"Things that cannot be measured"

[j_j]

Exactly the right analogy! A Kia Ceed will get 4 people from a to b in 30 minutes, as will a BMW 1 Series, and both will “measure” the same. However, the driving experience is quite different.

That's nonsense, really. One can measure the handling of an automobile. Your measurement is faulty, not the idea of measurement.

 

[j_j]

. We found out that the brain was easily fooled by just about everything.

Ugly color added for emphasis.

 

[j_j]

No, the driving experience is about a range of qualitative aspects such as how the car handles, its response to steering, accelerator and braking inputs, the balance of suspension spring and damping rates, or the way the car behaves when pushed to the limit. None of this is evident from the various measures we use for cars.

I'm sorry, but that's also quite false. You just don't read the right "measures".

 

[j_j]

tens of thousands of dollars of equipment in absolutely horrible rooms would make a good thread...

AMEN

 

[q3cpma]

Sorry, "how it handles" would be like the feel of the knobs, or how easily the menus are navigated.

Unless you're supposed to (literally) constantly use those knobs/menus, the comparison isn't really apt. If he had said "how the handbrake feels", you might have had a point.

 

[RayDunzl]

BOSE came out soon after with the magic cubes. Those 2.5 inch cubes that marketing said sounded like a full range tower.

Well, they did come with a sort-of woofer.

I bought one to have something while I was on the road.

I returned it after two days.

The first day was a holiday and the store was closed.

 

[solderdude]

1. That even the simplest circuit never becomes fully transparent / completely flawless.

It's often the simplest circuits that are the least flawless, no surprise there.

There does not exist a single audio component that is completely flawless nor do they have to be as long as the flaws stay below audible thresholds.

2. even very small insignificant measurable errors can have great significance for what is heard

Then either the measured errors are not insignificant but just low in 'number' or you are looking at the wrong measurement.

 

[mansr]

You've made a mistake.

That's very generous of you.

 

[audiophile]

In a recent heated discussion on the topic of “measurements vs listening” in the “Audiophiles on a budget” Facebook group here’s what famous speaker designer Andrew Jones' (KEF, Infinity, Pioneer, TAD, Elac) posted:

You have to be careful about discussing the merits of the different measurement tools that exist. I have spent a lifetime studying the different ways a speaker can be measured. Its not easy to get the right result. Almost no-one that measures uses a standard sound source to calibrate their measurements. EVERY measurement process and test signal interact with the device under test in some manner. As a result, you cannot just "measure" a speaker and have complete confidence in the measured result if you are new to the game.

For instance, there are very few fully accurate anechoic chambers. The very big ones are few and far between. Even the biggest I have used, in Copenhagen, is not accurate below 50Hz. Small ones such as the one you use in Canada, require some sort of correction curve below about 150Hz. The difficulty with this is that if the correction curve is microphone and speaker position dependent, which is mostly the case, you cannot accurately measure a vent box speaker with a rear vent. You cannot measure a large speaker with multiple woofers. A system like the Klippel makes measuring in a small space more automated, and most likely gives a better measurement than most people would know how to get, but it does not give more accurate results than other methods that are more tedious but possibly superior. Even nearfield measurements for bass response are very prone to error by the uninitiated, and gross error by those that follow some of the published procedures that purport to blend the nearfield and farfield responses.

Never mind that frequency response is a term that refers to a linear system, a class of which a loudspeaker does not belong. Its response is level dependent and time dependent and pre-conditioning dependent: sweep a slow frequency sweep into a some tweeters and at a particular input level, and their apparent frequency response will be different if you sweep from low to high as compared to high to low. What this means is that to experienced designers with a deep understanding of measurements is that a simple set of measurements is insufficient to fully determine how a speaker will sound; that is takes years of measurement and design and listening experience to make judgements on a speakers performance. It becomes easy to see consistent errors in the measurements that some folks show to "characterize" a speakers performance. Of course we can dive down a rabbit hole easily enough, or get into Floyds circle of confusion : how do we know how a speaker should sound on any particular recording? We don't have a clue how that recording was processed, and believe me most times it was heavily processed. Even audiophile standards are not necessarily recorded with no eq or compression or use laboratory mics. So that leaves us in a prickly position if we are to take our polar opposite viewpoints. We cant necessarily trust either the measurements nor the listening. We need experience. Then we have to argue with others over whose experience is "better".

