Step Response: Does It Really Matter?

[ctrl]

. you can see in csd image a big gap on this frequency. in your side by side view can see better that magnepan is faster. you can also resize so it is in y same length as the other
so the db values are so much diffrent because in 1 plot is cursor at maximum on other at minimum.

Not sure if you are reading the CSD plot correctly.

Suppose we want to compare the decay of the two speakers at 900Hz.
So we draw the 900Hz point (green lines) for the first frequency response slice in both diagrams (top red circle).

In the right diagram (the Magnepan speaker) the xz-plane is 16dB below the blue line (which represents around -2dB). The lower red circle shows at which point the 900Hz are attenuated by 16dB.

So in the left diagram (10'' woofer DeVore) we may also consider only a range with 16dB attenuation. This area is roughly represented by the blue plane I have drawn. Also there the lower red circle represents an attenuation of the 900Hz by 16dB.

The further the frequency response slices are apart, the faster the speaker decays.
The fewer frequency response slices there are in an attenuation range, the faster the speaker decays.

In Post#172 you said, "I think the important thing is, how fast the level is drop 12 db", if you now look at the decay at 900Hz for the first 12dB of attenuation, you see that the Magnepan speaker on the right decays much slower.

 

[bennybbbx]

Not sure if you are reading the CSD plot correctly.

Suppose we want to compare the decay of the two speakers at 900Hz.
So we draw the 900Hz point (green lines) for the first frequency response slice in both diagrams (top red circle).

In the right diagram (the Magnepan speaker) the xz-plane is 16dB below the blue line (which represents around -2dB). The lower red circle shows at which point the 900Hz are attenuated by 16dB.

So in the left diagram (10'' woofer DeVore) we may also consider only a range with 16dB attenuation. This area is roughly represented by the blue plane I have drawn. Also there the lower red circle represents an attenuation of the 900Hz by 16dB.

The further the frequency response slices are apart, the faster the speaker decays.
The fewer frequency response slices there are in an attenuation range, the faster the speaker decays.

In Post#172 you said, "I think the important thing is, how fast the level is drop 12 db", if you now look at the decay at 900Hz for the first 12dB of attenuation, you see that the Magnepan speaker on the right decays much slower.
View attachment 114883

I do not understand how you come to 900 hz -2 db. see my lines that i perspective draw from 6 db and 12 db. the 6 db is a little less as your blue line and the 12 db is more as your blue line. so the 900 hz on the magnepan have at least 6 db and not -2 db. but you are right the 10 inch speaker have at 900 hz more level. but it look even with the low level the magnepan is faster. at 400 hz both speakers have simular level and the magnepan go not so much in front. so it isfaster here is original https://www.stereophile.com/content/magnepan-lrs-loudspeaker-measurements https://www.stereophile.com/content/devore-fidelity-orangutan-o96-loudspeaker-measurements and when look at the step response of both speakers the magnepan is faster too.. 10 inch start time 4 ms, end time (to reach 0) 6 ms. so 2 ms duration.magnepan start time 3.8 ms end time 5.2 ms. 1.4 ms duration. to see how much influence bass have i do a step response of eris 3.5 without eq so bass is much lower

 

[bennybbbx]

to see how much diffrent frequency change the step response i do eris with and without EQ measures. there can see there is not much changes. I do with and without EQ with same microphone position(on stative). see also older step response of eris. there is not much change.it can clear see which speaker is faster

 

[ctrl]

I do not understand how you come to 900 hz -2 db.

I can't show you in detail how to evaluate a 3D graph, you can teach yourself these basics.

I have drawn you the graphical evaluation for three frequencies (see the pink lines), which lie on the first frequency response slice.

The sound pressure level at the frequency 5993Hz. From this point we know that it is about -2dB (-1.74dB) - see the blue A.
And the sound pressure levels at 900Hz and 400Hz. The 900Hz point on the first frequency response slice is on the blue straight line which we already know represents -2dB. The alternative graphical evaluation in pink is also shown.

at 400 hz both speakers have simular level and the magnepan go not so much in front. so it is faster

Even at 400Hz, the frequency response slices of the Magnepan speaker are much closer together than those of the 10'' woofer. The decay behavior for the first 10-13dB attenuation is thus significantly worse than that of the 10'' woofer - always provided that the measurements and parameters were comparable when creating the diagrams
.

