measurement and Review of Schiit BiFrost Multibit DAC

[March Audio]

Ignore button guys, ignore button.

 

[Thomas Savage]

Thomas, honestly, you do not think this happens here?

If you read what I wrote you might find that questions redundant.

 

[sssn]

Multibit/ladder DACs bring out the most passionate pleas from their advocates. I made a contribution to one the most popular threads on WBF before my number got retired there. I was the first US owner of a TotalDAC. I upgraded to an even more gigantic TotalDAC. My time with the TotalDac reminds me of the time spent with a turntable. It was filled with highs and lows. But I never felt comfortable that I was getting everything out of my system. Via open minded exploration I learned what it was that I loved about it. And I learned how I had made another expensive mistake.

There are distortions that some DACs can produce which are quite pleasing to most listeners. For example, the TotalDac has a rolled off high end because it lacked a digital anti-aliasing filter. The DAC had FIR filter to flatten out the top end. I preferred the less accurate setting. Later on in my audio journey I learned about listening test done by Harman which show listeners prefer a rolled off high end. That was about the same time I began to investigate DSP solutions.

So why do we do all these measurements if we know people don’t necessarily prefer the most accurate reproduction of the recording? I think it’s still helpful. In the end, the listener can choose to inject whatever distortion/nonlinearity he/she chooses. To do so, it’s helpful for the listener to know what types of distortions are preferred and which ones are not. It’s also important to have a standard or baseline in any endeavor. Without foundational basics, we can hardly communicate with each other to express our own preferences and desires. We can build from the baseline and move into more accurate ways to “distort” the signal in a way that is consistently pleasing.

I don't get this at all. Why are people so adamant about spending hundreds or even thousands of Dollars on deliberately inaccurate DACs and amplifiers when a puny, simple software equalizer can achieve the same thing and much more even?!
Especially a reduction of top end/high frequencies is trivial for even the cheapest audio software's equalizer. And this approach is no less impure than a DAC internally rolling off/distorting.

 

[Blumlein 88]

Ignore button guys, ignore button.

Bingo!

 

[Candlesticks]

@amirm

http://superbestaudiofriends.org/in...echnical-measurements.5770/page-3#post-190762

Response by Atomicbob on the Bifrost MB linearity.

I have put a lot of those graphs in the measurement section on several forums. Most of mine are labeled Device Model Techincal Measurements. "Really bad" is relative. Mimby and Bifrost MB are using a 16 bit ladder DAC intended for instrumentation use which require level accuracy. As BiMB has 2Vrms output at 0 dBFS, -78 dBFS is the last 3 bits to produce approximately 244 uVrms. This is about 0.2 nanowatts for an HD650 or HD800. at 300 ohms.

 

[Blumlein 88]

Will do this one more time with you.
At lease you have found a home as you really cannot do this at CA.

"Eric Auer said: ↑
You guys at Schiit have to work on Sundays?" Please put it in google translate, thank you.

As far a your word "naivety" give yourself another 40 or 50 years and if you are lucky you might have half the knowledge I do. I am not sandyk so I will not be lashing back at you as this is your MO, provoking people. Have a wonderful life.

So you are responding to me about shills based on someone else's post implying you are a shill. Not my post. Okay.

If you have some knowledge, then maybe in the next 50 years you'll share it instead of playing this hide and seek game you no doubt think makes you appear very clever.

 

[Candlesticks]

Can we just move Rebbitputzmaker's "contributions" to another thread at this point. Throughout this entire thread they have said nothing of relevance to the measurements made and I don't believe they have any intention to start soon.

 

[Purité Audio]

Half the knowledge you have that will be a teeny tiny amount won’t it?
Keith

 

[Thomas Savage]

So you are responding to me about shills based on someone else's post implying you are a shill. Not my post. Okay.

If you have some knowledge, then maybe in the next 50 years you'll share it instead of playing this hide and seek game you no doubt think makes you appear very clever.

I’m onboard with this, I’m fed up with the posturing egotism too..

@rebbiputzmaker theres nothing in your posts in this thread to suggest you have any knowledge of Audio whatsoever.

Iv no clue what your issue is with member @Blumlein 88 , he’s one of the least antagonist forum members Iv ever come across and imo has been extremely patient with you.

If you have something to contribute beyond information devoid protests and a insistence you know more than everyone else please feel free to do so but if not I’d suggest you don’t bother as more of the same from you is going to end up with you being shown the door sooner rather than later.

