I understand that ideally, each driver’s membrane ought to be positioned at exactly the same distance to the listener. But if you want to combine a compression driver/horn (from somewhere around 500-800 Hz and up) to one or several direct radiating driver(s) below, you either get horns ‘sticking out’ or will have to deviate from the abovementioned ideal.

(I would sure love to use horns all the way down, but at least in the deepest bass, I’m afraid they get too big – WAF).

Are there any rules of thump regarding “how much” (if at all) one can deviate from the ideal at, for instance, (a) 100 Hz, (b) 300 Hz, (c) 500, and (d) 800 Hz without compromising the sound too much?

I thank you very much in anticipation!

Regards
pk

 Altec AN-9 is available for reading at - Sam P. 13:10:53 04/17/02 (0)
 Thanks to you all! - PK 04:22:30 04/17/02 (0)
 Re: How important is alignment? - Hornlover 13:45:55 04/16/02 (0)

...

Show full thread outline!

 Altec AN-9 is available for reading at - Sam P. 13:10:53 04/17/02 (0)
 Thanks to you all! - PK 04:22:30 04/17/02 (0)
 Re: How important is alignment? - Hornlover 13:45:55 04/16/02 (0)
 Time domain vs. frequency domain - Wayne Parham 11:30:21 04/16/02 (7)
 taking a step back from time alignment - darkstar 05:02:48 04/17/02 (0)
 Re: Time domain vs. frequency domain - PK 00:15:19 04/17/02 (2)
 Re: Time domain vs. frequency domain - Wayne Parham 02:58:20 04/17/02 (1)
 Thanks! - PK 04:18:47 04/17/02 (0)
 with the VOT, it DOES show up in the response - Sam P. 14:31:26 04/16/02 (2)
 Adjacent second-order filters and first-order slopes - Wayne Parham 15:49:14 04/16/02 (1)
 nope, my feeble brain say 120 degrees "out" - Sam P. 16:22:53 04/16/02 (0)
 Re: How important is alignment? - Robert Hamel 12:07:19 04/15/02 (1)
 all that altec was trying to do was - Sam P. 15:59:09 04/15/02 (0)
 Very ! - djk 02:19:38 04/15/02 (2)
 I guess I'd disagree - darkstar 11:14:02 04/15/02 (0)
 Re: Very ! - PK 03:24:30 04/15/02 (0)
 Re: How important is alignment? - Tom Brennan 10:28:01 04/14/02 (2)
 Further alignment question - SLM 06:46:37 04/15/02 (1)
 End Game - wgeiger 21:00:28 04/15/02 (0)

Thank you very much for your reply. Please excuse me for asking questions in the following, you probably consider trivial, but I'm new to these matters.

You wrote:

>All you need is an all pass delay network.

Can an "all pass delay network" be implemented in an active or passive crossover without using a digital delay unit, and if yes: Do you happen to know where, I can find information about such circuits?

>If the off-set is less than 1/4 wavelength at the crossover point >the amplitude levels can be mis-aligned so that the phase lines up >at the crossover point.

Could you please elaborate on this part of the reply?

Thanks!

Regards
pk

r

The objective is to match the time domain of acoustic signals in the far field (listening position area) that emanate form a LF and HF driver pair.
This matching is critical particularly within the signal spectrum that is combined in the c/o overlap region.
Note that this region is typically several octaves wide and spans frequencies where maximum ear sensitivity occurs.
For best results, strive to make performance here as seamless as possible in regards to not only signal delay, but also phase, amplitude and dispersion pattern. This is not the only issue to be addressed in 'good' loudspeaker design, but if you are using horns, it is a particularly important one because of the large physical displacements that inevitably occur between the drivers used.

Regards,

WHG

You wrote:

>> How important is [time] alignment?

There are a few challenges for a person who would attempt to make loudspeaker systems "time aligned." Two things should be understood first:

1. A filter that performs well in the frequency domain performs poorly in the time domain, and vice versa. What this means is that filters that are designed to kep things aligned in time suffer from being terribly non-linear. You might use such a time domain filter to isolate two events at different frequencies for comparison of when each event occured. So if you want frequency response to be linear, you must choose a frequency domain filter, which has poor performance in the time domain.

