Matrix Audio X-SABRE Pro MQA: Best Audio DAC in the World?

[boXem]

Thanks for the link! So he's using an older analyser and blaming the analyser for being old and thus not up to the task of properly measuring his design.

That doesn't make his creation distortionless at least we know where the "not measurable" statement originates from exactly.

My issue is that you seem to assimilate this product with snake oil while even if not perfect, and very expensive, there is some serious engineering behind it.
https://www.hypex.nl/img/upload/doc/an_wp/WP_AES120BP_Simple_ultralow_distortion_digital_PWM.pdf

 

[Veri]

My issue is that you seem to assimilate this product with snake oil while even if not perfect, and very expensive, there is some serious engineering behind it.
https://www.hypex.nl/img/upload/doc/an_wp/WP_AES120BP_Simple_ultralow_distortion_digital_PWM.pdf

It's not the first DAC to use PWM processing. The NAD M51 did it, too. Heck the $60 PowerDACv2 by @IVX does it.

My issue is with the very high price, and the intentional "so good its distortion can't be measured" tactic while using what looks like an older analyser. Time will tell if some independent measurements will show its performance and classify it as snake oil or not, but considering even Chord is affordable next to it, I reserve my doubts.

 

[boXem]

It's not the first DAC to use PWM processing. The NAD M51 did it, too. Heck the $60 PowerDACv2 by @IVX does it.

You should read the paper I linked.

My issue is with the very high price, and the intentional "so good its distortion can't be measured" tactic while using what looks like an older analyser. Time will tell if some independent measurements will show its performance and classify it as snake oil or not, but considering even Chord is affordable next to it, I reserve my doubts.

Expensive equals bad is as wrong as expensive equals good.
Concerning the "tactics" I am not sure that the marketing team knows that the AP has now a better model than the 2722.
Anyhow, I agree with you that only independant measurements would tell the truth, as with any product.

 

[bidn]

I can think of quite a number of reasons to choose a particular product over another that might be equally measurably transparent/audibly transparent when blinded and level matched:

-Cost
-Aesthetics
-Features (some may prefer minimalist devices, others may want lots of extra features for their use)
-Compatibility with other components in one's system (input/output types, balanced or not, compatibility with formats you own or services you stream from, etc. Some of this falls under "features" but there are things like output impedance etc. that might matter depending on what gear is being used as well.)
-Build quality
-Reliability
-Customer and product support
-Company history (I am more likely to support a company whose overall history I like, versus an unknown or one that I don't respect due to past products, behavior by principals in the company, etc.)
-Marketing (you might like no-nonsense marketing from one brand, and another that makes an equally great product might have marketing you hate using pseudoscience or ridiculous claims)
-Country of manufacture/parts origin/etc.

A lot of the above are interrelated, such as build quality, reliability, product support, company history, etc. And obviously, some people give more consideration to some of these factors than others.

A lot of excellent points here, Hugo9000!

I can relate a lot to the one re. the company history (e.g. I am a believer in open-source, I don't like that Sony always kept the DSD format closed, preventing people from doing copies of their own disks --which contributed ironically to DSD being a commercial flop, and above all I hate them for having introduced DRMs into regulars music CDs. Several years ago I bought a few CDs from Sony, they would not play in my linux computers nor in my (not recent) CD player, then I found out Sony had been putting DRMs in their CDs, so these could only be played in MS-Windows computers or in recent CD-players. Likewise I hate Meridian for pushing their lossy, inefficient, closed and full of DRMs proprietary MQA format).

I would add another aspect:
-Practical aspects of form-factor
Take for example the very expensive Chord Dave: it has a round display protruding above the top, and the button for volume control quite protruding somewhere on the top, preventing from stacking another device like an amp (or their M-Scaler) above (well they offer a solution, a very expensive rig for stacking their own devices...).
Or I checked out this Mola Mola (I can't get used to this weird name) device: its top is wave-shaped, you won't be able to stack something on top of it in a stable manner.
Funny that the more expensive the device, the worse the design for me. (actually here I would even prefer TotaleARNAC's very cheap metal sheet, at least it is practical...)

 

[bidn]

I would bet that no difference is audible when level-matched and listened to blind with normal material (not test tones). The question is, can we go lower in price/quality than the D30? What is the real-world threshold of transparency?

Well I would nuance this in several respects.
I hope the following will not be shocking...

(I am speaking below only of studio, i.e. "ideal", recordings).

