Step Response: Does It Really Matter?

[ernestcarl]

to show that it change alot when change the phase.

At the distance measured (somewhat anechoic at 1.5cm), phase does not change "a lot" in the mid-frequencies of the woofer even after using the "minimum phase version" of both filtered and non-filtered IRs.

not high-passed

high-passed

Maybe there are speaker developers that have faster speakers for a good price that give a step response around 0.7 ms (headphone reach 0.3 ms) this bring even more much better sound i think.

Good luck in your quest for "good price" headphone-like, full-size, full-range speakers.

 

[bennybbbx]

At the distance measured (somewhat anechoic at 1.5cm), phase does not change "a lot" in the mid-frequencies of the woofer even after using the "minimum phase version" of both filtered and non-filtered IRs.

not high-passed
View attachment 116103

high-passed
View attachment 116104



Good luck in your quest for "good price" headphone-like, full-size, full-range speakers.

Wy you not answer if you hear the position diffrence in the video i link ?. did you get ASMR feelings ?. for ASMR is know use headphones. I can get ASMR feelings but not with speakers. maybe work only on precise mid speakers. this bring me to idea i ask in asmr forum if they hear positions in my testvideo


I notice now when enable Export min phase of IR then i can not export with EQ.
you show only phase from 450 to 1.25 khz. but the HP filter give massive phase shifts in range 450 to 1.3 khz too.this unnatural phase shift of the HP filter do reduce the precision of step response. the spectrum with the slice view is resistant to such unnatural phase shifts step response not.

i have screenshots with and without HP filter. without HP diffrence from 400 hz to 1300 is 107 degree (1.3 khz is - value) so need add. from 400 -1300 hz on the version without HP it is only 50 degree. so the phase shift in this range is over 100% in the HP version. I hope it is clear now that not use a HP. it change measure alot. when use a peak filter and reduce or enhance to other speaker is much much better.

I think fast mid range is not a question of price. it depend if speaker developer and customer hear this. i know many that do not like speakers and hear more over headphones because they think it sound better. they thought it is the room and distance. but now i have the small speakers i know, it is not the room. sure room have influence but not so much.

neumann speakers are very expensive in compare to others but they have as can see on your impulses not so fast speed as your large sceptre 8 inch speaker. Maybe Kevlar is better for fast speakers. my 99 eur eris 3.5 (pair) have kevlar. sceptre as i see have no kevlar. in compare to mtm i think the midrange of eris sound more clear. but the high freq from tweeter of the MTM i think sound more clear. the only i see in measure is that the eris have a very large overshot in tweeter. the Kali have a little less and the MTM much less. this can only see in step response. the overshot in step response is not change with HP filter.

so to check if a speaker is fast and neutral should also look that it have not much overshot. and overshoot can only see in step response. because of normalized the overshot in rise time can not see, but the overshoot in fall time can see

 

[bennybbbx]

Oh no, I missed "44.1 kHz and 22 µs time resolution" party?

Here's an impulse, bandwidth limited to 2 kHz (so maybe I should have put "impulse" in quotes) and shifted in one channel by around 1.4 µs, 32 times. I bandwidth limited it because Audacity does a poor job at properly rendering the waveform, it just connects samples with straight lines.

I admit though, that I did not start at 44.1 kHz, upsample, trim and downsample but already at 705 kHz and just trim and downsample. AFAIK it shouldn't make much difference if any at all, but I'm ready to be told otherwise

View attachment 116071

we talk since longer time not about this. but because you write to this and do a graphic, then maybe do the same post at a frequency of 11 khz. this are around 4 samples for 1 11 khz wavecycle . how it look then when you not shift in sample steps. there can hear when upsample a sample from 44 khz to 48 khz that it loss some high freq when use a worse linear interpolation. maybe sinc interpolation avoid this. but shift a sample not in sample steps cause many errors. maybe some hear it some not. 1. interpolation is upsampling . 2. interpolation is shift. 3. interpolation need is downsampling. and keep in mind in music delays near always have a feedback parameter. this mean the delayed result is mix with a reduced value to the input again and is delayed again. so the interpolation errors are exponential increase

 

[ernestcarl]

Wy you not answer if you hear the position diffrence in the video i link ?. did you get ASMR feelings ?. for ASMR is know use headphones. I can get ASMR feelings but not with speakers. maybe work only on precise mid speakers. this bring me to idea i ask in asmr forum if they hear positions in my testvideo

Irrelevant.

