An Enticing Marketing Story, Theory Without Measurement?

[Hephaestus]

Just out of curiosity: what speakers you have Cosmik? I understood from some of your previous posts that you pay attention to the time domain performance?

I would like to see a loudspeaker with accurate time domain performance, excellent directivity and low distortion, very low latency and no reliability issues. Unfortunately I haven´t seen one.

Now I have to stick with two systems, maybe after 5 years some company can offer everything in one package?

 

[Cosmik]

Just out of curiosity: what speakers you have Cosmik? I understood from some of your previous posts that you pay attention to the time domain performance?

I would like to see a loudspeaker with accurate time domain performance, excellent directivity and low distortion, very low latency and no reliability issues. Unfortunately I haven´t seen one.

Now I have to stick with two systems, maybe after 5 years some company can offer everything in one package?

My speakers are housed in re-used 1970s monkey coffins (which possibly sacrifice sturdiness for looks; but they do look very nice), and using pretty much the cheapest drivers that might be termed hi-fi - that was part of the proof, pudding, etc.

They're sealed, three-way (8", 3", 1"), active, with DSP crossovers and are reasonably large. The DSP aims to neutralise the drivers and perform time alignment, so there's latency.

Directivity-wise they are what they are, but being three-way they're not atrocious. They do actually sound like hi-fi.

 

[PaulD]

I would like to see a loudspeaker with accurate time domain performance, excellent directivity and low distortion, very low latency and no reliability issues. Unfortunately I haven´t seen one.

I think the JBL M2 fulfils what you are asking for. It's not a domestic product though as the amps are noisy and must be in another room, and the price is more than most want to pay, although I think they are good value - I have 2 pairs at work... At home I use Quad63s (rebuilt) with good subs. This combo also fulfils your requirement although the question would be around reliability of the ELS panels. I've only needed one rebuild in 10 years, so I think they are ok, without subs they are more prone to failure trying to reproduce bass...

 

[Purité Audio]

Just out of curiosity: what speakers you have Cosmik? I understood from some of your previous posts that you pay attention to the time domain performance?

I would like to see a loudspeaker with accurate time domain performance, excellent directivity and low distortion, very low latency and no reliability issues. Unfortunately I haven´t seen one.

Now I have to stick with two systems, maybe after 5 years some company can offer everything in one package?

Dutch&Dutch 8C or Kii T`hREE, both have adjustable latency.
Keith

 

[Hephaestus]

Thanks everyone for the input.

By excellent directivity I meant relatively uniform directivity indeed.

Dutch&Dutch and M2 are interesting. Kii Three is made from low quality parts and has high latency (my subjective opinion). I would not drop that sort of money in a planned obsolescence product.

I hope that Dutch&Dutch will introduce slightly smaller nearfield option at lower cost in the near future (As Keith mentioned in a another thread) - software might be more mature in it as well. Interesting times.

 

[UliBru]

Farther at the back of our 14 meter long room where our dining table sits, the sound still wanted a little extra. With the app at hand, we dialed the bass level up by just 3dB. Wow, that was all the sound needed.

I'm still wondering, under pure logical aspects, if lifting the bass of the "neutral" Dutch&Dutch by 3 dB in the playback room deteriorates the speaker sound and belongs to the age-of-the-universe stuff. Definitely the neutrality is hurt. Of course the personal enjoyment may be increased but this does not count because of the "never hurt the neutrality"-postulation claimed for in this thread.

If this logic is wrong then why and what is the correct logic here?

 

[UliBru]

1. Let's think about an ideal speaker. It would perfectly reproduce the input signal, it transfer behaviour is 1:1. A FIR representation would consist of a Dirac pulse, one single sample has the value 1, all other samples are 0. The Fourier transform leads to a straight horizontal magnitude response from DC to fs/2. The group delay is constant for all frequencies.

