• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

The ASR "objectivists tribe" ;) has to read the Keynotes at the AES160th by Lars Risbo

Status
Not open for further replies.
amirm said:
The time separation occurs due to nature of the test signal. It is not at all how distortion gets created otherwise, i.e. with music. And at any rate, distortion shows up as "negative time" (prior to linear impulse) which naturally does not exist.
Click to expand...
yes by design of the stimulus in this very specific case. the time separation is simply how we can measure harmonics in a continuous sweep. it does of cause not break causality. This is why the separation of harmonics collapses at low frequencies (we need a very slow sweep to resolve down below 30Hz)
 
Which is not an issue. My current sweeps have 0.7 Hz resolution, plenty to resolve the low frequency response and distortion. And I could lengthen that even more if needed.

I am still not following why this is a reason to think distortion products in music reproduction are temporally dislocated.
 
MattHooper said:
Let’s remember that you don’t need to have a technical explanation for a phenomenon before testing for it. Observations often come first; technical or scientific explanations later.




This shouldn’t be that hard.

Take the two photos below. And imagine we have no technical explanation for the difference in these photos yet.

But there is an observable difference.

View attachment 536611

So how do we describe the difference?

Well, it’s fairly natural to apprehend the left photo as having the type of granular, speckled texture we would associate with something like grains of sand. So it’s not surprising that the type of distortion on the photo on the left has come to be known “grainy” images.

And you could run tests to see whether people perceive this texture or can distinguish photos via this grainy texture.

The same issues carry over to identifying different sonic traits.

So we could say, in sound quality terms:

Grain is the perception of a fine-textured roughness superimposed on sounds that should be smooth and continuous.

Like the audible equivalent to the difference between a piece of polished glass versus fine sandpaper.

Or in sonic terms it can mimic the quality of lightly rubbing sandpaper on an object, lightly pouring fine sand, or grains of sand being agitated over a sheet of paper. Ultimately when this is imposed on a recording it makes for a slightly rough, granular texture.

Grain imposed on an otherwise smooth recording can cause the perception that, for instance: Violins sound slightly scratchy rather than silky. Cymbals might sound sandy rather than shimmering. Quiet details may seem surrounded by a subtle haze.

And there’s nothing in principle against the idea that one can do controlled listening tests for whether an individual or groups can distinguish between sounds that have this added textural quality and those that don’t.
And then we can seek an explanation for what is causing that specific type of Sonic texture.

Or…conversely… one may technically identify a distortion via measurements… and then test whether the distortion is audible, at which point the character of that distortion can be described as “grain.”

(I would certainly describe some of the distortions I can hear from a vinyl record as grainy…. And I would describe sometimes the effect of close reflective wall surfaces to a loudspeaker as adding a hashy or grainy/granular distortion in upper frequencies. Also…in my DAW I have plug-ins some of which are used to degrade or low fi the sound or add sci fi character… and some of them have names like “granulation” and…yup… they make the sound “grainy” … so I guess those who designed such plug-ins were able to arrive at the technical features required to produce the sense of granulation).
Click to expand...
That is noise. Measurable in vision and sound.
 
amirm said:
Which is not an issue. My current sweeps have 0.7 Hz resolution, plenty to resolve the low frequency response and distortion. And I could lengthen that even more if needed.
Click to expand...
Using ESS or stepped sine? ESS does not separate harmonics at very low frequencies. It requires a very slow sweep and it approaches a stepped sine. The 0.7 Hz is that from the FFTwindow length ? that is another thing.
amirm said:
I am still not following why this is a reason to think distortion products in music reproduction are temporally dislocated.
Click to expand...
I don’t recall ever saying that. It’s the case for an ESS sweep. A glissando or bird chirp might share some properties. This is all the open questions.
 
Lars Risbo said:
The 0.7 Hz is that from the FFTwindow length ?
Click to expand...
No. It is computed from the CHIRP length. I really don't understand where you are going with this. Why do you need such specific, ultra high resolution distortion measurement in bass???
 