First,
there is the special case of people like me (I assume I must not the only one?)
who listen with headphones at very, very low volume levels:
I listen with my RME ADI-2 Pro with headphones directly connected to the dual jack headphone outs-> XLR out in average at about -60dB, (and less with some headphones, e.g. below -70 dB with the closed-back Focal Stellia), i.e. in power areas where the devices, DAC included, are much less performing and will not have such good numbers.
So having a device having the best numbers at the best point can matter for people who listen at power areas for which they are not optimized.

Second,
this is not my case, but I can imaging it could matter:
each stage in the audio chain will further degrade the SNR and IMD, so starting with the best possible situation would provide more margins, lower the risks.

Third, it depends a lot on which transducers are used.
I pass on speakers with all the problems of room acoustics, many quite difficult to resolve.
I listen mainly with headphones. Most are indeed not resolving enough to let hear differences.
But now take super resolving (at least at very low level volume levels) headphones like the Focal Utopia.
You can check here several people finding the same differences in level-matched double blind tests of cables with the Utopia:
(click on the button "afficher" below "Journée du 30 mai 2019"):
http://www.tellementnomade.org/forum/viewtopic.php?f=92&t=24039
This confirms the subjective impressions most people (incl. me) hearing (rather consistent) differences with cables with the Utopia (although I didn't use to believe in cable differences). Now the same people subjectively hear also differences between devices like DACs and amps with the Utopia.
Oops, I hope this is not too subjectivist...

 

[30 Ounce]

I would bet that no difference is audible when level-matched and listened to blind with normal material (not test tones). The question is, can we go lower in price/quality than the D30? What is the real-world threshold of transparency?

I think test tones and white noise are extremely revealing of sonic differences. Fact is music is so dynamic, fluid and complex that pinning a ABX comparison on music only is faulty. Amir’s test equipment uses test tones to find differences in DAC’s because it plainly lays out the differences. Why shouldn’t humans do the same?

 

[Bamyasi]

I think test tones and white noise are extremely revealing of sonic differences. Fact is music is so dynamic, fluid and complex that pinning a ABX comparison on music only is faulty. Amir’s test equipment uses test tones to find differences in DAC’s because it plainly lays out the differences. Why shouldn’t humans do the same?

Because in humans, sound signal processing by the brain includes as one of its steps a highly selective dynamic filtering and signal recognition of what you actually hear. This auditory perception function of the brain has been trained and tweaked by millions of years of evolution to filter out exactly that kind of background noise (i.e., pink noise and loud pure tones) while at the same time preserve and analyze at great detail and depth harmonic content of the sound.

If you are interested, here is some good science describing the concept.

 

[signalpath]

Exceptional specifications. However, would suggest that THD numbers are not a good indicator of sound quality. Years ago, there was a large double-blind ABX study which concluded that program distortion under 0.1% THD was inaudible. And most listeners couldn't detect abberation under 1%. There is in fact no consistent industry correlation between sound quality and THD numbers below 0.1%.

That said, it's nice to see ultra-low THD numbers, but more importantly this DAC also has exceptional linearity, signal-to-noise / dynamic range, which ARE solid indicators of sonic performance. Kudos to the Chinese team. I hope more manufacturers focus on making their DACs more linear, quieter, and less jittery.

 

[digicidal]

NIce analogy, too bad it isn't really relevant. I could put a hyper-sophisticated carbon fibre and titanium wheel on my bike, and it might be a conversation starter, but I won't go any faster.

Actually, since rotational mass is a significant factor in acceleration - it very well might go faster, provided that your existing wheels are heavier (which I would assume they are compared to CF & titanium). One could argue that the analogy is relevant on those terms - as it's similar to the difference between a 90dB SNR and a 110dB SNR... likely audibly irrelevant despite being on either side of 16bit range cutoff - but still considered at least measurably significant. And yes, I am picking your nits on that.

Indeed it would be quite challenging to create a signal generator that is better than these DACs. The one in Audio Precision analyzer gets there but it is very hard.

Not to mention the AP has an MSRP higher than 95% of even the most expensive, exotic audiophile DACs.

 

[MatrixAudio]

DAC+Pre+DSP w/output for 2+ subs...

At least it's impossible currently...

 

[digicidal]

At least it's impossible currently...

Well, when it becomes possible... I'm there. Regardless, great job on a beautiful and extremely well-engineered product!

 

[restorer-john]

Dedicated "rooms" in the CNC billet. So upmarket. I have gear with copper plated metal cases around the PCBs but in comparison, they could only be called "sheds" I guess....

