Does step/impulse response matter?

[bachatero]

I'm wondering if there is an audible effect of the step/impulse response of a speaker based on what kind of damping it has and how long the response lasts. I would assume the perfect impulse response is the dirac function (H(0) = 1, H(t) = 0 otherwise) and the step response is a perfect stair step, but no speaker I've seen comes even close to that. It looks like most speakers have an underdamped response with a couple specific ones being critically damped and none overdamped. Why is this?

 

[Another Bob]

To get a perfect step response would require bass extension to DC. Short of that, you end up with sort of a right triangle shape. The best step response I have seen is the Kii Three. (Stereophile's measurements here: https://www.stereophile.com/content/kii-audio-three-loudspeaker-measurements). Speakers are generally not designed with overdamped bass because that results in a drooping low end response, and frequency response is the most audible parameter, i.e., the relative lack of bass will be more objectionable (to most) than the increased ringing of a critically damped (maximally flat) alignment.

 

[bachatero]

From that article, it looks like the "perfect" step response comes with an absurdly high latency, while the underdamped one has low latency. Is this the reason why other speakers like studio monitors with low latency always look underdamped?

 

[Another Bob]

The latency is a result of the signal processing necessary to make the time domain response so nearly perfect.

 

[sam_adams]

Tweeter from the RS Minimus 77:

Morel MDT30s:

Morel MDT39:

The Morels are ferro-fluid damped and the RS is not.

 

[Palladium]

What frequency is the step response tested at?

 

[Beave]

What frequency is the step response tested at?

What do you mean by this?

 

[MaxwellsEq]

What frequency is the step response tested at?

All frequencies!

 

[Keith_W]

All frequencies!

He is referring to @sam_adams post showing the step response of the tweeters. So those are tweeter frequencies.

Anyway, to answer the OP's question: does step response matter? Depends on whom you ask. If you ask Amir, he does not see much value in it. He has said on a few occasions that he only posts step response for the ASR people who want it.

If you ask anybody who DSP's, then yes it matters. It is very good for spotting pre and post ringing. The second peak should not be higher than the first peak, and the rise time of the step response should be as short as possible. The length of the step response is a good indicator of post-ringing.

 

[Multicore]

Anyway, to answer the OP's question: does step response matter? Depends on whom you ask. If you ask Amir, he does not see much value in it. He has said on a few occasions that he only posts step response for the ASR people who want it.

Since step response and frequency response are tautological, they must surely both matter equally. My understanding is that Amir prefers frequency response to present his formal standardized measurements. I completely agree with him. Otoh, I think questions like the OPs are valuable in teaching us about or reminding us to bear in mind the relationships between time and frequency domain response. The well-known mechanical engineering concept of damping is relevant in loudspeaker design and made an appearance in this thread thanks to step response, usefully related to bass performance by @Another Bob .

 

[sktn77a]

Is this the reason why other speakers like studio monitors with low latency always look underdamped?

Yes, one is the corollary of the other.

 

[Another Bob]

Yes, one is the corollary of the other.

I beg to differ. Although in the case of the Kii, the processing is both causing the latency and modifying the low end response/damping, one might have processing in the system that creates latency with no effect on damping whatsoever. Conversely, one could build a highly damped, passive speaker that would have essentially no latency. Here are the Stereophile measurements of the Thiel CS2.4 that shows a pretty good step response: https://www.stereophile.com/content/thiel-cs24-loudspeaker-measurements Since it is a passive design, there is no latency.

It should be noted that there are multiple causes of wiggles in the step response. The wild swings one sees near the leading edge are usually due primarily to the lack of time alignment between the drivers. This is what you see in the Kii Three "Minimum" setting. There are plenty of debates over whether time alignment is audible. Lack of driver damping generally appears later, and in fact woofer ringing will be most visible beyond the narrow time window provided in the Stereophile measurements.

 

[sam_adams]

Some woofers to compare against the tweeters. Measurements done on a baffle unless otherwise noted.

Silver Flute W17RC38:

GRS 6PR-8:

Heppner 4846-SD:

Faital Pro 5FE120-8:

Faital Pro 5FE120-8 (Same driver as the above but in an enclosure with lots of poly-ether foam damping.):

 

[Cbdb2]

I would assume the perfect impulse response is the dirac function (H(0) = 1, H(t) = 0 otherwise)

Actualy H(0)=infinity. Its the integral (area, or H(0)xH(t)) thats zero. So like a square wave its a physicaly impossible signal to create. So like all approximations you have to use it properly.

 

[Cbdb2]

Some woofers to compare against the tweeters. Measurements done on a baffle unless otherwise noted.

