Woofer + Tweeter: Physical time alignment

[ramiro77]

Hi!
Recently built a pair of near field monitors with Peerless 830875 and Seas h1189.
I'm so pleased with them that I decided to add another woofer and make them floorstanders.
Given that I have to build everything back, I want to complicate the design. And for this, I thought to make a time alignment for the drivers.
The problem is I do not know exactly how. I read that magnetic motors are placed in the same vertical plane. Is this true? What should I consider if it is not true?

[sergeloc]

That should be the center of each voice coil

[CharlieLaub]

Some people have reported in this forum that they successfully time aligned drivers by designing the speaker with a "sloped" baffle, at about 15 degrees. The goal is to align the acoustic centers, but determining exactly where that is without some measurements (e.g. impulse arrival times) is not likely to be very precise.

-Charlie

[Tarasque]

I would suggest to build an enclosure without time alighnment and build each section of the filter for each speaker separately.
Drive each section of the filter with a separate amplifier and add a digital delay using your computer with a 6 channel sound card as a source.
Play with the delay's between the various drivers until you are satisfied with the result and then calculate the phisical offset of teh drivers from the digital delay that you ended up with.

When you look at commercial designs, you will find that the offset of the drivers is not exactly the alighnment of the drivers, nor the alignment of the voicecoils. Also the delay in combination of the electrical parameters is part of the acoustical delay.

[ramiro77]

Thanks to all of you.
The slopped baffle design is interesting. But i´m trying to align them with a regular frontal shape.
I was thinking about using something like double or triple front panel for the woofers, but not for the tweeters.
What do you think about?

[5th element]

While the computer delay is an option, it may not be possible or even desirable as part of the system. Indeed neither may be the addition of more power amps.

Time aligning the drivers is a little more complicated then physically aligning the voice coils, it just doesn't work like that, at least not in the way you'd think.

According to what I've learnt the acoustic centre for a driver moves in position relative to the frequency it is reproducing. Therefore any 'aligning' be it physical or electrical is only relevant to that one frequency, or rather you could say frequency band because the acoustic centre will most likely remain within acceptable limits over certain frequency bands.

The same is also true for the crossover employed. These add in electrical phase shifts that correspond to physically moving the tweeter through the Z axis and the amount to which this occurs is highly dependant on the target transfer function. That is to say the 'delay' imparted by the electrical crossover component is different for different xover frequencies and slope types.

The only true time alignment can come from digital filters that posses the ability to compensate for any phase shift across the entire audio band. I believe this is what the FIR proponents are getting at with regards to the 'point' or the holy grail of why you'd go digital in the first place. A bit like saying, if you've got a digital crossover and haven't implemented a phase linearising FIR filter then why bother at all? Of course there are many reasons why one should bother, but some would say you've missed half of the point by ignoring it.

Ignoring the digital FIR solution, you cannot simply place the tweeter in one position and be done with it.

For a start the wave fronts that the different drive units are producing can only be truly coherent at one listening height. This is obvious when you think about it because moving up and down alters the relative distance that the sound waves have to travel from the tweeter and the woofer. This is why properly done coaxial drivers attract such a large following, as their vertical off axis response mirrors their horizontal off axis. That is (if properly designed) remaining flat and coherent as you go off axis, with only high frequency, beaming related droop showing. Rather then a non coaxial system that has the two drivers going in and out of phase with one another as you change the vertical listening axis, creating dips and troughs all over the place.

As we're talking about a non coaxial driver, this means you have to design your system around one preferred listening height.

Next on the agenda you have to choose what xover point and slopes you'd like to use. The H1189 (27TDFC) is a capable device and can operate at low xover points without a problem, this is a very good thing.

The peerless driver being a 6.5" mid/bass will start to beam quite early on and should ideally be crossed over as low as the tweeter can go. Also if you use Zaph's measurement of it's smaller brother, the Peerless 830873, as a guide line, it shows that the driver will probably exhibit a rising trend in it's third harmonic too.

To this end the two drivers are a good match for one another. Something like a 4th order around 1800hz or so, maybe lower if you don't mind pushing it some.

Keeping in mind your desire to 'time align' the design by altering the position of the tweeter in the Z axis, what I'd suggest you do is this,

Build a test box and mount the mid/bass at the top of the enclosure. Then place the tweeter on top of the enclosure, such that it can be moved forwards and backwards.

