tweeter set-back vs all-pass delay

[Bigun]

I've noticed a few cases where time alignment is achieved by setting back the tweeter relative to the woofer. The Willson Watt/Puppy (sloping baffle), the FAST speaker from Xrk971, some designs from Troels are examples. And there are XO designs, active or otherwise that use an all-pass filter to produce the same result.

But are they equivalent - the physical set-back only provides the correct physical delay when listening on-axis. The electrical delay works over all listening angles.

Why would anybody prefer a physical set-back over an all-pass (or similar) delay ?

 

[Sonce]

the physical set-back only provides the correct physical delay when listening on-axis. The electrical delay works over all listening angles.

No, the electrical delay works only along one axis, the same as physical set-back.

 

[Lynn Olson]

Let's look at what happens for the 90-degree off-axis condition (well outside the listening window, but significant for early reflections).

If the tweeter is physically displaced behind the mid or woofer, that displacement disappears for the 90-degree off-axis radiation. Not only that, the woofer is acoustically lowpassed by the severe off-axis angle (compared to on-axis radiation), so that adds an additional time delay (compared to the on-axis condition).

This isn't hard to measure in the real world. Look at the impulse response of the entire speaker at 0, 15, 30, 45, and 90 degrees off-axis. A physically time-aligned (with displaced tweeter) system will be approximately OK over +/- 20 degree listening window, but gets really wonky outside of that. The time response behind the speaker, for example, will look nothing like the forward response.

It's a matter for separate discussion whether the early reflections need to be time-coherent ... some would say yes, others no. There is general agreement that early reflections should have a spectral similarity to the on-axis spectra.

If time-coherent early reflections are important (to you), the only kinds of speakers that deliver that are Walsh, German Physiks, MBL, or similar types of speaker (or an upwards-pointing full-range driver with a diffuser). Electrostats are bidirectional and well-behaved close to the centerline (front and back), but look pretty strange close to the 90-degree off-axis null.

 

[Bigun]

No, the electrical delay works only along one axis, the same as physical set-back.

I can't see why is that ?

 

[Davey]

I can't see why is that ?

All all-pass electrical delay is a frequency-dependent delay. It's not the same as a pure electrical delay or physical setback of the driver itself. Those are frequency-independent delays.

Dave.

 

[Bigun]

So the all-pass delay is set up to provide the optimal delay at the XO frequency where contributing drivers overlap but further away the time alignment won't be as accurate but at these points one driver will dominate the output - a tradeoff but not a bad one perhaps.

But doesn't it work off-axis just as well as it works on-axis ?

 

[eriksquires]

Just to muddy the water even further, most designs with staggered tweeters are still not time coincident.

Mind blown? 🙂

E

 

[Bill poster]

Talking of time aligned set back, i like raidho's solution, which is more of a 'set forward'

 

[freddi]

fwiw i got a more "coherent" sound from a 12 inch pro Eminence coax (Martin sound 1114) using "this" (on my woofer)

(DJK mentioned PAS using 12uF and 0.35 in their cx)

 

[Sonce]

So the all-pass delay is set up to provide the optimal delay at the XO frequency where contributing drivers overlap but further away the time alignment won't be as accurate but at these points one driver will dominate the output - a tradeoff but not a bad one perhaps.

But doesn't it work off-axis just as well as it works on-axis ?

As Davey pointed out, all-pass filter delay is different from pure time delay (digital), in a view of frequency dependence. Because of frequency dependence all-pass filter may have better off-axis frequency response, but only in one vertical direction, say up. In other direction (down) it will be much worse than pure time delay. So, no clear winner here.

 

[Sonce]

Just to muddy the water even further, most designs with staggered tweeters are still not time coincident.
Mind blown? 🙂

Marketing is more important than engineering. If loudspeaker looks like a time-coincident one, it will sell better.

 

[AllenB]

So the all-pass delay is set up to provide the optimal delay at the XO frequency where contributing drivers overlap but further away the time alignment won't be as accurate but at these points one driver will dominate the output - a tradeoff but not a bad one perhaps.

But doesn't it work off-axis just as well as it works on-axis ?

Even if it were a perfect delay, delay is one dimensional, as is physically pushing back the tweeter. Horizontal offset won't be fixed this way, and the response may have issues off-axis.