I enjoyed reading the article in general, but that section about the Farina Log Sweep is extremely misleading, actually outright wrong. I file that under "brain fart" that can happen even to the brightest minds. There is something missing between what we read and interpret and what Lars actually wanted to bring across. The wording didn't work out.
 
amirm said:
Then what are you saying??? How a measurement protocol separates distortion from the signal has nothing to do with audibility of said distortion.
Click to expand...
The ESS point is specifically about what it reveals about the ear’s temporal sensitivity — not a claim that music always has the same temporal structure as a sweep. Those are two different things.
 
amirm said:
No. It is computed from the CHIRP length. I really don't understand where you are going with this. Why do you need such specific, ultra high resolution distortion measurement in bass???
Click to expand...
the chirp length has nothing to do with the spectral resolution. you can use the shortest chirp and deconvolve 10 sec IR at get 0.1Hz resolution. The ability to resolve harmonics in the bass is a different story.
 
KSTR said:
I enjoyed reading the article in general, but that section about the Farina Log Sweep is extremely misleading, actually outright wrong. I file that under "brain fart" that can happen even to the brightest minds. There is something missing between what we read and interpret and what Lars actually wanted to bring across. The wording didn't work out.
Click to expand...
maybe you could elaborate more on why you think it’s wrong?
 
Lars Risbo said:
maybe you could elaborate more on why you think it’s wrong?
Click to expand...
Quoting, "The exponential sine sweep, the standard method we use to generate THD plots, works by spreading harmonics out across time. The harmonics of the sweep arrive at the listener’s ear at different times."
(Bold mine)

When you listen to a slowly swept sine -- or constant sine for that matter -- all the harmonics of the current frequency are generated at the same time and we hear the corresponding "chord pattern". How could it not be like that, by what magic? It is not like when the signal passes, say, 100Hz fundamental you hear the 200Hz overtone at a different time (later, earlier?). But that's what you wrote, the way people are reading it.

What I think was going on in your mind might have been this: After the convolution with the sweep's inverse to obtain the impulse response from the recorded log sweep the harmonic pulses are separated from the main pulse and separated from each other, which is the beauty of the Farina method. Depending on sweep direction the harmonic pulse land before, or after, the main pulse. The second harmonic is spaced from the main pulse determined by the octaves-per-second sweep rate, H2 pulse is some time apart from the main pulse (and H3 is separated from H2 by a smaller time interval, and so on.. we all know Farina's seminal paper where this is explained). That also poses the restrictions on how far the block of data used for an FFT can stretch out before adjacent pulses bleed in.

However, we don't listen to that impulse response and If we did, we'd just notice that there are several pulses spaced apart, which don't have a distortion quality to them. They represent the distortion, of course. That's why your original statement does not make sense to the skilled reader and my thinking is you happened to mix things up, confusing the domains.
Therefore we hope you can clarify what you actually meant to say.
 
MattHooper said:
“ an elephant in my kitchen” is a ‘something” claim that can be evaluated.
Click to expand...
I'd like to see you prove it. You may have an observation or two, but how reliable are your senses in reality? Do you really see what's in the kitchen, or is your mind playing tricks? And are you sure that everybody who has ever lived, is living, or will live in the future will agree with you that there is no elephant in your kitchen? And if someone doesn't agree, then is your observation valid or is theirs? And how do you prove that?
PristineSound said:
A long time ago, I have also thought about the possibility of such "temporal distortion" and wondered to myself, are there any tests that samples the input signal over a period of time at a high resolution, I don't know, say at 44.1kHz, and compare it with the output signal over the same period of time with the same sampling rate.

Does such test exist? If so, I think that can be a candidate measurement such "temporal variation" as reference to by Lars. If such test does not exist, I can't imagine it being hard to create such a test.

In fact, @pkane, can you use Claude to add this code to multitone?
Click to expand...

Not sure there's anything new here from a measurement perspective. Delayed harmonics/IMD will show up in normal THD/IMD FFT plot just with different phase angles. I think it's really a psycho-acoustic claim that delayed harmonics generated by the driver are not masked by the original signal. This has more to do with the measurement and audibility testing of speaker drivers than any electronics, since any delayed echo in electronics will contain the signal and its harmonics, and so should be subject to the same critical bandwidth masking.
 
test1223 said:
I think the pendulum started with many believers in every snake-oil audio product. As a result, it has swung to the other extreme on this forum: believing only in standard measurements and denying almost all potential improvements that fall outside that scope. This may not be your position, but it is certainly a viewpoint that exists here.

