• 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.

Musical Fidelity MX-DAC Review (Balanced DAC)

pkane

Major Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
2,576
Likes
4,274
Location
North-East
your statement is rigorously true, but again, that wasn't the question.

That’s because the question is flawed. There are thousands of ways to mess up a reconstruction filter in a way that its distortions will become audible, in thousands of ways. But since the way to reproduce a perfect waveform has been known for well over a century, any device that allows audible aliasing is simply broken.
 

MaxBuck

Addicted to Fun and Learning
Forum Donor
Joined
May 22, 2021
Messages
563
Likes
625
Location
SoCal, Baby!
It's amazing how people revert to treating amir like a cult leader or conspiratorial mastermind instead of some dude who provides data and his own opinions for free.

I have said it before and will say it again; mere consumers are not capable of having productive technical discussions the way hobbyist diyers and actual researchers are.

Besides maybe cables, there is no less consequential category of audio equipment than a DAC. None of them sound different. These ratings which draw so much ire are only interesting to people who take pleasure in seeing how close to the theoretical minimum noise and distortion manufacturers can approach. What are we at now, -124db? That's pretty cool, but that's all it is. Sort of sad to think that we may never see actual 24 bit audio in our lifetimes.

The problem with including filter measurements is that it perpetuates a fiction that they matter, in the same way that speaker measurements matter. Frankly I think Amir should stop including these insane measurements (has anyone heard the ESS hump??) Or at least include a caveat that everything here is inaudible.
I have my problems with the "Computer Audiophile" and his site, but he has reported on his blind ABX comparison of the three new flavors of Schiit Yggdrasil DACs, claiming that he heard distinct differences among them even though they all measure very well (SINAD above 110 dB). Do you doubt him? What biases could lead to problems with such blind ABX tests and lead to false positives when it comes to ability to distinguish among them? As I understand it, he rigorously equated output levels.

Also see https://goldensound.audio/2021/09/21/schiit-yggdrasil-more-is-less-measurements/ for some interesting measurements at different input frequencies for the Yggdrasil More-Is-Less. A single-frequency SINAD value may be misleading when discussing "sound signature differences" among DACs, for example. Bottom line is that measurements can fully describe the audio output of components; I'm less certain that we are entirely sure how to comprehensively interpret those measurements, nor indeed that we're necessarily measuring everything that can affect that output sound. But I acknowledge that I'm quite ignorant on the science of this stuff. Just pondering some things here.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
35,336
Likes
135,185
Location
Seattle Area
I have my problems with the "Computer Audiophile" and his site, but he has reported on his blind ABX comparison of the three new flavors of Schiit Yggdrasil DACs, claiming that he heard distinct differences among them even though they all measure very well (SINAD above 110 dB). Do you doubt him?
He didn't run an ABX test. That aside, what he did is of zero significance and showed how little knowledge folks have on what a proper audio test looks like. I created a recent video just on this topic:


It has become quite fashionable these days to create these pretend blind tests. Please rise above them and realize the marketing stunts they are.
 

pjug

Major Contributor
Forum Donor
Joined
Feb 2, 2019
Messages
1,271
Likes
1,057
He didn't run an ABX test. That aside, what he did is of zero significance and showed how little knowledge folks have on what a proper audio test looks like. I created a recent video just on this topic:


It has become quite fashionable these days to create these pretend blind tests. Please rise above them and realize the marketing stunts they are.
Looks like you need to update your own video with another way to do it wrong.
 

PeteL

Major Contributor
Joined
Jun 1, 2020
Messages
1,500
Likes
1,428
That’s because the question is flawed. There are thousands of ways to mess up a reconstruction filter in a way that its distortions will become audible, in thousands of ways. But since the way to reproduce a perfect waveform has been known for well over a century, any device that allows audible aliasing is simply broken.
Yep, audible is the keyword here.
 

MaxBuck

Addicted to Fun and Learning
Forum Donor
Joined
May 22, 2021
Messages
563
Likes
625
Location
SoCal, Baby!
He didn't run an ABX test. That aside, what he did is of zero significance and showed how little knowledge folks have on what a proper audio test looks like. I created a recent video just on this topic:

It has become quite fashionable these days to create these pretend blind tests. Please rise above them and realize the marketing stunts they are.
What elements of a valid ABX test were missing? Or alternatively, what did he do that invalidated the comparison test?
 

pjug

Major Contributor
Forum Donor
Joined
Feb 2, 2019
Messages
1,271
Likes
1,057
If I read it correctly, he always knew which of X, Y, or Z were playing. He just didn't know which DAC matched with these letters. I'm sure he thought he heard a difference, but there is no way to know if he did or not. Wow, a plane trip to do that.
 
