Topping D90SE Measurements (DAC)

[VintageFlanker]

Can you run same 44.1k multitone test with your Cosmos running at 384kHz and show the full 192k bandwidth of multitone result? I am seeing something with my Cosmos and wanted to see if it is anti aliasing filter related or something else.

There you go :

PS: It is the Matrix, not the D90SE.

 

[Music1969]

There you go :View attachment 198128

PS: It is the Matrix, not the D90SE.

Thanks. Interesting that signals only go to 22kHz but there is lots of stuff very close to that, still high in level (relative to signal level).

But graph stopping at 20kHz wouldnt show it.

 

[bennetng]

Not the first time. Likely MQA related.

Matrix Audio Element X - issue

I thought I'd start a new thread documenting the issue I have with my new Element X (and hopefully its resolution), rather than corrupting the review thread any further. I was playing around with my newly-received X over the weekend, and noticed that all 7 internal filters sounded identical to...

www.audiosciencereview.com

 

[theREALdotnet]

I wonder why we’re still seeing anti-aliasing (the slope starting around 20kHz going down towards the left until it sinks into the noise floor around 600Hz) with 384kHz sampling rate?

 

[bennetng]

IMD caused by ultrasonic artifacts can reflect into passband as well.

 

[theREALdotnet]

IMD caused by ultrasonic artifacts can reflect into passband as well.

Ok, but starting at 192kHz, shouldn’t they be way, way down at 22kHz? The plot above looks very much what I’d expect from a 44.1kHz sampled signal. Isn’t the point of higher sampling rates that those artifacts are much further up high and therefore much easier to filter (i.e without having to resort to aggressive filters)?

 

[Music1969]

Would be nice if @amirm 's multitone's for DACs could always show up to 100kHz, to see this kind of crap (crap = not part of test signal) that gets hidden when plots stop at 22kHz

Right around the corner from 22kHz... can be nice surprises like this.

 

[bennetng]

Ok, but starting at 192kHz, shouldn’t they be way, way down at 22kHz? The plot above looks very much what I’d expect from a 44.1kHz sampled signal. Isn’t the point of higher sampling rates that those artifacts are much further up high and therefore much easier to filter (i.e without having to resort to aggressive filters)?

Pay attention to the Matrix plots from both VintageFlanker and manisandher. The MQA filter is much more relaxed than some of the "fast" ESS filters. For example the ESS linear phase fast filter has 120dB attenuation beyond 24.1kHz for 44.1kHz sample rate.

192kHz is the ADC measurement bandwidth, not the input sample rate to the DAC.

 

[theREALdotnet]

192kHz is the ADC measurement bandwidth, not the input sample rate to the DAC.

Exactly! And exactly! The rate that the DAC is operating at would likely be much higher than 192kHz, probably at 384 or 768 (or whatever the equivalent to the SDM rate is the DAC is actually working at). Anti-aliasing artifacts at this amplitude should not occur during high-rate D/A conversion, they should have died down way up in the ultra-sonic. Perhaps @Miska can be bothered to weigh in, he seems to understand these things.

 

[theREALdotnet]

Then again, zooming in on the Y axis it appears that the artifacts top out at -135dB or so. So perhaps this is what we should expect from a 384kHz test signal after all. I’m not going to lose sleep over something at -135dB

 

[pma]

The multitone result in post #81 is at least strange. PC re-sampling?

 

[bennetng]

Anti-aliasing artifacts

There is no such thing as "Anti-aliasing artifacts". We are talking about Cosmos ADC's anti aliasing filter. DAC artifacts beyond 22kHz for 44.1kHz sample rate are called imaging artifacts, not "Anti-aliasing artifacts".

If the DAC is infected by a botched MQA firmware, then the issue should be fixed first before looking into other upsampling methods.

 

[bennetng]

The multitone result in post #81 is at least strange. PC re-sampling?

An MQA issue:

Matrix Audio Element X - issue

I thought I'd start a new thread documenting the issue I have with my new Element X (and hopefully its resolution), rather than corrupting the review thread any further. I was playing around with my newly-received X over the weekend, and noticed that all 7 internal filters sounded identical to...

www.audiosciencereview.com

 

[pkane]

There you go :View attachment 198128

PS: It is the Matrix, not the D90SE.

Could it be some sort of NOS or a very lazy reconstruction filter in use? Seems very unusual (and unhealthy )

 

[bennetng]

Seems very unusual (and unhealthy )

Yes, need to get rid of glitched MQA firmware to restore health.

Mytek DAC infected with MQA:

One of the default ESS filters in the Okto:

 

[pkane]

Yes, need to get rid of glitched MQA firmware to restore health.

Mytek DAC infected with MQA:
View attachment 198262


One of the default ESS filters in the Okto:

View attachment 198263

That MQA filter shape looks almost exactly like the shape of the ultrasonic images in @VintageFlanker 's multitone with Matrix, so most likely the cause.

 

[theREALdotnet]

DAC artifacts beyond 22kHz for 44.1kHz sample rate are called imaging artifacts, not "Anti-aliasing artifacts".

Ah, I’m talking about the stuff below 20kHz, not above.

 

[bennetng]

Ah, I’m talking about the stuff below 20kHz, not above.

Then they are IMD, not "Anti-aliasing artifacts".

 

[theREALdotnet]

Then they are IMD, not "Anti-aliasing artifacts".

Pardon my muddled terminology, if anything it would be aliasing, not anti-aliasing artifacts. But yes, from what I understand, aliasing is the result of intermodulation between the signal and the sampling frequency.

 

[pma]

Yes, mirror images should not get below 0.5 Fs. Even worst case should not. (No aliasing in case of DAC). Nonlinear distortion, as mentioned by @bennetng