A Broad Discussion of Speakers with Major Audio Luminaries

[Sokel]

I think the impulse or step response is the least thing to review when evaluating the sound quality of a loudspeaker. Obviously many people do not even understand the theory behind it. I for myself focus on other - more relevant - performance metrics.

That's what this thread all about though, so we can read people like @j_j and the others and get a closer picture.
For the obvious like FR (and I mean by eyeballing, FR can be very,very detailed explained in ways we will never learn) we have all the other usual threads.

 

[Keith_W]

I think the impulse or step response is the least thing to review when evaluating the sound quality of a loudspeaker. Obviously many people do not even understand the theory behind it. I for myself focus on other - more relevant - performance metrics.

You're right. But it's very important when you have speakers and multiple subwoofers and want to study the integration, the room effects, and look for side effects of DSP.

 

[Jack B]

Glad you brought this up. Supposedly no, it does not, and driver material is immaterial. Frequency response and directivity is king.

If that is the case, I would love to know why Revel use Be and Ti for their drivers on their high end models. I have a feeling I know what the answer is, but I want to hear it directly.

In my ever-so-humble opinion, the beryllium and titanium make for very sexy advertizing copy. How you gonna get people to spend much bigger bucks, if your speaker drivers are made out of compressed cow plop?

 

[kyuu]

In my ever-so-humble opinion, the beryllium and titanium make for very sexy advertizing copy. How you gonna get people to spend much bigger bucks, if your speaker drivers are made out of compressed cow plop?

To be fair, beryllium does have technical advantages. Enough to make it worth the extra outlay and the toxicity problems? Not in my opinion, no, but it certainly does lend an air of luxury that, unfortunately, seems to be necessary to sell high-end speakers. Myself, I would be quite happy with Revel's F206 or F208 if I was shopping for floorstanders and that line was still readily available.

I'm really hoping Revel refreshes their non-Be lineup in the near future (and particularly their in-walls, but that's my own personal bugbear).

 

[Jack B]

To be fair, beryllium does have technical advantages. Enough to make it worth the extra outlay and the toxicity problems? Not in my opinion, no, but it certainly does lend an air of luxury that, unfortunately, seems to be necessary to sell high-end speakers. Myself, I would be quite happy with Revel's F206 or F208 if I was shopping for floorstanders and that line was still readily available.

I'm really hoping Revel refreshes their non-Be lineup in the near future (and particularly their in-walls, but that's my own personal bugbear).

Agreed on all counts (though the toxicity issue affects only the Mfr. assembly operation, not the end user, barring some extremely unlikely scenario). But the sexy/exotic/luxury factor motivates the buyer to pay the big bucks...look at the ?Italian? speakers coated in fancy leather. If you poke it and listen very hard, can you hear it moo?

 

[kyuu]

(though the toxicity issue affects only the Mfr. assembly operation, not the end user, barring some extremely unlikely scenario)

Well it affects the end-user if and when the speaker reaches end-of-life and disposal has to be considered.

 

[Jack B]

Granted, you're right. But in the meantime, can we hear the moo?

 

[Everett T]

Well it affects the end-user if and when the speaker reaches end-of-life and disposal has to be considered.

Darn, those tennis rackets I used in the 80s should have been better disposed of.

 

[RegularStereoGuy]

It is resonances that are mainly responsible for lack of neutrality in loudspeakers, so the first task for competent speaker designers is to minimize them. After that spectral balance and directivity control need to be attended to. The industry standard measurement method, used by Amir, which had its origins in Canada while I was at the National Research Council, is excellent at revealing resonances. It seems that, historically, this company has studiously ignored the science and gone the route of designs with "personality". It is a choice, and clearly there is a market for "colored" loudspeakers. This may make the loudspeaker sound "distinctive" in a demo, but the trouble is that everything played through them is colored in the same way, which eventually gets tiresome. Neutrality wins, and simple tone controls or equalization can compensate for most problems in recordings or add coloration for personal preferences. The advantage of starting with neutral loudspeakers is that one can go back to listening to what was recorded.

Agreed. 40-50 years ago, when I and my cousins received our first 'stereo systems' as kids or early teens, this is typically what we got:

I would seriously estimate that as much as half of what we were listening to was coming from the speaker cabinets, as much from the drivers themselves.

No wonder our parents were always yelling at us to "turn it down!".

Just the simple act of transplanting the drivers and crossover(if those cheap stereo things had crossovers) into solid wood cabinets (or event thicker particle board) would reduce most of the resonances that they were really complaining about.

Thankfully most of us grew up and now realize that the only place sound should be coming from is those speaker drivers.

 

[j_j]

Well, I had some time, so I wrote a bit of code to explain how impulse responses add up.