Silver Flute W17RC38:

View attachment 383507

GRS 6PR-8:

View attachment 383508

Heppner 4846-SD:

View attachment 383509

Faital Pro 5FE120-8:

View attachment 383512

Faital Pro 5FE120-8 (Same driver as the above but in an enclosure with lots of poly-ether foam damping.):

View attachment 383514

Are those raw drivers, no xover/eq? A LPF on a woofer can make a huge difference on attack time, and those graphs,

 

[ZolaIII]

It's time and space across the spectrum. You want alignment to start with (perfect placement dosent exit), you don't want pre ringing and you don't want it to ring over second peak as much as possible and in the declining manner. If it post gings much it will be masking and preringing is even worse. If you have very good timing you can add (simulate) space. Where you can't have that you pick a lesser evil and try to respect what it can't do (very high SPL PA huge sub's...) and in generally compromise. Al do difference is there (audible) you will be well trained in well known space you are accommodated to neglect it. Clever use of crossovers has a same function to dump what could bleed to mids and keep the bass rather flexible (as it needs to be like that for ELC on more serious compensation levels with a lot of low and sub bass boost ~-30 dB [of 88 dB SPL stereo cal point] to rather work as psy we hear).

 

[sam_adams]

Are those raw drivers, no xover/eq? A LPF on a woofer can make a huge difference on attack time, and those graphs,

No XO or EQ. Raw drivers on an baffle. 2.83 V at 8 Ω drive level. Examine the plots carefully to see the 'attack' time is on the order of a few µs. Why would one use a LPF on a driver if one was attempting to characterize the native response of the driver?

 

[test1223]

Since about 90% don't deeply understand what the step response is it will never be a good discussion. So if you aren't 100% sure to know the fundermentals please read the following:

The step or impulse response is "only" another representation of the frequency response. The fundermental data is the same (with the assumption of a linear time-invariant system which is a good assumption to start with).

The term "frequency response" should be used for a plot which contains the "amplitude frequency response" and the "phase response". Be aware that "frequency response" is often incorrectly use for the plot of the "amplitude frequency response" which has no phase information.

With a time frequency transformation and its inverse transformation like the fourier transform and the inverse fourier transform you can get from the impulse response to the frequency response and back.

Since all humans are not that good in using a fourier transformation in their head it is hard to correctly interpret a step response or impulse response. Since it is the overlayed response of all frequencies. It is very hard to do a fair comparison since the frequency content is "hidden" in the data and can hardly be seen. So a deeper bass response or a wiggle in the mid frequencies will both alter the step response, but it is very hard to see the cause of the alternation, maybe it is a out of useful bandwidth behavior or it is a very audible resonances, you will have a lot of trouble to find the cause.

Since all humans are not good at using a inverse fourier transform in their head the frequency response "hides" most time related distortion of the loudspeaker. E.g. it is hard to see the pre- and post-ringing properties of a frequency band. You can see some problems when you know how to read the phase response but it is not intuitive.

So there is another representation of the same data which is called watterfall plot. Where you got band limited impulse responses over all frequencies. This plot is ideal to start the investigation of the timing behavior of a loudspeaker. Here you can easily see if the woofer has a time gap compared with the tweeter or at which frequencies long post ringing is comming from.

REW room eq wizard and many other programs offer such or similar plots. But be aware that all measurements especially in the bass frequencies are distorted by the room acoustics. So it isn't easy to correctly compare the timing of deep bass with the timing of all other frequencies. Since you typically need at leas two measurements one very close to the woofer and one at a meaningful distance which is related to your listening position.

 

[olieb]

Since all humans are not that good in using a fourier transformation in their head it is hard to correctly interpret a step response or impulse response. Since it is the overlayed response of all frequencies. It is very hard to do a fair comparison since the frequency content is "hidden" in the data and can hardly be seen. So a deeper bass response or a wiggle in the mid frequencies will both alter the step response, but it is very hard to see the cause of the alternation, maybe it is a out of useful bandwidth behavior or it is a very audible resonances, you will have a lot of trouble to find the cause.

Quite so.
The biggest problem here is that in the step response effects from magnitude and phase are combined and very small phase rotations can have a huge visual impact on the step response.
Here is a comparison for simulated (perfect) speakers (Calculated with REW) FR and step response

Above: lowpass with -6dB at 25Hz
Below: the same lowpass and additionally some very gentle phase rotation from an all pass filter
(< 60 degree/oct ... group delay is way below what is considered audible)

It looks like a big difference but it isn't.

EDIT: For comparison here is an example of the effect of magnitude variation on the step response.
The step response does not look that bad, but have a look at the magnitude in lower mid!
This is absolutely audible and not a small effect at all!

 

[Keith_W]

REW room eq wizard and many other programs offer such or similar plots. But be aware that all measurements especially in the bass frequencies are distorted by the room acoustics.

Excellent post. FWIW, I look at the step response the other way round - the step response is a good indicator of how well a speaker/subwoofer combination has been dialled in to the room. It is less relevant for a speaker or subwoofer under anechoic conditions.