 

[bennybbbx]

@ctrl I do it this way in REW and REW allow in perspective settings bring the slice to front so can verify. rew have a perspective grid. i do a example and show 2 screenshots. for me it look in 3d with all slices around 85 db. but showonly 2 slices and more exact the point is at 83.5 db. waterfall and i give no friends. i think 2D slice view better . I prefer a additional slice view

 

[skyfly]

DSP crossover and crossover-less put aside, the only pro audio brand, I know, that values nice looking step response is PSI audio.

I never saw their off axis response curves. They publish directivity factor, though.

 

[ernestcarl]

DSP crossover and crossover-less put aside, the only pro audio brand, I know, that values nice looking step response is PSI audio.

I never saw their off axis response curves. They publish directivity factor, though.

The graphs are interesting to look at... but very difficult to interpret on their own in isolation.

What is the point of having a "fast" looking decay at 1kHz if all other variables are compromised one way or the other -- not only of the loudspeaker, but think about the listening room space condition for instance.

...

Personally, it would be nice to see the phase response of every speaker published side-by-side along with the FR. Along with directivity & beamwidth plots.

 

[bennybbbx]

The graphs are interesting to look at... but very difficult to interpret on their own in isolation.

What is the point of having a "fast" looking decay at 1kHz if all other variables are compromised one way or the other -- not only of the loudspeaker, but think about the listening room space condition for instance.

...

Personally, it would be nice to see the phase response of every speaker published side-by-side along with the FR. Along with directivity & beamwidth plots.

of course all variables need to look. the most important is FR. but this can correct good with EQ. another thing is the directivity. But can you confirm that a speaker with better FR and better directivity and Phase have good stereo width and depth of field ?. I think not, the Kali have very good direcivity in compare to Canton. phase shift of canton(a passive speaker) is simular to kali. But the canton have much better precision in stereo width and transients.

If you really want verify the trueth you can do it easy. I have audio records of all speakers i test upload and measure *.wav i upload soon. Then you can correct the kali testsong n Phase and FR as you do your speakers. play my testsong corrected with your other speakers(when you play original testsong) or my other speakers records. and when the Kali testsounds sound still with less room width but you correct phase and FR, then you can be sure, that there is another variable that is important for stereo width and transients.

and this my tests show is the speed of midrange, that is important. and the worst thing this can not easy correct. so it is usefull to see the step response in measurements too or better a spectrum slice at 1000 hz. I think so it is possible to find speakers that have good stereo precision and width and good transients

sure you can hear loud so the room reflections get much and it sound more stereo and roomy. then what you hear is not the record room. its your room. additional loud hear damge ears.

 

[KSTR]

Personally, it would be nice to see the phase response of every speaker published side-by-side along with the FR. Along with directivity & beamwidth plots.

Even better would be (at least the on-axis) impulse response data (not just the plots one can derive from it) so everybody can make their own analysis. That's what the world is really missing, a database of anechoic impulse responses. The ZIP with the spin data @amirm provides is nice, but it totally escapes me why no source data is published. This kills any attempts to phase-linearize speakers with reliable source data.

 

[ernestcarl]

If you really want verify the trueth you can do it easy.

There is no need for you to convince me of the truth of what you hear.

It's the rather broad and seemingly concrete conclusions that you are making that I must consider with caution. I consider everything that I think I understand as provisional anyway.

But there really is no argument from me that for good spatial reproduction in stereo and surround sound an accurate transient response is very much desired.

Even better would be (at least the on-axis) impulse response data (not just the plots one can derive from it) so everybody can make their own analysis. That's what the world is really missing, a database of anechoic impulse responses. The ZIP with the spin data @amirm provides is nice, but it totally escapes me why no source data is published. This kills any attempts to phase-linearize speakers with reliable source data.

One could ask the companies directly, but I suspect the replies will be similar to Klipsch's canned e-mail reply to @MZKM here (edited below for brevity):

As an engineering-driven company, we consider _________ measurements proprietary and do not publish this information.