Obviously this requires some explanation. The first row is a single, DC to FS/2 impulse, in column 1, and it's spectrum in colum 2
Row 2 limits the bandwidth to low frequencies, and shows the impulse response and bandwidth that result
Row 3 limits the bandwidth to mid-frequencies, and the resulting impulse response
Row4 limits the bandwidth to high frequencyes, and the resulting impulse response
Row 5 ***ADDS*** the 3 bandwidths, and in column 1, ***ADDS*** the impulse responses in rows 1-4. Oh, look, right back to a single impulse.
Linear superposition still works.

Notice that because I used constant-delay frequency limiting, the impulse response tails are symmetric in all cases.

Were I to use IIR type filters to make the bandwidth limiting (i.e. use high level RLC crossovers) everything would be a great deal messier.

 

[j_j]

Thankfully most of us grew up and now realize that the only place sound should be coming from is thos speaker drivers.

(cough) Bass Reflex systems are not wrong, you know. That's sound from the acoustics of the box.

Now, yeah, the box itself vibrating, that's almost never going to be good. And I say "almost never" while thinking "never ever" but also "never say never".

 

[Tim Link]

Start with a definition of what you mean by "distortion", and at what point in what space you measure the speaker.

Issues around what point in space you measure the speaker have been weighing big on my mind lately. I've been measuring speakers as an amateur for quite a while now. But certain issues only became apparent to me more recently. Like the distortion issue. If your speaker is increasingly directional in the high frequencies and you measure off axis, you get really low distortion measurements. Many of you may be thinking, "Duh!" But I had never thought it through before until I stumbled upon it doing off axis measurements.

So how about I EQ it flat off axis, listen off axis, and enjoy the low distortion?

Doesn't sound so good. That bright on axis beam gets back to you viciously after a few bounces around the room.
But what if I put an absorber strategically placed on axis to reduce that high frequency energy before it bounces around too much? I don't know, haven't gone there yet. Seems like a precarious setup.

Another issue with measuring that's really been hitting me hard lately is how far back from the speaker you measure. If you have a speaker that gets highly directional as frequency goes up, like I do, there's an issue with fall off rate, and it should not be ignored! A speaker like this that measures flat on-axis at 1 meter will be tilted up on axis at 2 or 3 meters, and this is even if you use a long IR window that includes all the off axis lower frequency that bounces back to the measuring position. The highs don't fall off as fast as they get further from the speaker, including when they start bouncing. So the off axis energy gets strange, and despite the lower frequencies going out to hit more surfaces sooner, they're all falling off faster and they're losing the fight, at least in my room.

I've got an exponential horn that was naively designed (by me and my friend) to cover the range from 200 to 2000 Hz, with a tweeter horn coaxially mounted in the middle. After years of experimenting with this contraption I wrote it off as hopeless above about 600 Hz, which meant it had a really small useful range. The dispersion was the problem, I assumed. By 2k the mid horn was beaming like a laser, and when the tweeter took over it was going wide and bouncing off everything.
I got to thinking about this again recently and decided to take a long IR measurement at the listening position of the midrange horn by itself without EQ or crossover filters to see what it did. With smoothing it gave a remarkably straight line result from 300 Hz to 6000 Hz tilted at about 3 dB / Octave. Measuring up close it shows a 6 dB / Octave roll off over the same range, verifying the fact that indeed the frequency response is increasingly tilted up as you move back.

So what should the measured response be at the listening position with a speaker like this? If you make it flat at the listening position it'll be getting brighter further back when it starts reflecting off walls. 3 dB/ Octave at the listening position is too much, but it doesn't sound as bad as I expected. I think it's hard to calculate what the slope should be with speakers like this so it's good to try different slopes using EQ and listen. I've been doing that and I'm astounded how much better this horn is sounding than I've ever heard it before when I was trying to EQ it flat at 1 or even 2 meters using various methods to try to get a quasi-anechoic measurement. I actually really like it now! And, at 6000 Hz the tweeter is pretty beamy too so the coaxial horns actually work together nicely, beaming in harmony and producing a remarkably smooth, relaxed and detailed effect.

 

[j_j]

Issues around what point in space you measure the speaker have been weighing big on my mind lately. I've been measuring speakers as an amateur for quite a while now. But certain issues only became apparent to me more recently. Like the distortion issue. If your speaker is increasingly directional in the high frequencies and you measure off axis, you get really low distortion measurements. Many of you may be thinking, "Duh!" But I had never thought it through before until I stumbled upon it doing off axis measurements.

...

this horn is sounding than I've ever heard it before when I was trying to EQ it flat at 1 or even 2 meters using various methods to try to get a quasi-anechoic measurement. I actually really like it now! And, at 6000 Hz the tweeter is pretty beamy too so the coaxial horns actually work together nicely, beaming in harmony and producing a remarkably smooth, relaxed and detailed effect.

You are indeed beginning to understand the issues surrounding interaction between loudspeakers and listening environments.

