Time Domain measurements?

[thorvat]

Doesn't matter to me if you believe any of my graphs or subjective listening test results.

Science is about objective proofs, religion is about beliefs.

 

[lowkeyoperations]

When I’m producing music on my Amphion One15s I am able to more clearly hear the differences between compressor settings than on my KH120s. These monitors are being used in the near field about 75cm from my ears.

SSL compressors have an attack knob that ranges from 0.1ms up to 30ms in steps. Minute adjustments are more easily perceived on the One15s.

I don’t put the difference between the two monitors in highlighting timing changes down to their frequency response.

 

[dasdoing]

When I’m producing music on my Amphion One15s I am able to more clearly hear the differences between compressor settings than on my KH120s. These monitors are being used in the near field about 75cm from my ears.

SSL compressors have an attack knob that ranges from 0.1ms up to 30ms in steps. Minute adjustments are more easily perceived on the One15s.

I don’t put the difference between the two monitors in highlighting timing changes down to their frequency response.

right. if the bass region is 10ms late, low atacks will be hard to hear

 

[haraldo]

Most of the phase correction is due to the phase deviation caused by passive XO. Luckilly, those deviations are practially inaudible, at least in a room with "normal" reflections.

I believe the phase shifts by passive xovers are audible, or rather the lack of phase shifts is audible

 

[dasdoing]

I believe the phase shifts by passive xovers are audible, or rather the lack of phase shifts is audible

I think he is wrong and wright.
A typical crossover at 1000Hz will only cause 1ms of delay in the lower part. that would be hard to hear.
BUT typical speakers have more like 10ms delay, due to woofer being slower I guess. everybody can AB that

 

[Frgirard]

I believe the phase shifts by passive xovers are audible, or rather the lack of phase shifts is audible

Believe is not the reality. Abx succeeded and you will be famous.
You can try with rePhase.

 

[dasdoing]

Believe is not the reality. Abx succeeded and you will be famous.
You can try with rePhase.

it doesn't matter, because there will be much more (audible) group-delay in a system.

 

[dasdoing]

check it out:

this is the group delay of the filter

 

[thorvat]

I believe the phase shifts by passive xovers are audible, or rather the lack of phase shifts is audible

Passive XO phase shift has been proven by blind tests to be inaudible,

 

[ernestcarl]

Passive XO phase shift has been proven by blind tests to be inaudible,

Truth is, I don't really expect to hear any difference in the majority of room and speaker setups including my untreated open plan living room where the speakers are close to four meters away. Also need to consider the amount/severity of phase shift... where in most cases it probably really isn't significant.

If one could do a blind test with something like the KH80 in the very nearfield (fully treated room) switching very quickly between active LP and passive xo mode back and forth for unlimited number of times similar to the way Amir does with his ABX tests in Foobar, I kind of expect this to be audible by more than a few people. Very subtle, perhaps, but audible. Even in my own post phase-only passive xo FIR linearization ABX tests I could, in fact, hear a difference. Otherwise, this kind of absolutist statement makes it sound as if the folks at Neumann, Genelec, Hedd, Meyer etc. all a bunch deaf and deluded morons too in the way they promote or provide as optional, modifiable active linear phase corrected systems. In the studies that have looked at phase distortion (the ones I've seen anyway), often mentioned is the effect (when actually heard) as something that can be very subtle, for sure. In that sense, and considering the reality of most room setups out there, making the goal "linear phase" or close to it obviously is not going to be as important as other criteria like the frequency response magnitude and directivity.

 

[dasdoing]

again, the crossover is not the problem.
this is the group delay of a typical system, and anybody can hear the diference in the video, because I can even on a smart-phone

check it out:

this is the group delay of the filter

View attachment 156212

 

[ernestcarl]

again, the crossover is not the problem.
this is the group delay of a typical system, and anybody can hear the diference in the video, because I can even on a smart-phone

In so far as the passive xo is the cause of the added group delay, one cannot really separate or exclude it anyway.

 

[dasdoing]

In so far as the passive xo is the cause of the added group delay, one cannot really separate or exclude it anyway.

is it realy the cause? cause a traditional XO at 1000Hz adds only about 1ms

 

[ernestcarl]

is it realy the cause? cause a traditional XO at 1000Hz adds only about 1ms

I don't really care too much if it is the xo itself that is the direct cause of the added group delay -- maybe it's already inherent in the driver. But yes, even the xo of the KH120 itself, for example, at 2kHz GD is less than a ms yet correcting for it is audible to me (ABX test results I attached in previous page).

 

[ctrl]

But yes, even the xo of the KH120 itself, for example, at 2kHz GD is less than a ms yet correcting for it is audible to me (ABX test results I attached in previous page).

One should be careful not to create new myths.