It creates an environment that is not particularly friendly to innovation. For example, Purifi has delivered genuine audio innovations, as have many other audio companies that tend to fly under the radar in this regard. Why are so many of these companies under the radar? Because they do not want to fight a battle with the "objectivist tribe", reveal their innovations to competitors in the process, and lose anyway, since many of you are unwilling to buy a slightly better product if it costs much more due to the research and development behind it.

As a result, these companies often blend in with the "subjectivist tribe" to some extent and try to convince people through the perceived sound quality of their products instead.

Needless to say, the current state of the audio community is not particularly satisfying for these kinds of engineers and companies.
Click to expand...
I don’t agree with many of your points.

People on this forum trust measurements because they have been backed-up by rigorous science. The science involved both objective measurements and subjective listening. They found extremely strong correlations between the two. That is how we now know we can use objective measurements to predict subjective preference.

If some properly done double blind listening tests were done that demonstrated certain speakers consistently sounded better than others and could not be explained by existing measurements, I don’t think people here would dismiss that at all. They would be extremely curious and want to understand the results.

We now know what your agenda is. This objectivist label is misrepresented nonsense IMO.
 
Last edited:
Keith_W said:
To be clear, I am not suggesting that we should muddy up our systems by injecting noise into our music. There is no evidence for that either way.
Click to expand...
From Izotope:

Clarity​

Overview​

Clarity uses hundreds of bands to maximize the spectral power of your audio. This is achieved by adaptively contouring the spectrum towards flat noise. Clarity will tame resonances and unmask background content. This can effectively “pull the blanket off” dull mixes without sounding harsh.

In scientific, blind listening tests we have also found that the Clarity processing can increase the perceived loudness of a master at the exact same integrated LUFS.

Controls​

The Clarity module includes the following controls:

  • Amount: Scales the gain of Clarity’s spectral processing. At 100, the maximum gain that Clarity can possibly boost or cut is 6 dB.
  • Tilt: Adjusts the flat target that Clarity is contouring towards. At the default 0 dB/oct, the target is pink noise. This can be adjusted in dB/oct where positive values are brighter and negative values are darker. For example, tilting +3dB/oct will target white noise. A Tilt of -3 dB/oct will target brown noise.
  • Attack: Adjusts the amount of time (in milliseconds), for Clarity to react to the incoming signal.
  • Release: Adjusts the amount of time (milliseconds) for Clarity to return to a baseline of no gain change.
Click to expand...
 
KSTR said:
Quoting, "The exponential sine sweep, the standard method we use to generate THD plots, works by spreading harmonics out across time. The harmonics of the sweep arrive at the listener’s ear at different times."
(Bold mine)

When you listen to a slowly swept sine -- or constant sine for that matter -- all the harmonics of the current frequency are generated at the same time and we hear the corresponding "chord pattern". How could it not be like that, by what magic? It is not like when the signal passes, say, 100Hz fundamental you hear the 200Hz overtone at a different time (later, earlier?). But that's what you wrote, the way people are reading it.

What I think was going on in your mind might have been this: After the convolution with the sweep's inverse to obtain the impulse response from the recorded log sweep the harmonic pulses are separated from the main pulse and separated from each other, which is the beauty of the Farina method. Depending on sweep direction the harmonic pulse land before, or after, the main pulse. The second harmonic is spaced from the main pulse determined by the octaves-per-second sweep rate, H2 pulse is some time apart from the main pulse (and H3 is separated from H2 by a smaller time interval, and so on.. we all know Farina's seminal paper where this is explained). That also poses the restrictions on how far the block of data used for an FFT can stretch out before adjacent pulses bleed in.