This will create all sorts of diffraction issues, so you will need to flush mount the tweeter in it's own baffle. The baffle should be the same width as the main enclosure and at least as high as the tweeter is wide.

You will then need to measure the drivers and design a filter for them, but rather then aiming for a phase coherent design (like you would if the tweeter was fixed in place on a flat baffle) you want to aim for getting the xover acoustic slopes as accurate as possible. That is as close as they can say to the 1800hz LRW 4th order I mentioned before.

After having done this you need to measure the drivers with the filter in place and measure for a reverse null. This should be done at your typical listening distance and typical listening height. You should then move the tweeter backwards by say 5mm and measure again. Keep moving and measuring until you've come up with the position that yields the best reverse null. Flip the polarity and you're done. The response should be flat, but could show some irregularities due to diffraction around the 'step' created by moving the tweeter backwards.

All that is left is to construct the final box.

Troel's Gravesen has done a study on tweeter and baffle mounting with regards to 'stepped baffles' and it can be found here.

There are of course more then one way to go about this, but if it was me, I'd start out by trying what I've suggested here.

[ramiro77]

That article is EXCELLENT!
I´ve read it and now i understand better how this works.
I´ll try some things from that article.
Thank you 5th element!

[twinter]

5th Element's reply is one of the most accurate and concise responses I've seen in quite a while on a complicated speaker design topic.

A couple of comments.
1) Generally, the tweeter offset is usually 1 1/4" to 1 1/2" when all is said and done.
2) Ramiro77, are you building a TWW configuration? You stated "....I decided to add another woofer and make them floorstanders." If so, time aligning two woofers with one tweeter is more complicated, as the second woofer's relative location will have to taken into account, along with lobing concerns at the cross region. You may want to consider a WTW configuration, which then brings up significant diffraction issues. Mounting the Seas 27TDFC tweeter in an MCM waveguide, like John did at Zaph Audio, would help both the time alignment and diffraction issues. Also, he used a 2.5 TWW design, as shown on his website, with the lower woofer just playing bass frequencies to avoid lobing between the tweeter and the lower woofer.

[5th element]

Oh, I just noticed this again, you're most welcome and good luck with your design!

The wider the baffle, with a TMM, and the high/steeper the slopes are from the tweeter to the primary mid/woofer, the less the 0.5 woofer and the tweeter will interact with one another. Twinter is quite right in saying that the 0.5 woofer and tweeter could have some odd interaction issues that you'd rather avoid.

If you were using 4" wide cabinets with small drivers, baffle step would start as high as 1500hz and as the 0.5 driver is generally rolled off with a shallow slope it would most certainly contribute to output in the main crossover range. How much it would contribute would be determined by the cross between the main mid/bass and the tweeter of course.

If you're using the 830875 you've probably got cabinets around 8" /20cm wide, where baffle steps starts around 700hz, by then the 0.5 woofer should already be down 6dB, and by 2000hz, should be around 20dB down in level. This should be significantly well attenuated to not cause too many issues with the tweeter, if steep slopes are used.

Another way to configure the 0.5 woofer is to drive it and it's inductor after the crossover that feeds into the primary mid/bass. Naturally you need to reconfigure the xover for the primary driver for the halving of the impedance. What this does is provide the gentle slope of the 0.5 woofers inductor to compensate for baffle-step. But as the 0.5 woofer is being driven through the primary drivers crossover too, it adds in the extra attenuation that the primary crossover provides, helping even more to keep the 0.5 drivers contribution out of the way during the main crossover region

Of course you really need to simulate these for the best effect.

The wave-guide idea is a nice one and if you could design everything perfectly, so that the physical offset provided by the wave-guide would perfectly time align the system, for your preferred listening axis, then I think you'd be on to a winner.

[CharlieLaub]

Originally Posted by twinter

5th Element's reply is one of the most accurate and concise responses I've seen in quite a while on a complicated speaker design topic.

Hmmm, 5th Element made some interesting points, but some that I would disagree with - see below:

Originally Posted by 5th element

While the computer delay is an option, it may not be possible or even desirable as part of the system. Indeed neither may be the addition of more power amps.

