Step Response: Does It Really Matter?

[AdamG]

“Profoundly weird” is the exact term that springs to mind when I catch a glimpse of my aging bullocks in the mirror after a shower. But I digress . Permanent Forum Ban was issued. This particular Troll shan’t be back.

 

[DonH56]

“Profoundly weird” is the exact term that springs to mind when I catch a glimpse of my aging bullocks in the mirror after a shower. But I digress . Permanent Forum Ban was issued. This particular Troll shan’t be back.

Good for you for ditching the troll, but the reference to your aging bullocks after a shower is an image I simply did not need. TMI. Way, way, way TMI!

 

[j_j]

Good for you for ditching the troll, but the reference to your aging bullocks after a shower is an image I simply did not need. TMI. Way, way, way TMI!

Amen to that!

 

[skyfly]

If I listen to the fundamental (sine functions as bases) of a violin today, and the overtones of the violin tomorrow, I will not fee I heard violin sound.

 

[j_j]

If I listen to the fundamental (sine functions as bases) of a violin today, and the overtones of the violin tomorrow, I will not fee I heard violin sound.

That's quite some allpass filter, but it can be done.

 

[tomtoo]

Is a speaker a time invariant system?

 

[DonH56]

Is a speaker a time invariant system?

No, though it can be treated as such for routine analysis. Most LTI systems are simplified versions of the real world, but the whether the simplifications matter is strongly application-dependent.

 

[tomtoo]

No, though it can be treated as such for routine analysis. Most LTI systems are simplified versions of the real world, but the whether the simplifications matter is strongly application-dependent.

Just like to learn. So iam surly wrong. But if its not time invariant you can not simply use fourier?

 

[DonH56]

Just like to learn. So iam surly wrong. But if its not time invariant you can not simply use fourier?

I am not a speaker designer so prefer to defer to those with more experience. Things like thermal effects (heating), material stiffness (mainly during the first few seconds of operation), the effect of atmospheric pressure changes on the acoustics and suspensions, and so forth vary over time. But during a test run that may be only seconds or minutes you can consider it time-invariant.

The applicability of the FFT depends upon its length and sampling rate, i.e. how long a time is captured, and the extent to which time-invariance matters.

I would not say you are wrong; I am looking at boundary conditions beyond typical use. In general I would expect most of us to treat a speaker as time-invariant for all practical purposes. The main exception I can think of off-hand is power compression tests, where tests repeated back-to-back may show shifts due to voice coil heating. But for a typical test such as Amir and others perform I would expect it to be time-invariant.

Never ask an engineer a yes/no question; there are too many cases on the fringes of the bell curve that we have to be cognizant of but have little applicability in the real world...

 

[redshift]

you have not answer the question. you write about time delay. I mean time delay is the result of too slow speaker speed. Can say time delays(and overshoots amd undershoots) depend on speaker speed and level changes the speaker must do. this i mean. and the result is that a mid bass woofer that reach only 6 khz -3 db(which many big not better do) have problem to reach good enough time for ITD in correlation to the other speaker when the waveform level changes of left and right speaker is diffrent at same time(which happen near allways in stereo signals. confirmed in wave screenshots)

What compounds a bad temporal behavior of loudspeakers is the ringing of sharp antialiasing filters.

I dislike the jarring “sizzle” artifacts from decompression and poorly mixed/mastered recordings shoved through a DAC with brick wall IIR filters.

https://en.m.wikipedia.org/wiki/Ringing_artifacts

Obviously the same thing occurs in a loudspeaker that has a poorly dampened cabinet and drivers.

So, yes, the time domain responses matter. Specially if some ringing is hidden in the “averaging effect” of a FR.

