Is this normal behaviour for dacs?

[VNandor]

If you look at the graph, you can see that Amir implies that lower=better. You keep saying that straight line is good for some reason, I still can't figure out why though.

It seems like measuring the IMD of this DAC is not easy due to the noise produced by the DAC which is something I've just noticed when I tried to look for the place where amir says straight line is good. A high amount of noise could explain why the measured IMD goes down with the test level increasing but it's hard to tell what's really going on from a single image with a bunch of curves and 1.5 half sentences by amir.

 

[Ghoostknight]

If you look at the graph, you can see that Amir implies that lower=better. You keep saying that straight line is good for some reason, I still can't figure out why though.

well imo a bit more noise is less harmful than harmonic distortion

i guess this graphs shows non-linearitys specially once harmonic distortion creeps "above" the noisefloor, its quite logical why other dacs with less noise show earlier IMD

what striked me tho was the relationship this happens in

-3,6db to avoid clipping and we got H3 at -102db (so 99db of "nothing" between signal and H3)
-10db to avoid excessive IMD and we got H3 UNDER the noisefloor at -125db (so we got !!! 115db !!! of "nothing" between signal and noise floor/H3)

if i go by what objectivist say "keep it at near full volume minus 1-3db for intersample clipping" i dont get the best performance.... i hope its clear now what im saying

 

[Ghoostknight]

Here is the full review if you wanna look up "how bad this dac is" https://www.audiosciencereview.com/...-revolution-2x2-review-audio-interface.22141/

 

[VNandor]

if i go by what objectivist say "keep it at near full volume minus 1-3db for intersample clipping" i dont get the best performance.... i hope its clear now what im saying

That depends on what you mean by best performance as it's a fairly meaningless by itself. If you mean lowest amount of THD, maybe you don't get the best performance, I'm not sure. If you mean lowest amount of THD+N or SINAD, you are wrong for most cases. I don't think THD alone is a good metric for "performance". THD is only a measurement of harmonic distortion, so it's possible to have a good THD but bad overall performance. THD+N is a bit more encompassing but the same still applies of course.

Something else to note with these tests is that they are done with test signals. For example the THD is very likely measured with a full scale sine wave which has a -3dB RMS. Music is between -10 and -20dB RMS for most cases. So if you want to figure out how a DAC performs with typical music, a test signal with a -10 ~ -20 dB RMS is very likely more telling than a full scale -3dB RMS signal even though the full scale sine wave would have the same peak level as the music.

 

[Ghoostknight]

Something else to note with these tests is that they are done with test signals. For example the THD is very likely measured with a full scale sine wave which has a -3dB RMS. Music is between -10 and -20dB RMS for most cases. So if you want to figure out how a DAC performs with typical music, a test signal with a -10 ~ -20 dB RMS is very likely more telling than a full scale -3dB RMS signal even though the full scale sine wave would have the same peak level as the music.

How is RMS a good indication here? we also hear peaks, no? and usual music peaks can peak to -1 to -3db (of course frequency dependend or it would sound lets say "not great"

i guess this is also where a 1khz test tone is actually more telling than for example a 10khz one in terms of IMD, music usually has quite reduced highs in comparisons to bass

That depends on what you mean by best performance as it's a fairly meaningless by itself. If you mean lowest amount of THD, maybe you don't get the best performance, I'm not sure. If you mean lowest amount of THD+N or SINAD, you are wrong for most cases. I don't think THD alone is a good metric for "performance". THD is only a measurement of harmonic distortion, so it's possible to have a good THD but bad overall performance. THD+N is a bit more encompassing but the same still applies of course.

Well we would need to ask the question what is more important/audible, specially at these low levels, noise or harmonic distortion

for me its just striking that i basicly get a 16db improvement (IF i value noise and distortion just the same depending on the level) when i turn the volume down by 10db

We probably could argue endlessly again of how audible this theoretical improvement is, but imo its a improvement

the question is whether we value 6-8db of higher noisefloor more than 25db of H3, but 6-8db higher noisefloor gives you a theoretical "ideal" (atleast with the BLA dac) in terms of harmonic distortion, there basicly is none, but the higher noisefloor of course drags the THD+N value way up

 

[Ghoostknight]

a good question probably would be if the additional harmonic distortion of the dac shows up on some loudspeaker in room measurements ... tho i think the umik-1 has far to huge noisefloor for that

 

[bigshot]

The distortion in speakers is orders of magnitude above distortion in DACs. Could you hear a mouse squeak next to a jack hammer? If you want to reduce distortion, you’re barking up the wrong tree. And if you want to know how much distortion is too much, you should do some research into the audibility of distortion and exactly what -70dB means.

