A history of Time Aligned Loudspeakers

[Salt]

I want to be certain that I understand what you're saying here. It looks like you have implemented a 2nd-order Linkwitz-Riley crossover, but instead of inverting the polarity of one of the drivers, you're offsetting it by one-half wavelength. Correct?

If so, then the attached graphic shows what happens to the frequency response.

Simulation with ideal parameters is often not what is measured in wild life.

 

[Duke]

I want to be certain that I understand what you're saying here. It looks like you have implemented a 2nd-order Linkwitz-Riley crossover, but instead of inverting the polarity of one of the drivers, you're offsetting it by one-half wavelength. Correct?

If so, then the attached graphic shows what happens to the frequency response.

Like I said, there is a juggling of tradeoffs. Some of that shows up as additional filtering being required. In practice the crossover slopes accelerate to something closer to 4th order at some distance to either side of the crossover frequency. Ime quite a bit of trial-and-error is involved. So in the end it's a non-standard and asymmetrical second-order filter, and not a textbook Linkwitz-Riley (nor textbook anything else) by any means.

 

[gberchin]

Simulation with ideal parameters is often not what is measured in wild life.

The very same argument applies to the original concept.

 

[Salt]

The very same argument applies to the original concept.

So here we go...

 

[norman bates]

uncharacteristically, I'm staying out of this one.

 

[Cote Dazur]

Mission 780 Argonaut loudspeaker
The drive-units are time-aligned, the woofers being mounted forward of the tweeter on sub-baffles that look to be made from structural foam (though this doesn't mean that the speaker is phase-linear, this being dependent on the crossover configuration)

 

[norman bates]

Diffraction nightmare.....

I'd love to see a freq graph......

But extra credit for trying to line up the acoustic centers.

 

[Salt]

Diffraction nightmare.....

I'd love to see a freq graph......

But extra credit for trying to line up the acoustic centers.

Don't see a grand problem horizontal, more vertical, but may be not by MTM configuration.

 

[gnarly]

Here's two of my latest DIY time-aligned speakers.

First, a 5-way. Designed for use with a sub below 40-50Hz.

It's measurement.
Bottom two small panels, summed mag and phase, and impulse.
Middle panel, each of the 5 sections mag response (and summation)

Top panel, the interesting one....each of the 5 sections individual impulse responses.
Note how all the peaks are time-aligned. It's a property of linear-phase time-alignment (as opposed to minimum phase time alignment that requires impulses initial rises to be aligned). Makes time (impulse) aligment soooo much easier than min-phase/IIR.

2nd one, a 4-way. Designed for use with sub below 100Hz.

Instead of repeating the same flat mag, phase at zero, and impulse traces that it has too, here is a burst decay that also includes a sub.
I think it illustrates how flat mag and phase, with time alignment, translate into clean decay.
Measurement taken outdoors off a deck at 4 meters.

My 74 yr old arse has been using DSP and linear-phase FIR filters on every DIY box I've built over the last ten years. (over forty of them)
Old dogs can learn new tricks Lol.

 

[Waveform Fidelity]

Wow, I find your design fascinating! I am not a horn expert, but I thought horns are fed with a compression driver at the end of the waveguide. What the basic/theory and geometry? Is it your own design?

 

[Keith_W]

Wow, I find your design fascinating! I am not a horn expert, but I thought horns are fed with a compression driver at the end of the waveguide. What the basic/theory and geometry? Is it your own design?

It's a type of horn called "MEH" - Multiple Entry Horn. Google "Danley MEH" or "Synergy Horn" and you will find a lot of discussions. You can buy horns like that, e.g. the Red Spade PSE-144.

 

[Mr. Widget]

My 74 yr old arse has been using DSP and linear-phase FIR filters on every DIY box I've built over the last ten years. (over forty of them)
Old dogs can learn new tricks Lol.

Very impressive work to be sure. I do wonder about the subjective sound though. In my experience when I have really gone to the limit to make a speaker extremely linear with DSP, the sound is subjectively less enjoyable than a less corrected version.

It is very difficult to quantify or to use words to express the effect, but allowing most of the naturally occurring minor peaks and dips above the Schroeder frequency to remain uncorrected has generally sounded more pleasing when I toggled between the corrected version and the less corrected version.

Have you experimented with anything along these lines?

 

[RetroStereo]

I have a friend who has DSP'd and PEQ'd his system to near mathematical perfection. He likes it. To me, it sounds somewhere between sterile and dead (as in R.I.P. - not the sonic meaning, although, that, too, somewhat).

 

[gnarly]

Wow, I find your design fascinating! I am not a horn expert, but I thought horns are fed with a compression driver at the end of the waveguide. What the basic/theory and geometry? Is it your own design?

As Keith said, they are called a unity / synergy / MEH by DIY communtity (& a sucky name lol).
They come from the inventive genius of a guy named Tom Danley.

And yes, there is a compression driver at their end (throat), along with all the rest of the spectrum's drivers mount on the horn walls as well.
Main benefit is it compacts drivers' center-to-center spacings beyond what is normally available with conventional designs.
They are tricky though, because the driver sections all interact and effect each other acoustically.

 

[gnarly]

Very impressive work to be sure. I do wonder about the subjective sound though. In my experience when I have really gone to the limit to make a speaker extremely linear with DSP, the sound is subjectively less enjoyable than a less corrected version.