Even if it is only the on-axis impulse response, it probably feels like giving away their actual "master reference" source data naked for the whole public to openly scrutinize.

Not sure about @amirm, but probably has to do with consideration in network bandwidth/or disk space? Although, I suppose, if it's just a single wave file impulse response then it shouldn't really take all that much space.

 

[bennybbbx]

There is no need for you to convince me of the truth of what you hear.
But there really is no argument from me that for good spatial reproduction in stereo and surround sound an accurate transient response is very much desired

https://www.researchgate.net/publication/16552180_Localization_of_sound_in_rooms

The experiments indicate that the localization of impulsive sounds, with strong attack transients, is independent of the room reverberation time, though it may depend upon the room geometry. For sounds without attack transients, localization improves monotonically with the spectral density of the source.

so can be sure faster transients in speaker sound more real as slow transients and the instrument seperation in a mix is enhance too when transients are not slow. maybe the kali is extrem worse with the 0,5 ms slower. I do examples that delay signals left with 44 microseconds 80 microseconds 120 microseconds. then you can all hear. i use a rimshot stick in a much and large room reverb, this have no high freq and no bass freq, so the FR that is need for this is very low.

Even if it is only the on-axis impulse response, it probably feels like giving away their actual "master reference" source data naked for the whole public to openly scrutinize.

when speaker developer fear to give on axis impulses. a 60 degree impulse or a woofer only impulse is good too. that transients are important for localisation can read here

 

[bennybbbx]

@ernestcarl here is a video with diffrent delay times on left channel. hear with headphones or speaker. at which 0.x ms delay time do hear that the signal is not in center ?. the level of left and right channel are always the same.there can see that fast mid range speaker is important

remember 1 khz is a period time of 1 ms. the test use a lowpass cutoff at 1.2 khz and also cut off all lower freq as 280 hz
if you can hear diffrent postion then your brain detect the position by phase shift between left and right transients

 

[KSTR]

One could ask the companies directly, but I suspect the replies will be similar to Klipsch's canned e-mail reply to @MZKM here (edited below for brevity):

As an engineering-driven company, we consider _________ measurements proprietary and do not publish this information.

Even if it is only the on-axis impulse response, it probably feels like giving away their actual "master reference" source data naked for the whole public to openly scrutinize.

Not sure about @amirm, but probably has to do with consideration in network bandwidth/or disk space? Although, I suppose, if it's just a single wave file impulse response then it shouldn't really take all that much space.

In the speaker companies I've worked for, in-house and 3rd-party measurement source data was considered classified material, for the reason you mention and "because no one else does it". This why sites like this great ASR are so important, providing independent top-class measurements under controlled/frozen circumstances which makes comparisons much easier and more reliable.

As for the IR's, IIRC Amir said it would be possible but not a straight-forward procedure with the Klippel? Anyway, that data is still there and could be made available at some point later, maybe under a subscription model (access for forum donors only or something similar).

 

[ernestcarl]

As for the IR's, IIRC Amir said it would be possible but not a straight-forward procedure with the Klippel? Anyway, that data is still there and could be made available at some point later, maybe under a subscription model (access for forum donors only or something similar).

If it takes some extra manual effort/processing then I can see why. Adding it later as a subscription feature sounds pretty reasonable. Although, as I recall, some of his speaker measurements were a bit off center.

 

[onion]

I have impulse response measurements for two rooms, one with M&K cinema speakers that are in-wall (room correction by RoomPerfect); the other with Genelec 8341s (room correction by GLM).

The in-wall speakers have a 'poorer' impulse response that looks jagged, probably due to vibrations emanating from the plane of the wall. This translates to poorer stereo imaging and a lower cross-talk cancellation (XTC) at the ear when running a BACCH filter.

 

[bennybbbx]

I have impulse response measurements for two rooms, one with M&K cinema speakers that are in-wall (room correction by RoomPerfect); the other with Genelec 8341s (room correction by GLM).

The in-wall speakers have a 'poorer' impulse response that looks jagged, probably due to vibrations emanating from the plane of the wall. This translates to poorer stereo imaging and a lower cross-talk cancellation (XTC) at the ear when running a BACCH filter.

