Odd Order Butterworth's and MTM Lobing errors

[spintonik]

Hi,

I am working on MTM design, where I want to get rid of lobing errors around XO point, in order to make a speaker with no sweet spot. For that purpose I use Seas DXT Tweeter ( THE ART OF SOUND PERFECTION BY SEAS - H1499-06 27TBCD/GB-DXT ) which has almost identical response all the way to 60 degrees off axis. Woofers are 6" Scan-Speak.

I've been reading many threads about this, few books and of course Joseph D'Appolitos original white paper. Most of them agree, that type of XO chosen has impact to lobing errors. This relates to phase difference XO introduces, in combination to phase difference which comes from distances sound needs to travel from different speakers. For instance, Even order LR (for instance 2 order LR on both) should be quite directional, while odd order Butterworth should not.

However, what different sources do not seem to agree, is what should be the relation between XO frequency wavelength and distance between tweeter and midwoofer.

D'Appolito white paper has example graphs where distance between tweeter and midwoofer is 3/4 of XO frequency. Even order LR is very directional, while odd order butterworth is not. I've seen also many other figures, but decided to believe the Man himself...

With Seas DXT 3rd order high pass makes sense, so I've been working with 2/3 Butterworth design, with 14cm distance between my tweeter and woofers. With 3/4 wavelength this relates to 1842 Hz XO, so that's what I squeezed out of my components - In simulation I think I did reach 1844 Hz. Close enough?

All well in theory, yes?

In practice, life is not that easy. Off axis response is always attenuated around XO point. So I am getting directional design despite my intentions to design something else. I've done various versions, tried also 2/1 and 2/2 Butterworth, every time the same. All along the another axis is ok.

(MTM impacts only to the direction where speakers are located, for instance, if you have Vertical MTM (speakers on top of each other), XO modifications does not impact horisontal directivity)

To demonstrate the case, I did quick / dirty measurement. In this case I have 2/1 Butterworth, with XO point around 1,6kHz or so, hard to say. This is not optimal design, just something I had spare components for.

Light Yellow is On Axis response
Light Purple is about 50 degrees off axis vertical
Green is about 50 degrees off axis horisontal.

Of axis is a bit attenuated as I was too lazy to start measuring exact distances, On axis is measured from 50cm, off axis cases are closer to one meter. Off-Axis frequencies below 1kHz are attenuated because woofer is directional, but as soon as DXT kicks in right after 1kHz (1st order high pass, remember), you can see that all is ok - Off axis horisontal.

MTM issues 2-1 butterworth | Flickr - Photo Sharing!


So does someone know for sure, what relation between speaker distances and XO Wavelength would give me non-lobing design, if I am using 2/3 Butterworth?

Thank You in advance!
All the best,
Mikko

[speaker dave]

Hi Mikko,

You have a number of factors here and they all add together for the final system performance. They are: MTM spacing, driver directivity and crossover phasing.

The driver directivity is evident in your green curve. It is measured horizontally 50 degrees off. We assume the crossover is having little effect as we move sideways, so most of the drop is simply due to the drivers. In the end anything you do will have this directivity as a minimum.

The second issue is spacing and the best way to think of that is how many wavelengths apart the mids are at the crossover point. (Above the crossover the tweeter dominates, below the crossover the wavelengths are longer and the spacing is less of an issue). You will get vertical lobing purely from the spacing, even if your drivers have no directivity. If the drivers were 1/2 wave apart you would have complete cancelation at 90 degrees up or down. If you could get them less than 1/2 wave apart you would have no lobing. In practice this is very hard to achieve. Even with a tiny neo tweeter the mids will be over 1 frame diameter apart. The only hope is to get the units tight together and keep the crossover frequency low. (and don't worry about having a null at 90 degrees).

The third issue is crossover order and phasing. Good that you have read the D'Appolito paper. Everyone assumes that Joe "invented" the array when in fact the paper is about the relative phasing of outer to inner units. Joe found that if the drivers have a consistant 90 degrees between units then the lobing nulls are minimized. How to achieve that? Some will tell you that Butterworth networks will inherently have a 90 degree transistion but that totally ignores the added phase shift from the relative depth of the acoustic centers of the 3 units. In truth it is a crap shoot and you must measure phase with potential networks and see if you can get the 90 degree difference with any particular combination.

My advice? Squeeze the drivers tight together, use the lowest practical crossover, either measure phase to see if you can get 90 degree summing, or just experiment with various orders while watching both on and vertical off axis. Also, be realistic. If you can get +-30 or even +- 20 degrees with no lobes and minimal response variation then you will have covered a broad listening window. Beyond that angle your driver's inherent directivity will degrade response anyhow.

Regards,
David S.

[Soundminded]

This strikes me as a horrible tweeter. Its on axis FR has a rolloff starting just above 6 khz and is down around 7 db at around 15 khz. Its 60 degree off axis response is even worse starting above 2 khz and down around 17 db at 15 khz. That's around 10 db at 15 khz 60 degrees. By way of comparison, the AR3a 3/4" tweeter, the second best I know of is down 5 db at 15 khz 60 degrees off axis. The tweeter also has a very nasty undamped resonant peak just above the audible range. This could cause problems in the audible range through IM products if it isn't suppressed. Both issues should be dealt with by active and passive equalization with a compensating boost below 20 khz and a very sharp cutoff above it.

