Dr. Klaus Heinz of HEDD Audio (ex ADAM Audio) - measuring speakers, in particular speaker dynamics

[thewas]

can you please post step response of KEF LS50 ?

Here you go:

Source: https://www.stereophile.com/content/kef-ls50-anniversary-model-loudspeaker-measurements

PS: Stereophile always shows the step response of all loudspeakers they measure.

 

[ernestcarl]

Here you go:

View attachment 102656
Source: https://www.stereophile.com/content/kef-ls50-anniversary-model-loudspeaker-measurements

PS: Stereophile always shows the step response of all loudspeakers they measure.

Thank you! I myself have been too lazy to look.

This isn't exactly comparable but here's the step of another coax (sceptre s8):

The 0 deg on-axis step is pretty much near identical to the vector avg of all step responses at all angles.

Assuming the two graphs are even comparable, I'd say the Sceptres looks somewhat "faster" in this area. What would be interesting to me personally is to have a side-by-side listening comparison of specific tracks that have a ton of fast HF/MF transients. Would one speaker sound "faster" or "brighter" (my own experience) -- in just the transients, mind you-- if both were equalized to a similar overall neutral curve(s). My hunch (just nothing more than a hunch!) is probably yes.

 

[thewas]

In my experience step like impulse response are just an inconvenient method to show the complex frequency response and usually lead to questionable optical shape based interpretations which usually don't hold blind testing and the theory behind.

Personally I prefer other representations like the group delay to which known psychoacoustic audiblity limits have been studied.

 

[ernestcarl]

In my experience step like impulse response are just an inconvenient method to show the complex frequency response and usually lead to questionable optical shape based interpretations which usually don't hold blind testing and the theory behind.

Personally I prefer other representations like the group delay to which known psychoacoustic audiblity limits have been studied.

I do agree looking at the step at it's own is simply not enough. But I would expect the group delay to show a similar gradual change in delay in the graphs -- why would it not? Blind listening is very much desirable too, but there are speaker tonal differences that may be just too obvious for one not to notice which speaker is playing. I'm speaking here specifically of when I was ABing between the KH120 and S8. It was so patently obvious which speaker was which just through the overall tonal differences between the two. Both speakers sounded neutral, but still very, very different.

Maybe if I could somehow change the "speed" or precise timing of the same speaker through DSP seamlessly, and AB against the same track on the exact same monitor... I do know Dave Gunness has performed something of the kind where temporal (time-domain) equalization was switched on and off in the same speaker (under development) where the immediate apparent change observed was an overall increase in "brightness". Now this is presuming his DSP didn't increase the magnitude of the response itself. And, of course, personal anecdotes aren't proof -- but theoretically, it does sound plausible to me.

 

[bennybbbx]

In my experience step like impulse response are just an inconvenient method to show the complex frequency response and usually lead to questionable optical shape based interpretations which usually don't hold blind testing and the theory behind.

Personally I prefer other representations like the group delay to which known psychoacoustic audiblity limits have been studied.

I see only this large diffrence on the LP6 and my other speakers that sound much better in stereo. here are my REW mdat files of the speakers. maybe you can tell if group delay is better and what bad thing in group delay bring the worse analytik and stereo width of LP6 ?

https://file.io/wMs4ifUXTuRg
https://file.io/6CyfuezuzUIF

 

[bennybbbx]

I do agree looking at the step at it's own is simply not enough. But I would expect the group delay to show a similar gradual change in delay in the graphs -- why would it not? Blind listening is very much desirable too, but there are speaker tonal differences that may be just too obvious for one not to notice which speaker is playing. I'm speaking here specifically of when I was ABing between the KH120 and S8. It was so patently obvious which speaker was which just through the overall tonal differences between the two. Both speakers sounded neutral, but still very, very different.

Maybe if I could somehow change the "speed" or precise timing of the same speaker through DSP seamlessly, and AB against the same track on the exact same monitor... I do know Dave Gunness has performed something of the kind where temporal (time-domain) equalization was switched on and off in the same speaker (under development) where the immediate apparent change observed was an overall increase in "brightness". Now this is presuming his DSP didn't increase the magnitude of the response itself. And, of course, personal anecdotes aren't proof -- but theoretically, it does sound plausible to me.

when you record them maybe with a stereo microphone or a mono microphone first left rthen right channel you can test me, if i hear stereo width diffrence. now your step response is better to see. the S8 have a real good response it need 1 ms to reach 0. when look at stereophile there are most 2 way speakers that need 1.5-2 ms to reach 0 and they cost thousands of $. this is simular to kali and worse for me. but 1 ms i think is good.