The audibility of group delay in the low bass should be quite indisputable, since one can expect 10-20ms group delay depending on the speaker design principle (CB, BR, PR,...) and crossover frequency. There are also here in the forum ABX examples of group delay change (almost) without FR change, where one can check this.

The audibility of group delay in the midrange and treble caused by crossover is quite another matter.
Current studies clearly contradict the statement that these are audible with common filter slopes.

A fourth order Linkwitz-Riley crossover (of a typical 2-Way speaker) shows a group delay well below 0.5ms down to 200Hz.


A recent study "AUDIBILITY OF GROUP-DELAY EQUALIZATION, LISKI et al." investigates the audibility of group delay in the range +-0.25ms to +-5ms in the frequency range 500-4000Hz.

The used group delay curves are much more challenging than the group delay curve of a LR4 crossover (see above, right diagram, black curve).


Artificial and natural signals were evaluated. Even if the lowest individual hearing threshold of an artificial signal is taken at 2 kHz (0.27ms), the group delay of an LR4 crossover (0.23ms) is below the audibility threshold. When using natural signals, the perceptibility threshold is well above the group delay of a standard crossover.
Someone who can hear the group delay in a crossover LR4@2kHz with normal music can, according to the current state of science, join the Avengers - even if the superpower is not likely to do much against Thanos

In most cases, phase equalization might also lead to a change in the frequency response of the loudspeaker. It is likely that this frequency response change is perceived and not the group delay change.

 

[ernestcarl]

One should be careful not to create new myths.

The audibility of group delay in the low bass should be quite indisputable, since one can expect 10-20ms group delay depending on the speaker design principle (CB, BR, PR,...) and crossover frequency. There are also here in the forum ABX examples of group delay change (almost) without FR change, where one can check this.

The audibility of group delay in the midrange and treble caused by crossover is quite another matter.
Current studies clearly contradict the statement that these are audible with common filter slopes.

A fourth order Linkwitz-Riley crossover (of a typical 2-Way speaker) shows a group delay well below 0.5ms down to 200Hz.
View attachment 156618

A recent study "AUDIBILITY OF GROUP-DELAY EQUALIZATION, LISKI et al." investigates the audibility of group delay in the range +-0.25ms to +-5ms in the frequency range 500-4000Hz.

The used group delay curves are much more challenging than the group delay curve of a LR4 crossover (see above, right diagram, black curve).
View attachment 156620 View attachment 156621

Artificial and natural signals were evaluated. Even if the lowest individual hearing threshold of an artificial signal is taken at 2 kHz (0.27ms), the group delay of an LR4 crossover (0.23ms) is below the audibility threshold. When using natural signals, the perceptibility threshold is well above the group delay of a standard crossover.
Someone who can hear the group delay in a crossover LR4@2kHz with normal music can, according to the current state of science, join the Avengers - even if the superpower is not likely to do much against Thanos

View attachment 156623


View attachment 156622

In most cases, phase equalization might also lead to a change in the frequency response of the loudspeaker. It is likely that this frequency response change is perceived and not the group delay change.

I don't believe I made extraordinary claims to know what the exact amount of phase shift from the mid and high frequency crossover induced GD does time distortion start to become audible for all individuals. I performed listening tests to hear for myself whether there is an advantage/improvement in correcting the passive xo induced phase rotation in my smaller (non-linear phase) monitors.

The ABX tests were performed using foobar. Pure time correction using rePhase's "minimum phase filters phase linearization" LR4 at 2 kHz plus some minor additional paragraphic phase EQs were used. There is no frequency response EQ applied as well as none seen in the measurements -- zero, nada. Microphone placed at the exact same listening position. Apply frequency dependent windowing, say, 6 cycles or less -- then and only then do we clearly see an increase in the magnitude SPL around the xo region. I have not tested and listened for the audibility the aforementioned pure phase correction in the far-field. The attached test results was performed from less than 1 meter in my nearfield desk setup where much of early first reflections are minimal to absent. I have said before that correcting for this phase/GD/time distortion made vocals and music sound (I did not test pure tones or other noise) a little cleaner and less fatiguing -- i.e. less rough/grainy/grating and where perception of volume is somewhat decreased rather than increased -- I presume this is likely from the reduced GD "time distortion" and maybe a little similar in what we experience when harmonic distortion / resonances are reduced in a speaker cabinet, for example. *AGAIN: the effect was very subtle.

I'll dig out my measurements later (I know it's here in my hard drive somwhere) to show what I mean about magnitude only visibly changing after applying FDW.

 

[ernestcarl]

Someone who can hear the group delay in a crossover LR4@2kHz with normal music can, according to the current state of science, join the Avengers - even if the superpower is not likely to do much against Thanos

Let's be clear, I haven't sought out to break any records or make it be known that I have extraordinary hearing. In fact, I do have slightly unequal hearing in some frequencies between the left and right ears -- not to mention my long standing tinnitus and inability to sleep without a noise machine turned on/running in the background.