I notice now when enable Export min phase of IR then i can not export with EQ.

So extract the MP version after applying the HPF.

I notice now when enable Export min phase of IR then i can not export with EQ.
you show only phase from 450 to 1.25 khz. but the HP filter give massive phase shifts in range 450 to 1.3 khz too.this unnatural phase shift of the HP filter do reduce the precision of step response. the spectrum with the slice view is resistant to such unnatural phase shifts step response not.

i have screenshots with and without HP filter. without HP diffrence from 400 hz to 1300 is 107 degree (1.3 khz is - value) so need add. from 400 -1300 hz on the version without HP it is only 50 degree. so the phase shift in this range is over 100% in the HP version. I hope it is clear now that not use a HP. it change measure alot. when use a peak filter and reduce or enhance to other speaker is much much better.

That amount of phase shift is negligible, not to mention irrelevant for the purpose of evaluating the woofer's decay in the midrange window looked at using wavelet analysis. If you use the sliced window (your preferred method) under the spectrogram view to slowly slide down the midrange FR, you hardly see any difference in the "speed of decay" in that step view. The time is shifted, obviously, but the actual amount of decay stays the same.

*The HP filter was used so that we could compare the Eris and Kali on more even terms -- reducing the bass level difference between the two. That's it.

neumann speakers are very expensive in compare to others but they have as can see on your impulses not so fast speed as your large sceptre 8 inch speaker. Maybe Kevlar is better for fast speakers. my 99 eur eris 3.5 (pair) have kevlar. sceptre as i see have no kevlar. in compare to mtm i think the midrange of eris sound more clear. but the high freq from tweeter of the MTM i think sound more clear. the only i see in measure is that the eris have a very large overshot in tweeter. the Kali have a little less and the MTM much less. this can only see in step response. the overshot in step response is not change with HP filter.

As already mentioned before in my post with the image of the KH120s spectrogram zoomed-in, the difference in apparent "speed" has more to do with xo causing some lag in the phase which also can be fixed after-the-fact via DSP. The KH80 fixes this for you if you consider it as an important issue; alternatively, one can get the KH750 and pair with the KH120s (for linear-phase) if preferred.

But you said you wanted cheap loudspeakers that sounded like headphones. I don't know of any. You mentioned the "speed" of headphones as the thing that makes them better than full-sized speakers. There's more to it than that... Try asking Sean Olive or any other people here who are more knowledgeable about headphones. Maybe you can audition a pair of these floor-standing speakers in person and tell me if they are to your liking: https://www.audioholics.com/tower-speaker-reviews/polk-legend-series

*forgot about this: https://www.audiosciencereview.com/...pro-edition-for-those-considering-bacch.7917/

 

[bennybbbx]

That amount of phase shift is negligible, not to mention irrelevant for the purpose of evaluating the woofer's decay in the midrange window looked at using wavelet analysis. If you use the sliced window (your preferred method) under the spectrogram view to slowly slide down the midrange FR, you hardly see any difference in the "speed of decay" in that step view./

the look at step response is only a easy way to see slow mid range too when there is no spectogram with a slice (in all speaker tests i know is this not). with phase shifts due to filters can fake the step resonse results.
remember my first measure with sine burst on the is kali defect thread. you write in this thread too. the burst records show that the speakers are not precise. All in thread told i should use speaker measure software. so i do this in REW.

for the step response the phase shift is of some influence. Can see the 100% phase shift addition with the HP result in ~ 25% shorter Step response. the wavelet and slice show it good, but the measure system speaker testers have there is no such slice view. they show (when have luck) unnormalized waterfall and this is not good to compare. the kali and the mtm @amirm do have no waterfall. this https://www.audiosciencereview.com/...rk-rokit-5-gen-4-review-studio-monitor.20711/ have a waterfall and can see look slow on mid with average step time of ~ 2.5 ms to reduce level with 10 db 1 khz. but its not so clear see. When other speakers have such a diagram too then also can compare. there is 3 way speaker but no waterfall https://www.audiosciencereview.com/forum/index.php?threads/nht-c3-review-3-way-speaker.20287/

if hear diffrence in the video answer is important when hear diffrent position, because if hear it not, then of course you need no precise mid range and you hear maybe not much diffrence between large 2 way speakers or small in mid so many speakers are good enough for you.

and if you hear diffrence then i think it should be clear that i hear it really and it is no hifi voodoo. and when such a small delay change position and reverb then should understand that a speaker need fast in mid range.

