Step Response: Does It Really Matter?

[bennybbbx]

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.


View attachment 115657


Filtered Full step response view shows the Kali as slighty "slower"
View attachment 115658


Filtered Impulse response view indicates that the Kali has a somewhat inferior transient response
View attachment 115659


Filtered Wavelet spectrogram view is the clearest indicator of which woofer has the better transient response
View attachment 115660
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.

intresting result but, i think change the measure wav result with EQ is wrong, when there need set EQ for speaker then only when measure. because when i measure the Eris with Eq or without EQ the step response is near same. need 1 ms to reach 0. here is the wav of the measure without EQ

So i think it is not a good idea to get correct values to modify with EQ the measured impulse. because the speaker membran reaction depend on input signal. if a slow wobbling membrane measure result is high pass filter the slow wobbling is reduce. you can see in my attached noeq wav that the eris need 1 ms with lower bass as kali or more bass as kali, and never reach the values you show in your diagram. here are all measure diagram eris with or without EQ

https://www.audiosciencereview.com/...oes-it-really-matter.1999/page-11#post-686831

Edit: i have attach noeq and eq measure. this measures are with same microphone position

 

[ernestcarl]

The ff. may actually be more fair for the step response view?

200Hz HPF equally same and no level offset.

We want the bass FR roll-off slope to be at least the same here to level things a little bit even if the mids are not the same in SPL level.

The difference in the step response looks more obvious -- same with impulse response. But the wavelet now is more difficult to compare/assess due to very different SPL levels in the mids between the two. If we are just comparing the mids using the wavelet, it's preferable that the level be close to the same throughout that range.

 

[bennybbbx]

if you filter i think you need also enhance the frequency from 500 hz -1 khz from kali. when you do this, then the overshot of kali get more and it get slower. but i still think EQ a impulse response did not work. for example can compare with a slew rate diagram. a speaker too have a slew rate and as see in this example phase shift happen. now when EQ this phase shiftzet result is a diffrence as when EQ the signal for Measuring. https://en.wikipedia.org/wiki/Slew_rate and of course on stereo music each speaker play diffrence to other so phase shfift can happen independent. what need for slow speaker is something that is used for TFT monitors. they called it overdrive. when maybe from a impulse can calc a overdrive table for a speaker then it is possible to boost the signal with a plugin that use the measure of the speaker.

 

[ernestcarl]

if you filter i think you need also enhance the frequency from 500 hz -1 khz from kali. when you do this, then the overshot of kali get more and it get slower. but i still think EQ a impulse response did not work. for example can compare with a slew rate diagram. a speaker too have a slew rate and as see in this example phase shift happen. now when EQ this phase shiftzet result is a diffrence as when EQ the signal for Measuring. https://en.wikipedia.org/wiki/Slew_rate and of course on stereo music each speaker play diffrence to other so phase shfift can happen independent. what need for slow speaker is something that is used for TFT monitors. they called it overdrive. when maybe from a impulse can calc a overdrive table for a speaker then it is possible to boost the signal with a plugin that use the measure of the speaker.

I don’t think it’s necessarily “wrong”. We only want to view a section of the impulse responses above a certain point; mainly the mids where the mid-woofer is active. The change in phase after the HP filter is not gigantic. Again, not really doing this for publication in a paper so don’t care that the measurements fit your specifications to a T.


Though I think it’s already quite clear at this point that the full step response of a loudspeaker is almost useless when viewed in isolation. One cannot draw broad conclusions about speakers by just looking at their step response... No wonder why we don’t always see it more often in professional reviews.

 

[bennybbbx]

I don’t think it’s necessarily “wrong”. We only want to view a section of the impulse responses above a certain point; mainly the mids where the mid-woofer is active. The change in phase after the HP filter is not gigantic. Again, not really doing this for publication in a paper so don’t care that the measurements fit your specifications to a T.