All multiway loudspeaker systems have interference patterns in their crossover region unless they meet one of the two following criteria. 1, their crossover slopes are infinite and two dirvers never operate at the same frequency. 2. They propagate their sound from the same point in space in precise synchrony in time which is compensated for if they are a different distance from the listener (Physical displacement and electrical to mechanical propagation delay must both be taken into account.) In all multiway designs I know of the interference patterns will occur at different points and to different degrees depending on their geometric relationship to each other, the point in space you are listening at, and the frequency. This exhibits itself in changes in steady state FR. The effect will range from a +3 db rise where there is perfect phase coherence (delta theta=0 degrees) to an infinite dip where there is complete cancellation (delta theta = 180 degrees.) Fortunately our ears are not point receivers so the effect is to a degree mitigated.

MTM configurations IMO attempt to create interference patterns that are symetrical above and below the tweeter so that they will cancel each other out. But that is likely only in the vertical plane at the same height as the tweeter. The Linkwitz Riley crossover has a clever idea. It does not negate what I said but its trick is that the worst interference effect is located in places in the room where the listener is not likely to be located. Unlike a Butterworth filter it will not exhibit the flattest combined electrical amplitude response.

[CharlieLaub]

The lobing in an MTM has a lot to do with the spacing of the "M" drivers, as well as the crossover type and order. The greater the center-to-center distance of the M's, the worse the lobing gets at a given frequency. It's best to make the MTM as tightly grouped as possible, and cross over as low as possible to the tweeter. I like to cross over at 1.5k Hz, and with special drivers (not a dome tweeter though) you can cross over as low as 750 Hz.

in an MT system, odd order Butterworth crossovers have lobing but flat power response. In an MTM, the combined output from the two M drivers almost eliminates lobing and still has flat power response, with a Butterworth 3rd order crossover.

-Charlie

[spintonik]

Hi Dave,

Thanks for your views. I did one 2/3 XO design after my post with XO point at 1,6kHz, which is about as low as you can go with that tweeter. In listening tests lobing issue is there, but it does not bother too much. I'll develop that design further and see if I can get speakers even more close, although 14cm is pretty extreme with these midwoofers.

All the best,
Mikko

[spintonik]

Originally Posted by Soundminded

This strikes me as a horrible tweeter. Its on axis FR has a rolloff starting just above 6 khz and is down around 7 db at around 15 khz. Its 60 degree off axis response is even worse starting above 2 khz and down around 17 db at 15 khz. That's around 10 db at 15 khz 60 degrees. By way of comparison, the AR3a 3/4" tweeter, the second best I know of is down 5 db at 15 khz 60 degrees off axis. The tweeter also has a very nasty undamped resonant peak just above the audible range. This could cause problems in the audible range through IM products if it isn't suppressed. Both issues should be dealt with by active and passive equalization with a compensating boost below 20 khz and a very sharp cutoff above it.

Hi Soundminded,

thanks for your comments and views. DXT lense is a special beast, that needs to be heard before you can appreciate it. But please check DXT basics from The DXT® Tweeter - a waveguide lens for dome tweeters and try to have a listen when you have a chance. Ugly on-axis response is side effect from lense, but like I mentioned this is pretty easy to fix with 1st order high pass at 8kHz.

Very aggressive peak above 20kz is pretty common to Seas aluminium tweeters - many other manufacturers stop their spec sheets at 20kHz with their aluminium tweeters, I wonder why? ;-)

Finally, point in this DXT tweeter is the way 30 degrees off axis curve follows on-axis curve identically all the way from 1kHz to 20kHz - power response is pretty and it sounds pretty non-loudspeaker-like. I admit, measuring at 50 or 60 degrees is overkill.

Finally, thanks for LR tip, I'll have a look at that too.

-Mikko

[spintonik]

Originally Posted by CharlieLaub

in an MT system, odd order Butterworth crossovers have lobing but flat power response. In an MTM, the combined output from the two M drivers almost eliminates lobing and still has flat power response, with a Butterworth 3rd order crossover.

Charlie, thanks for your comments.

One question, do you mean 3rd order for both M and T?

Again, I'll have a look as soon as I source some suitable coils

-Mikko

[planet10]

Originally Posted by CharlieLaub

The lobing in an MTM has a lot to do with the spacing of the "M" drivers, as well as the crossover type and order. The greater the center-to-center distance of the M's, the worse the lobing gets at a given frequency. It's best to make the MTM as tightly grouped as possible, and cross over as low as possible to the tweeter. I like to cross over at 1.5k Hz, and with special drivers (not a dome tweeter though) you can cross over as low as 750 Hz.

I went even further, XOing at 300-400 Hz, i was able to get away with 1st order XO. C-C is less than 1/4 wl at XO.

EL166 MTM ML-TL

dave

[bbggg]

Spintonik, your best bet for a smooth XO transition is a larger than usual tweeter with a small mounting plate. This way you may cross lower and pack the drivers close together. The problem is that larger tweeters have a drooping top octave response because of apex-to-rim cancellation. This can be ignored if you are older and have lost hearing acuity in that frequency range. Most men over 40 cannot hear much above 15 kHz, much as they are willing to deny it. Listening slightly off tweeter axis (say 15 deg. lateral) also helps in this respect.

[boris81]

Xdir is a free program that can simulate the directivity pattern of various driver combinations. It should help you get an idea of driver spacing.

Also this topic might be of interest.
Power5