• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. 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!

An Enticing Marketing Story, Theory Without Measurement?

Dialectic

Major Contributor
Forum Donor
Joined
Sep 26, 2017
Messages
1,738
Likes
3,090
Location
a fortified compound
Better a degree in one of the hard sciences than a Master's of Bullshit from the Ronald M. Popeil and Sally Struthers School of Marketing.
You really should read up before you continue. The pace of your posts suggest that you haven't done so, despite having had lots of time on your hands for the past month.
 

Bentoronto

Member
Joined
Jun 27, 2019
Messages
38
Likes
12
Location
Toronto and Delray Beach, FL
Well, at least we've established that your background isn't in EE, physics... or sound. :)
Correct. Bell Labs (Murray Hill*) recruited me for research in perception and cognition. Seems more relevant to Toole's critique than having an EE. Don't you think?

What did you do at Bell Labs?


* this will certainly date me, but Schroeder's office was just down the hall from mine
 

GrimSurfer

Major Contributor
Joined
May 25, 2019
Messages
1,238
Likes
1,484
Correct. Bell Labs (Murray Hill*) recruited me for research in perception and cognition. Seems more relevant to Toole's critique than having an EE. Don't you think?

The only thing that you've contributed to this thread so far is a bunch of made up terms and bafflegab related to photography and art. Now you trot out an affiliation with Bell Labs... and name dropping of guys who worked down the hall.

I'd could be trite and ask you whether your time at Bell Labs was influential enough to get "sound objects" and "Gestalt" into the technical lexicon, but we all know the answer to that.

I started reading your posts out of genuine interest from where Amir and others left off. While I found their posts to be informed and informative, yours didn't seem connected to anything but a platform for social science double speak... which is frequently a prelude to stupid audiophile BS.

Now psych or photography or German philosophy may be your bag... and good for you if it is. But if you look around, you'll see that this forum isn't about that.

So if you want to talk about phase, separation, frequency (audible and inaudible), spl, digital decoding, the Blumlein technique, etc., I'm all ears. But if you wish to "paint an aural canvas across the Gestalt which is ASR", you won't be taken as seriously as you think your background and work experience should confer.
 

Bentoronto

Member
Joined
Jun 27, 2019
Messages
38
Likes
12
Location
Toronto and Delray Beach, FL
In light of the Toole critique, with the subwoofer speaker representing a sound object away in a corner a few feet from the main speakers, would the frequency response experienced by the listener be (a) the sum of the subwoofer and the woofer outputs plotted/swept separately or would it be (b) the plot/sweep of the two played together and interacting within their crossover band?

(As a thought-experiment of the extreme case, if (a) were true, then polarity of the speakers wouldn't matter. Which sounds like an absurd conclusion in theory if not in practice.)

B.
 
Last edited:

edechamps

Addicted to Fun and Learning
Forum Donor
Joined
Nov 21, 2018
Messages
910
Likes
3,620
Location
London, United Kingdom
In light of the Toole critique, with the subwoofer speaker representing a sound object away in a corner a few feet from the main speakers, would the frequency response experienced by the listener be (a) the sum of the subwoofer and the woofer outputs plotted/swept separately or would it be (b) the plot/sweep of the two played together and interacting within their crossover band?

It's (b) of course. Which is precisely why subwoofer have phase controls to ensure the response sums up correctly with the main speakers near the crossover frequency.
 

UliBru

Active Member
Technical Expert
Joined
Jul 10, 2019
Messages
123
Likes
337
Hello,

I've stumbled across this interesting thread and decided to add some thoughts:

1. Room correction is a misnomer. Indeed the room is not corrected. But:
E.g. complex numbers have scared off generations of pupils and students because they combine 'numbers' with complexity. No one will change the name forever.
Now the wrong term room correction is used for more than a decade. Everyone knows what it is intended for despite the room does not corrected. So why to started heated discussions about the name?

2. A perfect sound reproduction system would result in the same left/right sound field close around the ear which is created by the real sound event (we ignore here head turns/movements or bass waves shaking the stomach). Will we ever reach this?
A more restricted view is to perfectly play at least the music track (we ignore the question if the stored track data represent the reality of the sound event). This means that there is a 1:1 transfer from track data to sound waves without any losses or added information/noise/distortion. In a linear time-invariant system a Dirac pulse represents such a transfer. And indeed a Fourier transform leads e.g. to a perfect flat frequency response without any phase changes. Thus a loudspeaker must measure perfectly flat under this assumption.
Unfortunately even perfect speakers would have to play in a real room. And here the sound waves arriving at the listeners ear are no longer perfect. We experience reflections, diffusion, crosstalk, frequency attenuations, frequency peaks and dips, phase changes. As each listening room is different there is also a different result of music perception. We hear the original environment on the track combined the the effects of the playback chain and the playback room.

