Article: Understanding Digital Audio Measurements

[pma]

Isn't SNR easy to measure?

In case of measuring very low noise audio amps or dacs, the best way to measure noise is to use a special low-noise measuring amplifier, to amplify the measured noise and finally to subtract the gain (dB) of the measuring amplifier from the measured amplified noise.

 

[trl]

How can I set the dbFS to 0 on the SU-9? The SU-9 has a volume knob, does that has anything to do with it?

dBFS means dB at Full Scale, so you need to max. out the volume potentiometer to achieve the Full Scale output.

 

[nhatlam96]

dBFS means dB at Full Scale, so you need to max. out the volume potentiometer to achieve the Full Scale output.

The Sanskrit 10th MK2 DAC has no volume knob, does it automatically set itself to 0dBFS? Is it possible to adjust the dbFS without volume pot?

 

[trl]

does it automatically set itself to 0dBFS

By default it will output sound at max. power, that means 0dBFS.

Is it possible to adjust the dbFS without volume pot?

Digitally control from the OS or from the application you're listening music with. Amir was probably adjusting the output power by adjusting the input signal power from his AP equipment.

 

[nhatlam96]

Digitally control from the OS or from the application you're listening music with.

If I set Windows Audio to 100% and Spotify to 100%, I doubt that I would have 0dbFS,
because Windows Audio 100% + VLC 200% is a lot louder. So OS and music application can't be it?

 

[trl]

Leave everything at 100%, not at 200%, then adjust the volume from the external amplifier.

 

[KTN46]

I was wondering why IMD is only done on DAC tests and not headphones/amps as well. My understanding is that headphones and amps are also able to generate IMD. Why is it only worth considering in DACs?

My understanding that a low THD (at least in amps) is generally related to low IMD, but I'm unsure how or why that would be the case, or if the correlation is true.

 

[yigitboy]

Which measurement correspond to which audible spec? For example linearity = resolution or high db in low freq = clean bass response ?
Forgive me if written before. I search for it but English not my native language.

 

[JoachimStrobel]

Amir, after reading this a second time, and having seen many of your Sinad scores, I have a question:
You write that a filter takes out the 1khz peak and the noise, and finally Sinad, comes from the leftovers.
But this means that the sharpness of this filter defines the noise that is left. I assume it takes the actual peak and not the ideal digital feed in the first place. So some logic centers the filter on the the actual peak, like 999.8 hz? And what is the peak width? What happens if the peak got a bit wider, and that is then all noise? I would imagine, that ideally you would have two graphs showing filter position vs Sinad and filter width vs Sinad. Does the system create those? Otherwise a DAC could produce a slightly wide and off frequency peak and therefore creating a huge Sinad value while the other noise is really low. Would that be audible in the same way as a super sharp peak with a spot on frequency that has more noise on the floor? Could a clever manufacturer not engineer a DAC that performs super Sinad wise that way but would not sound better than a wide&off-peak DAC?

 

[Blumlein 88]

Here is where I embedded a 997 hz tone at -1 dbfs in -64.78 db of white noise.

The upper value is from notching with a relatively wide notch having a Q=4.
The lower is what you get with a notch filter with Q=10.

I don't know how sharp the filter is in the AP (would imagine pretty sharp), obviously the difference in the true value of any residuals and that lost by notching out is going to be small, probably less than .1 db.

Here is the 32K FFT with Q=4.

 

[JoachimStrobel]

View attachment 115411

Here is where I embedded a 997 hz tone at -1 dbfs in -64.78 db of white noise.

The upper value is from notching with a relatively wide notch having a Q=4.
The lower is what you get with a notch filter with Q=10.

I don't know how sharp the filter is in the AP (would imagine pretty sharp), obviously the difference in the true value of any residuals and that lost by notching out is going to be small, probably less than .1 db.

Here is the 32K FFT with Q=4.
View attachment 115412

Interesting. Could you draw the filter over the peak to provide a visual?

 

[nhatlam96]

Interesting points have been made:
https://www.reddit.com/r/headphones/comments/ltwyz9

 

[JimB]

Interesting points have been made:
https://www.reddit.com/r/headphones/comments/ltwyz9

Well. I would agree that worrying about small differences in high (good) SINAD figures is not appropriate (including consideration of performance at low levels). I would also agree that SINAD is insufficient to characterize sound 'quality' when it is low (the component is definitely coloring the sound). HOW the sound gets colored will be important. But here, on ASR, the reviews generally give us much more that just a SINAD figure at max output, no? If I wanted a metric of how a significantly distorting component will sound, and whether I prefer the corresponding 'quality' (or lack thereof), I think I'd need to actually hear it. But I tend to want components that are transparent. In that case, what's the complaint with such figures of merit, when they far exceed the ability to hear? This argument seems targeted at how bad is acceptable - that high level technical performance is not required. And maybe not, in some specific cases. With transparent equipment now readily available, I don't need to worry about such tradeoffs.

 

[JoachimStrobel]

Interesting points have been made:
https://www.reddit.com/r/headphones/comments/ltwyz9

I am not sure to which post you are relating your answer. But I want to make clear that I do not question SINAD and the other parameters at all and their value for benchmarking DACs. I was simply looking into the measurement as such and try to understand more about it.

 

[JimB]

I am not sure to which post you are relating your answer. But I want to make clear that I do not question SINAD and the other parameters at all and their value for benchmarking DACs. I was simply looking into the measurement as such and try to understand more about it.

The one I included in my response - the one referenced by @nhatlam96.

 

[Wes]

I don't see anywhere in the OP where it says that SINAD includes, or is influenced, jitter.

That might be improved with an explicit stmt. on what SINAD really measures, and to what extent.

 

[don'ttrustauthority]

Your fanboys like to call us fanboys AudioPhools because we describe sound subjectively, but you really haven't stated why a -85 db jitter vs - 110 db jitter matters. Can you hear a difference? I can't. Maybe you're the golden eared audiophools?

 

[BlackH20]

How about a simple Table of Contents for tested products?

I'm exhausted looking for 100 DAC streamers reviewed in an article.

"For best performance, you want products in the blue bucket and avoid those in red." But they are blurred at the bottom with a link to the site's homepage? Where do I find any product category with all the products tested and ranked? Streamers, Dacs, all tested, click where? Amplifiers...all tested....click where? Speakers, all tested; click where? Seem to be on a merry-go-round of the whole website, finding recent reviews, reader forums, donations (which I have made), but how about a decent table of contents on all the actual product tests?.

Not quite as old as like Clint here in the picture, but I respect the site and want to read the results of Amir's work, or at least be able to find it.

Turn one of the Excel graphs sideways, and let me see how the names on the DAC/Streamers brands you test rate out (you state one a 100) versus blurring them at the bottom? What good does the do for the reader? Where is the list of over the hundred listed players you tested?

 

[amirm]

Please see: https://audiosciencereview.com/forum/index.php?pages/Audio_Equipment_Reviews/

 

[BlackH20]

Please see: https://audiosciencereview.com/forum/index.php?pages/Audio_Equipment_Reviews/

Thanks for the reference and all the options to drill down to the specifics.