It is very difficult to quantify or to use words to express the effect, but allowing most of the naturally occurring minor peaks and dips above the Schroeder frequency to remain uncorrected has generally sounded more pleasing when I toggled between the corrected version and the less corrected version.

Have you experimented with anything along these lines?

Yes indeed. I wonder greatly about the amount of correction, its measurement improvements vs its sonic effect. And have run near countless experiments comparing degrees of correction, trying to assess subjective sonic results.

I've found tunings having what folks would consider identical transfer functions and impulse responses, to have not so identical sound.
Tunings where nothing changed other than the degree of correction, in terms of how finely the PEQs/ filters were applied. Tunings where on-axis and spins, all look identical at 1/12th smoothing, and sometimes even at 1/24th. And still sound different....(frankly it is quite frustrating.)
It's one of those imponderables for me.....

We know in theory any minimum-phase correction of a minimum phase device is a valid correction, correcting both magnitude (frequency) and phase at the same time. And that drivers are predominantly minimum-phase devices. So I make all my corrections at the driver level, using a DSP/amp channel for each driver section.
I figure if a minimum-phase correction of a driver holds up spatially, it is a valid correction that is likely to have audible (and certainly measurable benefits.)
I also figure the quality of the measurements has a ton to do with the quality of the corrections. (I use dual channel FFT coherence to assess measurement quality)

Using that logic, I've tested different degrees smoothing of applied filters. With FIR, using min-phase EQs embedded into the filter for driver correction, it's very easy to change filter resolution simply by varying the number of taps (coefficients) used. With IIR , I vary the restriction on Q/bandwidth being applied.
In both cases, I'm strongly led to believe higher Q filters, no matter how well they seem to correct drivers, are best avoided.

(Worth noting, avoiding high Q/bandwidth PEQs does not mean avoiding high order crossovers. Their rate of change continuity is such even a LR 96 dB/oct lpf or hpf does not have a high Q property. IIR high order imparts too much phase rotation/ freq dependent group delay, but complementary linear-phase FIR high order crossovers do not. If we can live with the fixed latency of lin-phase FIR crossovers, it's all win ime/imo).

Anyway, I've rambled and digressed enough i guess.. I could go on for along time about all this lol
fwiw, here's the same response for the big white 5-way i posted, but now at 1/48th measurement resolution.
You can see a bunch of good sounding whiskers/stubble

 

[gberchin]

I'm strongly led to believe higher Q filters, no matter how well they seem to correct drivers, are best avoided.

@gnarly this might be the most important statement in your entire message. In my experience, trying to correct a high-Q anomaly with a similarly high-Q filter almost never works. I can only take an educated guess at the reason, but I believe that even very small errors in the center frequency or the Q result in very audible artifacts. Add to that the fact that the characteristics of the anomaly being corrected can change with temperature, humidity, phase of the moon, etc. (because a speaker is a physical system, existing in a physical environment), and what was "right" yesterday might not be "right" today.

 

[gnarly]

Absolutely Greg ! 1000% on board with all that...including the variability my damn mood

 

[Multiway]

You don't have to over DSP / correct.
I've had great fun and results both with IIR near zero order XO Multi Entry Horns and more recently FIR.

So easy to A/B compare each iteration of new trials with the last, go back to several earlier to compare and ditch a setting or tweak that over does it.
No need to DSP the things to within an inch of their lives - auto systems like miniDSP and what's in the Wiims (I've tried both) seem to do just that.

MEHs get so much right.
Single point and clarity / scale of sound capability / ability to handle very varied recording techniques and quality is how I'd describe them.

Question does gnarly = Mark?

 

[Mr. Widget]

Absolutely Greg ! 1000% on board with all that...including the variability my damn mood

In all seriousness, I think our moods have a significant subjective impact on perception which is just another reason why subjective evaluation is unreliable. That said, I am not one who "listens" with my eyes looking at beautiful graphs and plots. After all, listening is the ultimate goal... at least for most of us.

But I digress... yes, I also agree that high Q filters are almost never a good idea, though I had a room with a very narrow band reinforcement node that required a high Q cut filter. Inserting that filter was absolutely necessary for quality LF playback.

 

[Waveform Fidelity]

@gnarly this might be the most important statement in your entire message. In my experience, trying to correct a high-Q anomaly with a similarly high-Q filter almost never works. I can only take an educated guess at the reason, but I believe that even very small errors in the center frequency or the Q result in very audible artifacts. Add to that the fact that the characteristics of the anomaly being corrected can change with temperature, humidity, phase of the moon, etc. (because a speaker is a physical system, existing in a physical environment), and what was "right" yesterday might not be "right" today.

The problem that I find in my work on FIR/IIR equalization is trying to determine if the driver has a small, narrow band resonance, or its a diffraction effect (and sometimes its a combination of both) - I’ve also stopped chasing/compensating High Q resonance because of that reason. In most cases anyway resonances are highish Q.

My DSP is limited to 512 taps and this seems about the number I actually need to compensate for the broader band resonances. I have flat response (+/-1dB) from 300-18KHz on my KEF UniQ on axis and the quality/realism of speech is very good, but I agree with other observations that a flat response (amplitude and phase) sounds a bit “thin” or hollow” - maybe because there is no pre/post ringing.