3. Neglecting non-linear effects the playback system, especially the loudspeaker and the playback room, heavily modifies the result. We may think about speaker and room as filters. So we would get closer to the original content if we could apply inverse filters of speaker and room. The only chance to do this is to apply measurements (even eq'ing the system by listening uses the ear as a measurement device). Thus we try to get pulse responses (completely describing the behaviour of the LTI system, of course only at the measurement point) by using a microphone in combination with appropriate excitation signals. From these pulse responses we can derive much infomation about frequency and phase response and many others. The big question is now: how to get the proper inverse filters from the measured pulses?

4. Unfortunately an inversion of a measured pulse quickly results in numerical instabilities. It is also a question if it is wise to cancel a reflection by the inverse filter. A reflection is like an echo and the science of echo cancellation is not a new topic. But think about an applied cancellation and then simply move around in the room. The reflections and their times will change and the cancellation filter will even lead to a worse result.
So the game is to create inversion filters which do not harm but improve the enjoyment of music listening.

5. There are tons of discussions about frequency responses. A frequency response as a FFT result on a pulse response is a good source of information. BUT: a frequency response is a steady-state result, it does not contain anymore time informations! Feed the system with a sine wave of a given frequency and after some time the amplitude (and phase) will settle down to the result shown in the frequency response. The problem:
MUSIC IS NOT A STEADY-STATE SIGNAL ! A nice lesson: take a short piece of music and calculate the frequency response by FFT. Reverse the music and run the FFT again. It will show up the same frequency response. But of course both tracks will sound very different.

6. The creation of inverse filters includes many different methods and solutions. A simple example is a parametric EQ to reduce a bass boost. It belongs to the same category with misname "room correction". At the end it is the resulting quality that counts. Amongst the methods you find gating or windowing (concentration on the direct sound) which is also equivalent to smoothing (1/n octave, ERB ...) and averaging (e.g. multiple measurements at different positions). This leads of course to different solutions and vendors of room correction as there are different solutions and vendors of loudspeakers and equipment for room treatment.

7. A good room correction filter corrects both speaker and room. Typically crossovers introduce excessphase distortions which can be corrected. So IMHO a correction below the Schroeder frequency is only a part correction. And if we accept that an uncorrected system sounds different at different room positions we should also accept that a corrected system does not sound perfect everywhere. If the overall sound is good and especially improved at the sweet spot we have already done a great job.

7. It is of course wise to select a good speaker and to optimize the playback environment. But it is also interesting how many "good" speakers measure bad. And how many music lovers (and their spouses) are willed or capable to modify their living room into a studio? Today we have a new (considering audio history) and powerful tool (mis)named 'room correction'. Any tool helping to improve our listening enjoyment should be welcome.
 

Cosmik

Major Contributor
Joined
Apr 24, 2016
Messages
3,075
Likes
2,180
Location
UK
1. Room correction is a misnomer. Indeed the room is not corrected.

...So why to started heated discussions about the name?

You may have had me in mind as one of those naughty people. But then you make the case for me:

7. A good room correction filter corrects both speaker and room....

People really do believe that by changing what comes out of the speaker, they have 'corrected the room'. Or 'removed the effect of the room' (which boils down to the same thing). But they haven't. The room remains exactly the same, and the effects of the room remain exactly the same, but now applied to a weirdo sound from the speaker.
 

UliBru

Active Member
Technical Expert
Joined
Jul 10, 2019
Messages
123
Likes
337
People really do believe that by changing what comes out of the speaker, they have 'corrected the room'. Or 'removed the effect of the room' (which boils down to the same thing). But they haven't. The room remains exactly the same, and the effects of the room remain exactly the same, but now applied to a weirdo sound from the speaker.
:)

Think about an audio setup as a chain of elements. If an element has a "fault", e.g. some attenuation of a frequency range, you can of course correct this element if possible. If not, you may correct the fault by another element.
Examples:
- a vinyl has a rising frequency response due to the RIAA eq during the cutting. The cartidge as the next element does not correct this but the following element = phono preamp applies an inverse RIAA eq. Thus the correction is following the fault
- a speaker driver plays too loud due to its higher efficiency. The wise speaker designer corrects this fault by attenuating the voltage for the driver in the speaker crossover. This correction is preceding the faulty element.