 

[CMOT]

Ugly color added for emphasis.

I feel I have to defend the brain. Characterizing it as being "easily fooled" I think kind of misses the point. The human brain implements a "black box" transfer function between inputs and experience. But this transfer function is knowable. Indeed @j_j your posted lectures elucidate some of the well-established principles of how inputs are processed. We could add in other factors - appearance, cost, etc and still build a transfer function - it might vary across individuals, but it would capture a lot. And many nominal illusions are really just leveraging the sort of inferential processing our brains do in the face of noisy and variable inputs. Perception is a hard problem, but not an unknowable one, even considering the messiness of human preference and context effects. So maybe if we are audio SCIENCE review, we should keep with science across all domains of interest! Let's understand things!

 

[j_j]

I feel I have to defend the brain. Characterizing it as being "easily fooled" I think kind of misses the point.

Because the brain incorporates all senses, pretty much seamlessly, yes, it is quite easily fooled when one one sense is being tested, if there is any teeny-tiny hint from any other sense.

So the statement is true.

The brain is great at integrating all of the senses, all of the time. That's how it can be fooled.

 

[CMOT]

Because the brain incorporates all senses, pretty much seamlessly, yes, it is quite easily fooled when one one sense is being tested, if there is any teeny-tiny hint from any other sense.

So the statement is true.

The brain is great at integrating all of the senses, all of the time. That's how it can be fooled.

I think my objection is to the use of the term "fooled" as in "stupid brain..." - but I totally agree that our brains/experience is integrated across modalities. Experience is never modality specific - it is experience. So our perceptual experiences reflect inferences from a variety of inputs and can, sometimes, create experiences that integrate things in a manner that deviates from the world specified by the incoming physical data. But because perception is inferential and there are lots of good reasons to integrate in this manner. Much of the time, this sort of inference is what keeps us alive, but it doesn't always align well with specific perceptual goals, for example, accurate judgments of different audio signals. So maybe "The brain is great at integrating all of the senses, all of the time. That's how it can lead to non-veridical experiences"? or "inferences that deviate from the physical facts of the inputs"?

 

[NTK]

In a recent heated discussion on the topic of “measurements vs listening” in the “Audiophiles on a budget” Facebook group here’s what famous speaker designer Andrew Jones' (KEF, Infinity, Pioneer, TAD, Elac) posted:
...

It seems that AJ is in the camp that "measurements can tell us a hell of a lot, but only when I am the one doing the measurements". So, he alone is the measurement guru. Interestingly that Elac isn't at all forthcoming with measurement data of their products. I can't find a single FR graph in their website.

Also interesting is that he talked about the accuracy of FR measurements in the bass frequencies. Doesn't he know that a flat FR curve is not that important for bass? The room dominate the FR below ~200-300 Hz. EQ based on room measurements is a must if you want good bass. For bass, quantity is quality (think ANSI/CTA-2010 type measurements). Why should I care if the anechoic FR is flat there?

Clip from a Positive Feedback interview.
https://positive-feedback.com/Issue30/andrew_jones.htm

 

[solderdude]

In a recent heated discussion on the topic of “measurements vs listening” in the “Audiophiles on a budget” Facebook group here’s what famous speaker designer Andrew Jones' (KEF, Infinity, Pioneer, TAD, Elac) posted:

