Importance of impulse response

[gnarly]

Nice.

So if the impulse response is perfect, you get text book frequency response?

Yes, you will.

But a little deeper digging is worthwhile...
Different usages of term "frequency response" are the source of great confusion, imo.

The common usage of frequency response often does not include the phase response.
The correct technical definition of frequency includes phase response

When people speck of frequency response and transfer functions, etc; they are almost always using the correct definition , and call common usage frequency response magnitude. Phase thank goodness, is just phase.
So freq response = both mag and phase


Here's why digging deeper matters....

If you have a perfect impulse, then you will have both perfect mag and perfect phase curves.
If you have both perfect mag and phase curves, you will get a perfect impulse.

You can have a perfect mag curve (again, the common definition of frequency response), but that does not equal a perfect impulse.
Phase must be perfect too, to have a perfect impulse.
Flat magnitude 20-20k can have almost any impulse...even ones really whacked out.....

And what other benefits appart from receiving the sound from different frequencies at the same time in your hear?

Let's take flat magnitude as a given for a well designed speaker.
Then what difference does a perfect impulse, or rather flat phase, bring to the sound?
To my ears, cleaner stronger transients are easiest to hear.
I'd say improvements to timbre, rhythm, and overall clarity/definition also occur.

But that's just me.....ymmv.

 

[fineMen]

,,, speed of sound ... the medium ... the wave ...

And before somewhere else something about cabinet resonances and distortion and all. We won't get into a three blind mice situation here.

Let's assume we have a so called linear transmission line. (NOT the bass alignment, guys). It transfers a signal. Add to the signal, the sum comes out: A signal doesn't compromise B signal.

The signal may be what some call an impulse, that needle thing.

Mathematically it is relatively easily described as a combination of an infinite (!!) number of frequencies, layed over each other, summing the amplitudes up. But the phases of theses frequencies are just chosen so, that the resultant is that needle at some point in time and nil everywhen else. It is possible, believe me; You've got no chance other than to acknowledge that for the moment. Hint: take away some range of frequencies or alter the phase of a few, and You end up with all that wiggles.

That is what happens in a not so perfect transmission line. The phases are altered, the amplitudes are altered, and so the perfect 'impulse' as a signal isn't recognizable anymore.

Wait--we are talking 'signal'. What the audiophile wants is information! Again, believe me, the information is all there. The human ear is not, as some say, lacking something. To the contrary, the hearing is perfectly capable of detecting the full (!!) and all information even from a transmission line that has disturbances that, to the eye look terrible.

Evolution has proven the benefit of neglecting nearly all defects the signal may undergo. The only relevant information is the frequency content, the tone. The phase is by a long stretch irrelevant. Phase only comes into play when the signal at the two ears is different in this respect. But since the speakers are (nearly) identical left-right, no prob.

That is why an engineer doesn't care about phase, and consequently the 'impulse' with the end product. She only displays the frequency response, for a very good and quite honest (!) reason. The marketing, though, isn't willing to give up cheating. Some audiophools follow the fairytails of that provenience because ... tja, what do You think?

I don't want to bother anybody, but why is it, that basic mathematics, or communication theory isn't understood in consumer audio? Or in the general public to begin with ... I personally don't think it's anyway hard to grasp ...

 

[617]

The entire audiophile knowledge of 'phase' as an issue in loudspeakers comes, in my opinion, from a set of myths created by speaker manufacturers who were seeing phase behavior in their measurements with the advent of inexpensive IR based (as opposed to steady state) measurement systems.

They speculated, wrongly, that it was 'phase' that explained why speakers with similar frequency responses sounded different, and hypothesized about why this is the case in various ridiculous ways.

As it turns out, the relative phase of a speaker driver array is immensely impactful of sound quality, because drivers playing out of phase will create cancellations. Shallow crossovers create broad cancellation, higher orders create narrow cancellations.

What they were really seeing is directivity - where the speaker is radiating energy and where there are suckouts. But they attributed it to 'time domain' behavior. In reality, phase shift has to be really extreme and / or chaotic to be perceived; but a dipole speaker (with huge cancellation to the side) will always sound different from a monopole (which may not have many nulls at all.)

The speakers that reflected these design priorities look like anachronisms now - Vandersteen, DCM, Dick Sequerra, who am I missing? Thiel probably? Dunlavy? Their designs haven't held up because they generally have bad directivity behavior.

Now that measuring directivity is 'easy', loudspeaker manufacturers can (over) sell that.

So, long story short, blame MLSSA and CLIO.

 

[kimmosto]

The most lame, undynamic, scattered and disappointing speaker I have heard has almost perfect directivity and spinorama, but terrible timing; phase, step, ETC and excess group delay. Good directivity and good timing do not exclude each others so no need to underestimate manufacturers who target also good timing.