You can characterize the transfer behaviour of each element as a filter. The end result is thus the convolution result of the filter elelments. For LTI systems (indeed we truly hope our system belongs to this class) the commutative law is also applicable for convolution. This means we can correct a fault by the element, by previous elements or by following elements.
Example: you can place some porous but absorbing material near to your ears and thus the high frequency range get dampened. If a mastering engineer knows about this he can prep the music track accordingly and you may love to listen to this track.
THIS DOES NOT CORRECT THE SITUATION OF THE ABSORBER, IT IS STILL THERE.

Like the simple examples a room behaves like a filter. The only trouble: it behaves like a very complex filter. Thus it is not possible to correct everything. But anyway the sound gets better despite the room does not change.
Again an example: the room has some reverberation time. In case of a resonance (standing wave) you need little energy to get a loud signal. Whereas other frequencies behave well you get a bass boost at this frequency with same input voltage. When you stop the signal the reverberation time is pretty long. It simply takes time to dissipate the energy. Now a room correction reduces the voltage, it applies an attenuation at htis frequency. This also means that there is much less energy in the room. As the sound level is lower it also takes less time until it is below the listening threshold. So you experience a shorter reverberation time DESPITE the room is not changed at all.
So typically with room correction you can notice a cleaner sound instead of muddiness.

Last but not least it is possible to measure again with applied room correction. Despite the unchanged room the measurement clearly proves the effect of the correction filter.
If you now like to place your ears directly close to the speaker you will of course notice a "weirdo" sound. But this is only because now you take off the element 'room' out of the playback chain ;)
 

Cosmik

Major Contributor
Joined
Apr 24, 2016
Messages
3,075
Likes
2,180
Location
UK
:)

Think about an audio setup as a chain of elements. If an element has a "fault", e.g. some attenuation of a frequency range, you can of course correct this element if possible. If not, you may correct the fault by another element.
Examples:
- a vinyl has a rising frequency response due to the RIAA eq during the cutting. The cartidge as the next element does not correct this but the following element = phono preamp applies an inverse RIAA eq. Thus the correction is following the fault
- a speaker driver plays too loud due to its higher efficiency. The wise speaker designer corrects this fault by attenuating the voltage for the driver in the speaker crossover. This correction is preceding the faulty element.

You can characterize the transfer behaviour of each element as a filter. The end result is thus the convolution result of the filter elelments. For LTI systems (indeed we truly hope our system belongs to this class) the commutative law is also applicable for convolution. This means we can correct a fault by the element, by previous elements or by following elements.
Example: you can place some porous but absorbing material near to your ears and thus the high frequency range get dampened. If a mastering engineer knows about this he can prep the music track accordingly and you may love to listen to this track.
THIS DOES NOT CORRECT THE SITUATION OF THE ABSORBER, IT IS STILL THERE.

Like the simple examples a room behaves like a filter. The only trouble: it behaves like a very complex filter. Thus it is not possible to correct everything. But anyway the sound gets better despite the room does not change.
Again an example: the room has some reverberation time. In case of a resonance (standing wave) you need little energy to get a loud signal. Whereas other frequencies behave well you get a bass boost at this frequency with same input voltage. When you stop the signal the reverberation time is pretty long. It simply takes time to dissipate the energy. Now a room correction reduces the voltage, it applies an attenuation at htis frequency. This also means that there is much less energy in the room. As the sound level is lower it also takes less time until it is below the listening threshold. So you experience a shorter reverberation time DESPITE the room is not changed at all.
So typically with room correction you can notice a cleaner sound instead of muddiness.

Last but not least it is possible to measure again with applied room correction. Despite the unchanged room the measurement clearly proves the effect of the correction filter.
If you now like to place your ears directly close to the speaker you will of course notice a "weirdo" sound. But this is only because now you take off the element 'room' out of the playback chain ;)
It's a fantasy that you can 'correct' just a bit here and a bit there and make it better, but not fully better. It's a misunderstanding of dimensionality. Your hearing has evolved to solve 'sound' as a multidimensional problem where everything tallies: location, frequency response, time response, reverberation, echo. The human listener is more than capable of separating the frequency response of the source from that of the room - unlike your mic and laptop. The human has two ears and a head that turns.

The problem is that you have been seduced by the screen: you are imagining that a human hears how a mic and laptop 'hear'.

It is true that a listener is very discriminating when it comes to frequency response of a speaker. It is not true that the listener hears frequency response of room + speaker combined. Only the laptop is 'hearing' that.
 

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
It is true that a listener is very discriminating when it comes to frequency response of a speaker. It is not true that the listener hears frequency response of room + speaker combined. Only the laptop is 'hearing' that.