Though I think it’s already quite clear at this point that the full step response of a loudspeaker is almost useless when viewed in isolation. One cannot draw broad conclusions about speakers by just looking at their step response... No wonder why we don’t always see it more often in professional reviews.

the only thing you show and we know now is that when you change a measured impulse later if EQ the result change alot. but you see when use a EQ on speaker or no EQ on speaker the FR of the Eris 3.5is lots diffrent, but the step response is near simular. this show the impulse wav i upload too

but it is true that a good spectogram (that is at all frequency normalized to best compare) is better as a step response. step responses are always normalized and most CSD see are not normalized at each frequency. and this is the main reason i think step impulse better as a not normalized spectral decay or waterfall diagram to see speed

Your results confirm that it is bad to change a measured impulse later with a EQ. you see on your Eq ed impulse that the impulse go to 0 in around 0.7 ms. on my measure the eris need around 1 msec with and without EQ

And if you think the low frequency have influence of the impulse remember 50 hz have 20 ms periode time. 200 hz 5 ms.
So the question is wy the impulse is not 5 ms or 2.5 ms or 10 ms long and only 1 ms on my measure and 0.7 ms on your measure ?

I think the step response show the fastest time a speaker can reach 100% and drop to 0 and show the overshot of the speaker.

 

[ernestcarl]

the question is wy the impulse is not 5 ms or 2.5 ms or 10 ms long and only 1 ms on my measure and 0.7 ms on your measure ?

Which measurements are you talking about exactly?

 

[bennybbbx]

View attachment 115674

and from my measurements i mean this with and without EQ. see screenshots below.You see there is much frequency change in my measure with and without EQ. but the step response is near same. to reach 0 it need around 1 ms. on your above step response it need only 0.5 ms for the eris toreach 0. i try later to EQ the eris that it is in FR range near same as the Kali.

 

[ernestcarl]

and from my measurements i mean this with and without EQ. see screenshots below.You see there is much frequency change in my measure with and without EQ. but the step response is near same. to reach 0 it need around 1 ms. on your above step response it need only 0.5 ms for the eris toreach 0. i try later to EQ the eris that it is in FR range near same as the Kali.
View attachment 115878View attachment 115879

I did say that a 200Hz HP filter was applied. Magnitude of bass frequencies and extended decay was reduced.

 

[bennybbbx]

I did say that a 200Hz HP filter was applied. Magnitude of bass frequencies and extended decay was reduced.

Yes, but wy your changed impulse wav view is diffrent so much to my measured impulses with and without EQ. and when i use the EQ before measure there can not see diffrence

Which HP filter you use exactly ?. I use then the filter settings you use in my t.racks minidsp and measure the eris. so can see if it can work to change impulses after measure.

A high pass filter do more phase shift as peak filter. the t.racks show phase. here can see phases of diffrent flters. On output 3 is left Eris. The t.racks have also allpass filters. but they seem to can not correct phase shifts. can only add more phase shift. but this is maybe usefull to simulate more phaseshift and hear how it sound

 

[ernestcarl]

Yes, but wy your changed impulse wav view is diffrent so much to my measured impulses with and without EQ. and when i use the EQ before measure there can not see diffrence

Which HP filter you use exactly ?. I use then the filter settings you use in my t.racks minidsp and measure the eris. so can see if it can work to change impulses after measure.

A high pass filter do more phase shift as peak filter. the t.racks show phase. here can see phases of diffrent flters. On output 3 is left Eris. View attachment 115966View attachment 115967View attachment 115968View attachment 115969View attachment 115970

I just used REW's default HP filter -- I assume a Q of 0.707

Exported the impulse with the EQ filters enabled or the HPF and re-imported. That's it. Dunno what else happens in between the software, really.

 

[bennybbbx]

I did not know how can use filter in REW. I think the settings in EQ tab are for extern device and not work on impulses. Exporting in REW seem also change data. see how the importet wav look in rew and how the mdat look in compare. what export format you use ?. I use float 32 the default in REW. maybe thats not good ?. normaly float 32 can hold 16 bit exact. i measure with 24 bit but my microphone is not so much low noise as a expensive one. so i think more as 16 bit dynamic i can not get. at 6 khz the reimport wave is 3 db more. When i import the impulse wav with the arta Demo it seem recognize this not. maybe @ctrl can say more to this, whats happen ? . i click on fit and level is very low . look as only noise is import

the zip compress mdat files are larger as forum allow (need 1.4 megabyte each). so i upload here the eris mdat with eq and without eq both in this zip.

https://file.io/V8zcpsG8SFNk

FR and phase change on import but the step response stay same. I use for FR on both 1/6 smooth, but it is also visible with 1/1 smooth

 

[ernestcarl]

@bennybbbx

Use the EQ window button and apply filter(s).