2. Even a "single driver" loudspeaker having no electrical crossover is a filter. The motor/chamber system is an acoustic filter. So no loudspeaker system - including the most simple - is immune to this issue. Certainly those with more than one diaphragm, a crossover and several acoustic devices such as horns are systems containing several filters.

Here are the challenges:

1. Electrtical reactive filters change phase as a function of frequency. This means that the time difference between them - if expressed as a distance - changes with respect to frequency.

2. Acoustic filters such as horns and motor/cabinets (sealed or ported) also change phase as a function of frequency. The time offset of the generated signal changes as function of frequency.

3. The position on the baffle is a fixed position, so time offset generated by different diaphragm positions is not a function of frequency, it is fixed.

4. Parallax makes the difference between the listener and the drivers different as the listener or speaker moves. This means the offset in time is a function of position, and two listeners will hear a differet phase relation unless the listeners occupy the same space. Parallax is an issue between two different diaphragms, and it is also the issue across the diaphragm of a single driver.

5. Each diaphragm moves, so the offsets are moving targets.

To make a loudspeaker truly "time aligned", the system would need to:

1. Correct the difference cause by changing phase as well as fixed distance offsets. That's two different things, and must be corrected with two different compensation mechanisms.

2. Correct the offsets at different listening positions along the axis perpendicular to a line drawn between diaphragm centers.

These challenges are daunting, to say the least. Especially since a person cannot tell the difference between a sawtooth wave with the peak on the left from a sawtooth wave with its peak on the right. These are two different things, separated by the phase of the high frequency event. But a listener cannot tell the two signals apart, so it draws into question the whole issue of phase and our ability to discern events in the time domain.

I have always made these things my design goals:

1. Events in the time domain that express themselves in the frequency domain must be corrected. Concentrate on finding those issues that cause frequency anomalies because of phase issues, such as cancellation by diffraction or phase inversion.

2. Events in the time domain that displace the signal so far as to be perceived as an audible delay or echo must be avoided.

So the main importance of the time domain is how it expresses itself in the frequency domain.

Wayne Parham

Of course, with two sources, there will always be parallax in one dimension or another. With vertically aligned sources, adjusted to be the same distance from the listener, the physical distance to each source will remain pretty constant as you move from the one side of the room to the other.

Is it worth the trouble? Heck, this is the second $50 tweak for 511 altecs posted today. Cut them. Align them. Probably more improvement than any other $100 you might spend. It's not about timing, it's about getting a handle on the phase relationships between the drivers, after all the other reactive influences have had their say.
Is random positioning going to be better??? Sam

You wrote:

>> With the VOT, it DOES show up in the response at xover when the back to back BW's are in use.

You've expressed a situation where an anomaly in the time domain has expressed itself in the frequency domain. This is quite common with contiguous second-order networks. Low-order crossover slopes allow a wide band of overlap, so they are vulnerable to this kind of anomaly over a greater set of frequencies and listening positions. But the situation with adjacent second-order networks is evident on-axis, and is usually severe.

>> Is random positioning going to be better?

You wouldn't be referring to a specific system, would you? [grin]

Wayne

Oh yeah, did I mention that by using NEW components, the left and right speakers response looks like almost mirror images. Try that with a pair of mismatched, ebay 416's or 806's. Sam

Thanks alot for the thorough reply! Well, I must admit that it's a bit over my head, which you, of course, can't be blamed for.

You have convinced me that alignment is far from being a trivial issue (at least not for me!).

You have also made it clear that the problem of (mis)alignment consists of a lot of different although interrelated problems. Some of these can be solved (e.g., misaligned drivers), while others are almost impossible to deal with (e.g., misalignmen due to moving membranes). If I get it right, then some of these problems must be more damaging to accurate sound reproduction than others.

Please correct me, if I'm wrong, but if I understand you right, then "time delaying filters" is not really a good solution to misaligned drivers, as they "solve" one problem (the time delay) by making a just another one (unlinear frequence response).

Assuming that this is correct, then misaligned drivers really ought to be avoided in the first place, right?