View attachment 115421

View attachment 115418

View attachment 115418

the top 2 posts are the in wall speakers ?. there can see it have more noise at 25%. maybe wall vibration or from a grill. Have they a grill ?.the lower have some peaks at 25%. this seem reflektions from wall. better is you use a step response view. there can see better

I measure the iloud MTM at 1.5 cm distance. this have near same fast speed as the eris 3.5, but the front grill of the MTM seem produce reflection when measure very near. how much distance is your measure ?

 

[ctrl]

Here is the analysis of @bennybbbx's measurements from post#194 using Arta.
Since @bennybbbx didn't link the speaker models, I'm going to guess they are

Presonus Eris E3.5
Kali Audio LP-6

Preliminary remarks:

• For most plots (CSD and Burst Decay BC), downsampling of the impulse response from 44.1k to 16k was performed. This increases the frequency resolution in the CSD at low frequencies - see Arta Handbook section 7.3 for more information.
• To make the Eris sound more similar to the Kali in the bass frequency range, @bennybbbx has boosted the bass range of the Eris via miniDSP - quote: "I have boost the bass in the Eris(in t.racks mini dsp) so it is stronger as on Kali".
• Near-field measurements (best <1cm from woofer dome) of the woofers almost completely mask the BR port. Accordingly, possible problems caused by the BR port, such as port resonances, are not displayed (strong cabinet or port resonances may be transmitted in attenuated form).

The order is the same for all diagrams, Eris is shown first, then Kali.

1. frequency response of the near field measurement

The Kali shows an early frequency response drop, which doesn't make the comparison easy.
In the range 100-200Hz both LS show their sound pressure level maximum, so this frequency range is particularly well suited for a comparison.

Since the crossover frequency of the Kali is 1.5 kHz, the frequency response drops very early. With the Eris, the crossover frequency is 2.8 kHz.
Already above 400-500Hz, the frequency response drop of the Kali is so large that the comparison to the Eris is no longer fair.

With the Eris it seems as if no baffle-step correction is made (flat frequency response in the near field measurement), with the Kali the baffle-step is taken into account (slight frequency response drop from 150Hz).
This circumstance alone could explain the different sound impressions.


2. Culmulative spectral decay (CSD) waterfall (30dB attenuation)

Above 200Hz, the Eris shows better decay than the Kali, based on 30dB attenuation. Below 150Hz, the opposite is true.
We will take a closer look at how big the difference is in the decay behavior later.


3. CSD spectrogram
A much easier to read presentation of the decay behavior. The frequency range of occurring resonances can thus be determined precisely.

These are exactly the same measurements as in 2, only in an easier-to-interpret representation.
Here one can see that the resonances of the Kali, which are noticeable in the waterfall diagram and swing out for a long time, are already attenuated by 20dB or more - so they do not play a major role.

If you look at the boundary of the orange area (12dB attenuation), slight advantages for the Eris are also apparent here from 150Hz.
Below 150Hz, it becomes dramatic for the Eris in this representation as well.


4. Burst decay waterfall
It's a kind of CSD with a oscillation period based plot, instead of a time based plot. Impulse response is convoluted with a sine burst - more details see Arta user manual section 6.5

This type of visualisation makes the resonances at different frequencies directly comparable, since the number of oscillation periods serves as a measure.
But even with this visualization, the direct comparison is difficult due to the 3D plot. The visualization as a spectrogram is better suited for this purpose.


5. Burst decay spectrogram

If we look at the yellow area with 18dB damping, there are only minimal differences on average.
With the Kali, the range 600-1200Hz shows delayed decay (to be able to compare it better with the Eris, the sound pressure would have to be increased in this range).
The Eris shows a delayed decay in the 800-900 range. In addition, there is a resonance between 100-140Hz. The delayed oscillation around 50 Hz is certainly the most problematic.
The frequency response boost by the miniDSP in the low frequency range of the Eris is not such a good idea - but it may be that @bennybbbx has done this only for the measurement.