 

[ernestcarl]

step response is only a easy way to see slow mid range too when there is no spectogram

Can you tell me which among these four loudspeaker systems has the slowest mid-range?

They're all minimum phase versions of their IRs, too.

 

[bennybbbx]

Can you tell me which among these four loudspeaker systems has the slowest mid-range?

View attachment 116202

They're all minimum phase versions of their IRs, too.

Are this recordet impulses from speakers and not process with HP Filter or EQ ?. in my speaker measure i notice the slower speakers get more overshoot too. (see kali it reach -70%. the MTM and eris reach -50%) . I write earlier you need a look at the time the speaker reach 0 and how much % overhoot it have. the overshoot did not change when use HP i see. but can change with EQ. the fastest(blue) have step response(to reach 0) of ~ 8 ms. this is much slower as the kali. the yellow is a little slower and have less overshoot. I did not know which is more worse much overshoot or slower step response. for this i need hearing tests. but you want not do records of your sceptre and the KH120 with my demo song with 1 speaker in a DAW so the left and right seperate records are phase exact.

for step responses there should no EQ use. for frequency range measure too no EQ is use.
maybe step response did not change when a phase linear EQ is used. I notice when use peak eq with less phase shift that step response very few change. see my eris with and without EQ measure. but your HP measure give much phasesift.

 

[ernestcarl]

Are this recordet impulses from speakers and not process with HP Filter or EQ ?. in my speaker measure i notice the slower speakers get more overshoot too. (see kali it reach -70%. the MTM and eris reach -50%) . I write earlier you need a look at the time the speaker reach 0 and how much % overhoot it have. the overshoot did not change when use HP i see. but can change with EQ. the fastest(blue) have step response(to reach 0) of ~ 8 ms. this is much slower as the kali. the yellow is a little slower and have less overshoot. I did not know which is more worse much overshoot or slower step response. for this i need hearing tests. but you want not do records of your sceptre and the KH120 with my demo song with 1 speaker in a DAW so the left and right seperate records are phase exact.

for step responses there should no EQ use. for frequency range measure too no EQ is use.
maybe step response did not change when a phase linear EQ is used. I notice when use peak eq with less phase shift that step response very few change. see my eris with and without EQ measure. but your HP measure give much phasesift.

Right. You might as well be reading tea leaves, then.

 

[bennybbbx]

Right. You might as well be reading tea leaves, then.

this mean this speakers are not process with EQ and are so slow ?. which speakers are this ?. then please post spectrum slice so can see if there is really much diffrence. see the kali spectrum for settings. the step response and the spectrum slice have not much diffrence of all speaker and headphones i measure. I have 4 speakers and 4 headphones to measure and all are much faster as the results you post. Maybe this speakers you post are really worse and slow ? . I need hear this but you can not record.

did you know if i can export the whole recordet sine sweep in rew ?. Then i can process this easy with a phase linear EQ and so can see the change in impulse with your HP eq processing happen because the eq shift the phase in unatural way.

 

[bennybbbx]

@ernestcarl maybe this explain more what the step response show. please look at this wikipedia image https://en.wikipedia.org/wiki/Slew_rate#/media/File:Slew-rate.svg you see that on right side the input signal go immidiatly to 0. but the output signal need some time. and this is what step response show the time to reach 0 and the overshoot. a normalized step response set from the time when the input signal go to 1 and immidiate to 0 as 0 ms. it also set the peak to 100%. With this explain it is also clear to see, that when you use a EQ that do phase shift this measure can not correct, because the EQ shift the result

 

[ernestcarl]

All speakers use FIR filtering to linearize parts of their phase.