2. The ideal behaviour is valid for the movement of the driver coil. A battey connected to the coil will lead to a constant displacement of the coil. Thus it can transfer also DC. But we listen to sound waves in the air. The membrane oscillation causes changes in air pressure radiating into the room. With common speakers a constant displacement of the membrane does not lead to a constant change of air pressure (whereas moving the front wall in a totally sealed room would achieve this). Anyway the result is: we do not expect a transfer behaviour of a speaker down to DC.
furthermore we experience a drooping frequency response at the listening position because of a bigger distance between listener and speaker. Whereas a close-up measurement will show up the neutral behaviour of our ideal speaker.
An example of such an assumed ideal speaker is thus

It is not important of the lowest frequency is now 40 Hz, 25 Hz or 10 Hz. This depends on the size of the bass driver/s and the speaker design. The slope may be of 2nd or 4th oder dependent on closed or vented box. It also does not matter where the high frequency droop starts and how big it is. So the picture simply shows a possible example which includes the low frequency limitations and the high frequency droop. The speaker is still ideal.

As we live in a causal world the speaker will not play before it receives a signal. It will answer by a minimumphase behaviour which means that the given frequency response is achieved by the minimum phase changes in the speakers time behaviour. The derived step response of the ideal speaker thus is

Keeping this ideal step resonse in mind it makes sense to study published step responses of different speakers including ultra high-end speakers. A nice learing experience. Luckily our ears is quite tolerant for deviations but we are talking about the ideal case here.

Now let's take a real world step response example (a speaker in the class of about 15000 Euros), measured at the listening position:

We can identify the common behaiour of tweeter first, then midrange driver and then the bass. It's caused by the XO design chosen for this speaker.
With an applied room correction we get now:

Obviously the timing has improved pretty much. There are still jags in the curve caused by room reflections. The reflections are not corrected. It simply does not make sense to do this because the reflection times are not constant when you move around in the room. The reflections show up also the improved shape of the direct step response.

So what's wrong with a "room correction" here? Please note that we have not talked here about frequency response or target curve. Just about one dimension = timing.

 

[Purité Audio]

Thanks everyone for the input.

By excellent directivity I meant relatively uniform directivity indeed.

Dutch&Dutch and M2 are interesting. Kii Three is made from low quality parts and has high latency (my subjective opinion). I would not drop that sort of money in a planned obsolescence product.

I hope that Dutch&Dutch will introduce slightly smaller nearfield option at lower cost in the near future (As Keith mentioned in a another thread) - software might be more mature in it as well. Interesting times.

The Kiis have infinitely adjustable latency, useful if you are using them for video tracking , it is not the individual cost of the components that matter but the design and overall implementation, the Kiis are superb loudspeakers.
Keith

 

[Hephaestus]

The Kiis have infinitely adjustable latency, useful if you are using them for video tracking , it is not the individual cost of the components that matter but the design and overall implementation, the Kiis are superb loudspeakers.
Keith

I agree with you on technical performance - absolutely enjoyable speakers to listen to music. Even the visual design is living room friendly

However they are made from junk parts and wont last for 20 years and can’t be repaired after 8-12 years which is propably quite near their designed ”life cycle” - with M2 investment is much more safe in this sense.

I know one party which reverse engineered a pair - they were not impressed by longeavity of the design of ”Three”. Wont give any names here.

 

[Purité Audio]

In as much the m2’s are just two drivers in a box you are probably quite correct, to state that the Kiis are made from ‘junk parts’ just shows your ignorance.
Keith

 

[Hephaestus]

In as much the m2’s are just two drivers in a box you are probably quite correct, to state that the Kiis are made from ‘junk parts’ just shows your ignorance.
Keith

Didnt mean to offend anyone. I am seriously concerned in longeavity of loudspeakers in this price range - it is not ignorance.

My opinion is that true ignorance is to design expensive loudspeaker with relatively short life cycle and to make it non-repairable.

If I had an assurance of Kii Threes to last for 20 years or more (as high quality monitor should) - I would buy them with a smile on my face!

 

[Purité Audio]

Why would you say they are non repairable, from my own admittedly small representative sample I have encountered more issues with ‘traditional’ monitors than with either the Kii or Dutch&Dutch 8Cs.
Keith

 

[Hephaestus]

Keith, I dropped lengthy PM for you to avoid more off topic in this thread - and apoligies for the off-topic in the first place!

 

[Hipper]

Is digital digital room correction just another story of theory without measurement? Generally speaking, there are two kinds of information: There are observations and there are theories. So generally speaking you can believe the observations and you don’t need to believe the theories. But where are the observations – competently managed blind tests – that support the theory of digital room correction?