A minor quibble, none of this is "distortion", per se, you're talking about frequency response irregularities, and confusion between timbre of the direct vs. the reflected sound. These are indeed problems, but they are not "distortion". It's still a linear system, just not a system with a frequency response that's what you want in some fashion or other.

 

[Tim Link]

A minor quibble, none of this is "distortion", per se, you're talking about frequency response irregularities, and confusion between timbre of the direct vs. the reflected sound. These are indeed problems, but they are not "distortion". It's still a linear system, just not a system with a frequency response that's what you want in some fashion or other.

I was looking at the THD and its components as measured in REW. It seemed the harmonic distortion components were getting beamed straight ahead and not showing up as well off axis. So I could alter REW's distortion results by changing where I measured. The question then in my mind was if that meant I would also hear the driver as having lower distortion if I listened off axis and absorbed the on-axis output?

 

[j_j]

I was looking at the THD and its components as measured in REW. It seemed the harmonic distortion components were getting beamed straight ahead and not showing up as well off axis. So I could alter REW's distortion results by changing where I measured. The question then in my mind was if that meant I would also hear the driver as having lower distortion if I listened off axis and absorbed the on-axis output?

That's interesting in a couple of ways! The harmonic components will beam more, of course, because they are at higher frequencies, but the fact that they are beaming suggests that distortion is happening before the actual physical air-movement that creates the sound.

 

[gnarly]

Well, I had some time, so I wrote a bit of code to explain how impulse responses add up.

Nice !

To build on that, and give a real world example, here is a 5-way speaker's acoustic measurements,
that shows the passbands and impulse responses for each of the passbands.
Linear phase passbands of course.

The bottom two panels are the full speakers' impulse, and transfer function.
Sorry the impulse is so small to view. I would reopen the project and present a large summed impulse in the top panel, but the program has been updated and won't open older files.
The flat mag and phase in the bottom panel pretty much say it all for the passbands' summed impulses....a pretty clean single spike.

 

[Sokel]

Nice !

To build on that, and give a real world example, here is a 5-way speaker's acoustic measurements,
that shows the passbands and impulse responses for each of the passbands.
Linear phase passbands of course.

The bottom two panels are the full speakers' impulse, and transfer function.
Sorry the impulse is so small to view. I would reopen the project and present a large summed impulse in the top panel, but the program has been updated and won't open older files.
The flat mag and phase in the bottom panel pretty much say it all for the passbands' summed impulses....a pretty clean single spike.

View attachment 498259

That looks like if it was electrical!

That's the best I can do at MLP with no EQ:

wouldn't call it symmetrical.

..and that's with the EQ I apply mainly to knock-off this 32Hz mountain down there:

Not much of a difference.

(first one sounds like a guilty pleasure at late night low levels )

 

[gnarly]

That looks like if it was electrical!

That's the best I can do at MLP with no EQ:

Your traces look great ! They look devoid of room
What do you mean by no EQ? I guess I'm asking what is the speaker, how is it set up? And sorry to be a dummy, but what is MLP?

The traces I showed do almost look electrical...the power of FIR.
I took them when I was first learning the measurement program, so they are simple tune-to-a-spot, without regard to fully optimizing polars.
So the example was just to show a speaker measurement, in addition to a sim. (If others are like me, I often have an easier time connecting with measurements than sims.)

Here's a similar 5-way electrical sim, where I can show the 5 sections impulses precise summation....
Black lines in lower transfer function panel are summed frequency response (mag) and phase

Same thing but summed impulse response.
Ye ole Dirac

 

[Sokel]

Far, FAR from devoid, it's maybe the chart under the impulse that is deceiving because of the way it looks, that's a look at a normal chart with EQ and no EQ:

MLP=main listening position, the sweet spot, mine is about in the inner corner of a 2.8 meter triangle.
And no EQ means no EQ, just the speaker (a semi-active big-ish 3-way with analog el. x-over) and my room (also biggish, about 80m² and treated though)

 

[levimax]

Your traces look great ! They look devoid of room
What do you mean by no EQ? I guess I'm asking what is the speaker, how is it set up? And sorry to be a dummy, but what is MLP?

The traces I showed do almost look electrical...the power of FIR.
I took them when I was first learning the measurement program, so they are simple tune-to-a-spot, without regard to fully optimizing polars.
So the example was just to show a speaker measurement, in addition to a sim. (If others are like me, I often have an easier time connecting with measurements than sims.)

Here's a similar 5-way electrical sim, where I can show the 5 sections impulses precise summation....
Black lines in lower transfer function panel are summed frequency response (mag) and phase
View attachment 498395

Same thing but summed impulse response.
Ye ole Dirac
View attachment 498396

Do you have any actual speaker measurements showing that the pre-ringing with a complimentary crossover actually canels out? Linear phase crossovers are nice is many ways but I am a little skeptical exactly how well the "cancellation of pre-ringing" actually works with speakers in a room, especially when thinking about how it would work off access.