 

[mitchco]

In my experience step like impulse response are just an inconvenient method to show the complex frequency response and usually lead to questionable optical shape based interpretations which usually don't hold blind testing and the theory behind.

Personally I prefer other representations like the group delay to which known psychoacoustic audiblity limits have been studied.

I agree that group delay (and phase for that matter down low) is worth looking at, but I respectfully disagree about the step response leading to questionable interpretations. If the speaker system does not follow the the "ideal" step response, we can certainly draw conclusions. It is helpful to know who to read a step response in the first place:

There is a concept of preringing with linear phase filters, so we want to watch for this type of distortion, even though in listening tests I have conducted, large amounts are hard to audibly detect. I have some charts of preringing here on ASR. The tell-tale sign is a ramp up or oscillation of the signal before the actual signal. Most noticeable on sparse music transients, like a kick drum for example, where it sounds “reversed” in the extreme case. Most modern DSP correction software have preringing compensation as it is well understood mathematically, so this is no longer an issue.

See the vertical step itself starting at time 0 milliseconds? One can think of the vertical amplitude as the frequency scale with 20 Hz starting at the bottom and 20 kHz at the top, as this is what we specified in our ideal loudspeaker design. If I designed for flat to 30 kHz, then the vertical spike would be higher. If the drivers were not time aligned, then we would see horizontal offsets away from 0 ms of the straight vertical line representing parts of the frequency spectrum arriving at our ears at different times, and different between channels too. This is very important to keep in mind, the point being our ideal loudspeaker has all direct sound frequencies arriving at the same time for both channels, i.e. at 0 ms.

The slope of the roll off, or shape of the tail, after the initial vertical step and to where it crosses the 0 ms horizontal time axis, is based on the loudspeakers low frequency roll off and cabinet alignment (i.e. slope of roll off). A roll off at a higher frequency would push the “cross the 0 time” threshold towards the left and a lower than 10 Hz roll off will push the 0 crossing point to the right, say at 15ms or even 20ms, depending on loudspeaker design (e.g. subs or no subs) and size of room.

Finally, what I don't show on this chart is beyond 20 ms are the low frequency room reflections. This goes a bit beyond the ideal loudspeaker and mainly a concern with room correction. But excessive reflections or room build up or maximum phase peaks that are greater than the direct sound are easily detected with a step response chart.

So knowing how to interpret a step response can tell us many things, is there preringing, is there time domain distortion between the drivers, is there too much/too little high frequencies based on how well the speaker tracks to the target, is there lagging bass response or too much or too little bass looking at where the bass crosses the the horizontal axis. Here is an example of a speaker tracking just about perfectly to the target step response:

The ideal step is in black and the measures are in blue and red. From my article on accurate sound. Note that the step response is but one of many views of the transfer function we want to look at. But it is useful as it can tell use many things and to my ears, certainly audible.

Happy New Year!

 

[thewas]

I know that all these can be seen in the step response, but still I don't see much extra value compared to usual loudspeaker measurements for loudspeaker analysis but rather its application for special causes like digital room correction.

So knowing how to interpret a step response can tell us many things, is there preringing

As you know preringing is such a special case which can happen only when we use of FIR filters to linearise their phase, not on usual "causal" loudspeaker designs.

is there time domain distortion between the drivers

Which can still be better seen and interpreted from the group delay to which also psychoacoustic research limits exists, while the step response of a classic passive or IIR active 3-way loudspeaker may look "wild" and still be inaudible.

is there too much/too little high frequencies based on how well the speaker tracks to the target, is there lagging bass response or too much or too little bass looking at where the bass crosses the the horizontal axis.

Which can be better seen from the amplitude frequency responses.