6. Burst decay spectrogram (only 15dB attenuation)
Since @bennybbbx said that for his "theory" (a better stereo imaging of the Eris because it is "faster") it is the first 12dB of attenuation that matters, we look at the oscillation period based decay up to an attenuation of 15dB.

From 200-500Hz there is hardly any difference in the decay behavior. Above 500Hz, the frequency response of the Kali drops too steeply to be comparable. But one can see that the Kali LS should have slight decay problems in the 700-1100Hz range.

The Eris shows delayed decay around 100-120Hz and heavily delayed decay around 50Hz.


7. Step response Eris vs. Kali
Here is the step response comparison of Eris and Kali. I think it should be clear from the above diagrams that in most cases the interpretation of the step response is more guesswork than scientific evaluation.
What you can see very nicely, however, is the delayed decay of the Eris woofer in the low frequency range - Eris (yellow curve), Kali (green curve)

 

[ernestcarl]

Thanks for the detailed analysis @ctrl!

There's a lot to digest and learn here...

Seems like @bennybbbx is primarily interested in the midrange area (?) And so, in these two measurements the attenuation level of the Kali in that area makes it difficult to conclude even within the CSD & spectrograms how big of a problem it truly is (potentially) overall. While, as you say, the Kali do appear to have glimpses of some "slight decay problems" up there, him focusing on in it too much skirts other disadvantages the Eris might also have. After all, sustained resonance in the bass is very much still a transient response issue -- the extended step response & CSD spectrogram views makes it very clear to see.

However, since bennybbbx is doggedly focused in on the midrange decay as a window to stereo soundstage performance, I suspect this may not change his mind on how 'extreme' the Kali is compared to everything else.

 

[bennybbbx]

Here is the analysis of @bennybbbx's measurements from post#194 using Arta.
Since @bennybbbx didn't link the speaker models, I'm going to guess they are

Presonus Eris E3.5
Kali Audio LP-6

Preliminary remarks:

• For most plots (CSD and Burst Decay BC), downsampling of the impulse response from 44.1k to 16k was performed. This increases the frequency resolution in the CSD at low frequencies - see Arta Handbook section 7.3 for more information.
• To make the Eris sound more similar to the Kali in the bass frequency range, @bennybbbx has boosted the bass range of the Eris via miniDSP - quote: "I have boost the bass in the Eris(in t.racks mini dsp) so it is stronger as on Kali".
• Near-field measurements (best <1cm from woofer dome) of the woofers almost completely mask the BR port. Accordingly, possible problems caused by the BR port, such as port resonances, are not displayed (strong cabinet or port resonances may be transmitted in attenuated form).

The order is the same for all diagrams, Eris is shown first, then Kali.

1. frequency response of the near field measurement
View attachment 115525 View attachment 115526
The Kali shows an early frequency response drop, which doesn't make the comparison easy.
In the range 100-200Hz both LS show their sound pressure level maximum, so this frequency range is particularly well suited for a comparison.

Since the crossover frequency of the Kali is 1.5 kHz, the frequency response drops very early. With the Eris, the crossover frequency is 2.8 kHz.
Already above 400-500Hz, the frequency response drop of the Kali is so large that the comparison to the Eris is no longer fair.

With the Eris it seems as if no baffle-step correction is made (flat frequency response in the near field measurement), with the Kali the baffle-step is taken into account (slight frequency response drop from 150Hz).
This circumstance alone could explain the different sound impressions.


2. Culmulative spectral decay (CSD) waterfall (30dB attenuation)
View attachment 115527 View attachment 115528
Above 200Hz, the Eris shows better decay than the Kali, based on 30dB attenuation. Below 150Hz, the opposite is true.
We will take a closer look at how big the difference is in the decay behavior later.


3. CSD spectrogram
A much easier to read presentation of the decay behavior. The frequency range of occurring resonances can thus be determined precisely.
View attachment 115529 View attachment 115530
These are exactly the same measurements as in 2, only in an easier-to-interpret representation.
Here one can see that the resonances of the Kali, which are noticeable in the waterfall diagram and swing out for a long time, are already attenuated by 20dB or more - so they do not play a major role.

If you look at the boundary of the orange area (12dB attenuation), slight advantages for the Eris are also apparent here from 150Hz.
Below 150Hz, it becomes dramatic for the Eris in this representation as well.