Can we conclude that room correction is «a misnomer» and just an «enticing marketing story»? Why don’t we have more observations on this fascinating area?

I've read through this thread and it has deviated into a lot of interesting areas, but I seem to have missed the answer to the OP's question. These are my thoughts now:

1. It does seem to be more about theories which are still evolving.

2. There are only a tiny number of reports of experiments on the subject.

3. Room Correction is perhaps a misnomer but we're stuck with it.

4. We don't have more observations because, I would guess, the cost, and the reliance on personal experiences leading to unpredictable outcomes. It's not really an exact science.

5. There seems to be a market for room correction but to sell well it needs to be easy to understand and use. Reading this thread shows that that is very difficult to achieve.

This whole area is fraught with complications many of which have been discussed here:

1. The transition frequency - the sound coming from a source behaves differently dependent on its frequency. Furthermore the transition frequency itself varies with the size and shape of the room. And it's not an exact frequency but a range in which this transition takes place.

2. Psychoacoustics - the collection of the sound by our ears and the interpretation of that sound by our brain. Whilst there are some general observations that cover all of us there are lots of individual differences that mean results aren't necessarily the same for each of us. This includes age related and other hearing damage, ear and head shape, personal experience, equal loudness curves, precedence effect.

3. Measurements are done with microphones. Mics do not replicate how we perceive sound; not even how our ear collects sound.

4. The recordings we listen to are of inconsistent quality.

All this would suggest that a one size fits all DSP solution is not possible, and would not even work for one person who has various qualities of recordings.

Perhaps then it is an 'enticing marketing story' that appeals to our sense of exactness but doesn't, and can't, deliver.

 

[Cosmik]

I've read through this thread and it has deviated into a lot of interesting areas, but I seem to have missed the answer to the OP's question. These are my thoughts now:

1. It does seem to be more about theories which are still evolving.

2. There are only a tiny number of reports of experiments on the subject.

3. Room Correction is perhaps a misnomer but we're stuck with it.

4. We don't have more observations because, I would guess, the cost, and the reliance on personal experiences leading to unpredictable outcomes. It's not really an exact science.

5. There seems to be a market for room correction but to sell well it needs to be easy to understand and use. Reading this thread shows that that is very difficult to achieve.

This whole area is fraught with complications many of which have been discussed here:

1. The transition frequency - the sound coming from a source behaves differently dependent on its frequency. Furthermore the transition frequency itself varies with the size and shape of the room. And it's not an exact frequency but a range in which this transition takes place.

2. Psychoacoustics - the collection of the sound by our ears and the interpretation of that sound by our brain. Whilst there are some general observations that cover all of us there are lots of individual differences that mean results aren't necessarily the same for each of us. This includes age related and other hearing damage, ear and head shape, personal experience, equal loudness curves, precedence effect.

3. Measurements are done with microphones. Mics do not replicate how we perceive sound; not even how our ear collects sound.

4. The recordings we listen to are of inconsistent quality.

All this would suggest that a one size fits all DSP solution is not possible, and would not even work for one person who has various qualities of recordings.

Perhaps then it is an 'enticing marketing story' that appeals to our sense of exactness but doesn't, and can't, deliver.

I mainly agree, but I think it still starts from a place that gives credence to the wrong idea by default.

Can anyone here see how 'science' itself can miss the point? Many arguments seem to be along the lines of:
"Experiments show that by far the most important aspect in human listening is frequency response. Gross phase distortion is rarely noticed, and harmonic distortion can be at very high levels before it is noticed. But small changes in frequency response of +/-0.5dB are easily audible. There is reliable, repeatable data to show this, based on real listening tests with real people.

The room affects frequency response - often by many dB. Therefore, it follows that if the signal from the speaker can be modified to get a more accurate frequency response at the listener's ears, this will result in the best sound."

This argument is used all the time. And it fits the "You can't argue with data" idea that you see often.

The thing is, there are holes in the argument all over the place but it is almost impossible to counter a sequence of statements that seems to point to something 'obvious'. And playing around with a mainly wrong method will result in something that by definition gives the 'right' measurements, and kind of sort of, seems to 'work' - although not well enough to stop people experimenting with interminable new algorithms, different calibrations, vinyl, valve amps and passive speakers just to relieve the disappointment.