2x Neumann KH310 + KH750 (linearized via MA-1 & KH750)
1x Neumann KH120 + Rythmik F12 (linearized via miniDSP)
1x Presonus Sceptre S8 + Rythmik F12 (linearized through internal DSP, and external DSP via JRiver)

All IRs were recorded in different rooms and/or listening positions -- also different setup & equipment settings -- distances ranging from 0.8m - 2.0m

Looking at the full step response & (to a much lesser degree) wavelet between different loudspeaker setups is not very informative and far from optimal unless some kind of filtering and controls are applied. I think just about all step responses I've seen posted in reviews have no such pre-processing done. So asking people to post their loudspeaker step response as a quick way to determine the "speed" of their system is largely pointless.

I may have mentioned before (or not), but the way one exports the IR in REW can cause different results. It seems that the best way is to go to the ALL SPL tab and use the export options there rather than in the main file menu.

 

[ernestcarl]

Filtered IRs of same

MDAT files:

200Hz HPF, level offset, FDW 5, est. IR delay (Min phase version)

200Hz HPF, level offset, FDW 5, est. IR delay


Which loudspeaker system has clearly the "slowest" mid-range, then?

Personally, I'm not so sure if there is even a significant difference among the four samples just looking at the step responses of the pre-processed IRs.

 

[bennybbbx]

Filtered IRs of same

View attachment 116352

View attachment 116353

View attachment 116354

View attachment 116355

View attachment 116356


MDAT files:

200Hz HPF, level offset, FDW 5, est. IR delay (Min phase version)

200Hz HPF, level offset, FDW 5, est. IR delay


Which loudspeaker system has clearly the "slowest" mid-range, then?

Personally, I'm not so sure if there is even a significant difference among the four samples just looking at the step responses of the pre-processed IRs.

please see my last post with the wiki image. that step response show usefull you should not add a EQ or linearize the phase. when you lineraize the phase it make the speaker not really faster. only the step response look nicer. So please post measures when EQ and any phase correct is in mini dsp switch off. also the Eq knobs of the speaker should be in the middle . when your both speakers use diffrent EQ you can of course not compare. The CSD slice of the HP 200 eris noeq show no diffrence because in the 1 khz range is nothing eq. but when speaker is boost with Eq around 1 khz it give phase shift too.

 

[ernestcarl]

please see my last post with the wiki image. that step response show usefull you should not add a EQ or linearize the phase. when you lineraize the phase it make the speaker not really faster. only the step response look nicer. So please post measures when EQ and any phase correct is in mini dsp switch off. also the Eq knobs of the speaker should be in the middle . when your both speakers use diffrent EQ you can of course not compare. The CSD slice of the HP 200 eris noeq show no diffrence because in the 1 khz range is nothing eq. but when speaker is boost with Eq around 1 khz it give phase shift too.

You can't see the forest for the trees...

Thanks, I appreciate your enthusiasm (or relentless persistence? ), but I really don't think I can add much more to this discussion.

 

[bennybbbx]

You can't see the forest for the trees...

Thanks, I appreciate your enthusiasm (or relentless persistence? ), but I really don't think I can add much more to this discussion.

I dont know wy you not understand that to test a speaker it is important to use no EQ. because Eq change phase. so wy do you not show a no eq step response of your speaker ?. I dont know wy you not just show from the spectrum the slice at 1000 hz. then can see how fast your speakers are and how this is in compare to step response.

 

[bennybbbx]

I find on stereophile a fast speaker step response https://www.stereophile.com/content/manger-s1-active-loudspeaker but they have no near field desktop speakers and they are very expensive.

i look on manger homepage, and they write this.

Particular emphasis was given to a study of the human direction sensing mechanism. This showed that in addition to the well established system of pitch recognition on sustained notes, the ear has an evolutionarily older mechanism by which it locates a sound source through transients.

This mechanism is very powerful because it is the descendant of a highly evolved survival technique from the dawn of mankind, where the slightest noise, such as the snapping of a twig, would represent a threat to survival. Noises of this kind were useful information for all species throughout the world. Speech and music specific to the locality came much later, the noise of machinery later still.