 

[amirm]

@amirm

Please reference the intro which appears at the top of many of my technical measurement posts:
If you are unfamiliar with audio measurements please use a search engine with the query:
"audio measurements" or "audio measurement handbook"
Look for publications by Richard C. Cabot and also by Bob Metzler, both from Audio Precision. There are other useful publications as well. These will provide basic knowledge.

So much for him documenting his work. He is sending people to google for themselves? No wonder no one notices or even pays attention to much of what he measures.

Response by Atomicbob on the Bifrost MB linearity.

Thanks. He is trying to paper over the data unfortunately. Let's parse what he is saying:

"I have put a lot of those graphs in the measurement section on several forums. Most of mine are labeled Device Model Techincal Measurements. "Really bad" is relative. Mimby and Bifrost MB are using a 16 bit ladder DAC intended for instrumentation use which require level accuracy. As BiMB has 2Vrms output at 0 dBFS, -78 dBFS is the last 3 bits to produce approximately 244 uVrms. This is about 0.2 nanowatts for an HD650 or HD800. at 300 ohms."
​

Now that there is agreement that his measurements are the same as mine, this is what the DAC designer Mike Moffat had to say about those results: https://www.head-fi.org/threads/wha...n-robert-hunter.784471/page-449#post-14028674

"The linearity reported measurement is atrocious and far below the performance level of any 16 bit Bifrost ever sold. This means either the linearity measurement or the Bifrost is broken. "​

Seeing how our 3-year apart measurements match, we can rule out the DACs being broken. That leaves us with atomicbomb saying the device is "excellent" but both the DAC designer and I saying it is unacceptable. That should be an untenable situation for him.

Now let's address his point of these DACs being used for instrumentation use. He doesn't expand on that so let me do that. There is a need to generate accurate voltages for reference in many applications. Furthermore it would be good to do that under the control of a microprocessor/software so that it can be programmed quickly and is stable over time unlike potentiometers used in olden days. That is one of the main uses of this Analog Devices DAC that is used in Schiit BiFrost Multibit (and Yggdrasil/Modi Multibit). In this application the value is programmed and if a glitch is generated it doesn't matter because a short period later the glitch goes away and the instrument is ready for use.

This is not the case in audio. In audio our waveforms are dynamic and there is no opportunity to wait for anything to settle down. The sine wave goes up and down and you better produce every step of it accurately. Importantly if the waveform values are increasing the output voltage better do the same thing and not go backward (it is called "monotonicity"). This ladder DAC as is cannot do this and hence the reason its own manufacturer, Analog Devices, does not recommend its use for audio.

Techniques exist for example by using two of them in either interleaved mode or one handling the positive and the other the negative cycle of the waveform to get around the worst case situation that is zero crossing (problem remains elsewhere but it is not as extreme).

None of these mitigation techniques exist in this DAC so no wonder we see the problems in linearity.

Now let's address his "this is too small of a number" argument. He is improperly positioning the numbers. First of all there is no headphone amplifier in this device. The 2 volt RMS is the output from its analog RCA jacks. Using 2 volt and impedance that rises to 500 ohms in low frequencies and we get to a wattage of 0.008 or 8 milliwatts. So in the case of headphones we are already starting with small numbers in milliwatt region. Headphones block ambient noise so it is not hard at all to reach threshold of hearing with them where our hearing is most sensitive: https://audiosciencereview.com/forum/index.php?threads/dynamic-range-how-quiet-is-quiet.14/

Research shows that live music can have peaks exceeding 120 db (even in unamplified cases). Taking that we see that the required dynamic range for transparency in mid frequencies is a whopping 130 db. Nothing remotely close to 78 dB which he is declaring "excellent."

Now we can backtrack and pick a lower number but 78 is not it. In my book if you can't reach CD's 16-bit dynamic range value, you should not play.

Let's put aside all of this. If the results he is getting for Schiit BiFrost Multibit DAC are "excellent," what room does it leave for him to rate the Topping DX7?

Conclusions
We are nicely converging on everyone agreeing that my data is the same as his data in this regard. All the specious arguments put forward before are just that: specious.

When we wear the hat of measurement guy, we need to convey wisdom with our data. JA in stereophile does that. So do I. It is that wisdom and knowledge that is important. None of that is present in his data which he generates with automated scripts and posts them. The multiplicity of what he posts hides problems. And lack of comparison data on the same graph makes any attempts at such an analysis tedious at best. I think even he is not paying attention to his own data because of such reasons.