 

[danadam]

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

It's only an indication that the distortion is lower than the noise. And because the noise is so high, we can't tell what exactly the value of distortion is. Well, maybe right at the end where the graph flattens, which I think indicates that the distortion more or less equals the noise.

-10db peak level, harmonic distortion under -120db, so 110db of "clear dynamic range, without any harmonic distortion"

110 db of "clear dynamic range"? You realise that the noise floor is at about -95 dBFS? You cannot just visually read it from the graph, because you can make the graph however low you want by increasing the FFT size.

And btw, you have the SINAD vs Level graph in the review, a bit before the IMD graph.

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

Really? So you've seen that Black Lion is also not a straight line?

straight line = good

straight line = N+D is dominated by noise. Not sure why it should necessarily mean "good".

When/if the graph flattens then N+D becomes dominated by distortion.

If the graph starts rising, well, then it's not especially good but AFAICT it doesn't happen often.

well imo a bit more noise is less harmful than harmonic distortion

And according to Amir it's the opposite From the review:

Either way, I am not sure how noise floor can be compromised for some sonic ability when it is so audible unlike distortion.

so we got !!! 115db !!! of "nothing" between signal and noise floor

Again, the noise floor is at -95 dBFS, so there's only 85 dB of "nothing".

 

[danadam]

When/if the graph flattens then N+D becomes dominated by distortion.

If the graph starts rising, well, then it's not especially good but AFAICT it doesn't happen often.

Actually I got it slightly wrong. After reading The Signal Chain: How do noise and distortion propagate through my system?, it's more like:
If the graph flattens then D matches N.
If the graph starts rising then N+D becomes dominated by distortion.

And I don't think such deviation from straight is anything wrong.

 

[gregorio]

The third-order products, 2f2 – f1 and 2f1 – f2, are a major concern because of their proximity to the fundamental frequencies

If your peak playback level is say 85dBSPL then the IMD of the topping, even where it rises at 0dBFS, would be at -10dBSPL. How are you ever going to hear -10dBSPL even in a world class anechoic chamber with no other sound, let alone when you’ve got music playing at 85dBSPL? Your HPs wouldn’t even be able to reproduce that IMD and how is no IMD at all any sort of concern, let alone a “major concern”? That’s ridiculous!

straight line = good

No it’s not, it’s good for a frequency response graph but not for IMD. Don’t you know the difference between a freq response graph and an IMD graph?

a good question probably would be if the additional harmonic distortion of the dac shows up on some loudspeaker in room measurements ... tho i think the umik-1 has far to huge noisefloor for that

Sure, IMD that your speakers probably aren’t even reproducing will show up in your room with a noise floor 100-200 times higher, when reproducing music recordings with a noise floor 100-200 times higher and when using a measurement mic that’s inaccurate by about 50dBSPL. Again, that’s ridiculous!

-3,6db to avoid clipping and we got H3 at -102db (so 99db of "nothing" between signal and H3)
-10db to avoid excessive IMD and we got H3 UNDER the noisefloor at -125db (so we got !!! 115db !!! of "nothing" between signal and noise floor/H3)

What “nothing”, do you live in a world class anechoic chamber where you never replay music recordings?

How is RMS a good indication here? we also hear peaks, no?

You mean apart from the fact that what we hear is far more closely related to RMS level than Peak level?

if i go by what objectivist say "keep it at near full volume minus 1-3db for intersample clipping" i dont get the best performance.... i hope its clear now what im saying

Yes, it’s clear what you’re saying and has been all along. You’ve purchased a $400 ADC/DAC that has worse IMD performance than a $9 Apple dongle, no better performance than a prosumer ~$80 ADC/DAC and a way worse performance than a $200 Topping DAC. But, you can overlook/justify the fact you’ve been scammed by a snake oil company because it has more of a “straight line” looking IMD graph, which you think is prettier to look at!