Transient noises produce a one-off pressure step whose source is accurately and instinctively located. If these transients are not accurately reproduced, the subconscious direction finding mechanism is defeated. The sound becomes unrealistic, the stereophonic image is impaired and the result is listening fatigue.

This research led not just to a confirmation of why the traditional loudspeaker is so poor, but also to the specification of an ideal loudspeaker. From then on, the problem was simply how the specification could be met. I was told it was impossible, but I argued that if the human hearing mechanism could do it, then a man-made device could also do it. After a great deal of development, the required result was obtained.

https://mangeraudio.com/en/discover/about/acoustical-reality

 

[ctrl]

I find on stereophile a fast speaker step response https://www.stereophile.com/content/manger-s1-active-loudspeaker but they have no near field speakers and they are very expensive.

Your way of looking at it ("after how many milliseconds the x-axis is intersected during the step response") is, as has been said several times now, much too simplistic and naive and tempts others to adopt this shortened representation as well.

With @bennybbbx I have already given up hope, but perhaps the following shows others that the step response often does not tell you much and often the wrong thing.

Let's start from the Step Response in the Stereophile.
According to @bennybbbx, an extremely fast speaker:

Source: Stereophile.com

The crossover frequency to the woofer is 330Hz and the step response shows that the woofer is connected phase inverted.
Thus, it is obvious to assume a 2nd order crossover.

Now we simulate the crossover with ideal drivers:

It does look quite similar. So, now let's deactivate the woofer and just look at the Manger driver:

OMG, if the woofer is deactivated, then the manger driver slows down and crosses the x-axis later!!!
This is of course total nonsense, nothing has changed at all in the decay behavior of the Manger driver.

Now we get completely cocky and look at the Manger driver's step response without a crossover:

Source: troelsgravesen.dk

WTF, what is this lame sh..t! By @bennybbbx's definition, this would be a driver for the trash can.


Unfortunately, the usual CSD is missing from the Stereophile Manger S1 review. But you can find a CSD of the Manger driver at troelsgravesen.dk:

In the frequency range above 3-4kHz the Manger driver shows a good decay behavior, the resonances visible in this range decay very quickly.
Below 3 kHz there are severe decay problems. In the range 600-800Hz there is a bad resonance that needs at least three to four oscillation periods up to 25dB damping (maybe even significantly more, this can not be said clearly, since several resonances run into each other).

There is no use in making up nice advertising stories for "audiofools" and hobby sound engineers about transients when the speaker does not deliver on these promises. To stay in the advertising language, how should such a loudspeaker cleanly separate transients in the midrange, if they still resonate for a very long time?

For comparison, a loudspeaker that shows very good decay under the same evaluation conditions.
After about 1.4ms the speaker is damped by 25dB down to 400Hz.
The Manger driver needs 2-3 times as long in many frequency ranges.
Even if the numerical values should not be taken too precisely, the differences are extreme.

Source: troelsgravesen.dk

Who would have expected such a "fast" behavior from this speaker when looking at the step response of the woofer

Therefore again, the interpretation of impulse response and step response which quite a few do is very often "esoteric science".

A good book like "Testing Loudspeaker" by Joseph D'Appolito, Arta Handbook and Manual and the simulation of different filters and crossovers with free software saves one from false simplifications (but one must plan a few months of study time).

 

[bennybbbx]

Your way of looking at it ("after how many milliseconds the x-axis is intersected during the step response") is, as has been said several times now, much too simplistic and naive and tempts others to adopt this shortened representation as well.

With @bennybbbx I have already given up hope, but perhaps the following shows others that the step response often does not tell you much and often the wrong thing.

Let's start from the Step Response in the Stereophile.
According to @bennybbbx, an extremely fast speaker:
View attachment 117004
Source: Stereophile.com

The crossover frequency to the woofer is 330Hz and the step response shows that the woofer is connected phase inverted.
Thus, it is obvious to assume a 2nd order crossover.

Now we simulate the crossover with ideal drivers:
View attachment 117021
It does look quite similar. So, now let's deactivate the woofer and just look at the Manger driver:
View attachment 117022
OMG, if the woofer is deactivated, then the manger driver slows down and crosses the x-axis later!!!
This is of course total nonsense, nothing has changed at all in the decay behavior of the Manger driver.