However, we don't listen to that impulse response and If we did, we'd just notice that there are several pulses spaced apart, which don't have a distortion quality to them. They represent the distortion, of course. That's why your original statement does not make sense to the skilled reader and my thinking is you happened to mix things up, confusing the domains.
Therefore we hope you can clarify what you actually meant to say.
Click to expand...
hi, thanks yes i agree with your criticism. It is not as simple as i put it and I am probably to used to look at the deconvolved response. however, the deconvolution is simply a filtering operation which averages the chirp it self out save for one pulse. the harmonics and their time encoding is in the signal we hear.

Cheers

Lars
 
Buckchester said:
If some properly done double blind listening tests were done that demonstrated certain speakers consistent sounded better than others and could not be explained by existing measurements, I don’t think people here would dismiss that at all. They would be extremely curious and want to understand the results.
Click to expand...
Oh no I don't think so.

There would be accusing of cheating if no third party is integrated into the test process and so it would be very expensive and time consuming.

Revealing the name of the competitor speaker would cause bad blood. A comparison agains older speakers would not be considerd a propper comparison and can also cause bad blood nevertheless. A comparison with other speakers of the same company is more or less pointless.

It would be very likely a disaster marketing wise. A lot of money. Endless discussion which is typically good, but in this case it would be discussion within the scientific community only and all scientific standards would be checked rigorously, since this is how science works but average joe doesn't know this. The average joe thinks these kind of tests where done all the time and will take away the skepticis from the reaction of the audio science community.

No one has the money and resources to repeat the test. So there would be no third party validation...

Sorry but I think it would be expensive and really hard to get a good outcome out of it.
 
I’m surprised at some of the replies by folks who otherwise appear to appreciate science.

pkane said:
I'd like to see you prove it. You may have an observation or two, but how reliable are your senses in reality? Do you really see what's in the kitchen, or is your mind playing tricks? And are you sure that everybody who has ever lived, is living, or will live in the future will agree with you that there is no elephant in your kitchen? And if someone doesn't agree, then is your observation valid or is theirs? And how do you prove that?
Click to expand...

So…let’s just abandon any pretence to scientific knowledge then?

What you’re doing here is exactly what I already headed off in my earlier response to Karmacoma HERE.

Perhaps you didn’t see it so I will just repeat it (with emphasis).

Remember that I was responding to the claim “it's impossible to prove that something doesn't exist, you can only prove that something exists,”

So in the context of empirical knowledge, the word “proof” was used.

As most people here would know,
in empirical fields “proof” is usually shorthand for “demonstrated to a high degree of evidential confidence,” not logical certainty.
Science doesn’t deal in deductive proof. It deals in evidence that reaches a level where the claim becomes the rational conclusion to accept. We’re not omniscient so empirical inferences have more to do with justifying confident levels in the context of possible error. In that sense pragmatism is a component of empirical claims to knowledge. (And everyday rational empirical inferences are continuous with this).

That’s why I replied:

“Under any practical understanding of empirical proof, I can prove to myself or to anybody who joins me that there isn’t a full-size bull elephant in my kitchen right now: By simply walking into the kitchen to observe the elephant is not there.

(to deny this conclusion, you’d have to place such high burdens on empirical proof that you also couldn’t prove something exists).”

So….when it comes to my (or anyone else’s ) ability to determine there is no elephant in my kitchen, the type of questions you imply need answering raises the burden of proof to an impractical degree. The same burden would undermine the ability to determine something exists in my kitchen. They are philosophy class 101 sceptical questions that are answered by the pragmatic version of empirical reasoning we use in science and to get through our daily life.

When scientists determine through observation/tests that the hantavirus is the pathogen responsible for the outbreak on the cruise ship, do you reject this because they haven’t established their senses as reliable?
“ how do we know all of you scientists aren’t hallucinating? And if you present evidence to us how do we know we are not hallucinating?”

Worse: are the scientists not justified in their conclusion unless “everybody who has ever lived, is living, or will live in the future will agree with you that hantivirus is on the cruise ship?”

Silly, right? No empirical conclusions could survive that burden for justification . There’s a good reason we don’t use untenable burdens of “proof” for our empirical conclusions.

Cheers
 
Status
Not open for further replies.
Back
Top Bottom