 

[UliBru]

The room affects the frequency response. Thus it seems logical to correct the modified frequency response. But this idea is putting the cart before the horse. A frequency response does not contain time information anymore.

Whereas we are perceiving sound waves as time information. As already written before in this thread music played reversely sounds totally different despite it has the same frequency response.

Music starts to play by the direct sound wave from the loudspeaker. The room sound is following behind the direct wave. Reflections, standing waves and diffuse sound now start to add on the continuously played direct sound. Luckily the precedence effect helps us to grab the information of interest from the complex signal.

This means that a room correction must operate similarily. It has to distinguish between the direct sound and the up-following sound. It is a good question of course how to separate the direct sound from a recorded room pulse response. There must be some gating or windowing method and the quality of a correction depends on how clever the separation is carried out.

Now it is worth to think about a connection between time domain and frequency domain. Shortening of a room pulse response by gating or windowing (= removal of later room effects) leads to a smoother frequency response. Applying an inverse FFT on a smoothed frequency response results in a shorter time domain pulse. Thus simple correction algorithms take the frequency response, apply a more or less proper smoothing (including of averaging multiple measured frequency responses), calculate the difference between a target and the smoothed response and finally calculate a minimumphase correction filter by IFFT.
Indeed this way even simple approaches result in filters which can work astonishingly well.

Of course it is possible to refine the applied methods and algorithms.
And of course a basic difficulty remains as there is a sliding transition between direct sound and direct sound superimposed by the room sound. There is no precise border line in between. Thus there is no one-button solution for a correction. The user should be able to adjust and finetune the correction.

Because of preferencing the direct sound part of a room measurement the room correction indeed also takes care about the behaviour of the loudspeaker and thus corrects the speaker too. It may be worth to note that many loudspeakers benefit from a correction filter.
From this point of view room correction definitely is a misname. But it is common now. There is a difference though: a measurement of a speaker in an echoless chamber results in a speaker correction - a measurement of a speaker in a listening room results in a correction of a speaker including a correction of some but not all room effects .

 

[Snarfie]

1. Let's think about an ideal speaker. It would perfectly reproduce the input signal, it transfer behaviour is 1:1. A FIR representation would consist of a Dirac pulse, one single sample has the value 1, all other samples are 0. The Fourier transform leads to a straight horizontal magnitude response from DC to fs/2. The group delay is constant for all frequencies.

2. The ideal behaviour is valid for the movement of the driver coil. A battey connected to the coil will lead to a constant displacement of the coil. Thus it can transfer also DC. But we listen to sound waves in the air. The membrane oscillation causes changes in air pressure radiating into the room. With common speakers a constant displacement of the membrane does not lead to a constant change of air pressure (whereas moving the front wall in a totally sealed room would achieve this). Anyway the result is: we do not expect a transfer behaviour of a speaker down to DC.
furthermore we experience a drooping frequency response at the listening position because of a bigger distance between listener and speaker. Whereas a close-up measurement will show up the neutral behaviour of our ideal speaker.
An example of such an assumed ideal speaker is thus

View attachment 29371

It is not important of the lowest frequency is now 40 Hz, 25 Hz or 10 Hz. This depends on the size of the bass driver/s and the speaker design. The slope may be of 2nd or 4th oder dependent on closed or vented box. It also does not matter where the high frequency droop starts and how big it is. So the picture simply shows a possible example which includes the low frequency limitations and the high frequency droop. The speaker is still ideal.

As we live in a causal world the speaker will not play before it receives a signal. It will answer by a minimumphase behaviour which means that the given frequency response is achieved by the minimum phase changes in the speakers time behaviour. The derived step response of the ideal speaker thus is

View attachment 29372

Keeping this ideal step resonse in mind it makes sense to study published step responses of different speakers including ultra high-end speakers. A nice learing experience. Luckily our ears is quite tolerant for deviations but we are talking about the ideal case here.