The Schiit BiFrost Multibit due to its design is simply not a performant DAC on the bench. It has unacceptable performance as Mike Moffat mentioned. I don't care what architecture is used. The end goal should be accurate representation of input PCM samples. Let's be on side of consumers than manufacturers and point out issues as we see them. Not fight and try to paper over them.

 

[drconopoima]

@amirm

http://superbestaudiofriends.org/in...echnical-measurements.5770/page-3#post-190762

Response by Atomicbob on the Bifrost MB linearity.

So, it's inaudible. Great to know. Great for my wallet to know I mean, that won't be paying 650$ for inaudible differences towards the 'less accurate' site of the spectrum (although, if you required accuracy, which is not the case for music, I could see as justifiable paying 650$ for the accuracy)

 

[Candlesticks]

@amirm as this thread is now 20 pages I've compiled a set of links so that anyone new entering this thread. If you want could you edit them into your original post please.

Below is an abridged series of events for people new to the thread.

Measurements posted by SBAF user Atomicbob:

• Schiit Bifrost Multibit - Loaner Unit
• Schiit Bifrost Multibit

Measurements of Schiit Bifrost Multibit after warm up overnight by amirm.

Response by amirm after seeing Atomicbob's measurements as linked above. The second part. The third part.

First post by amirm in the Mike Moffat (Schiit engineer) blog on Head-Fi. Further down are more posts by amirm in that thread. Response from Mike Moffat. Reply by amirm.

Post written by Atomicbob on measurement procedure. Response by amirm to that post.

Post by Atomicbob on non-linearity of Schiit Bifrost Multibit. Response by amirm.

 

[Cosmik]

Glitches can be greatly reduced using a sample-and-hold or similar.
http://www.ti.com/lit/ug/tidu022/tidu022.pdf

But no use if you believe that such circuits contain evil spirits. sound bad in ways that cannot be measured. If you didn't believe that, you obviously wouldn't be messing about with this method at all, and would simply invest in a $1 audio DAC chip that measures perfectly.

 

[amirm]

@amirm as this thread is now 20 pages I've compiled a set of links so that anyone new entering this thread. If you want could you edit them into your original post please.

That's great. I went ahead and added it to the first post.

 

[Jimster480]

Multibit/ladder DACs bring out the most passionate pleas from their advocates. I made a contribution to one the most popular threads on WBF before my number got retired there. I was the first US owner of a TotalDAC. I upgraded to an even more gigantic TotalDAC. My time with the TotalDac reminds me of the time spent with a turntable. It was filled with highs and lows. But I never felt comfortable that I was getting everything out of my system. Via open minded exploration I learned what it was that I loved about it. And I learned how I had made another expensive mistake.

There are distortions that some DACs can produce which are quite pleasing to most listeners. For example, the TotalDac has a rolled off high end because it lacked a digital anti-aliasing filter. The DAC had FIR filter to flatten out the top end. I preferred the less accurate setting. Later on in my audio journey I learned about listening test done by Harman which show listeners prefer a rolled off high end. That was about the same time I began to investigate DSP solutions.

So why do we do all these measurements if we know people don’t necessarily prefer the most accurate reproduction of the recording? I think it’s still helpful. In the end, the listener can choose to inject whatever distortion/nonlinearity he/she chooses. To do so, it’s helpful for the listener to know what types of distortions are preferred and which ones are not. It’s also important to have a standard or baseline in any endeavor. Without foundational basics, we can hardly communicate with each other to express our own preferences and desires. We can build from the baseline and move into more accurate ways to “distort” the signal in a way that is consistently pleasing.

I think that it also depends on what kind of music you listen to.
Being as I listen to many different kinds of music, I can tell you that if you go somewhere to listen to rock music (like classic rock) then its unlikely you will experience music in a way that is 0% distorted. Infact you may be expereincing music in a way that is 10% or more distorted based on how they have the equalizer adjusted and how high the gains are on the amps, aswell as the environment they are being played in.
So I think that some people will prefer sound based on their live experiences.
I can tell you though that if you go somewhere to listen to Live classical music, such as a concert hall.... that you will likely have the purest possible experience and when tones are missing you will surely notice.
For me I personally prefer a bit more bassy of a headphone to enjoy my Electronic musics, but I prefer my Aeon for classical and most jazz... especially live editions.

