How much difference in sound quality will you hear between qutest and tone board ?

[maxxevv]

Do you understand though that sibilance is not a bad thing in itself, that it’s a normal part of all speech? Let’s just clear that up first.

Then please send me a recording that you find to be excessively sibilant.

I’ll apply de-essing compression to it to reduce the sibilance.

That will then satisfy you, surely?

Pardon the slightly delayed reply.

Of course I do understand the "SSSSss" part of speech/intonation.

But the question here isn't the recording but rather the observed DAC/amp and output transducer phenomena.

How do you measure / represent the phenomena of sibilance in measurements ?

I get sibilance being exceedingly evident (to the point of irritation) in some earphones while sounding completely "benign" sounding in others.
Frequency response curves of output transducers do not seem to represent this anywhere.

Likewise how the output stage changes in amplifiers alter the prevalence of such phenomena.

 

[Blumlein 88]

Here is a different way to do FR besides a sweep. Stepped sine waves every 1/3 octave. This one was at -63 dbFS and the fuzzy black line below is the noise floor. So still quite flat. No harmonic distortion peaking up from the noise floor.

 

[Blumlein 88]

Pardon the slightly delayed reply.

Of course I do understand the "SSSSss" part of speech/intonation.

But the question here isn't the recording but rather the observed DAC/amp and output transducer phenomena.

How do you measure / represent the phenomena of sibilance in measurements ?

I get sibilance being exceedingly evident (to the point of irritation) in some earphones while sounding completely "benign" sounding in others.
Frequency response curves of output transducers do not seem to represent this anywhere.

Likewise how the output stage changes in amplifiers alter the prevalence of such phenomena.

Sibilance is going to be in the 4-8 kHz range. If a sharp enough peak even 10 kHz can make it happen. Guess what range almost all large condenser mics favored for vocals accentuate? Here is a cardioid LDC which is maybe average on the upper end response.

And some are worse, some recording folks will EQ a voice to get it to cut through a dense mix, and then try and dynamically de-ess it. So that and some distortion can cause harmonics to pile up there so on and so forth.

So I've found it common that some 2 or 3 way speakers with 1st order crossovers are bad to be sibilant. Why? Because they have some real off axis lobing that can include peaks up in that region. Which can splatter all around a room if it isn't an absorbent room.

So maybe its possible to measure a tendency for sibilance. I couldn't do it myself. But things that cause sharp and transient sound levels in that 4-10 kHz range will end up sibilant if the basic FR gets out of control.

 

[maxxevv]

Sibilance is going to be in the 4-8 kHz range. If a sharp enough peak even 10 kHz can make it happen. Guess what range almost all large condenser mics favored for vocals accentuate? Here is a cardioid LDC which is maybe average on the upper end response.

View attachment 26911

And some are worse, some recording folks will EQ a voice to get it to cut through a dense mix, and then try and dynamically de-ess it. So that and some distortion can cause harmonics to pile up there so on and so forth.

So I've found it common that some 2 or 3 way speakers with 1st order crossovers are bad to be sibilant. Why? Because they have some real off axis lobing that can include peaks up in that region. Which can splatter all around a room if it isn't an absorbent room.

So maybe its possible to measure a tendency for sibilance. I couldn't do it myself. But things that cause sharp and transient sound levels in that 4-10 kHz range will end up sibilant if the basic FR gets out of control.

Thanks for trying to explain it in terms of frequencies. It goes some way in sorting that query as there is some coherence with what I have encountered.

From what I have encountered is that it seems to be a combination of what you describe here and also transducer distortion / damping deficiencies at specific frequencies / energy levels.

But frequency response curves of earphones / amps do not seem to hint at where / what to look for "sibilance", which is why I'm asking.

 

[solderdude]

I found headphones that show peaks between 6 and 8kHz are likely to have sibilance.
The level of the peak compared to the mids seems to have a direct relation to how bad it is.
The funny thing about this is that I found that sometimes those peaks are not seen in dummy head measurements but do show on my flatbed rig.
I assume, as this is a difficult part of the FR to measure in headphones due to the wavelength and HRTF is most active in this area, that either the fake Pinna/ear canal or its 'compensation' (which is usually obtained using speakers) lowers peaks or applied smoothing so they may not show up in all FR plots.
I do see them show up (with clear relation) in my plots. Notching at those frequency cures the problem but sometimes you get other issues in return.
It's why I do not smooth and measure what is coming from the driver.