 

[Killingbeans]

I don't want to bother anybody, but why is it, that basic mathematics, or communication theory isn't understood in consumer audio? Or in the general public to begin with ... I personally don't think it's anyway hard to grasp ...

The vast majority of the human population do find it hard to grasp. You can't fight the normal distribution.

Besides, math, physics and chemistry aren't sexy. In a world driven by 'eat, sleep, f##k', thinking logically doesn't get top priority.

I remember how most of my classmates at school saw physics and chemistry as a form of torture they had to endure.

One of the guys stole a big chunk of sodium and blew up one of the school toilets. I think "sodium + water = boom!" was the only piece of information that was retained in any of their brains.

 

[617]

The relationship between time and frequency is sort of interesting in that it's totally obvious but also not intuitive.

 

[617]

The most lame, undynamic, scattered and disappointing speaker I have heard has almost perfect directivity and spinorama, but terrible timing; phase, step, ETC and excess group delay. Good directivity and good timing do not exclude each others so no need to underestimate manufacturers who target also good timing.

What speaker was that Kimmo?

 

[NTK]

What speaker was that Kimmo?

The Fujitsu TenTD712z? Measurement from Stereophile of the Mk 1. Ouch.

Fujitsu Ten Eclipse TD712z loudspeaker Measurements

Sidebar 3: Measurements The Eclipse TD712z's voltage sensitivity was to specification but significantly lower than average, at an estimated 83.5dB(B)/2.83V/m. However, its impedance remains above 6 ohms at almost all frequencies (fig.1), and rises in value above the midrange, this due, I assume...

www.stereophile.com

White paper from manufacturer for the Mk 2.

https://www.denso-ten.com/business/technicaljournal/pdf/33-2.pdf

[Edit] Looks like there are some newer models.

Japanese Speaker Manufacturer Eclipse Launches Its First New Model In Seven Years

The new Eclipse TD307MK3 is a compact, full-range speaker with some advanced technologies and is a direct replacement for the company’s much-loved TD307MK2.

www.forbes.com

 

[ppataki]

I have seen a lot of people here say that impulse response is not relevant, and that basically only frequency response matters.
This confused me because I happen to use Eclipse single driver speakers, which I like a lot and I indeed think they sound really transparent like they claim, and their whole philosophy is basically designing for good impulse response. Is it just BS?
Also I always thought that it makes sense that a smaller driver would be faster and thus sound faster, due to it being lighter.
Some people say that if it can reproduce 20Khz its fast enough which also makes sense. Is that actually true? and if it is, is fastness really FR related too, because it really feels like a different "sensation" if that makes any sense.

The gents before me already shared some really useful information; I just would like to share some personal experience

The speaker in the above picture is a single-driver one-way speaker - those have a characteristic that will have an impact on the impulse response (IR) that you measure in your room when you measure such a speaker: they tend to beam, meaning that they will have a very small sweet spot, especially in the higher frequencies.
Due to this behavior they will also cause less reflections in the room - and that is very much visible in the IR curve

Let me show you the IR curve of a multi-way speaker (Nubert nuVero 140) in my living room (with DSP correction):

The spikes I marked with the highlighter are actually room reflections - you can even measure their distance on the IR curve so you would know in your room what caused those reflections.

Now let's see the IR of another system with a pair of 12" full range drivers in the same room and the same position (also DSP corrected)

No spikes! (or at least they are well under 10%)

So if you measure that one-way Eclipse speaker in a room and then measure another multi-way speaker in that same room you will likely have similar differences in the IR curve that I demonstrated above

 

[gnarly]

A point that should be made when talking about a "perfect" impulse response, is that it's in reference to the audible spectrum.
The best it can look using the 20Hz-20kHz spectrum.
Infinite bandwidth is neither needed, nor relevant. Heck, if folks want to make infinite terms relevant, then there's no such thing as a friggin sine wave either.


You know, there's only two domains in audio...the frequency domain and the time domain.
We know a great deal about how the ear works in the frequency domain. Most all hearing research seems to be how the ear responds in the frequency domain.
Not much to be found about how the ear works in the time domain.

I don't think we really know how to assess audibility of the time domain yet, particularly keeping audibility results in the time domain and not how they effect the frequency domain.
I also don't think we've had the tools to study phase until rather recently with cheap FIR. For example, a phase linearized full-range speakers capable of delivering low powerful bass, along with source material mastered on the same type speaker. Imo, that's the ruler our speakers need to be compared to to assess phase audibility. (and outdoors)

I certainly am not about to make claims about phase audibility.
On my DIY speaker builds, it's become pretty easy to achieve near perfect mag, phase, and impulse. Along with pretty good directivity ala the spinorama.
Now how much of the excellent sound I feel i achieve, is due to phase I dunno....it could just be due to it is so easy to get near perfect frequency response with my flat phase tuning process.