4. Burst decay waterfall
It's a kind of CSD with a oscillation period based plot, instead of a time based plot. Impulse response is convoluted with a sine burst - more details see Arta user manual section 6.5
View attachment 115531 View attachment 115532
This type of visualisation makes the resonances at different frequencies directly comparable, since the number of oscillation periods serves as a measure.
But even with this visualization, the direct comparison is difficult due to the 3D plot. The visualization as a spectrogram is better suited for this purpose.


5. Burst decay spectrogram
View attachment 115533 View attachment 115534
If we look at the yellow area with 18dB damping, there are only minimal differences on average.
With the Kali, the range 600-1200Hz shows delayed decay (to be able to compare it better with the Eris, the sound pressure would have to be increased in this range).
The Eris shows a delayed decay in the 800-900 range. In addition, there is a resonance between 100-140Hz. The delayed oscillation around 50 Hz is certainly the most problematic.
The frequency response boost by the miniDSP in the low frequency range of the Eris is not such a good idea - but it may be that @bennybbbx has done this only for the measurement.


6. Burst decay spectrogram (only 15dB attenuation)
Since @bennybbbx said that for his "theory" (a better stereo imaging of the Eris because it is "faster") it is the first 12dB of attenuation that matters, we look at the oscillation period based decay up to an attenuation of 15dB.
View attachment 115535 View attachment 115536
From 200-500Hz there is hardly any difference in the decay behavior. Above 500Hz, the frequency response of the Kali drops too steeply to be comparable. But one can see that the Kali LS should have slight decay problems in the 700-1100Hz range.

The Eris shows delayed decay around 100-120Hz and heavily delayed decay around 50Hz.


7. Step response Eris vs. Kali
Here is the step response comparison of Eris and Kali. I think it should be clear from the above diagrams that in most cases the interpretation of the step response is more guesswork than scientific evaluation.
What you can see very nicely, however, is the delayed decay of the Eris woofer in the low frequency range - Eris (yellow curve), Kali (green curve)
View attachment 115537

you think it is not clear visible that the Eris have better step response ?. I think it is clear but it seem arta miss the feature REW have. "Normalize to peak at each frequency". this is important to enable. the kali does not reach max level at 1 khz because it is so slow. the kali have crossover frequency at 1.5 khz so at 1 khz need full level or maybe 1 db - see- Because kali reach not max level it can of course faster decay. so look not so worse as it really is. but if you find a option in arta "Normalize to peak at each frequency" then you can better compare. and the much more bass boost on the Eris i do because i want show that also a bass enhance not change the step response much.

you need keep in mind. at 1 khz period time is 1 ms. at 100 hz period time is 10 ms. so if a speaker is at 1 khz 1 ms faster it is very much in relation to period time. If a speaker at 100 hz is 1 ms faster is very few. also need remember that fast step response below 150 hz is not important because binaural hear begin only around 150 hz. have you hear my video with the delay example ?. did you hear diffrence in position and room ?.in this example are lower frequency as 280 hz cut.

If arta have no option "Normalize to peak at each frequency". i can do on Kali EQ too, so they look simular to eris. let me know. for example the spectrum of kali look with and without this option enable as this.

EDIT: In your step response there is too no normalisation done. normalisation is important to compare. i think step response can show more than if there is no step response .

best to show speed i think is your arta burst decay BUT only when you find a option that it is nornalized, so all peaks in all frequency reach 100% or 0 db

 

[ernestcarl]

Applying a high pass filter so that the roll-off in the bass is equal between the two (and disregarding the Eris's bass EQ completely) might make things more easy to visually compare.

Filtered Full step response view shows the Kali as slighty "slower"

Filtered Impulse response view indicates that the Kali has a somewhat inferior transient response

Filtered Wavelet spectrogram view is the clearest indicator of which woofer has the better transient response

1/3 resolution (favoring time resolution a bit more), 25dB scale

Can't say this is the absolute best (or even valid) way to do the transient response comparison, but additional HP pre-filtering (idea I got from Dave Gunness) does seem to make the process of visual analysis much easier.