Concerning the other issues addressed here: Trying to convince people that are already convinced, but in another way, seems to be futile. Their mind is not open for change nor for evidence on the contrary of their thinking.
I try not to bother but sometimes I can't help myself (depends on my mood)

 

[Calexico]

As noted, it is a theorem, not theory. A theorem has mathematical proof. It means it is correct for all variables and all times. It has no exceptions by definition. Sampling theorem is called that because it indeed has a mathematical proof. Just google for the proof and you will find many examples. Here is the first one I found: https://www.tutorialspoint.com/signals_and_systems/signals_sampling_theorem.htm

View attachment 26901

A theory is composed of some maths and theorems to model the reality no?
I think dac combines signals sampling theorems and not only one that make it difficult to predict exactly the result of the math for a giving approximation without knowing their personal algorithm used.
@Blumlein 88 The flat curve is nice looking at -60db but seems to roll off from about 15 khz (not sure can you give the starting point) of rolling off?. We could check differences of this low level roll off between dacs to see if they all behave the same.
Do you think the roll of is changing with the level?
It's quite -15db roll off it's not nothing.
@amirm did your theory was predicting this low level earlier and quicker roll off in the freq response? If yes why no tests show this ?

 

[Blumlein 88]

A theory is composed of some maths and theorems to model the reality no?
I think dac combines signals sampling theorems and not only one that make it difficult to predict exactly the result of the math for a giving approximation without knowing their personal algorithm used.
@Blumlein 88 The flat curve is nice looking at -60db but seems to roll off from about 15 khz (not sure can you give the starting point) of rolling off?. We could check differences of this low level roll off between dacs to see if they all behave the same.
Do you think the roll of is changing with the level?
It's quite -15db roll off it's not nothing.
@amirm did your theory was predicting this low level earlier and quicker roll off in the freq response? If yes why no tests show this ?

No the roll off is right about 20 khz. That is a logarithmic scale not linear. The steep near vertical line is 22.05 khz. I could pull it up and put a pointer there showing you the values, but it is around 20 khz. The roll off is consistent at various levels. So no tests are needed of this. There are some DACs with different filters that roll off slightly differently than that. Most are audiophile DACs touting minimum phase filters which is mostly a wrong headed idea in the first place.

As for DACs combining signal theorems, sure it is and can be usefully done. There is no difficulty with the prediction. You seem intent on complicating it as touchy to all sorts of issues. Instead it isn't touchy at all. One of the rock solid, easily predictable, reliable, repeatable and accurate things you can run across.

 

[Calexico]

No the roll off is right about 20 khz. That is a logarithmic scale not linear. The steep near vertical line is 22.05 khz. I could pull it up and put a pointer there showing you the values, but it is around 20 khz. The roll off is consistent at various levels. So no tests are needed of this. There are some DACs with different filters that roll off slightly differently than that. Most are audiophile DACs touting minimum phase filters which is mostly a wrong headed idea in the first place.

As for DACs combining signal theorems, sure it is and can be usefully done. There is no difficulty with the prediction. You seem intent on complicating it as touchy to all sorts of issues. Instead it isn't touchy at all. One of the rock solid, easily predictable, reliable, repeatable and accurate things you can run across.

If it's not complicated why do designers always improve the oversampling Technics and filtering in dacs? It's only marketting?
There is no room for improving??
I don't want to complicate things but i wanted to show it's oversimplifying to say that dac apply truthfully the theory.
The truth is that there are technical choices limited by the technology.
No test show where the technical trade off are. Maybe now there is no revealing trade off and you're right i don't know.
But there is Always research i think as we see new Technics on new dacs.
Maybe you re right it's pointless that akm and ess have research for improving their filters.

 

[solderdude]

If it's not complicated why do designers always improve the oversampling Technics and filtering in dacs?

To avoid patents of other manufacturers.
To improve MEASURABLE performance (so they are higher spec'ed than competition)
To allow more format support (higher bitrates/depths)
To compete with other manufactures in price, size or feature set
To give the chip(set) buyer more filter options (the market asks for it)
To have functionality other manufacturers lack
To be able to 'boast' about their technique used (in publications) with fancy words and lots of sales pitches.
To seek out the limits of their capabilities (with financial decisions weighed in).
To seek out new life and civilisations... to boldly go where no designer has gone before ?
To be 'different' in their solutions in order to generate more sales.