Then under the file menu, export the data with the EQ filters -- I believe I used the export as wave option.

Also, I have noticed that depending on how one exports the IR, the response can change.

 

[ernestcarl]

I don't see anything out of the ordinary.

Applied 200Hz HPF to the files:

Expected shift in phase

and step response

I still don't see what the whole big issue is...

 

[bennybbbx]

I see in rew only menu "export filters impulse response as wav" . this does not save the measured impulse. this save only the impulse of the filter. this feature is used for convolution VST plugins and contain EQ data to correct speakers. here i do a HP with 200 hz

then save with this menu "export filters impulse response as wav"

when load the export impulse then it show frequency response of a 200 hz low pass filter and not the eris.

I did not see how i can change with EQ the eris impulse in REW

 

[ernestcarl]

I did not see how i can change with EQ the eris impulse in REW

After you've applied the filter in the EQ window, close that window. Make sure the measurement you want to export is active.

Step one:

Step two:

Step three:

Just select the exported file and open it.

 

[bennybbbx]

ah ok, now i understand wy many think step response is worse. I dont know wy REW have such a option. if want filter it need a linear phase filter. see in the phase response. the phase response is change alot after HP filter. and step response need the phase to calc correct in some parts. when do Frequency range measure and change with Eq give too wrong results. so get wrong results when change the phase of a speaker much with a filter. there is option Export min phase Version of IR. then step response 1.5 ms to reach 0. but i try what happen when i do this with my DSP and HP 200. maybe it fail because i do not record impulse from 10 to 20 khz. need examine closer.

"export EQ filtered with min phase version of IR"

 

[bennybbbx]

@ernestcarl I try the 200 HZ HP on my DSP before measure. the result is same you get with calculation. Here are measures with diffrent larger unnatural phaseshifts due to EQ. stepresponse change but the spectrogram slice i use stay same. see screenshot. but on all measures can see that the kali speed is much worser and when i do this with headphones, that the headphones are much better. problem is only that arta and i guess the klippel have no waterfall normalize at each frequency function so speed is not easy to compare. Maybe you can do the spectogram with the settings i use 1/1 and Amplitude Linear % peak so can compare more easy your speakers ?

 

[ernestcarl]

"export EQ filtered with min phase version of IR"

Now why would you do that?

if want filter it need a linear phase filter

I just find this whole prescriptive approach of yours rather strange.

Maybe you can do the spectogram with the settings i use 1/1 and Amplitude Linear % peak so can compare more easy your speakers ?

The conditions my monitors were measured in will not really exactly match yours. The KH120’s 'sliced' step at 15cm will also appear different on a measurement taken at 1m — drop in decay looks steeper.

Also, that's just not how I prefer to use the wavelet spectrogram view, quite frankly.

 

[bennybbbx]

Now why would you do that?

to show that it change alot when change the phase. in fact when measure a speaker there is no such a phase change situation as happen when use large phase shift with a HP filter on 2 way speaker

I just find this whole prescriptive approach of yours rather strange.

Wy think you this strange ?.

maybe you can answer the question first on which delay time you can hear diffrent position of the rimshot of my testvideo ?.

when you hear at 0.100 ms or 0.200 ms a large diffrence(as i do) then i think it should be clear for you too, that a mid range speaker that is so slow as current speakers is not optimal. but of course when you can not hear a faster speaker in compare then you did not know how much quality loss you live with.

I notice with my ears and brain that large 2 way speakers sound very worse. I not send back the kali and think because there is nothing about too slow 2 way speakers read that i can correct it with EQ or phase. But then the 30 day moneback was over and i live with "bad" sound, hear often headphone until i compare direct with the 5 inch cantons. kali or a tannoy nfm 8 speaker i have years before sound for me as i hear music thru a door. and not as sit in the room.

I still think speakers not so good as headphones but i notice that the small speakers get much better transients sit in a room feeling as all my larger speakers i have all the years. so i am more happy now in hesr music with speakers. 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. its thrn 2* faster as kali and 50% faster as eris. kali are 50% slower as Eris or MTM

 

[danadam]

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