• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Review and Measurements of Matrix Audio Element X DAC/Streamer/Amp

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
That sounds like option (A) from my prior post, assuming that "suitable attenuation" is unity gain. It's a reasonable assumption since the analog ouputs on a DAC chip are typically around 2 Vrms. In this case, when you digitally attenuate within the DAC to 50 mV, you're passing a small signal (attenuated by -32 dB) through an analog stage at unity gain. If most of the noise & distortion comes from the analog path (not the DAC), then you expect 32 dB worse SNR when you shrink the original signal by 32 dB. Essentially, you shrink the signal while noise level remains constant.

Ideally, when you attenuate the signal you want to attenuate the noise too. Since some of the noise comes from the amount of gain in the analog stage (lower gain = lower noise), why not take the full-scale DAC signal (say 2 Vrms) and pass it to an analog stage with less than unity gain -- that attenuates it by -32 dB (gain of 1/40 or 0.025). Now you have the same 50 mV output, but your analog stage has attenuated a full scale signal, so noise & distortion should be lower.

Would that work?
If there is extra output buffer, we need to assume that the output buffer has less noise than the dac. So attenuate the signal and the noise the same time using step attenuator will be able to preserve the SNR. Opa1612 is easily 0.3uV noise. It will be better than input noise of most devices.

If you use opamp as attenuation then the least noise is 1 times the noise of the opamp itself which works similarly but a bit higher noise in every volume setting.

The way to preserve the SNR is to track the digital volume. Lets say we have a step for every 6db of attenuation. So at -6db it switches from digital attenuation to step attenuator and set back the digital volume to 0db. 6 steps will give roughly 36db of extra dynamic range.

However when the dac has noise of lets say 0.6uV, it doesn't really matter in the entire system for the most part as the next component in the signal chain will likely have higher input referred noise. Wantting higher SNR is not really necessary. It can be done to just as stated above.
 

Julf

Major Contributor
Forum Donor
Joined
Mar 1, 2016
Messages
3,004
Likes
3,998
Location
Amsterdam, The Netherlands
However when the dac has noise of lets say 0.6uV, it doesn't really matter in the entire system for the most part as the next component in the signal chain will likely have higher input referred noise. Wantting higher SNR is not really necessary. It can be done to just as stated above.

And again, as a reminder, if you are using a 24 bit DAC (with 20 or 21 effective bits), you have 24-48 dB of digital attenuation before it starts affecting the SNR.
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
And again, as a reminder, if you are using a 24 bit DAC (with 20 or 21 effective bits), you have 24-48 dB of digital attenuation before it starts affecting the SNR.
That's not true. The noise doesn't decrease with the signal. One db of decrease in digital level will result in -1db of snr. The output noise is mostly independent from the digital input level.
 

Julf

Major Contributor
Forum Donor
Joined
Mar 1, 2016
Messages
3,004
Likes
3,998
Location
Amsterdam, The Netherlands
That's not true. The noise doesn't decrease with the signal. One db of decrease in digital level will result in -1db of snr. The output noise is mostly independent from the digital input level.

My point is that the analog noise floor is somewhere between -90 and -126 dBFS. 24 bits gives you 144 dB. Thus you have at least 18 dB, more likely up to 54 dB attenuation range before you get above that noise floor.
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
My point is that the analog noise floor is somewhere between -90 and -126 dBFS. 24 bits gives you 144 dB. Thus you have at least 18 dB, more likely up to 54 dB attenuation range before you get above that noise floor.
Let's say 0dBFS is 2V and noise is 2uV. So you have 120dB of SNR. When decrease the level to -20dBFS to 200mV. Now you only have 100dB of SNR. The noise doesn't decrease with the signal. Only way to preserve the SNR is in analog. Stepped attenuator, Potentiometer, inverting amplifier etc are possible to preserve most of the SNR.
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
...The noise doesn't decrease with the signal. Only way to preserve the SNR is in analog. Stepped attenuator, Potentiometer, inverting amplifier etc are possible to preserve most of the SNR.
That's what I thought. When we see the SNR drop by 32 dB going from 2 Vrms to 50 mV, we're seeing the SNR drop by the same amount the signal dropped. So noise is constant, and we're probably seeing the effect of digital attenuation in the DAC chip upstream of the analog stage.

