Is this normal behaviour for dacs?

[gregorio]

thats what i said myself .... it still made a difference

That’s not even vaguely what you said! Where did you say that your assertions were hilarious and even worse than ridiculous? What you actually said/claimed is the IMD produced by your DAC was somehow audible, even though it’s so low in level it probably isn’t even being reproduced by your speakers to start with!

i also love the fact objectivists keep arguing that everything is inaudible but you guys still care for measurements lol, thats some logic i dont have to understand...

And I love the fact that subjectivists love to make up false facts about what objectivists state/argue! Strange that you believe you don’t have to understand the logic of the facts that we’re actually stating but you do somehow believe the logic of making up false facts about what “objectivists keep arguing”. Maybe that’s the root cause of the arguments; what you think is logic is actually the opposite, completely illogical!

As i started out i still cared for this, the topping d10 was my first standalone dac a few years ago as i was going by amirs recommendation, but the difference between it and the BLA interface is so obvious...

The difference is obvious, they look different, their price is different and that’s just to start with!

i just feel sorry for the people that cant trust their ears

And I feel sorry for the people who say they trust their ears but never bother to actually find out what their ears are telling them!

some objectivists should tell people how they can improve performance for free instead (by reducing gain), but funny enough i never saw it mentioned "anywhere"

Yes, you can improve performance and reduce IMD, all noise and other types of distortion “by reducing gain” to nothing, better still, don’t even turn on you DAC, amp or speakers, no distortion whatsoever of any type and therefore perfect performance. Thanks for your advice and for mentioning it here for the first time! Lowering gain obviously lowers everything, the signal level and the distortion. Duh!

G

 

[Ghoostknight]

Lowering gain obviously lowers everything, the signal level and the distortion. Duh!

but its still a unlinear relationship which you dont seem to get, better to make some BS arguments

 

[gregorio]

but its still a unlinear relationship which you dont seem to get, better to make some BS arguments

As there’s no such thing a “unlinear relationship” then obviously I don’t get it and also obviously you are taking your own advice and making a BS argument! Well done!

G

 

[Ghoostknight]

not sure why you feel like you have to "play this topic down"

i just thought i would share some measurements that show a clear unlinear relationship between signal and harmonic distortion since this objectively(!) improves performance

 

[Ghoostknight]

isnt this what all the measurement crap is all about, to see how something performs (worst case i guess) no matter how audible it is? .... i thought its interesting to share something that actually improves performance "from the worst case", because my guess is thousands of people are not aware of this, including me till recently

 

[gregorio]

i just thought i would share some measurements that show a clear unlinear relationship between signal and harmonic distortion

Again, as there’s no such thing as an “unlinear relationship”, how do you think you’ve shown a clear example of one?

isnt this what all the measurement crap is all about, to see how something performs (worst case i guess) no matter how audible it is?

No, but don’t let that stop you from making up nonsense!

G

 

[Ghoostknight]

I just noticed this after watching this: because i wasnt aware of "Phase Noise" but i always wondered how the thicker bottom part of a sinetone is "expressed", its phase noise

As you see, even "Phase Noise" is improved by lowering gain, but i guess i have to line up the top of both signals to make it a true comparison

 

[bigshot]

I think he is being genuine. He’s just incapable of applying logic and reason and his solipsist experience prevents him from considering any other possibility.

 

[danadam]

One more thing... could someone explain why a dynamic range measurement shows less dynamic range then there actually is? (see the multitone test with 115db)

i know other dacs show even more in the multitone test but i cant wrap my head around why multitone shows 115db and the DR measurement 95db

For anyone who actually wants to read I'll repeat what I wrote earlier:

You cannot just visually read it from the graph, because you can make the graph however low you want by increasing the FFT size.

Here's 1 kHz tone in 16-bit / 44.1k sampling rate, using FFT size 8K, 64K and 512K:


To get an accurate noise level value, all FFT bins have to be summed up. That's what REW is showing in the "distortion" overlay (which I suggested to use already in the first reply to this thread). The noise is -93.6 dBFS regardless of how low it looks on the graph.

i always wondered how the thicker bottom part of a sinetone is "expressed", its phase noise

It most probably isn't. You probably use 1000 Hz tone and Hann window for FFT and that combination will result with such "thicker" bottom. Here's 1000 Hz and Hann window compared to:

• 1001.294 Hz and Hann window
• 1000 Hz and Blackman-Harris 7 window

And btw, REW documentation says about HD measurements:

To obtain accurate results the window should have low side lobes. Good choices in order of reducing side lobe level are Dolph-Chebyshev 150 (side lobes 150 dB down), Blackman-Harris 7 (side lobes 180 dB down), Dolph-Chebyshev 200 (side lobes 200 dB down) and Cosine sum 9-235 (side lobes 235 dB down). Hann, Blackman-Harris 4 and Flat-Top are not recommended.

 

[Ghoostknight]

To get an accurate noise level value, all FFT bins have to be summed up. That's what REW is showing in the "distortion" overlay (which I suggested to use already in the first reply to this thread). The noise is -93.6 dBFS regardless of how low it looks on the graph.

Oh! i think they dont have to be summed up but taking more multiple measurements just cancels more noise out right?

let me do some more measurements with your recommended settings later, thanks

 

[VNandor]

https://www.ap.com/news/fft-scaling-for-noise
This effect is sometimes called the "FFT gain".

 

[Ghoostknight]

-3.6db dac gain:

-10db dac gain:

ADC without Input:

 

[Ghoostknight]

Can i take from this that noise is "always" higher than any distortion component? and distortion just shows up because of the FFT processing?

 

[VNandor]

In the first measurement it seems like the noise level is higher than the THD. The noise (-85dB) and THD+Noise (-84.7dB) are basically the same. This implies the THD is an order of magnitude lower than the noise.

In your second picture with the reduced gain, it looks like the noise level remains the same (-85.3dB) but THD+Noise rises noticably (-78.7dB). I'm not sure why that is since if I look at the individual harmonics, they are supposed to be far below the noise, so even after they are summed together, it shouldn't increase the THD+N so much. Maybe I'm reading some of these numbers wrong? I'm not familiar with REW and I'm not about to get familiar with it.

Can i take from this that noise is "always" higher than any distortion component? and distortion just shows up because of the FFT processing?

Fundamentally, the FFT lets you "see" the signal even if the signal is below the noise floor in some cases. The FFT window must be long enough, should be weighted properly to reduce spectral leakage, the noise spectrum should be reasonably even, and the signal should not be wide-band, it should consist of just some distinct harmonics. When these conditions apply, it's possible to detect the signal even if its level is below the noise floor's level. How well it can be detected depends on how well these conditions apply.
In this case, the distortion is very likely always there but way below the noise floor. You use a long window with weighting, the signal you want to see is just a couple of harmonics and the noise is even so the harmonics show up on the graph.

 

[Ghoostknight]

In your second picture with the reduced gain, it looks like the noise level remains the same (-85.3dB) but THD+Noise rises noticably (-78.7dB). I'm not sure why that is since if I look at the individual harmonics, they are supposed to be far below the noise, so even after they are summed together, it shouldn't increase the THD+N so much. Maybe I'm reading some of these numbers wrong? I'm not familiar with REW and I'm not about to get familiar with it.

the -78.7db are also SNR, maybe this gets calculated into the thd+noise measurement?

Here with distortion lowpass at 20khz gain -10db:

-92db noise is probably more in line with what amir measured
it makes sense that SNR goes down with reducing volume... i guess the thd+n measurement takes the signal level also into account, which makes kind of sense