I can't see why "computer delay" would not be desirable, if executed properly. There are some interesting DSP crossover solutions like the MiniDSP that can provide any desired digital delay for loudspeaker crossover work. In my opinion, having one amplification channel per driver is optimum. Amplification is cheap, and coupling the driver directly to the output stage without passive components in between does have its advantages, as outlined in this article:
Active Vs. Passive Crossovers

Originally Posted by 5th element

Time aligning the drivers is a little more complicated then physically aligning the voice coils, it just doesn't work like that, at least not in the way you'd think.

According to what I've learnt the acoustic centre for a driver moves in position relative to the frequency it is reproducing. Therefore any 'aligning' be it physical or electrical is only relevant to that one frequency, or rather you could say frequency band because the acoustic centre will most likely remain within acceptable limits over certain frequency bands.

What you have written above is certainly something new to me. Can you expand on "what you've learnt". I do not agree that there is some frequency dependence to the acoustic center. I believe we are not talking about doppler distortion type issues, so I am not sure how frequency comes in to play - it's the "at rest" position that is of interest, since audio signals are AC and therefore the transducer's diaphragm is moving about this position.

Originally Posted by 5th element

The same is also true for the crossover employed. These add in electrical phase shifts that correspond to physically moving the tweeter through the Z axis and the amount to which this occurs is highly dependant on the target transfer function. That is to say the 'delay' imparted by the electrical crossover component is different for different xover frequencies and slope types.

OK, now you seem to have switched to talking about phase alignment, which is different from aligning acoustic center. Both will have an effect on the interference of the driver wavefronts in space, but in different ways so you cannot really lump them together. The acoustic center misalignment is frequency independent and this is why correcting it with digital delay (also frequency independent) is a good approach. Analog delay can be designed to have flat group delay up to some maximum frequency, but then the group delay falls back to zero. However, analog delay can still give very good results when it is designed and applied correctly.

But in the end of the paragraph above, you are absolutely correct in that each filter function that makes up a crossover has an associated phase response, as do the drivers themselves, and this can lead to phase leads or lags that cause undesired perturbations (peaks or nulls) in the frequency response. You need to model the entire system together to see how all the sound sources will combine at the listening position (and for power response everywhere).

Originally Posted by 5th element

The only true time alignment can come from digital filters that posses the ability to compensate for any phase shift across the entire audio band. I believe this is what the FIR proponents are getting at with regards to the 'point' or the holy grail of why you'd go digital in the first place. A bit like saying, if you've got a digital crossover and haven't implemented a phase linearising FIR filter then why bother at all? Of course there are many reasons why one should bother, but some would say you've missed half of the point by ignoring it.

Ignoring the digital FIR solution, you cannot simply place the tweeter in one position and be done with it.

Now I can't claim to know much about the particulars of implementing digital filters, but I have to disagree again about how one can properly align and cross over two drivers. There is not need to use esoteric digital signal processing as both can be done using analog circuits. The effect of driver phase can also be removed pretty easily in the analog domain. The only effect left to deal with is the phase response of the crossover itself.

Originally Posted by 5th element

For a start the wave fronts that the different drive units are producing can only be truly coherent at one listening height. This is obvious when you think about it because moving up and down alters the relative distance that the sound waves have to travel from the tweeter and the woofer. This is why properly done coaxial drivers attract such a large following, as their vertical off axis response mirrors their horizontal off axis. That is (if properly designed) remaining flat and coherent as you go off axis, with only high frequency, beaming related droop showing. Rather then a non coaxial system that has the two drivers going in and out of phase with one another as you change the vertical listening axis, creating dips and troughs all over the place.

As we're talking about a non coaxial driver, this means you have to design your system around one preferred listening height.

Next on the agenda you have to choose what xover point and slopes you'd like to use. The H1189 (27TDFC) is a capable device and can operate at low xover points without a problem, this is a very good thing.

Again, you hit the nail on the head above. The system has to be modeled including all sources of phase and amplitude deviation/response at the listening position (at a minimum). Then compensation and crossover networks must be designed to make the overall response of the system as flat and linear phase as possible.

So, I'd definitely like to see another post from 5th Element, since it seems like we have some different ideas about how to go about things. I imagine he has some perspective that he can lend to the thread, and this topic is one that I am currently studying. I am open to different ideas, even if I don't agree with them in their entirety.

-Charlie