 

[bennybbbx]

I do now new measurements with -8 db level and -18 db level 1 cm away. I use the DSP to limit

speakers from 150 hz to 1.1 khz. there can clear see JBL have faster raise time and decay time as the L6. see on both step response and impulse response. .intresting thing, on the kali, distortion get 3* larger between -8 db and -18 db volume at 399 hz. this much change distortion depend on level can be large problem for ITD precision. because on stereo left and right speakers move allways diffrent. I measure all 1 cm away and do always 2 or more measure to see measure diffrences. there is show for -8 db 92 db in FR, this is only because measure is so near. at 1 meter distance -8 db is around 75 db and -18 db around 65 db. so no stress for speaker.

399 hz 2. nd JBL 104 BT -8 db 0.45% -18 db 0.339% distortion diffrence 0.11%
399 hz Kali LP6 -8 db 0.53% -18 db 0.239% distortion diffrence 0.29%

so kali have near 3* more distortion change in compare to JBL

the group delay also change depend on volume. this show that the signal can not produce exact. kali reach on some frequency over 90 microseconds diffrent group delay between -8 db and -18 db measure.

on 399 hz it is not much more on LP6 but there are other frequency group delay change about 90 us. also the group delay on LP 6 is much more diffrent depend on frequency.

edit: here can see much diffrent distortion when record at -8 or -18 db

I see clear kali is slower. maybe some strong neodyn magnets put on kali woofer can make them faster ?

 

[NTK]

I see clear kali is slower. maybe some strong neodyn magnets put on kali woofer can make them faster ?

The different step response shapes can be easily explained by comparing the frequency responses. Here is an overlay of the 2 FR curves. The more extended FR of the JBL (red curve) at the higher frequencies gives the "faster" looking step. Also, the more extended FR of the Kali (green curve) at the lower end gives a "slower" decaying step.
[Note: A perfect step response does not decay to zero. It stays constant at the step height. A speaker with better low end extension therefore gives "slower" decaying steps.]

 

[bennybbbx]

the question is more wy the LP6 have much level loss at 1 khz. from 180 hz upto 1 khz(the highest peak levels) they look very simular. maybe the strange FR(it look shifted to left) of the kali happen because it is too slow and get on the reduce level over 1 khz much undershoot. I measure with 256 k sweep and repetitions 2. the FR of the JBL is more as the curve that is in the DSP show. i think the changes in FR happen only on 6 db lower than max level this have not so much influence. on both speakers the nearest wall(desk) is more than 30 cm away from speaker.this mean(because the wave need move to desk and back) 60 cm and because i use 1 cm measure distance the level is much much lower. simualar as when measure 1 meter away 60 meter.

FR and phase are both inside step respone. a slow speaker show then also problems in FR

But i try measure in rew limit from 190 to 1 khz. then the FR is very simular of both and can not cause so much diffrence because the frequency over 1 khz are not measure

 

[bennybbbx]

maybe the best solution is a look on group delay at diffrent volumes ?. I have add screenshots of a cheap 9 eur headphone for compare. it is measure too at 1 cm distance. but there can see on LP 6 GD change lots more than on JBL. and on LP6 group delay change much more depend on volume.

 

[Music1969]

https://www.aalto.fi/en/news/the-human-ear-detects-half-a-millisecond-delay-in-sound

 

[bennybbbx]

https://www.aalto.fi/en/news/the-human-ear-detects-half-a-millisecond-delay-in-sound

in this example they use only 1 speaker but intresting. in step repsonse can often see 0.5 ms delay of tweeter and mid/woofer. When have a stereo system the delay between speakers that can hear is much smaller. but how much people this can really hear need verify. this is a new link i find with many books about ITD.

