Is this normal behaviour for dacs?

[Ghoostknight]

Hello,

i just did some measurements of my interface which made me think....

(BEFORE i did measurements i always had a feeling that dacs sound "better" with a little bit of headroom)

First, take a look what i did:

Dac gain was at -3.6db (just enough to avoid clipping) black graph:

As you see, there is some harmonic distortion going...

Now the same measurement with -10db (maybe it was -12-14db, i forgot) black graph

As you see, peak volume is at -10db and the 25db higher than noisefloor H3 completely dissappeared, i always thought harmonic distortion is some relative thing to the input signal, this measurements looks like this assumption was wrong?

i mean in hindsight it makes sense, since engineers talk about "overdriving amps" and such, but in my case i think i even notice a audible difference?

Also i thought it was quite interesting that on amir measurements, where this interface "really performs bad in his opinion" has in this graph the most linearity:

Is this kind of linearity up to -5db a indicator for low distortion (or this sort of overdriving)?

Let me know what you think, since imo this is the first real hint why i like headroom on my dac, even when all objectivists say it doesnt matter....

 

[Ghoostknight]

Oh, and just for completletness, my tests were done with DAC routed to ADC with zero gain

Here is the ADC without input with zero gain and max gain:

 

[Ghoostknight]

What i think of this..... isnt amir giving a completely different picture with testing dacs without headroom if this sort of non-linear harmonic distortion is a thing?

 

[danadam]

Why don't you show distortions computed by REW? (first button at the top: "Show distortion")
And then I would suggest to configure REW to show distortion in dB: Settings (the gear icon at top right), then "Distortion settings" button, then "Show distortion in dB":

 

[gregorio]

BEFORE i did measurements i always had a feeling that dacs sound "better" with a little bit of headroom …
Let me know what you think, since imo this is the first real hint why i like headroom on my dac, even when all objectivists say it doesnt matter....

What objectivist has ever said that headroom doesn’t matter in a DAC? As a supposed objectivist, can I play that game too and state that subjectivist audiophiles “say it doesn’t matter” if a DAC is broken?

Maybe you’re just confused about where headroom matters? Headroom definitely matters when recording, where 18dB or more of headroom is sometimes required. Headroom doesn’t really matter in the audio files distributed to consumers, although it is wise (and has always been industry standard practice) to allow a tiny bit, roughly 0.3dB with CDs but up to around 1.0dB. Headroom in a DAC definitely does matter though, to avoid distortion arising from inter-sample peaks (ISP) when oversampling. An absolute minimum of 3dB headroom and ideally a few dB more. However, pretty much all DACs, going back 35 years or so ago, provide at least 3dB of headroom, certainly all the professional ones do and even all the cheap consumer DACs do, as far as I’m aware. It’s unclear exactly what you meant by “DAC gain” but if I assume it refers to the digital level then it would appear your DAC has little/no headroom, which would be an exceptionally serious design flaw/error!

but in my case i think i even notice a audible difference?

Look at the graph you yourself posted: If your playback level were just over 100dB SPL peak, then the peak distortion level would be at 0dB SPL. How do you think that would be audible?

Also i thought it was quite interesting that on amir measurements, where this interface "really performs bad in his opinion" has in this graph the most linearity…

But Amir did NOT state that (IMD) is “where this interface performs bad in his opinion”, you just made up that false assertion, just like you made up the false assertion that objectivists say that headroom in a DAC doesn’t matter!

According to Amir’s measurements there are quite a few places where this DAC performs badly and some aspects even worse than it’s IMD performance: The noise floor is pretty poor, especially the artefacts from noise shaped dither, which presumably contributes to the poor IMD performance. Also the jitter noise/distortion is not good by current standards, the THD+N in response to freq sweep is particularly terrible (-84dBFS, really?) and harmonic distortion peaks are also poor (roughly -108dB). All this is bad but almost certainly inaudible, however what would be audible is its output level, which is atrocious! This ADC/DAC is marketed as a professional unit but is incapable of producing a professional output level, in fact it’s nearly a whole volt down on pro-level, although that can of course be mitigated by higher amp gain. All in all, there are better performing consumer/pro-sumer units at nearly a quarter of the price of this so called professional unit!