I agree with you partially. The extent to which a human hears what a mic “hears” varies, most notably with frequency.
 

Cosmik

Major Contributor
Joined
Apr 24, 2016
Messages
3,075
Likes
2,180
Location
UK
I agree with you partially. The extent to which a human hears what a mic “hears” varies, most notably with frequency.
It boils down to dimensions (or 'variables') and 'quality metrics', or 'scores'. The room correction person thinks that they can tweak a couple of dimensions out of the 27 (or whatever) that are in play and assess the result with a quality metric based on that couple of dimensions. It's all they have available to them or can understand.

The listener, on the other hand, has a different score based on the 27 dimensions. They 'hear' the 27 dimensions, and their 27-dimensional accuracy metric points back to what can only be a weird source.

It's a generally-applicable idea.
 

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
It boils down to dimensions (or 'variables') and 'quality metrics', or 'scores'. The room correction person thinks that they can tweak a couple of dimensions out of the 27 (or whatever) that are in play and assess the result with a quality metric based on that couple of dimensions. It's all they have available to them or can understand.

The listener, on the other hand, has a different score based on the 27 dimensions. They 'hear' the 27 dimensions, and their 27-dimensional accuracy metric points back to what can only be a weird source.

It's a generally-applicable idea.

TBH you’ve lost me here :)

If the steady-state in-room response is perceptually dominant in a particular range of frequencies, the listener will perceive correction in this region as less “weird” than no correction.
 

Bentoronto

Member
Joined
Jun 27, 2019
Messages
38
Likes
12
Location
Toronto and Delray Beach, FL
If the steady-state in-room response is perceptually dominant in a particular range of frequencies, the listener will perceive correction in this region as less “weird” than no correction.
The core of this thread is to examine ways in which that isn't the correct way to look at sound reproduced in a room (Toole's critique). You're just making up stuff that seems logical to you but with no basis in a knowledge of human perception.
 

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
The core of this thread is to examine ways in which that isn't the correct way to look at sound reproduced in a room (Toole's critique). You're just making up stuff that seems logical to you but with no basis in a knowledge of human perception.

Did you read the part of my post that said “in a particular range of frequencies”?

I completely agree with Toole on this topic.

PS let’s please be nice here, I enjoy this forum a lot. Instead of accusing me of “making stuff up”, how about asking for references instead? :)
 
Last edited:

Cosmik

Major Contributor
Joined
Apr 24, 2016
Messages
3,075
Likes
2,180
Location
UK
TBH you’ve lost me here :)

If the steady-state in-room response is perceptually dominant in a particular range of frequencies, the listener will perceive correction in this region as less “weird” than no correction.
Given a selection of monochrome photographic prints with colour casts, do you think you could identify the one that was closest to neutral grey? In tungsten light, sunlight, fluorescent light? I reckon I could. But do you think that that print combined with the lighting would measure absolute neutral grey in every condition? It wouldn't.

The key to identifying it would be to adjust for the context of the lighting; effectively the eye/brain's 'auto white balance'. An extra dimension.

In yellow-ish lighting, the print that measured neutral grey would have a blue colour cast - and be perceived as blue. The true neutral grey print would measure as yellow, but would look neutral in the context of everything else illuminated in the same light.

The laptop and mic is the gadget that picks out the print with the most neutral in-room colour measurement - but looks blue. The human hearing system is what picks out the most neutral print and perceives it as neutral.

The latter doesn't contradict that colour is "dominant" (over what?), but it gives a different result from the former.
 

UliBru

Active Member
Technical Expert
Joined
Jul 10, 2019
Messages
123
Likes
337
I do not claim that we can correct everything. But we can also enjoy partial corrections.

Let's do a thought experiment:

You listen to music but then you leave the room. Does the music stop tp play? It does not matter to the playback system if you are there or not.
Now place the mic at the listening position and do a recording. The mic does not know anything about the playback system, about your ears and about 27 dimensions of brain interpretation. It simply records the arriving sound waves as a time signal (btw typically we listen to music recorded by microphones and not by nerve signals of ears, is the music bad because we have lost 27 dimensions?). The time signal characterizes the sound waves that have passed the mic at this single point.

Now we can do some recording of proper test signals by the mic. We can do a recording with a nice playback environment and we can do the same with removed absorbers and diffusors. Obviously the recordings will be pretty different.
Now the question is: is it possible to find a filter which connects the two recordings. The target is to get the same time signal recording or at least a very similar one when the filter is applied. Remember: the ears are out of the game at this moment. Now let's assume that such a filter exists. So another mic recording with applied filter can prove this.
You come back into the room and place your ear at the mic position. Will your ear notice a difference as the sound wave is the same as before?