Now let's take a real world step response example (a speaker in the class of about 15000 Euros), measured at the listening position:

View attachment 29373

We can identify the common behaiour of tweeter first, then midrange driver and then the bass. It's caused by the XO design chosen for this speaker.
With an applied room correction we get now:

View attachment 29374

Obviously the timing has improved pretty much. There are still jags in the curve caused by room reflections. The reflections are not corrected. It simply does not make sense to do this because the reflection times are not constant when you move around in the room. The reflections show up also the improved shape of the direct step response.

So what's wrong with a "room correction" here? Please note that we have not talked here about frequency response or target curve. Just about one dimension = timing.

I'm really a novice regarding Room correction an measurements. But if i see your latest green corrected graph is see still in efficiencies. Basically i guess that we are looking for a corrected graph where the correction calculated is totally flat between 20 to 20.000 Hz. Only reason to tweak this flat line is because of personal taste (if your room correction software has this functionality). I have 2 examples using Mathaudio Room EQ.


First example is my current setup a JK Acoustics Optima 3 full range loudspeaker around 1 meter high.

Measurements shows this Grey line is found inefficiencies sometimes almost 15 db's ofset. Basically the white flat line is all i need, don't want to enhance anything because it sound good for my (subjective) taste.

Now have a look at the IMF Compact II monitor which are truly monitors so no full range speaker. Do know this graph the B&W (used as subwoofers) that you see are not used. Graph are only the IMF speakers unfortunately i could not find another picture whitout the B&W's.

Measuring shows this. As you could see i draw a preferd slope form around 300Hz to 20.000 Hz an boosted some db's my lows because it suited my (subjective) taste an eventually had the same sort of characteristics as the JK Optima 3 where no change for my subjective taste had to been made.

So IMO people will always listen to music as they Like or worse (and that happened also with me) as they could be conditioned to listen to music in a bad acoustic room/condition. Basically I liked after so many years high tones. I had to go in rehab an had to learn to listen neutral to music to find a balance between instruments an voices.


With this roomcorrection I could manage that. I'm 60 an love my music but I rediscovered almost all my old music again. If you look for instance to the gray line of the Optima 3 basically you could almost not hear on the Abbey Road album Paul McCarthy playing his bass . With the room correction not only the bass came back it was balanced with the other instruments an voices it became suddenly almost intimate. But ok my experience is based on my own room characteristics. I guess use your experience ha ha if Barry Whit has a high pitch voice pronouncing everything with an ssss like an high-hat than something is wrong with the recording but most likely with your acoustics.

 

[UliBru]

But if i see your latest green corrected graph is see still in efficiencies. Basically i guess that we are looking for a corrected graph where the correction calculated is totally flat between 20 to 20.000 Hz. Only reason to tweak this flat line is because of personal taste ...

1. The last green corrected graph does NOT show a frequency response but a step response (time domain). Do not mix it.

2. Your frequency response shown above has a falling slope above 1 kHz. This is a normal behaviour with measurements. The target should also have such a falling slope, otherwise you are boosting the higher frequencies by the correction.
The frequency response of the second measurement does not look very good. It looks like the tweeter has too much energy.

At the end it is always wise to inspect the correction graph. Then you see if you boost something that you shouldn't do. If the target is flat but the correction has a raising slope at the high frequencies you are boosting the treble.

 

[Juhazi]

UliBru, did you really try to fix a 3-way loudspeaker's step response by FIR'ing the input signal? What is the point there? The step response you showed is typical for a 3-way.

 

[Snarfie]

1. The last green corrected graph does NOT show a frequency response but a step response (time domain). Do not mix it.

2. Your frequency response shown above has a falling slope above 1 kHz. This is a normal behaviour with measurements. The target should also have such a falling slope, otherwise you are boosting the higher frequencies by the correction.
The frequency response of the second measurement does not look very good. It looks like the tweeter has too much energy.

At the end it is always wise to inspect the correction graph. Then you see if you boost something that you shouldn't do. If the target is flat but the correction has a raising slope at the high frequencies you are boosting the treble.

As far as i know Mathaudio room eq does not work with a found correction graph it only find the inefficencies of the room you are measuring than you are free to use the slider to correct the found inefficensies (white line) to your liking. Have a look at https://mathaudio.com/room-eq.htm i have not enough knoledg to explain there used theory.