Lastly, what has the IMD decreasing in a straight line got to do with anything, why do you think it’s “quite interesting”? Continuing:

Is this kind of linearity up to -5db a indicator for low distortion (or this sort of overdriving)?

No, it’s the actual level of IMD that indicates the amount of distortion, not whether the IMD increases or decreases in a straight line!

Again, you have claimed to be, or at least implied, you’re a professional audio engineer but not even a student engineer wouldn’t know any of the above, let alone a professional one! Incidentally, Black Lion Audio have been a snake oil company for many years, they survived in the pro-audio world by selling “mods” targeted at newbies/prosumers.

G

 

[Ghoostknight]

What objectivist has ever said that headroom doesn’t matter in a DAC? As a supposed objectivist, can I play that game too and state that subjectivist audiophiles “say it doesn’t matter” if a DAC is broken?

Maybe you’re just confused about where headroom matters? Headroom definitely matters when recording, where 18dB or more of headroom is sometimes required. Headroom doesn’t really matter in the audio files distributed to consumers, although it is wise (and has always been industry standard practice) to allow a tiny bit, roughly 0.3dB with CDs but up to around 1.0dB. Headroom in a DAC definitely does matter though, to avoid distortion arising from inter-sample peaks (ISP) when oversampling. An absolute minimum of 3dB headroom and ideally a few dB more. However, pretty much all DACs, going back 35 years or so ago, provide at least 3dB of headroom, certainly all the professional ones do and even all the cheap consumer DACs do, as far as I’m aware. It’s unclear exactly what you meant by “DAC gain” but if I assume it refers to the digital level then it would appear your DAC has little/no headroom, which would be an exceptionally serious design flaw/error!

Look at the graph you yourself posted: If your playback level were just over 100dB SPL peak, then the peak distortion level would be at 0dB SPL. How do you think that would be audible?

But Amir did NOT state that (IMD) is “where this interface performs bad in his opinion”, you just made up that false assertion, just like you made up the false assertion that objectivists say that headroom in a DAC doesn’t matter!

According to Amir’s measurements there are quite a few places where this DAC performs badly and some aspects even worse than it’s IMD performance: The noise floor is pretty poor, especially the artefacts from noise shaped dither, which presumably contributes to the poor IMD performance. Also the jitter noise/distortion is not good by current standards, the THD+N in response to freq sweep is particularly terrible (-84dBFS, really?) and harmonic distortion peaks are also poor (roughly -108dB). All this is bad but almost certainly inaudible, however what would be audible is its output level, which is atrocious! This ADC/DAC is marketed as a professional unit but is incapable of producing a professional output level, in fact it’s nearly a whole volt down on pro-level, although that can of course be mitigated by higher amp gain. All in all, there are better performing consumer/pro-sumer units at nearly a quarter of the price of this so called professional unit!

Lastly, what has the IMD decreasing in a straight line got to do with anything, why do you think it’s “quite interesting”? Continuing:

No, it’s the actual level of IMD that indicates the amount of distortion, not whether the IMD increases or decreases in a straight line!

Again, you have claimed to be, or at least implied, you’re a professional audio engineer but not even a student engineer wouldn’t know any of the above, let alone a professional one! Incidentally, Black Lion Audio have been a snake oil company for many years, they survived in the pro-audio world by selling “mods” targeted at newbies/prosumers.

G

This isnt about what you assume, i wanna know if this kind of unlinear harmonic distortion behaviour is normal for all dacs\


i repeat my observation ....

-10db peak level, harmonic distortion under -120db, so 110db of "clear dynamic range, without any harmonic distortion"
-3,6db peak level, harmonic distortion sits (H3) sits at -100db, so in this case its only 96db of "clear dynamic range, without any harmonic distortion"


So you get better performance with -10 peak level, IS THIS NORMAL FOR MOST DACS? if yes, it just shows that measurements of nearly zero headroom are pretty much worthless, unless you wanna listen to overdriving a dac

 

[Ghoostknight]

Look at the graph you yourself posted: If your playback level were just over 100dB SPL peak, then the peak distortion level would be at 0dB SPL. How do you think that would be audible?

it just is, i definitely prefer -10db headroom and i did so before the measurements, also with my previous dac where it was worse, i had to use -15db of headroom...

these measurements are just a good hint i got now, why i prefer actually more headroom than the usual -3db to avoid intersample distortion

but its not about audbility claims, i wanna know whether this is actual behaivor of most dacs

 

[Ghoostknight]

No, it’s the actual level of IMD that indicates the amount of distortion, not whether the IMD increases or decreases in a straight line!