And I understand that some people don't like treble and therefore they like warm sounding DAC's or any gear with rolled off highs. Sometimes they pair bright headphones with rolled off DAC's and this is why they think it "sounds the best" when they are actually losing detail.
So it is really about what people like, but objectivity does have a hand in it. And selling a $600 piece of equipment which is actually worse than a $75 piece of equipment doesn't have much going for it.

 

[tusing]

Let's put aside all of this. If the results he is getting for Schiit BiFrost Multibit DAC are "excellent," what room does it leave for him to rate the Topping DX7?

Savage. I think this highlights the issue with basically every DAC review. How many other reviewers immediately hailed the Bifrost as "excellent", or even "without issue"?

 

[amirm]

Better question would be how many reviewers ever give a negative nod to anything they review?

 

[svart-hvitt]

Better question would be how many reviewers ever give a negative nod to anything they review?

«Reviewers don’t waste time on poor products. Why should we waste time and space on mediocrity?»

This is actually what a couple of reviewers told me...

 

[tusing]

«Reviewers don’t waste time on poor products. Why should we waste time and space on mediocrity?»

This is actually what a couple of reviewers told me...

Or rather, reviewers have their own form of Midas' Touch: everything they review instantly turns to something within the range of "good" to "excellent".

But in all seriousness, I think there are a few dimensions to this.

Humans are pretty bad at gauging sound: NwAvGuy has this excellent writeup, "What We Hear", that I think really illustrates this point. Our perception constantly morphs to match our input; we discard far more data than any digital signal processor or measurement device. "Good" and "bad", then, are likely easily distinguishable only when one device is truly abhorrent and the other at the top of the game. Most of the reviews of the Bifrost are along the lines of "it's alright, not worth the price" or "it's amazing." Given these measurements, it seems like the former are the ones to pay attention to. Regardless, this really highlights the limits of our hearing. If some of the worst reviews for the Bifrost receives constitute "it's still alright I guess", I think that says more about our hearing ability than the Bifrost itself.

Furthermore, I think many reviewers understand that they can't really tell the difference between, say, an ODAC and a Yggdrasil, and so they need to find something to fill up 4 pages worth of review with. What are they going to say? You're sitting here with a beautifully designed piece of equipment and you genuinely can't tell the difference between any other piece of equipment. I suppose reviewers tend to just say good things at this point, because they can't really say anything bad about it, being physically unable to hear the difference. Sitting with neutrality and saying "it's a DAC and sounds like all my other DACs" isn't a substantive review and won't really help. As it seems here, the Bifrost is an engineering disaster, but I can understand the good reviews, regardless of their validity.

At this point, I'm fairly certain that the overwhelming majority of subjective reviews have more to do with the reviewer and their state of mind than with the actual equipment. And, well, I can't really blame them. If placebo affects our perception of sound so much, clearly placebo has some value regardless.

 

[svart-hvitt]

Or rather, reviewers have their own form of Midas' Touch: everything they review instantly turns to something within the range of "good" to "excellent".

But in all seriousness, I think there are a few dimensions to this.

Humans are pretty bad at gauging sound: NwAvGuy has this excellent writeup, "What We Hear", that I think really illustrates this point. Our perception constantly morphs to match our input; we discard far more data than any digital signal processor or measurement device. "Good" and "bad", then, are likely easily distinguishable only when one device is truly abhorrent and the other at the top of the game. Most of the reviews of the Bifrost are along the lines of "it's alright, not worth the price" or "it's amazing." Given these measurements, it seems like the former are the ones to pay attention to. Regardless, this really highlights the limits of our hearing. If some of the worst reviews for the Bifrost receives constitute "it's still alright I guess", I think that says more about our hearing ability than the Bifrost itself.

Furthermore, I think many reviewers understand that they can't really tell the difference between, say, an ODAC and a Yggdrasil, and so they need to find something to fill up 4 pages worth of review with. What are they going to say? You're sitting here with a beautifully designed piece of equipment and you genuinely can't tell the difference between any other piece of equipment. I suppose reviewers tend to just say good things at this point, because they can't really say anything bad about it, being physically unable to hear the difference. Sitting with neutrality and saying "it's a DAC and sounds like all my other DACs" isn't a substantive review and won't really help. As it seems here, the Bifrost is an engineering disaster, but I can understand the good reviews, regardless of their validity.

At this point, I'm fairly certain that the overwhelming majority of subjective reviews have more to do with the reviewer and their state of mind than with the actual equipment. And, well, I can't really blame them. If placebo affects our perception of sound so much, clearly placebo has some value regardless.

Maybe Trump was right? FAKE NEWS!