  • Like
Reactions: 617

617

Major Contributor
Forum Donor
Joined
Mar 18, 2019
Messages
1,543
Likes
3,304
Location
Somerville, MA
I have my problems with the "Computer Audiophile" and his site, but he has reported on his blind ABX comparison of the three new flavors of Schiit Yggdrasil DACs, claiming that he heard distinct differences among them even though they all measure very well (SINAD above 110 dB). Do you doubt him? What biases could lead to problems with such blind ABX tests and lead to false positives when it comes to ability to distinguish among them? As I understand it, he rigorously equated output levels.

Also see https://goldensound.audio/2021/09/21/schiit-yggdrasil-more-is-less-measurements/ for some interesting measurements at different input frequencies for the Yggdrasil More-Is-Less. A single-frequency SINAD value may be misleading when discussing "sound signature differences" among DACs, for example. Bottom line is that measurements can fully describe the audio output of components; I'm less certain that we are entirely sure how to comprehensively interpret those measurements, nor indeed that we're necessarily measuring everything that can affect that output sound. But I acknowledge that I'm quite ignorant on the science of this stuff. Just pondering some things here.

SINAD is only happens to be a useful metric for line level devices because it's generally line level devices which approach a degree of perfection such that their performance can be summarized with a single number. Power amps are sufficiently nonlinear that the single number isn't that descriptive; ASR provides a SINAD at something like 1khz/5 watts but it doesn't tell the whole story.

It turns out that it is very easy to make a transparent DAC, as evidenced by the fact that of the hundreds that amirm has measured for his own perverse reasons, the vast majority have noise and distortion below 80db of the fundamental tone. 80db is a very very bad performing DAC: it's easy nowadays to find one which is many times better. Remember that 6db is about half as loud, so 86 is twice as good, 92 is four times as good, 98 is eight times better, 102 is sixteen times better and so on. However, -80db is also an incredibly quiet signal. -80db is the equivalent of talking to someone one meter away, and then them walking 5 or 6 miles away. You can't talk to someone 6 miles away.

When someone says they can hear differences between 110db SINAD devices they are saying they can hear someone speaking or not speaking who is over 150 miles away. I think it is reasonable to assume that they are either mistaken, their test is flawed or there is something seriously wrong with our measurements, which go down to something like 130db with noise averaging.

So basically we are talking about devices which all do an essentially perfect job, which have slight variations in HF distortion caused by having a variety of filters.

I don't know about every test people have done on the audibility of cutting edge digital audio phenomena. I am not a scientist. However, I do have enough intuition about the db scale to have a sense of proportion about these devices.

The issue of the audibility of HF filters hinges on how HF distortion is weighted in the SINAD calculation. Perhaps someone can weigh in on that. Slow filters increase distortion at the top of the musical bandwidth, but let's say for example the HF distortion was incredibly bad- -60 db for example. -60db is incredibly damn quiet. I've done tests to demonstrate this. The idea that correlated distortion which 'blends in' with the signal would be audible at -60db is just not believable to me.

We all want the best, but DACs are something not worth worrying about. There is a lot for consumers and enthusiasts to learn about as they educate themselves about audio and I am a big fan of that. ASR has done an outstanding job educating people about headphone and speaker performance; both incredibly subtle and important areas.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
35,336
Likes
135,185
Location
Seattle Area
What elements of a valid ABX test were missing? Or alternatively, what did he do that invalidated the comparison test?
Did you watch the video? It answers all but the "ABX" part. ABX requires A&B and then a way to represent "X." You can't self administer such a thing since at all times you know A or B. A computerized switcher is normally used to implement ABX which clearly is not present here.
 

MaxBuck

Addicted to Fun and Learning
Forum Donor
Joined
May 22, 2021
Messages
563
Likes
625
Location
SoCal, Baby!
Did you watch the video? It answers all but the "ABX" part. ABX requires A&B and then a way to represent "X." You can't self administer such a thing since at all times you know A or B. A computerized switcher is normally used to implement ABX which clearly is not present here.
I understand now what you're saying, and you're obviously right. Thanks for better explaining ABX to me; this wasn't it.