Of course this thought experiment is limited: you are not nailed to the mic position. The time signal is dependent of the position in the room. The filter is not ideal. The measurement may have flaws (e.g. mic calibration). There are many obstacles.
But does this mean that you cannot improve something by a filter? A wrong filter can destroy a lot, but not every filter is a bad one.

There is a nice lecture of Amar Bose at MIT, see YouTube starting from about 1:00 to 11:30


It is a very nice experiment which shows how a room modifes the sound. It is worth to repeat it.
But it is also another nice experiment to repeat this with a 'room corrected' system. What is wrong if you can distinguish between speech, music and noise at the end of this experiment?
 
Last edited:

Cosmik

Major Contributor
Joined
Apr 24, 2016
Messages
3,075
Likes
2,180
Location
UK
I do not claim that we can correct everything. But we can also enjoy partial corrections.
.. which assumes there is something to correct. There might be, but it is the speaker not the room. The speaker can be corrected to get it closer to a neutral speaker. And even a non-neutral speaker can be adjusted in the room to get a compromise that works better. But I would claim that this adjustment (or correction if you will) requires some knowledge of the speaker's deficiencies. The simplest, and most important example: baffle step compensation where the frequency is defined by the speakers baffle width, and the depth is set with reference to the room.

Let's do a thought experiment:

You listen to music but then you leave the room. Does the music stop tp play? It does not matter to the playback system if you are there or not.
Now place the mic at the listening position and do a recording. The mic does not know anything about the playback system, about your ears and about 27 dimensions of brain interpretation. It simply records the arriving sound waves as a time signal (btw typically we listen to music recorded by microphones and not by nerve signals of ears, is the music bad because we have lost 27 dimensions?). The time signal characterizes the sound waves that have passed the mic at this single point.

Now we can do some recording of proper test signals by the mic. We can do a recording with a nice playback environment and we can do the same with removed absorbers and diffusors. Obviously the recordings will be pretty different.
Now the question is: is it possible to find a filter which connects the two recordings.

The target is to get the same time signal recording or at least a very similar one when the filter is applied. Remember: the ears are out of the game at this moment. Now let's assume that such a filter exists. So another mic recording with applied filter can prove this.
You come back into the room and place your ear at the mic position. Will your ear notice a difference as the sound wave is the same as before?
Would the same principle apply to, say, a solo performer? Could the performer be made to sound better by the addition of cotton wool bits taped onto their mouth, bits of balsa wood to create acoustic filters, etc.? You could keep tweaking until the singer's in-room frequency response matched the anechoic version. But would it sound as good, or might it sound a bit strange?

Without the additions you are simply accepting the extra 'acoustic ambience' of the room without even making any measurements. A singer in a room sounds how they sound, and everyone accepts that. There is a perfectly sensible reason for thinking that we should do the same for a hi-fi speaker - as long as it is neutral in terms of frequency response and dispersion to start with. If it isn't, you need to make some adjustments, but you need to know about the speaker's dispersion before you start; it can't be done solely from an in-room frequency response measurement because this conflates the room and the speaker's dispersion. A 'target curve' makes no sense - it is the tail wagging the dog.

A more fruitful way of thinking about it is this: a listener does not try to match absolute signals; they may hear as identical two very different signals when veiwed on a laptop, because they are listening to musical and acoustic 'objects' that are contained within the signal. The aim of a hi-fi system is to maintain the integrity of the objects as much as possible, and to maintain their separation as clearly as possible.

If I take a signal and add some 'room' then on the laptop screen it's going to look very, very different. With 'room correction' you may think you are "connecting" the in-room sound to the recording by getting it closer in one or two variables, but you're not: you're breaking the integrity and separation of the objects. The listener would not even have noticed the addition of the room, whatever the laptop display says, but they will notice the ham-fisted theory-free modification of the source.

If you want to listen to the signal directly then wear headphones, or sit closer to the speakers, or make yourself an anechoic chamber. Much easier.
 

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
You may have had me in mind as one of those naughty people. But then you make the case for me:



People really do believe that by changing what comes out of the speaker, they have 'corrected the room'. Or 'removed the effect of the room' (which boils down to the same thing). But they haven't. The room remains exactly the same, and the effects of the room remain exactly the same, but now applied to a weirdo sound from the speaker.

But with certain issues we can minimise the effect of the room. The minimum phase modes can be addressed (at a listening position). What else works or doesn't work?

@UliBru good to see you here. It would be great to have an Acourate thread. I need to discuss latency with you again :)
 
Top Bottom