The real problem is that 99% of all people using DAC chips (end customers) don't care about the specs nor technique used. They don't care if there are audible differences. They just care about convenience, looks and price of a finished (end) product.
It's only a very piece of the market that actually 'cares' and only a very small percentage of them really understand/comprehend what's in there.
Also a very small amount of people test 'properly' and know their hearing limits. More people THINK they know their limits and usually greatly overestimate their hearing capabilities.

You need to understand your hearing limits first .. Not just think you understand or assume nobody (and certainly technical folks) understands.

 

[andreasmaaan]

I get sibilance being exceedingly evident (to the point of irritation) in some earphones while sounding completely "benign" sounding in others.
Frequency response curves of output transducers do not seem to represent this anywhere.

I see.

Frequency response measurements of headphones and earphones are made using a dummy head. At high frequencies, these measurements are heavily influenced by the geometry of the dummy pinnae. Headphone and earphone measurements are therefore not accurate at high frequencies, as different pinnae will result in different resonances at different frequencies.

The most likely explanation is that the earphones are interacting differently with your pinnae than they are with the pinnae of the dummy heads used to take the measurements you're looking at.

 

[solderdude]

Frequency response measurements of headphones and earphones are made using a dummy head. At high frequencies, these measurements are heavily influenced by the geometry of the dummy pinnae. Headphone and earphone measurements are therefore not accurate at high frequencies, as different pinnae will result in different resonances at different frequencies.

That's my theory (not theorem ) as well (see 5 posts above)

 

[Calexico]

To avoid patents of other manufacturers.
To improve MEASURABLE performance (so they are higher spec'ed than competition)
To allow more format support (higher bitrates/depths)
To compete with other manufactures in price, size or feature set
To give the chip(set) buyer more filter options (the market asks for it)
To have functionality other manufacturers lack
To be able to 'boast' about their technique used (in folders) with fancy words and lots of sales pitches.
To seek out the limits of their capabilities (with financial decisions weighed in).
To be 'different' in their solutions in order to generate more sales.

The real problem is that 99% of all people using DAC chips (end customers) don't care about the specs nor technique used. They don't care if there are audible differences. They just care about convenience, looks and price of a finished (end) product.
It's only a very piece of the market that actually 'cares' and only a very small percentage of them really understand/comprehend what's in there.
Also a very small amount of people test 'properly' and know their hearing limits. More people THINK they know their limits and usually greatly overestimate their hearing capabilities.

You need to understand your hearing limits first .. Not just think you understand or assume nobody (and certainly technical folks) understands.

I agree.
You carricaturate my mind. I know lot people here understand better than me the théorie or theorem whatever.
But i don't like seeing arguments like what you hear is always placebo because of the theory.
The theorie of sampling and the shanon theorem stipulate an infinite frequency for delta sigma to reconstruct the analog wave perfectly. And dacs operates at a finite frequency.
And there are certainly other trade off like this.
Ok it's not necessarily hearable but we shouldn't assume/presume it's always true without validation.

 

[M00ndancer]

But i don't like seeing arguments like what you hear is always placebo because of the theory.

There is no "like". Our brain is really easy to fool. Like is a subjective word. We can't trust our ears. Finally, try to understand that a theorem is not a theory. IT's proven facts.

The blurb about

The theorie of sampling and the shanon theorem stipulate an infinite frequency for delta sigma to reconstruct the analog wave perfectly. And dacs operates at a finite frequency.

Watch the video until you understand.

 

[SIY]

The theorie of sampling and the shanon theorem stipulate an infinite frequency for delta sigma to reconstruct the analog wave perfectly.

No.

Get that refund.

 

[solderdude]

The theorie of sampling and the shanon theorem stipulate an infinite frequency for delta sigma to reconstruct the analog wave perfectly.

Nope it says double the required bandwidth is enough to reconstruct perfectly.
A question could be what is the required bandwidth ... but even here there are bandwidths that far exceed our hearing capabilities.

And dacs operates at a finite frequency.

Yes, the do and this frequency is more than enough.

Ok it's not necessarily hearable but we shouldn't assume/presume it's always true without validation.

Measurements and knowing about hearing capabilities is more than enough validation.
Maybe just not for you.
In this case learning your hearing capabilities and get worthwhile hands-on experience with (test) equipment is what is needed.