“You have to be careful about discussing the merits of the different measurement tools that exist. I have spent a lifetime studying the different ways a speaker can be measured. Its not easy to get the right result. Almost no-one that measures uses a standard sound source to calibrate their measurements. EVERY measurement process and test signal interact with the device under test in some manner. As a result, you cannot just "measure" a speaker and have complete confidence in the measured result if you are new to the game. For instance, there are very few fully accurate anechoic chambers. The very big ones are few and far between. Even the biggest I have used, in Copenhagen, is not accurate below 50Hz. Small ones such as the one you use in Canada, require some sort of correction curve below about 150Hz. The difficulty with this is that if the correction curve is microphone and speaker position dependent, which is mostly the case, you cannot accurately measure a vent box speaker with a rear vent. You cannot measure a large speaker with multiple woofers. A system like the Klippel makes measuring in a small space more automated, and most likely gives a better measurement than most people would know how to get, but it does not give more accurate results than other methods that are more tedious but possibly superior. Even nearfield measurements for bass response are very prone to error by the uninitiated, and gross error by those that follow some of the published procedures that purport to blend the nearfield and farfield responses. Never mind that frequency response is a term that refers to a linear system, a class of which a loudspeaker does not belong. Its response is level dependent and time dependent and pre-conditioning dependent: sweep a slow frequency sweep into a some tweeters and at a particular input level, and their apparent frequency response will be different if you sweep from low to high as compared to high to low. What this means is that to experienced designers with a deep understanding of measurements is that a simple set of measurements is insufficient to fully determine how a speaker will sound; that is takes years of measurement and design and listening experience to make judgements on a speakers performance. It becomes easy to see consistent errors in the measurements that some folks show to "characterize" a speakers performance. Of course we can dive down a rabbit hole easily enough, or get into Floyds circle of confusion : how do we know how a speaker should sound on any particular recording? We don't have a clue how that recording was processed, and believe me most times it was heavily processed. Even audiophile standards are not necessarily recorded with no eq or compression or use laboratory mics. So that leaves us in a prickly position if we are to take our polar opposite viewpoints. We cant necessarily trust either the measurements nor the listening. We need experience. Then we have to argue with others over whose experience is "better".
You have to be careful about discussing the merits of the different measurement tools that exist. I have spent a lifetime studying the different ways a speaker can be measured. Its not easy to get the right result. Almost no-one that measures uses a standard sound source to calibrate their measurements. EVERY measurement process and test signal interact with the device under test in some manner. As a result, you cannot just "measure" a speaker and have complete confidence in the measured result if you are new to the game. For instance, there are very few fully accurate anechoic chambers. The very big ones are few and far between. Even the biggest I have used, in Copenhagen, is not accurate below 50Hz. Small ones such as the one you use in Canada, require some sort of correction curve below about 150Hz. The difficulty with this is that if the correction curve is microphone and speaker position dependent, which is mostly the case, you cannot accurately measure a vent box speaker with a rear vent. You cannot measure a large speaker with multiple woofers. A system like the Klippel makes measuring in a small space more automated, and most likely gives a better measurement than most people would know how to get, but it does not give more accurate results than other methods that are more tedious but possibly superior. Even nearfield measurements for bass response are very prone to error by the uninitiated, and gross error by those that follow some of the published procedures that purport to blend the nearfield and farfield responses. Never mind that frequency response is a term that refers to a linear system, a class of which a loudspeaker does not belong. Its response is level dependent and time dependent and pre-conditioning dependent: sweep a slow frequency sweep into a some tweeters and at a particular input level, and their apparent frequency response will be different if you sweep from low to high as compared to high to low. What this means is that to experienced designers with a deep understanding of measurements is that a simple set of measurements is insufficient to fully determine how a speaker will sound; that is takes years of measurement and design and listening experience to make judgements on a speakers performance. It becomes easy to see consistent errors in the measurements that some folks show to "characterize" a speakers performance. Of course we can dive down a rabbit hole easily enough, or get into Floyds circle of confusion : how do we know how a speaker should sound on any particular recording? We don't have a clue how that recording was processed, and believe me most times it was heavily processed. Even audiophile standards are not necessarily recorded with no eq or compression or use laboratory mics. So that leaves us in a prickly position if we are to take our polar opposite viewpoints. We cant necessarily trust either the measurements nor the listening. We need experience. Then we have to argue with others over whose experience is "better".

speakers (in rooms) and headphones... yes. For electronics this all does not apply.
The funny bit is that while this is true for transducers most folks also think it applies to cables and electronics. That's where they are wrong and double wrong (like your post is 2x the same WOT) when they think their ears are 'better instruments' and triple wrong when they believe sighted tests are even 'better'.

 

[kristiansen]

You've made a mistake. You must look at short-term error spectrum vs. short-term signal spectrum. Then error matters.