Wasn't sure which saved MDAT file I had before so I re-measured my left KH120 (ports are sealed to match its phase with my sealed sub easier). Monitor is played solo with no EQ and the microphone less than a meter from the acoustical axis using 90 deg mic orientation and calibration file.

no difference whatsoever

Actually, a difference is visible if FDW is applied.

BUT, paradoxically:

perception of volume is somewhat decreased rather than increased

... with FIR convolution as it seems to remove a bit of the raw "graininess" or harshness to the sound ever so slightly.


And here's the zoomed-in wavelet spectrogram:

The peak energy time is essentially just another way to view the "group delay".

I'm going to say it again:

I don't really care too much if it is the [passive] xo itself that is the direct cause of the added group delay

There is no reason for me to say that "I heard zero audible changes/improvement" in this particular test setup when, in fact, I did. It may be only be a small subtle improvement -- I presume practically worthless to most people here -- but it's there alright.

 

[Frgirard]

I don't believe I made extraordinary claims to know what the exact amount of phase shift from the mid and high frequency crossover induced GD does time distortion start to become audible for all individuals. I performed listening tests to hear for myself whether there is an advantage/improvement in correcting the passive xo induced phase rotation in my smaller (non-linear phase) monitors.

The ABX tests were performed using foobar. Pure time correction using rePhase's "minimum phase filters phase linearization" LR4 at 2 kHz plus some minor additional paragraphic phase EQs were used. There is no frequency response EQ applied as well as none seen in the measurements -- zero, nada. Microphone placed at the exact same listening position. Apply frequency dependent windowing, say, 6 cycles or less -- then and only then do we clearly see an increase in the magnitude SPL around the xo region. I have not tested and listened for the audibility the aforementioned pure phase correction in the far-field. The attached test results was performed from less than 1 meter in my nearfield desk setup where much of early first reflections are minimal to absent. I have said before that correcting for this phase/GD/time distortion made vocals and music sound (I did not test pure tones or other noise) a little cleaner and less fatiguing -- i.e. less rough/grainy/grating and where perception of volume is somewhat decreased rather than increased -- I presume this is likely from the reduced GD "time distortion" and maybe a little similar in what we experience when harmonic distortion / resonances are reduced in a speaker cabinet, for example. *AGAIN: the effect was very subtle.

I'll dig out my measurements later (I know it's here in my hard drive somwhere) to show what I mean about magnitude only visibly changing after applying FDW.

i used rePhase and his Linearization filter with my K+H O300 and my KH 420. The Mid on the both models are in front and quickly unbearable. But it gives a nice step response curve.
With the K+H O300 and a KH810 (sub) there are not audible effect.

Strange Brain world

 

[ctrl]

Wasn't sure which saved MDAT file I had before so I re-measured my left KH120 (ports are sealed to match its phase with my sealed sub easier). Monitor is played solo with no EQ and the microphone less than a meter from the acoustical axis using 90 deg mic orientation and calibration file

no difference whatsoever

Actually, a difference is visible if FDW is applied.

Slight deviations can also be seen in your first measurement. However, these are almost undetectable due to the room resonances.



For measurements in normal living rooms, it is usually useful to use the average of several measurements, as this can improve the S/N ratio (noise floor).

To exclude falsifications of the measurement by room influences, these must always be removed from the measurement by gating ("windowing").

REW tries to improve this automatically with the FDW.
It would be even better if you could completely eliminate the room resonances by gating.

Assuming the comparison of the FDW measurements would be correct, you would have the perfect explanation for the perception differences between the "normal" and phase-corrected version (FIR) of the KH120 that you describe.
The FIR version shows a hump of about 0.7dB in the mid-high range between 900Hz and 4kHz. This change in timbre should be audible.

 

[Frgirard]

Slight deviations can also be seen in your first measurement. However, these are almost undetectable due to the room resonances.

View attachment 156683

For measurements in normal living rooms, it is usually useful to use the average of several measurements, as this can improve the S/N ratio (noise floor).

To exclude falsifications of the measurement by room influences, these must always be removed from the measurement by gating ("windowing").

REW tries to improve this automatically with the FDW.
It would be even better if you could completely eliminate the room resonances by gating.

Assuming the comparison of the FDW measurements would be correct, you would have the perfect explanation for the perception differences between the "normal" and phase-corrected version (FIR) of the KH120 that you describe.
The FIR version shows a hump of about 0.7dB in the mid-high range between 900Hz and 4kHz. This change in timbre should be audible.

a difference but what is the cause: eq calculation vs the phase correction ? When i used rePhase i had clipping ...