By "inverted amplifier" are you referring to the fact you can't get less than unity gain with a non-inverting op amp? So inverting amplifier means an analog "gain" stage with less than unity gain. I would guess the following from cleanest to least clean: (1) inverting amplifier (less than unity gain), (2) stepped attenuator (unity gain, then attenuated with metal film resistor voltage divider), (3) potentiometer.
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
That's what I thought. When we see the SNR drop by 32 dB going from 2 Vrms to 50 mV, we're seeing the SNR drop by the same amount the signal dropped. So noise is constant, and we're probably seeing the effect of digital attenuation in the DAC chip upstream of the analog stage.

By "inverted amplifier" are you referring to the fact you can't get less than unity gain with a non-inverting op amp? So inverting amplifier means an analog "gain" stage with less than unity gain. I would guess the following from cleanest to least clean: (1) inverting amplifier (less than unity gain), (2) stepped attenuator (unity gain, then attenuated with metal film resistor voltage divider), (3) potentiometer.
The second one you need an extra buffer AFTER the attenuation. Because you don't want the output impedance to be super high, which can cause frequency response change or distortion or unwanted attenuation when interfacing the cable and the next stage.
The third is equivalent to the second one.
The first one inverts the signal which isn't necessarily desirable. And using the same parts, inverting amplifier always has higher noise than non-inverting for the same amount of gain (extra 1time of noise gain).
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
...
The third is equivalent to the second one.
...
I don't think so, as metal film resistors are quieter than pots.

using the same parts, inverting amplifier always has higher noise than non-inverting for the same amount of gain (extra 1time of noise gain).
Even so, the gain is so much lower (-32 dB) it might be quieter overall.
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
I don't think so, as metal film resistors are quieter than pots.
No they aren't. They are the same if the resistance is the same. You can use 1kohm pot and it will give very good noise performance (equivalent of 250ohm noise at maximum). Less than 0.3uV for 1kohm pot.
And I was comparing to with stepped attenuator and potentiometer with non inverting buffer.
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
To restate the question: suppose you want to build a state of the art cost no object DAC with variable analog output level, optimized for highest possible SNR at the low to moderate levels used to drive headphones (say 50 mV). You're using a typical modern DAC chip so you can use its built-in digital attenuation, or use its full scale output and attenuate it downstream.

How would you design it?
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
To restate the question: suppose you want to build a state of the art cost no object DAC with variable analog output level, optimized for highest possible SNR at the low to moderate levels used to drive headphones (say 50 mV). You're using a typical modern DAC chip so you can use its built-in digital attenuation, or use its full scale output and attenuate it downstream.

How would you design it?
That depends on what I want. Personally I just won't allow digital volume control for simplicity reason. And add high performance preamp in the next stage.

And what I love to see are, if possible like the one Benchmark did, stepped attenuator resistor array. High performance, fine tune attenuation. Low noise.
Also more practically, combination of digital volume control and simple voltage divider is pretty good. This probably has SNR fluctuating within 0db to -6db range. This is also very good.
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
...metal film resistors are quieter than pots. ...
No they aren't. They are the same if the resistance is the same. ...
If we consider the thermal noise of a resistor, sure it depends only on the resistance and temperature. But that is the theoretical minimum noise. In the real world most resistors are noisier than this, depending on their type. Wire wound are closest (lowest noise), followed by metal film, then other types like carbon. Most volume pots use carbon and are noisier than metal film. But your response made me look into this further and I discovered "conductive plastic", a type of resistor that I had not heard of before. Apparently high quality pots are using these because they're quieter than carbon. Are they as quiet as metal film or wire wound?

...
And what I love to see are, if possible like the one Benchmark did, stepped attenuator resistor array. High performance, fine tune attenuation. Low noise.
...
I agree this seems the best approach. But if you think metal film resistors don't have lower noise than a pot, what is the benefit of a stepped attenuator?
 

w1000i

Active Member
Forum Donor
Joined
Aug 31, 2019
Messages
260
Likes
138
Location
Jubail SA
When I look back and compare these measurement to Chord Qutest, the 3rd,4th, etc.. harmonic distortion is way lower with Qutest.
I'm I reading that correct ?!