 

[thewas]

I see only this large diffrence on the LP6 and my other speakers that sound much better in stereo. here are my REW mdat files of the speakers. maybe you can tell if group delay is better and what bad thing in group delay bring the worse analytik and stereo width of LP6 ?

https://file.io/wMs4ifUXTuRg
https://file.io/6CyfuezuzUIF

To my limited knowledge I haven't seen research on such a correlation. The problem is that things like imaging perception are very complex and not fully understood topics and depend rather on directivity, diffraction and interaction with room acoustics. It is problematic comparing different loudspeakers which differ on so many parameters and implying a correlation just on one of them. If you want to experiment in that direction you could try using only one loudspeaker and changing its time response by filtering or other modifications to keep all other parameters the same.
I have participated in few comparisons of loudspeakers with a classic minimal phase vs linearised phase and the differences are by far not as big as your describe them but rather very small and audible with only some recordings.

 

[bennybbbx]

To my limited knowledge I haven't seen research on such a correlation. The problem is that things like imaging perception are very complex and not fully understood topics and depend rather on directivity, diffraction and interaction with room acoustics. It is problematic comparing different loudspeakers which differ on so many parameters and implying a correlation just on one of them. If you want to experiment in that direction you could try using only one loudspeaker and changing its time response by filtering or other modifications to keep all other parameters the same.
I have participated in few comparisons of loudspeakers with a classic minimal phase vs linearised phase and the differences are by far not as big as your describe them but rather very small and audible with only some recordings.

that there is a correlation on this need of course confirm with more tests. currently it is only a early theory that need confirm with more speakers than the 4 i have. so it is good when more people record their speakers as i do seperate left and right. then upload the results and when i ALWAYS say correct which speaker is wider and have better step response at least 30% better to reach 0 , then this theory is confirm.


also hear my upload new examples in this thread.I have now done no EQ correction. now can hear more the speakers with slow step response sound muddy and dull https://www.audiosciencereview.com/...-synth-sound-in-the-attached-mp3.18401/page-2

 

[thewas]

that there is a correlation on this need of course confirm with more tests.

Not more put purposeful tests. As said you can't compare 2 things which differ on several parameters and then make out of it a correlation for just one, you need to keep all other parameters constant.

For example:

now can hear more the speakers with slow step response sound muddy and dull

What you mean with "slow step response" can just mean just deeper bass which of course activates more of your room modes and can sound muddy and dull, but this has nothing to with the step response (which is just a result) but of the a frequency amplitude response. If you want to compare different step responses at least the frequency amplitude response should be the identical.

 

[KSTR]

If you want to compare different step responses at least the frequency amplitude response should be the identical.

... which boils down to displaying only the step response of the excess phase (which has unity amplitude through out). Same goes for group-delay or phase plots, better plot excess group delay if we are concerned about colorations on top of the natural (minimum phase) frequency response. For the latter, obviously a 6th-order active ported design with the same f3, say 50Hz, than a passive sealed one will have "lagging bass" and different tonality. With an f3 at 20Hz, much less so.
Whereas, IME, intrucing a LR4 (acoustic) crossover at 80Hz to integrate a sub spoils perceived "quickness" of bass and not all of that can be attributed to increased and deeper exitation of room modes which of course often will be the dominant factor. Only way to tell is to correct the Mains-Sub XO to linear phase and compare with the standard excess phase version.

 

[thewas]

... which boils down to displaying only the step response of the excess phase (which has unity amplitude through out). Same goes for group-delay or phase plots, better plot excess group delay if we are concerned about colorations on top of the natural (minimum phase) frequency response. For the latter, obviously a 6th-order active ported design with the same f3, say 50Hz, than a passive sealed one will have "lagging bass" and different tonality. With an f3 at 20Hz, much less so.
Whereas, IME, intrucing a LR4 (acoustic) crossover at 80Hz to integrate a sub spoils perceived "quickness" of bass and not all of that can be attributed to increased and deeper exitation of room modes which of course often will be the dominant factor. Only way to tell is to correct the Mains-Sub XO to linear phase and compare with the standard excess phase version.

Thank you, couldn't been written better!

 

[KSTR]

As for the thread topic, anyone have tried subtracting a convolved (with a measured IR) music signal from its direct recording, which exposes everything that isn't covered by the linear equivalent of the true transfer function?
You'll be shocked... of course, in this brute-force experiment, even the slightest amount of power compression will quickly dominate the residual. Now even when the IR was measured at very low levels and thus hopefully linear and time-invariant conditions, I feel there is a lot more dirt than the simple static distortion products alone could explain...

 

[bennybbbx]

Not more put purposeful tests. As said you can't compare 2 things which differ on several parameters and then make out of it a correlation for just one, you need to keep all other parameters constant.