In cases where pushing the tweeter back gives a diffractive/reflective baffle surface, delay can become an option. Otherwise an assymetrical crossover might be made to work?

 

[Bigun]

It doesn't sound too promising, this delaying business, kind of puts me in a bad frame of mind regarding multi-way speakers.

The trade-offs are complicated, I'm not getting a feel yet for the relative merits of different approaches.

 

[manninen]

i have project with 143mm acoustic distance difference and been twiddled around passive xo and it looks bit tricky.
simulated all-passes and different slopes, looks really no go for good results
i think active delay is easiest way me to get to goal

 

[bentoronto]

The trade-offs are complicated, I'm not getting a feel yet for the relative merits of different approaches.

Guess I figured time alignment (and the soul-searching earnestness of this thread) was at best a secondary issue compared to real big issues with speakers.

But I used to have a Klipschorn bass (considered like 16 feet long) and adding to the distance, dipole electrostatic panels out from the walls. About 130 Hz XO.

Sooooo, being a newly converted adherent of DSP (Behringer DCX2496.... and I think everybody really has to have one), thought I'd see if I could move the panels back in time 20 feet. (Bigun, that is how you can experiment, eh.)

I was surprised to find the difference was detectable* and worthwhile. At least for my sub set-up. Bass "transients" were clearer. Nice although not a large improvement on an already good system**.

B.
*after decades of listening to this system, my subjective judgment had some modest validity
**if you have a BR box with gross "group delay" it would just be a waste of time to fuss about bass "transient" sound

 

[phase_accurate]

An allpass can introduce time-smear due ti the already mentioned frequency dependant group-delay. It works best when used on a woofer or midrange branch.

**if you have a BR box with gross "group delay" it would just be a waste of time to fuss about bass "transient" sound

The worst group delay distortion is in the area of the lower cutoff frequency. For a reflex box it does not necessariliy have to be large in the crossover area where the time-alignment is most effective. So the positive effect of time alignment might not be lost due to the use of a reflex box.

Regards

Charles

 

[system7]

There may be something to be said for a time-aligned array like this Technics design. You can certainly line phase up better on second order filters. Which is a good thing. TBH I favour negative polarity on a flat baffle with steeper slopes, possibly exploiting some asymettry on slopes.

I suspect these allpass networks are fools gold though. I redid the B&W Matrix 1 third order plus allpass flat baffle crossover with a straightforward 4th order filter, flipping tweeter polarity. It ended up in the same place! Not surprising since each section stores energy and hence introduces time delay by nature.

 

[phase_accurate]

I just saw that this Statement might be misleading:

An allpass can introduce time-smear due to the already mentioned frequency dependant group-delay. It works best when used on a woofer or midrange branch.

What I wanted to say with this is that the delay achieved by using an allpass is closest to a pure time delay when it can be used on a midrange or woofer. I did by no means try to talk into using it on the woofer channel when a tweeter has to be delayed. It just works better in the situation where a low-paased channel has to be delayed.

Regards

Charles

 

[eriksquires]

Marketing is more important than engineering. If loudspeaker looks like a time-coincident one, it will sell better.

Sonce,

I'm not sure. Troels Gravesen has a couple of designs that use physical offsets but are not time co-incident. I think it can help remove a pole or two in order to get accurate phase alignment.

AFAIK, in order to get true time / phase coincident in a multi-way speaker you not only need the physical offset, but you also need to use a 1st order HP on the tweeter.

Best,

E

 

[bentoronto]

These last few posts seem to clarify what a fools errand this is, at least as the question is posed in terms of phase-arithmetic and textbook diagrams.

Perhaps what's needed is a distinction between phase alignment and time alignment. When talking about multi-ordered crossovers, 5 mm physical misalignment, and unknown behaviour inside the drivers, silly to try to aim to get phase anywhere right.*

But when you have several cycles wonky, perhaps it does make sense to get the general time alignment within a just one time cycle or two, depending on the wavelength.

B.
*don't laugh, but I've often pointed that in setting the polarity of sub which are located anywhere but under the mids, often impossible to decide by ear what is the right phase. And with a mic, you need to pick between two curves that are not clearly with one curve better.