Possibly but a measurement will not tell you how the error sounds or what meaning it has and there will always be errors bigger and smaller, only your listening experience and others' listening experience can in some cases give the measurement an audio signature.

As Jonathan Novick show in his video, but notes that the audio signature he demonstrates is only something that sounds unpleasant, and is not what I would call audio signature, this is typical of errors that can be measured , they give inaccurate and unpleasant sound.
But not a more natural sound of violin, a more correct timing, a better soundstage / imaging etc.

I have built many Hifi devices of all types, and have gradually built up an experience for what works and what does not, And I can therefore say with reasonable certainty that a construction or modification will sound good before I have heard it.
This experience has been gained through me and others' listening experiences, and they are in some cases, but far from all, confirmed through measurement.
In the vast majority of cases, the better sound signature / better timbre / better soundstage cannot be confirmed by measurement.

The only measurement says that if you do and achieve better sound, it seems that the device will not measure worse than before, No one knows what's on the recording except music, so why not choose the solution that sounds best / and most lifelike. if it does not make the measurements significantly worse????.

 

[solderdude]

only your listening experience and others' listening experience can in some cases give the measurement an audio signature.

correction: only your listening experience and others' listening experience can in some cases show an audio signature when tested while knowing what to listen out for in 'blinded' and objectively verified and correct testing with statistical relevance.

This experience has been gained through my and others' listening experiences, and they are in some cases, but far from all, confirmed through measurement.
In the vast majority of cases, the better sound signature / better timbre/ better soundstage cannot be confirmed by measurement.

Then you are not measuring the correct things or you cheated and peeked. It is that simple.

so why not choose the solution that sounds best / and most lifelike. if it does not make the measurements significantly worse????.

Why choose including the eyes and knowing the change when you're only objective is sound ?

This last question was entirely rhetorical of course. The reason is clear. Not knowing removes the 'magic' the brain applies.

 

[Roland]

But those are not core functions of an automobile. Sorry, "how it handles" would be like the feel of the knobs, or how easily the menus are navigated. It's not about whether the car can get you from point A to point B at such and such a speed (which is analogous to the SOUND of your system). All the things you mention are INCIDENTALS to the primary purpose of the car, whether they make it more enjoyable or not. Plus, those are measurable as well.

For those who enjoy driving, a car is so much more than a means of transport. An M2 BMW is a very different car from the ordinary 240i, but it measures much the same. Once you have active damping and steering, carbon brakes and pure 50:50 weight distribution a car becomes quite different to drive. If you enjoyed driving you wouldn’t buy a people carrier, even though it does a lot of things quite well. Similarly an AVR does lots of things quite well, but you’d buy something built for purpose if you were serious about listening to music in stereo.

 

[rdenney]

... Once you have active damping

Measurable.

and steering,

Measurable.

carbon brakes

Measurable.

and pure 50:50 weight distribution

Measurable.

a car becomes quite different to drive...

Measurable.

All of these are objectively measurable, designable, and can be detected in controlled subjective testing. Their importance can also be measured in track times, though that requires statistical care just like controlled subjective testing.

Rick “if it measures ‘much the same’, why would anybody do it?” Denney

 

[audiophile]

your post is 2x the same

Thanks, fixed it.

 

[Roland]

Measurable. Measurable. Measurable. Measurable. Measurable.

All of these are objectively measurable, designable, and can be detected in controlled subjective testing. Their importance can also be measured in track times, though that requires statistical care just like controlled subjective testing.

Rick “if it measures ‘much the same’, why would anybody do it?” Denney

You can’t measure how well a car handles. It’s not about track times - that’s a crude metric - it’s about how it feels to drive, how much you feel in control of it in different road conditions and speeds. A car can be more than the sum of the parts if the parts are designed to work together beautifully.

Talking of more than the sum of the parts, if you take a bog standard Nad amp like a c300 and install Panasonic FM/FC power caps, Elna smoothing caps, Wima input caps, Wima / polystyrene driver board audio path caps, film / Elna Cerafine / Nichicon Muse other audio path caps it will sound much, much better than stock, although I doubt it measures any different. Just like a Hypex NC500 sounds quite different with a Sparkos SS2590 op amp compared with standard, but it’ll measure the same.