 

[Calexico]

Nope it says double the required bandwidth is enough to reconstruct perfectly.
A question could be what is the required bandwidth ... but even here there are bandwidths that far exceed our hearing capabilities.



Yes, the do and this frequency is more than enough.



Measurements and knowing about hearing capabilities is more than enough validation.
Maybe just not for you.
In this case learning your hearing capabilities and get worthwhile hands-on experience with (test) equipment is what is needed.

Shanon say that sampling a frequency at the double of its rate is enough to reconstruct it perfectly i agree. But it needs calculation to obtain it. If you don't make calculation the wave will be squarish at 20khz for a 40khz sampled signal. More steps you calculate between samples more close it becomes to the ideal wave.
The theory say that the math is known to reconstruct perfectly the wave.
Are the dac enough sophisticated to apply the whole math for reconstructing?
And there are different methods of discret interpolation and each one have are better when more calculation is done.
You can interpolate from one sample before and after or from averaging two sampes and make more or less iterations etc...

 

[SIY]

Shanon say that sampling a frequency at the double of its rate is enough to reconstruct it perfectly i agree. But it needs calculation to obtain it. If you don't make calculation the wave will be squarish at 20khz for a 40khz sampled signal.

No.

Get that refund. You deserve it.

 

[solderdude]

There are golden-eared guys who even listen to NOS DACs at 44/16 without any post filtering or digital filtering nor upsampling at all.
They use their ears and limited bandwidths of the used amplifier and transducers to smooth (thus no math).
These folks claim it sounds better to them than filtered DACs.
NO filtering real stairsteps is what they are listening to and they love it.
This is NOT closer to the sampled signal at all (further from it) as samples are points in time and NOS is sample-and-hold in time.
It can reproduce squarewaves and needle pulses perfectly (which are illegal signals and don't exist in music nor recordings) and that's their claim to fame.

The reality is that what they love is the inherent upper treble roll-off due to the lack of proper reconstruction filtering, NOT the absence of pre- or post-'ringing'. The 'ringing' is only at freq. close to half the sample rate and is not there in recorded music.

And yes different filters are used some audibly worse than others and measure different.
There is nothing new to this nor strange.

Measurements of the Qutest and toneboard reveal there are no concerning audible differences in performance and use similar types of filters.
The fact that some of us claim there is no discernable difference (aside from level differences) is based on that knowledge.

This does NOT mean all DACs sound the same... they don't.
Merely the ones that measure in a certain way will sound the same when level matched and under equal conditions.

 

[Calexico]

No.

Get that refund. You deserve it.

Here is the theory
http://vlab.amrita.edu/?sub=3&brch=166&sim=820&cnt=1

Read the reconstruction part.
Here an extract:
The interpolated value at any point is the sum of contributions from infinitely many weighted sinc functions.
Dacs cannot caculate the infinite weighted sinc functions and are limited to number of them. So they approximate the theory.

 

[Calexico]

There are golden-eared guys who even listen to NOS DACs at 44/16 without any post filtering or digital filtering nor upsampling at all.
They use their ears and limited bandwidths of the used amplifier and transducers to smooth (thus no math).
These folks claim it sounds better to them than filtered DACs.
NO filtering real stairsteps is what they are listening to and they love it.
This is NOT closer to the sampled signal at all (further from it) as samples are points in time and NOS is sample-and-hold in time.
It can reproduce squarewaves and needle pulses perfectly (which are illegal signals and don't exist in music nor recordings) and that's their claim to fame.

The reality is that what they love is the inherent upper treble roll-off due to the lack of proper reconstruction filtering, NOT the absence of pre- or post-'ringing'. The 'ringing' is only at freq. close to half the sample rate and is not there in recorded music.

And yes different filters are used some audibly worse than others and measure different.
There is nothing new to this nor strange.

Measurements of the Qutest and toneboard reveal there are no concerning audible differences in performance and use similar types of filters.
The fact that some of us claim there is no discernable difference (aside from level differences) is based on that knowledge.

This does NOT mean all DACs sound the same... they don't.
Merely the ones that measure in a certain way will sound the same when level matched and under equal conditions.

Any signal is composed of an infinite sum of weighted sine waves. When converting dacs cannot calculate an infinite number of sine waves to decompose the signal.
See my link above.