I also think anybody who claims to fully know about the audibility of phase...well, not somebody I'm gonna listen to. Kinda an audiophool in reverse.
We simply don't know imho.

I'll admit I hope it comes to be we find phase is indeed more audible than the mainstream apparent believes not.
Because it just doesn't make sense to me, that we throw out the audibility of time itself.
Anyway, there is a growing body that is reporting phase may be / or is, audible at low frequencies.
I'm all ears, and eagerly await further science to evolve.

 

[fineMen]

Besides, math, physics and chemistry aren't sexy.

That is, scientifically proven, plain wrong!

'Hope so ...

The terms are ripped out of their context, mixed up, understanding the meaning with just the audio-perspective is a hopeless attempt, but we still do it once and again. People who (kudos, guys!) ask for more relevant information finally are left in lonely desperation. I personally can't do anything about it.

Signal is what goes through the transmission line, starting when the sound engineer choses the very microphone position, mixing console ..., ending when the sound hits the ears, information for the ear is just the frequency response, phase is by the widest margin irrelevant, so 'impulse'--whatever that means is irrelevant, and finally the information for the mind is the music. If the latter doesn't mean anything to You, what would You expect--the signal? We are blind mice?

 

[kimmosto]

What speaker was that Kimmo?

Manufacturers are not allowed to criticize each other on forums so I cannot say that. You just have to believe that timing can be so bad that speaker become unacceptable no matter how excellent magnitude responses to on-axis and off-axis are.

 

[fpitas]

The Fujitsu TenTD712z? Measurement from Stereophile of the Mk 1. Ouch.

Fujitsu Ten Eclipse TD712z loudspeaker Measurements

Sidebar 3: Measurements The Eclipse TD712z's voltage sensitivity was to specification but significantly lower than average, at an estimated 83.5dB(B)/2.83V/m. However, its impedance remains above 6 ohms at almost all frequencies (fig.1), and rises in value above the midrange, this due, I assume...

www.stereophile.com

White paper from manufacturer for the Mk 2.

https://www.denso-ten.com/business/technicaljournal/pdf/33-2.pdf

[Edit] Looks like there are some newer models.

Japanese Speaker Manufacturer Eclipse Launches Its First New Model In Seven Years

The new Eclipse TD307MK3 is a compact, full-range speaker with some advanced technologies and is a direct replacement for the company’s much-loved TD307MK2.

www.forbes.com

The frequency response of that thing (Mk 1) kinda sucks. When it's that bad I'm not sure coherence of any kind can save it. The Mk 2 doesn't seem to be a lot better. These guys must like screechy speakers.

 

[kimmosto]

What speaker was that Kimmo?

Here is timing measurement. It's certainly one of the worst I have ever seen. Look for example impulse envelope and step. Transient dynamics is quite totally lost because energy is distributed from edges to ca. 5 ms periods.

 

[fpitas]

Here is timing measurement. It's certainly one of the worst I have ever seen. Look for example impulse envelope and step. Transient dynamics is quite totally lost because energy is distributed from edges to ca. 5 ms periods.
View attachment 242174

And they accomplished all that with a single driver.

 

[fineMen]

Here is timing measurement. It's certainly one of the worst ...
View attachment 242174

Please don't do that. This graph cannot be understood by the typical audio enthusiast, it only frightens people and by that subdues their reasoning. It cannot be understood by me, who is an educated scientist, not an engineer. I've seen so much wired data, many fancy graphs, heard about such abstruse concepts to explain them--in my scientific field, not audio But I cannot grasp any reasonable insight from that graph!

What You could do is to represent (a) amplitude response together with (b) group delay. Then we could discuss the case.

(Excuse me, but as You say, "the worst", I think a frank and open comment is in order.)

 

[kimmosto]

And they accomplished all that with a single driver.

No. That result requires 3-way with very steep IIR filters, but it's "easier to accomplish" with 4- or 5-way or with sub crossed >100Hz.

 

[fpitas]

No. That result requires 3-way with very steep IIR filters, but it's "easier to accomplish" with 4- or 5-way or with sub crossed >100Hz.

Ah. Makes more sense.

 

[kimmosto]

What You could do is to represent (a) amplitude response together with (b) group delay.

On-axis magnitude response is flat from 27 Hz to 24 kHz.
(Excess) Group delay is ~3.5 ms@100Hz by manufacturer. Time-windowed measurement shows a bit less of course.

 

[kimmosto]

It cannot be understood by me, who is an educated scientist

You can compare that to close to ideal timing of Neumann KH150:
KH150 post #786