You wrote:
Is there, all other things equal, some frequency areas that are more sensitive to misalignment than others?

And finally: are the drivers in your folded horns aligned? (Please forgive me, but I don't know how the drivers should be arranged in a folded horn in order to be aligned with a "straight" HF horn)?

Thanks a lot!

Regards
Peter

You wrote:

>> Please correct me, if I'm wrong, but if I understand you right,
>> then "time delaying filters" is not really a good solution to
>> misaligned drivers, as they "solve" one problem (the time delay)
>> by making a just another one (unlinear frequence response).

That's right.

I would not attempt to make a simultaneously frequency-true and time-aligned crossover with analog filters. Even digital filters would be problematic, because thay are really just quantified representations of analog filters. But at least with digital solutions, you could conceivably install the multiplicity of filters that would be required to approximate this simultaneous solution. And it would still only work for one target position.

>> Assuming that this is correct, then misaligned drivers really
>> ought to be avoided in the first place, right?

The point of my post is that this is impossible. It's like trying to make thrust without friction.

>> Is there, all other things equal, some frequency areas that are
>> more sensitive to misalignment than others?

Yes. The higher the frequency, the smaller the scale. That means for any given distance, the equivalent phase relationship between two signals shifted by this amount is greater as frequency goes up.

>> And finally: are the drivers in your folded horns aligned? (Please
>> forgive me, but I don't know how the drivers should be arranged in
>> a folded horn in order to be aligned with a "straight" HF horn)?

No. They are not aligned and as I said, this is impossible. The drivers in my systems are rarely far apart, considering the scale of the systems, and attention is paid to make them properly phased along the horizontal axis. But it is not possible for them to be completely in alignment.

It is not possible for a single-driver system to be phase accurate either. The wavelength of a 10Khz tone is just a little over an inch, so if the diaphragm is even 4 inches across, the the sound eminating from each edge is four cycles apart. Now imagine what happens when you are listening 45 degrees off axis, so that the closer edge is 1.65 inches closer to you than the further edge. The signal component reaching you from the closer edge is a cycle and a half ahead of the signal component reaching you from the further edge.

That's why cone drivers beam at higher frequencies just like horns do. Phase is all over the place, and naturally, there is a consequence that manifests itself in the frequency domain. There is cancellation from this parallax, and that's what you face in the simplest loudspeaker system there is - A single driver with no crossover at all. Even this simple system suffers from enormous malalignment in the time domain.

No electronics can help you with that one. Best thing to do is to split the band to appropriate subsystems, and to do it wisely.

My design efforts have been focused on ensuring that effects in the frequency domain are limited as much as possible. Said another way, I put my attention on making sure there are few positions in the listening environment where there is a frequency anomoly caused by the crossover, acoustic devices or driver positions. The lion's share of this is focused in choosing horn shapes that control dispersion and in using crossover networks that limit overlap between drivers, so that the system more nearly approximates a point source.

You can read exactly what I recommend and how my speakers are built in the 1979 "Pi Alignment Theory" whitepaper and the crossover document that is sometimes discussed on this forum and on the π Speakers forum. I put my greatest effort in ensuring that on-axis response will have no frequency aberration and that performance along the horizontal plane is good (for all listeners positioned from side to side).

Take care!

Wayne Parham

Thanks for sharing your knowledge, I really appreciate it!

Regards
Peter

Regards
Peter

Since Wayne (always thorough: Middle name 'Phone-it-in' :)) discusses even single element speakers as having diaphragm/horn time issues, similarly (I would think) musical instruments have the same conditions. The entire body of a violin produces sound, (OK maybe that's small, consider a double bass instead) and is far from a time-aligned point source.

Could the time issue be related to microphone use? The assumption then is that the microphone is 'god' and any reproductive system's job is to reproduce what the diaphragm of the microphone records.

So, in my minds eye I see this huge wash of disparate sound sources, all relative to each other and themselves, bouncing, phasing, etc. together like a group of happy children until they hit the microphone. Is the engineer chartered then with redistributing those happy children at the speaker end?

Is this even possible?

r