The thresholds for detection of ITDs in pure-tone stimuli by well-practiced human listeners are around 10–20 μs (Klumpp and Eady, 1956; Zwislocki and Feldman, 1956; Brughera et al., 2013). That is remarkable acuity when one considers that it is accomplished with neural machinery having time constants of several hundreds of microseconds or longer. Thresholds are minimal (i.e., sensitivity is greatest) for frequencies from 0.7 to 1.0 kHz, they increase slowly for frequencies < 0.7 kHz, and they increase dramatically above 1.0 kHz such that listeners are essentially insensitive to ITDs at frequencies 1.4 kHz and higher (Brughera et al., 2013).

https://www.sciencedirect.com/topics/neuroscience/interaural-time-difference

 

[Music1969]

in this example they use only 1 speaker but intresting. in step repsonse can often see 0.5 ms delay of tweeter and mid/woofer. When have a stereo system the delay between speakers that can hear is much smaller. but how much people this can really hear need verify. this is a new link i find with many books about ITD. https://www.sciencedirect.com/topics/neuroscience/interaural-time-difference

Nice to see Genelec still co-funding a lot of university research.

They take audio science seriously.

 

[ctrl]

Thanks to @Music1969 for the link.

in this example they use only 1 speaker but intresting. in step repsonse can often see 0.5 ms delay of tweeter and mid/woofer.

Please read through at least the most basic background. You are misleading newcomers with statements like this.

The group delay (this is what the study is about) is frequency dependent, it cannot be determined by the time distance of the peaks in the step response!

The study confirms once again that group delay is imperceptible in 2-way loudspeakers and many 3-way loudspeakers, if reasonably designed (especially active concepts have advantages), at least in the 500-4000Hz frequency range studied - even with artificial signals.

For example, here is the group delay for a typical (real) 2-way speaker with a fourth order crossover:

Group delay >320Hz is <0.35ms, even below or at the audibility threshold for artificial signals. With real music signals, the threshold is clearly undercut.

 

[j_j]

We need to be careful here. ITD detection can be surprisingly short (under one sample).

Variable time delay across one ear is detectable SOMETIMES, but only sometimes.

 

[bennybbbx]

Thanks to @Music1969 for the link.


Please read through at least the most basic background. You are misleading newcomers with statements like this.

The group delay (this is what the study is about) is frequency dependent, it cannot be determined by the time distance of the peaks in the step response!

The study confirms once again that group delay is imperceptible in 2-way loudspeakers and many 3-way loudspeakers, if reasonably designed (especially active concepts have advantages), at least in the 500-4000Hz frequency range studied - even with artificial signals.

View attachment 145729

For example, here is the group delay for a typical (real) 2-way speaker with a fourth order crossover:
View attachment 145730

Group delay >320Hz is <0.35ms, even below or at the audibility threshold for artificial signals. With real music signals, the threshold is clearly undercut.

View attachment 145731

I have done much tests with ribbon tweeter at addon on my speakers and use DSP crossover. i notice step response show best if the tweeter and the mid/bass have fewest delay together. when in stereo change the delay of both tweeter is set to 0.1 ms it have lots diffrent sound. with mono sound i hear not so much diffrence. magnitude and phase are in correlation, and influence Magnitude at a specific frequency too. both have together influence of a sound. there can not say when the result of a step response look faster it is only because FR diffrence.

because music is a alternating voltage and the Magnitude depend on a average between peak and valleys(effectiv value). if a speaker have delay or faster rise but slower fall time this change Magnitude in compare to other frequency. for example the effective value is diffrent if you play a sawtooth a sine or rectangle wave. and with slow mid/bass woofers happen same at much lower frequency as with human ears. human ears can even if use 192 khz sample rate not hear a diffrence between triangle, sine, sawtooth waves at frequency above 8 khz. because ears are unprecise to hear the waveform diffrence(this are also see as diffrent harmonics) at this high frequencies. sure the higher harmonics are play from tweeter but a slow mid/woofer do lots mistakes on produce the lower frequency and create wrong or large delayed other harmonics.

on 800 hz ears can good hear a diffrence between triangle sine and rectangle waves even if FR and magnitude is same

problem in measure chart is, that when play a 800 hz tone only the distortion is show. this does not show where the distortion is. maybe need for find a good mid woofer a rta with a fixed frequency and see how the harmonics are. simular i do with the bass at 50 hz here https://www.audiosciencereview.com/...ubwoofer-comparison.20494/page-13#post-842067 but with frequency of 1 khz