It was, however, blind ABC testing, which strikes me as a pretty good test in its own right.
 

MaxBuck

Addicted to Fun and Learning
Forum Donor
Joined
May 22, 2021
Messages
563
Likes
625
Location
SoCal, Baby!
When someone says they can hear differences between 110db SINAD devices they are saying they can hear someone speaking or not speaking who is over 150 miles away.
Actually they're not. They're actually saying that SINAD at a single frequency is inadequate to define the sonic character of a component.
 

pkane

Major Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
2,576
Likes
4,274
Location
North-East
I understand now what you're saying, and you're obviously right. Thanks for better explaining ABX to me; this wasn't it.

It was, however, blind ABC testing, which strikes me as a pretty good test in its own right.

If you want to conduct a blind differentiation test, then you must remove visual identification of A or B. It doesn’t matter if you know which device A or B represent. As long as you can visually identify between A and B position of the switch, the test is no longer about audible or inaudible differences but about what you might imagine they might be.
 

MaxBuck

Addicted to Fun and Learning
Forum Donor
Joined
May 22, 2021
Messages
563
Likes
625
Location
SoCal, Baby!
If you want to conduct a blind differentiation test, then you must remove visual identification of A or B. It doesn’t matter if you know which device A or B represent. As long as you can visually identify between A and B position of the switch, the test is no longer about audible or inaudible differences but about what you might imagine they might be.
I don't claim the test to eliminate the possibility of bias, but the notion that there is any such thing as a perfect experiment is absurd. Claiming that the test was "worthless" is not appropriate IMO.
 

pkane

Major Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
2,576
Likes
4,274
Location
North-East
I don't claim the test to eliminate the possibility of bias, but the notion that there is any such thing as a perfect experiment is absurd. Claiming that the test was "worthless" is not appropriate IMO.

The test doesn’t have to be perfect, but ignoring simple, well-known and easy to implement controls and then claiming the test results as valid is absurd.
 

pjug

Major Contributor
Forum Donor
Joined
Feb 2, 2019
Messages
1,271
Likes
1,057
If you want to conduct a blind differentiation test, then you must remove visual identification of A or B. It doesn’t matter if you know which device A or B represent. As long as you can visually identify between A and B position of the switch, the test is no longer about audible or inaudible differences but about what you might imagine they might be.
Or be able to identify A or B in any way.

Also, I can't imagine trying to take a proper test between 3 devices. It is hard enough to pass an ABX with similar sounding A and B.
 

MediumRare

Major Contributor
Forum Donor
Joined
Sep 17, 2019
Messages
1,516
Likes
1,747
Location
Chicago
I don't claim the test to eliminate the possibility of bias, but the notion that there is any such thing as a perfect experiment is absurd. Claiming that the test was "worthless" is not appropriate IMO.
@pkane is right: At n=1 the test is not blind as he explained, thus it has lost its value. If n=50 or more it would be different because the error would become randomized.
 

PeteL

Major Contributor
Joined
Jun 1, 2020
Messages
1,500
Likes
1,428
Actually they're not. They're actually saying that SINAD at a single frequency is inadequate to define the sonic character of a component.
That would be very true, still the way but SINAD at 1K is still is for this community the main indicator of DACs performance, that's how they are ranked from great to poor in the reference DAC chart. Yes we have other measurements but even considering those, it's often not enough to judge if a product will be sonically different, let alone better than an other one. I think many get's over enthusiasts about a few numbers, which as you say doesn't suffice to show if sonically there is a benefit to reach that high. Most manufacturers that have product reviews here are not making measurements devices, they are making music listening devices. Spending engineering ressources, processing cost, and hardware cost, to achieve numbers that will not bring any benefit, is a waste and it is actually smart design IF indeed it can be determined that it doesn't degrade the listening experience. This I don't know yet for sure. All effort should be on stuff that can be appreciated by human ears, and do it reliably, for a long time, and looking good doing so.
 

GaryH

Senior Member
Joined
May 12, 2021
Messages
384
Likes
413
Because intermodulation with aliased tones in the output stage has the potential to affect performance in the audible range. Many, but not all, DACs with poor filtering show rising distortion in high frequencies.
Exactly. That's why using, say, a laptop with ultrasonic junk and relatively high IMD for ABXing hi-res with 16/44.1 could well give false positives...
 
Top Bottom