For example:

What you mean with "slow step response" can just mean just deeper bass which of course activates more of your room modes and can sound muddy and dull, but this has nothing to with the step response (which is just a result) but of the a frequency amplitude response. If you want to compare different step responses at least the frequency amplitude response should be the identical.

the MTM have deeper bass as the LP6. and in this test there is no bass play. you hear only guitar and drums. but if you think its a bass problem i have do step response from 300 hz upto 3 khz. the results are the same. lp6 is lots slower. see setting for measure. it play only sine wave from 300 hz to 3 khz.

I

f you want to compare different step responses at least the frequency amplitude response should be the identical.

but with lots changing signals in mid range the speaker with slow step response is too slow to follow the peaks and smooth it. so it reduce some frequences. thats maybe the reason because the slow transient speaker that is measure with room correct software and correct stay in sound muddy and dull. EDIT: the more curvy in canton is because it stand more near to desktop as the Kali or MTM. but can see it is faster as Kali and sound wider

 

[j_j]

I don't see any difference between step response and impulse response. Both tell us exactly the same thing.

 

[ernestcarl]

... which boils down to displaying only the step response of the excess phase (which has unity amplitude through out). Same goes for group-delay or phase plots, better plot excess group delay if we are concerned about colorations on top of the natural (minimum phase) frequency response. For the latter, obviously a 6th-order active ported design with the same f3, say 50Hz, than a passive sealed one will have "lagging bass" and different tonality. With an f3 at 20Hz, much less so.
Whereas, IME, intrucing a LR4 (acoustic) crossover at 80Hz to integrate a sub spoils perceived "quickness" of bass and not all of that can be attributed to increased and deeper exitation of room modes which of course often will be the dominant factor. Only way to tell is to correct the Mains-Sub XO to linear phase and compare with the standard excess phase version.

I don't know how to interpret these graphs enough yet, but from what can be quickly gleaned below do you see any indication as to why I am able to hear dense, fast transients much more clearer (immediate colouration heard is "brightness") on the Sceptre S8 than on the KH120?

Difference between the single on-axis extreme nearfield swept sine measurement of the S8 and the actual MMM response at 1m distance can be explained easily by its peculiar family of directivity curves.

Group Delay

Magnitude & Phase Response

Step

I don't quite know how to display "only the step response of the excess phase" in REW -- or at least there doesn't seem to be any obvious option there.

 

[bennybbbx]

I don't know how to interpret these graphs enough yet, but from what can be quickly gleaned below do you see any indication as to why I am able to hear dense, fast transients much more clearer (immediate colouration heard is "brightness") on the Sceptre S8 than on the KH120?



Difference between the single on-axis extreme nearfield swept sine measurement of the S8 and the actual MMM response at 1m distance can be explained easily by its peculiar family of directivity curves.

Group Delay


Magnitude & Phase Response


Step
View attachment 102874

I don't quite know how to display "only the step response of the excess phase" in REW -- or at least there doesn't seem to be any obvious option there.

in REW you need check box on below that is name step response and disable window.step response and impulse response show are same but step response can more easy read. on light gray can see your step response. the S8 reach faster maximum as the KH120. The S8 reach faster 0% S8 1 ms KH120 1,5 ms

 

[ctrl]

If the speaker system does not follow the the "ideal" step response, we can certainly draw conclusions. It is helpful to know who to read a step response in the first place:....
One can think of the vertical amplitude as the frequency scale with 20 Hz starting at the bottom and 20 kHz at the top, as this is what we specified in our ideal loudspeaker design. If I designed for flat to 30 kHz, then the vertical spike would be higher.

There is a small error in the drawing and in the text.

The slope of the step response is a measure for the upper frequency limit, not the amplitude - AFAIK. The rise time that elapses until the tweeter peak, for example, is a measure of the frequency response limit of the tweeter. Your example shows a step response with an infinite frequency limit.
infinite upper FR: FR with BU2@20kHz: FR with BU2@10kHz:

 

[ernestcarl]

in REW you need check box on below that is name step response and disable window.step response and impulse response show are same but step response can more easy read. on light gray can see your step response. the S8 reach faster maximum as the KH120. The S8 reach faster 0% S8 1 ms KH120 1,5 ms

Well, I know how to disable or enable those traces. As @KSTR suggested, I don't see an obvious visible option to show the "step response of the excess phase" within REW. Maybe in other software this can be done?