Now we get completely cocky and look at the Manger driver's step response without a crossover:
View attachment 117027
Source: troelsgravesen.dk

WTF, what is this lame sh..t! By @bennybbbx's definition, this would be a driver for the trash can.


Unfortunately, the usual CSD is missing from the Stereophile Manger S1 review. But you can find a CSD of the Manger driver at troelsgravesen.dk:
View attachment 117028
In the frequency range above 3-4kHz the Manger driver shows a good decay behavior, the resonances visible in this range decay very quickly.
Below 3 kHz there are severe decay problems. In the range 600-800Hz there is a bad resonance that needs at least three to four oscillation periods up to 25dB damping (maybe even significantly more, this can not be said clearly, since several resonances run into each other).

There is no use in making up nice advertising stories for "audiofools" and hobby sound engineers about transients when the speaker does not deliver on these promises. To stay in the advertising language, how should such a loudspeaker cleanly separate transients in the midrange, if they still resonate for a very long time?

For comparison, a loudspeaker that shows very good decay under the same evaluation conditions.
After about 1.4ms the speaker is damped by 25dB down to 400Hz.
The Manger driver needs 2-3 times as long in many frequency ranges.
Even if the numerical values should not be taken too precisely, the differences are extreme.
View attachment 117032 View attachment 117033
Source: troelsgravesen.dk

Who would have expected such a "fast" behavior from this speaker when looking at the step response of the woofer

Therefore again, the interpretation of impulse response and step response which quite a few do is very often "esoteric science".

A good book like "Testing Loudspeaker" by Joseph D'Appolito, Arta Handbook and Manual and the simulation of different filters and crossovers with free software saves one from false simplifications (but one must plan a few months of study time).

Yeah i need this "esoteric science" because there is no good measure solution publish to see which speaker is faster in midrange. I have only confirm that step response show the fastest speaker with my 7 systems(inklusice headphones i have). REW i think have a good solution when use spectrogram and slice at 1 khz. but no speaker tester publish simular. when i only can look at tests, i need to judge with thats whats there. the waterfall from the manger look really not good. the troelsgraven DIY think look better. But can you tell how much better in % at 1 khz it is as the MTM or Eris or Kali ?. the measurements from my speakers you have.

that the waterfall is not publish from the manger i also think strange. but ok, now another extension of my esoteric science If the overshoot is large as see in the manger speaker then it is too no good speaker. my better sounding systems or speakers use also less overshoot. you write that the woofer look as phase inverted, but i think it is much overshoot. waterfall did not show overshoot but overshoot happen on speaker, and of course a speaker with less overshoot is better

EDIT: i look the step response from your link for the manger. this step response look very slow. dont know wy there is so much diffence. I better learn by doing than by books, because there is only rudementary theorie. and i learn step responses (time to reach 0 and overshoot) are not much room depend. so question is wy on the link you post is the manger step response much slower as on stereophile and wy stereophile did not post waterfall. maybe @John Atkinson can say something about that

 

[bennybbbx]

I find another test with manger and waterfall. https://www.i-fidelity.net/testberichte/high-end/manger-z1/labor-seite-6.html here the waterfall look better as at troeslgraven, but can not see how much better it is as mtm or eris or kali

 

[ctrl]

that the waterfall is not publish from the manger i also think strange. but ok, now another extension of my esoteric science If the overshoot is large as see in the manger speaker then it is too no good speaker. my better sounding systems or speakers use also less overshoot. you write that the woofer look as phase inverted, but i think it is much overshoot. waterfall did not show overshoot but overshoot happen on speaker, and of course a speaker with less overshoot is better

You are right, this is again pure esoteric.
If a fourth order filter were used with the Manger speaker, there would not be such a large overshoot in the step response, but due to the larger phase shift (360° instead of 180°), the group delay would increase.

To be able to interpret such a complex thing as a step response correctly, one needs technical knowledge. Without it's not possible to draw the right conclusions - filter theory, in particular all-pass filter, the step response as integral of the impulse response, phase shift, group delay, crossover design,...