Here's one of my old Sony's copper plated "sheds" to isolate a separate D/A converter in an ES preamplifer from the very late 80s. Note they also used a rock solid foundation for the preamplifier chassis made of a ceramic/glass fibre polyester resin type stuff.

The D/A converter board sits underneath the main board in the cavity, steel plate top and bottom. No modern USB crap here, just 32/44/48K 16bit. 4xO/S. Built with real CMOS too and a pair of BB D/As. Several trusty low noise spec 5532s. The good stuff!

Whole thing is rather heavy at >13kg for a preamplifer. Bigger power supply than many medium powered amplifiers too. Potted tx/f. Another "metal shed" for the uP next to the transformer can. And more relays than you can poke a stick at for input selection....

An audiophile's dream- it's full of Nichicon Great Supply, Muse, Film caps, Elna Duorex and FETs galore. Bowden cable switches, motorised die-cast gearbox Alps Blue etc. This is the main line gain stage:

 

[gvl]

Do you know who is the designer behind this right? As far as I know, he tends to measure his stuff, and he did.
https://www.mola-mola.nl/history.php
I agree that the marketing claims are a bit exagerated, but since a part of the potential customers tend to run away when seeing objective measurements, I could understand that these are not too easy to find.

Interesting he recommends notching out the fundamental. @amirm, any comments?

 

[restorer-john]

Interesting he recommends notching out the fundamental. @amirm, any comments?

Notching out the fundamental (leaving enough for the analyzer to lock onto) means also the front end ADC can work without the range shifting of having to deal with full output at many tens of volts down to the uV level. What distortion effects the notch filter itself contributes could become an issue.

Sending the output from a typical analog THD analyzer (notched out fundamental with just distortion components) to a soundcard based FFT does essentially the same thing and gives around 60dB additional depth to play with, and no front end level conditioning (attenuation) required. The beauty of that implementation is you have a precision variable frequency notch filter that can deal with uV to 100V+ at low inherent levels of distortion up front.

I know Tom from Neuorchrome/TCA is working on a precision notch filter for that very purpose.

 

[amirm]

Interesting he recommends notching out the fundamental. @amirm, any comments?

All my measurements use the same. The fundamental is notched out and measured with one ADC. Then another is used without the filter. Then with signal processing, the two are merged together so you still see the fundamental even though the noise floor and distortions reflect it not being there. The older AP he is using can't do the stitching so shows the waveform without the fundamental being there.

 

[Julf]

Actually, since rotational mass is a significant factor in acceleration - it very well might go faster, provided that your existing wheels are heavier (which I would assume they are compared to CF & titanium). One could argue that the analogy is relevant on those terms - as it's similar to the difference between a 90dB SNR and a 110dB SNR... likely audibly irrelevant despite being on either side of 16bit range cutoff - but still considered at least measurably significant. And yes, I am picking your nits on that.

I knew someone would.

 

[Purité Audio]

Mola’s ‘tambaqui’ is Being reviewed/measured in next months HiFi News and record review.
Keith

 

[Frank Dernie]

differences with cables with the Utopia

By what mechanism does this difference occur?

 

[trl]

Here is a teardown review from a chinese website. The internal design is very nice.
http://www.soomal.com/doc/10100007215.htm

Thanks for sharing! Somal usually does great pictures of the internals.

I see that all opamps are OPA551, in all stages; now this is puzzling for me, as this opamp is usually used as output buffer instead of I/V or LPF.

 

[BDWoody]

I can think of quite a number of reasons to choose a particular product over another that might be equally measurably transparent/audibly transparent when blinded and level matched:

-Cost
-Aesthetics
-Features (some may prefer minimalist devices, others may want lots of extra features for their use)
-Compatibility with other components in one's system (input/output types, balanced or not, compatibility with formats you own or services you stream from, etc. Some of this falls under "features" but there are things like output impedance etc. that might matter depending on what gear is being used as well.)
-Build quality
-Reliability
-Customer and product support
-Company history (I am more likely to support a company whose overall history I like, versus an unknown or one that I don't respect due to past products, behavior by principals in the company, etc.)
-Marketing (you might like no-nonsense marketing from one brand, and another that makes an equally great product might have marketing you hate using pseudoscience or ridiculous claims)
-Country of manufacture/parts origin/etc.

A lot of the above are interrelated, such as build quality, reliability, product support, company history, etc. And obviously, some people give more consideration to some of these factors than others.

Come on...you can come up with more than that...