Well we would need to ask the question what is more important/audible, specially at these low levels, noise or harmonic distortion

Not “we” but you, because apparently you can’t work out for yourself: “What is more important/audible” the say 30-40dBSPL noise floor of your listening environment plus the noise floor of your recording which is somewhere between 25-50dBSPL or the IMD at -10dBSPL that your HPs probably aren’t even reproducing and definitely is not audible even if your HPs were reproducing it?

Again, it is completely impossible to be an audio engineer that does not know any of the above, seemingly cannot work it out but instead, just makes mountains out of molehills (so tiny they can’t be seen or don’t even exist) and then keeps arguing about it despite the obvious facts. And, how is that anything other than trolling? Lastly I’m curious, why would an audio engineer, apparently at a professional/commercial studio, want a 2x2 prosumer ADC/DAC?

G

 

[Ghoostknight]

What “nothing”, do you live in a world class anechoic chamber where you never replay music recordings?

stupid argument, the dac adds his own distortion....

Yes, it’s clear what you’re saying and has been all along. You’ve purchased a $400 ADC/DAC that has worse IMD performance than a $9 Apple dongle, no better performance than a prosumer ~$80 ADC/DAC and a way worse performance than a $200 Topping DAC. But, you can overlook/justify the fact you’ve been scammed by a snake oil company because it has more of a “straight line” looking IMD graph, which you think is prettier to look at!

and i will because it sounds better

 

[VNandor]

How is RMS a good indication here? we also hear peaks, no? and usual music peaks can peak to -1 to -3db (of course frequency dependend or it would sound lets say "not great"

RMS is an average level. The signal spends most of its time around that level. If a signal has a -20dB RMS, there will be only few 0~-1dB peaks, the bulk of the signal will hover around -20dB. The distortion at the typical level where the signal spends most of its time is more important than at the peak levels where it spends a comparatively short amount of time.

 

[bigshot]

I think he’s not a troll. His thinking process is just out of whack and it makes it impossible for him to learn anything. He just keeps thinking the same thing in an endless loop. It is because it is because it is because it is because it is because it is because it is…

 

[gregorio]

stupid argument, the dac adds his own distortion....

No, the “stupid argument” is that far higher distortion is better than lower distortion! According to you, distortion is better if it follows a straight line on a graph, even though it far higher. Not to mention that the distortion of the comparison, far cheaper DAC (Topping) is so low it probably cannot even be reproduced and how is distortion that doesn’t exist not a supremely “stupid argument”?!

i will because it sounds better

Can you provide reliable evidence of any difference whatsoever in the sound at reasonable listening levels (due to IMD in the Topping), or are you just making up falsehoods again?

And lastly, the really “stupid argument” in a science discussion forum is to deflect from answering any of the points/questions put to you and instead just respond with it’s a “stupid argument” (without any reliable evidence), as that is indicative of trolling!

G

 

[Ghoostknight]

No, the “stupid argument” is that far higher distortion is better than lower distortion! According to you, distortion is better if it follows a straight line on a graph, even though it far higher. Not to mention that the distortion of the comparison, far cheaper DAC (Topping) is so low it probably cannot even be reproduced and how is distortion that doesn’t exist not a supremely “stupid argument”?!

because i differentiate noise an harmonic distortion...

Can you provide reliable evidence of any difference whatsoever in the sound at reasonable listening levels (due to IMD in the Topping), or are you just making up falsehoods again?

And lastly, the really “stupid argument” in a science discussion forum is to deflect from answering any of the points/questions put to you and instead just respond with it’s a “stupid argument” (without any reliable evidence), as that is indicative of trolling!

i plan todo a levelmatched youtube comparison with this in the future, imo this is the best way to tell (for others)

and yes i also did it before... imo the difference is pretty clear once you heared it .... thats why i advice other audiophiles to test with a bit of headroom since i suspect other dacs will behave the same near clipping

i also did this with the aune x8 i had before, same conclusion but i needed a bit more headroom