Qutest :

index.php



Matrix

index.php
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
When I look back and compare these measurement to Chord Qutest, the 3rd,4th, etc.. harmonic distortion is way lower with Qutest.
I'm I reading that correct ?!

Qutest :

index.php



Matrix

index.php
yes
 

Veri

Master Contributor
Joined
Feb 6, 2018
Messages
9,596
Likes
12,036
When I look back and compare these measurement to Chord Qutest, the 3rd,4th, etc.. harmonic distortion is way lower with Qutest.
I'm I reading that correct ?!

Qutest :

index.php



Matrix

index.php
As far as harmonics go: yes, but the qutest has no proper balanced output so you end up with comparatively much more noise, even in single ended the matrix still does better in regards to noise.

Those harmonics mean nothing if they get buried in noise in practical real-world situations, the SINAD simply means that distortion+noise wise the matrix does better objectively, harmonic distortion or not.
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
...
Those harmonics mean nothing if they get buried in noise in practical real-world situations, the SINAD simply means that distortion+noise wise the matrix does better objectively, harmonic distortion or not.
True, though equal SINAD doesn't mean they sound the same since we perceive noise & distortion differently. We can hear correlated signal (distortion) somewhat below the noise floor. And distortion can be perceptually more annoying, or in some cases euphonic. Consider 2 devices A and B having the same SINAD, where A's "NAD" is high distortion low noise, and B is low distortion high noise. They may sound different, and B is probably more neutral.
 

Veri

Master Contributor
Joined
Feb 6, 2018
Messages
9,596
Likes
12,036
True, though equal SINAD doesn't mean they sound the same since we perceive noise & distortion differently. We can hear correlated signal (distortion) somewhat below the noise floor. And distortion can be perceptually more annoying, or in some cases euphonic. Consider 2 devices A and B having the same SINAD, where A's "NAD" is high distortion low noise, and B is low distortion high noise. They may sound different, and B is probably more neutral.
A fair point. I wonder how much of the distortion matters at levels approximating -130dB though.
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
If it's -130 dB at actual listening levels, it's almost certainly transparent. But the SINAD at actual listening levels is usually lower, sometimes much lower, than what it measures at full scale output. This is what I like about Amir's 50 mV output SINAD test.

Hypothetical worst-case example: suppose the device has 130 dB SINAD at full output, 2 Vrms. 50 mV is 32 dB quieter, and suppose SINAD also drops by 32 dB, so it's 98 dB at actual listening levels. You're listening to uncompressed classical music having 60 dB of dynamic range. At the quiet parts, the noise might only be 38 dB below signal. That's 1.2%, which may be audible under critical listening.

It's surprising to imagine that 130 dB at full scale could end up being audible even under worst-case conditions. Does my rough estimate have a mistake?
 
Last edited:

w1000i

Active Member
Forum Donor
Joined
Aug 31, 2019
Messages
260
Likes
138
Location
Jubail SA
As far as harmonics go: yes, but the qutest has no proper balanced output so you end up with comparatively much more noise, even in single ended the matrix still does better in regards to noise.

Those harmonics mean nothing if they get buried in noise in practical real-world situations, the SINAD simply means that distortion+noise wise the matrix does better objectively, harmonic distortion or not.


Honesty I don't give SINAD much weight, and I prefer to check each measurement where my listing level and setting apply most.

For higher Harmonic distortion, I read that higher H.Dist will affect who long you listen to music and give more transparency.
 

kairos

Member
Joined
Sep 23, 2019
Messages
15
Likes
21
Location
Vancouver, Canada
Matrix also makes an Element M that sports the Sabre 9028, and less capable pre-amp function, but is $1200 less. Not sure what other compromises they made to get it into that price point, but if it measures as well as the X, I would be willing to buy it (can't afford the X, sadly).

I might even buy it and submit it for review, but I am not sure of their return policy (in case it measures poorly), any word from the manufacturer on their return policy?
 
Top Bottom