.... if its a straight line we can assume its noise that varys in level, if unlinearity happens then this is a indication harmonic distortion gets added up to the noise.... that was my thinking here

so "stuff" is happening inside the actual phone dongle from -25db and up... also the topping changes its behaviour as we go above -20db when i look at this graph

 

[Vamp898]

What is the difference when you test with Multitone?

 

[gregorio]

So you get better performance with -10 peak level, IS THIS NORMAL FOR MOST DACS?

It’s normal for pretty much everything! If you lower the level by 10dB, you don’t just lower the peaks, you lower everything, so yes of course you generally get less noise and distortion.

if yes, it just shows that measurements of nearly zero headroom are pretty much worthless, unless you wanna listen to overdriving a dac

Again, you don’t seem to know how to read a graph and then you’re making up nonsense based on that lack of understanding!

it just is, i definitely prefer -10db headroom and i did so before the measurements, also with my previous dac where it was worse, i had to use -15db of headroom...

So I ask “If your playback level were just over 100dB SPL peak, then the peak distortion level would be at 0dB SPL. How do you think that would be audible?” - your response in a science discussion forum is “it just is” and you wonder why people call you a troll?!

these measurements are just a good hint i got now, why i prefer actually more headroom than the usual -3db to avoid intersample distortion

What intersample distortion do you avoid with more than 3dB of headroom and how do you know -3dB is “the usual”?

so "stuff" is happening inside the actual phone dongle from -25db and up... also the topping changes its behaviour as we go above -20db when i look at this graph

Then “when you look at this graph” you need spectacles, because “as we go above -20dB” the Topping is still producing less IMD than at -25dB! As I stated, you clearly don’t know how to read a graph and how can we have any sort of rational discussion if you are unable to read a simple graph?

G

 

[Ghoostknight]

Then “when you look at this graph” you need spectacles, because “as we go above -20dB” the Topping is still producing less IMD than at -25dB!

but not linear... i know a straight line when i see one

It’s normal for pretty much everything! If you lower the level by 10dB, you don’t just lower the peaks, you lower everything, so yes of course you generally get less noise and distortion.

well i explained it before ... you reduce volume by -10 and harmonic distortion actually goes -25db down, thats the whole reason why i made this thread if you dont know yet..

 

[VNandor]

Is this kind of linearity up to -5db a indicator for low distortion (or this sort of overdriving)?

Let me know what you think, since imo this is the first real hint why i like headroom on my dac, even when all objectivists say it doesnt matter....

These graphs indicate that in terms of IMD, the DACs perform best with full scale or near full scale signals.

 

[Ghoostknight]

These graphs indicate that in terms of IMD, the DACs perform best with full scale or near full scale signals.

yea because volume obviously increases DR ... thats why distortion gets less weighted (unless harmonic distortion creeps in, making the graph going towards full volume less linear...)

either you guys dont understand what im saying or dont "wanna" understand it

 

[VNandor]

It's not a DR measurement or noise measurement, it's an IMD measurement.

 

[Ghoostknight]

... Well

Intermodulation distortion (IMD) is a measure of the nonlinearity of an amplifier. When two or more sinusoidal frequencies are applied to an amplifier, any nonlinear behavior of the amplifier will produce additional frequency components called intermodulation products. For an amplifier with input signals at f1 and f2, the output will contain signals at the following frequencies: nf1 + mf2, where n, m = 0, ±1, ±2, etc. The third-order products, 2f2 – f1 and 2f1 – f2, are a major concern because of their proximity to the fundamental frequencies; and the fact that their power levels increase by a factor of three, relative to an increase in the power level of the fundamental tones. Additionally, their proximity to the fundamental frequencies precludes their removal by filtering. The third-order intercept point (IP3) or the third-order intercept (TOI), often used interchangeably, are figures of merit for intermodulation distortion.

So i guess even Amir doesnt understand what this measurements tells

straight line = good