DIY Flat Panel Speaker Love

[bradley.s]

A single near field sub is the smart way to go. I won't go the smart way because I'll get caught up trying to figure out how JBL does it. Then I'll look at the monstrosity I built knowing I could have done it cheaper and SMDH ;-)

[Kornbread]

Build the sub into a coffee table and set it beside the couch or just use a butt kicker.

[bradley.s]

One more note about Toole and subwoofers. He said there are two alternatives to multiple subs and the JBL proprietary algorithm. 1) multiple subs and patient manual adjustment. 2) one subwoofer placed in the near field of the listener. Just take a single sub and locate it close to your listening position. He said a guy put a sub in his ceiling directly above his seat and it worked. Nothing special about the ceiling placement other than it was close to the listener's seating position.

[bradley.s]

Oh, btw, I'll need to find a book/info source on DSP equalization so I can flatten the response in the array. Unequalized, the rise around 2kHz is terrible. I hear it and wince. My dog hears it and winces. If I can flatten it and keep the axis response tight like in the image above I'll be able to build the type of system Floyd Toole says will work.

[bradley.s]

My grasp of the CBT sound field is influenced by the graphical representations Don Keele made in his papers. I'm going to verbalize but I might get it wrong. Relative to a point source speaker, a CBT array supplies a more controlled vertical sound field which results in: 1) less volume decrease as you move away from the speaker, 2) reduced combing and lobing at higher frequencies, 3) a more focused polar response, and 4) less ceiling reflection in small rooms. The CBT Keele designed also has wide directivity, however, Floyd Toole said that's just basic good speaker design. It isn't unique to CBTs.

Keele's straight array is delayed to approximate the physical shape of his curved array. His straight array is delayed at the top not the bottom. However, his physically curved array is not delayed with DSP. It's delayed because the top of the speaker is physically further away from you than the bottom (see attached photo.)

One of the reasons I tried to do a CBT with DML/MAP is to take advantage of the wide directivity that seems to be inherent in an excited panel. I wanted wide directivity because I was influenced by Floyd Toole's book Sound reproduction: the acoustics and psychoacoustics of loudspeakers and rooms. I'm trying to create the type of sound field Floyd Toole says is preferred through double blind testing. My ear hears what you see in the image I posted above. As I turn the array the high frequencies remain constant but the lower frequencies fall off as I approach 90 degrees. But they don't fall off very much. In my case, I'm ok with that because I planned to make a different speaker to handle lower frequencies.

Another thing Toole covered in his book was movie dialog for home theater. The center channel is huge for intelligibility. He said they got improved dialog intelligibility by turning down the right and left speakers relative to the center channel. They got the best intelligibility by turning the side channels completely off. However, they lost overall movie sound enjoyment quality. Special effects and soundtrack suffered. DML excited panels seem to have better dialog intelligibility because the reflections aren't as coherent as the direct field. For that reason I want to build a MAP center channel to compliment the MAP arrays. I'm guessing that setup will let me have improved movie dialog intelligibility without turning down the side channels. The nature of the DML will compensate. I'm also going to make the center channel into a CBT, except in the vertical plane. It will also be a full virtual 180 degree CBT instead of a 36 degree arc. According to Keele et al, 180 degrees focuses the center channel source right in the center. However, a 180 degree CBT built with normal speakers creates distracting side wall reflections. I'm guessing the MAPs won't run into that problem because their reflections aren't as coherent so our brains ignore the reflections. I'm just not sure about how to design their geometry yet: long and narrow, wide and tall?

In the end I'll have two CBT arrays similar to the one I just built located on the ceiling (upside down) at 60 degrees off listening axis. Then I'll place the center channel CBT below the TV screen. I haven't figured out how to do the subwoofers yet. Toole said JBL uses multiple subs and a proprietary DSP algorithm. I still have to hunt that information down and figure out how they did it before I build the subs.

Toole also said something else in his book. He suggested the future of speakers was active DSP instead of passive crossovers, and that each speaker driver would be driven by an independent amplifier. He said the inexpensive amplifiers of today are ruler flat. Expensive amplifiers are obsolete today. The array I built uses cheap 100 watt Texas Instruments chip amps. They were only $3.50 each. They're powered by an inexpensive Meanwell 300 watt switching power supply, $25. The 300 watt Meanwell power supply will power both right and left channels as well as the center channel and have plenty of power to spare.

The only semi-expensive part of the system is the DSP. But that's only because I'm creating a pretty complex array of speakers. I'd bet someone could build a basic 3-channel MAP (Right, Left, Center) that would sound as good as the complex system I'm building. In a double blind test a listener probably wouldn't be able to tell the difference between a standard four exciter MAP and a twelve exciter CBT MAP. It might be a case where a microphone could tell the difference but a human brain couldn't. Although, I think you'd still want to use one Mini-DSP to flatten the frequency response of the MAPs. Each MAP would have four exciters with one inexpensive TI chip amp power all four. Then that MAP of four exciters with one amp would be controlled by one of the four outlets on the Mini-DSP. And I'd place the four exciters on the MAP the same way Tectonics places theirs. I'd even use Tectonics' geometry and panel material. The one advantage a CBT array might have is low frequencies. My MAP CBT is five feet tall. I could make it seven feet tall and improve low frequencies. Especially using Keel's ground-plane design like in the image below (placing a CBT on a reflective surface like a hard floor or ceiling.)

[Kornbread]

Just thinking out loud. I though the purpose of delaying the channels was so all sound arrives at the listening position at the same time?

One exciter, w/spline, foamboard, w/small tabs holding panel in place.

[bradley.s]

I tried to measure speaker directivity. Placed mic six feet from the speaker at 28 inches elevation. I kept the mic in the same location but rotated the speaker for each measurement: on-axis, 30 degrees, 45 degrees, 60 degrees, and 90 degrees. Smoothed to 1/3.

[bradley.s]

I used the mic to make sure each of the five array sections were attenuated by -3dB relative to the previous section. Then redid the impulse response. It's easier to see the five sections now. Though section three is lower than I expected.

[bradley.s]

Interesting. I noticed the initial humps in the response correspond with the delays set in each of the five sections. However, the fifth section is -12db so it barely shows up. In fact, it might be less than -12db. I haven't confirmed the speaker is actually -12db than the first section of the array.

[bradley.s]

Wait, those first images were measuring the back of the speaker. That's probably why it dipped to negative 100 first. These photos are measured from the front of the speaker.

[bradley.s]

This is what I get when I select the Step Response checkbox. First photo with red line is zoomed in to show -500u to 10.0m. Second photo with green is zoomed out to show the whole line.

[Kornbread]

Make it look like this.

[bradley.s]

Here's the graph under the impulse tab. I don't know what I'm looking at or if I measured this correctly. I placed the mic one meter from the speaker at 28 inches in elevation.

[bradley.s]

A paper somewhat related: Implementation of a Dipole Constant Directivity Circular-Arc Array. Array is physically curved but it's dipole. In contrast, the MAP array I'm trying to learn how to measure is a physically straight delay-curved attempt at a CBT. I assume a physically curved array is better. On the other hand, a physically straight array might gain from near wall placement by angling the axis toward or away from the wall. With a curved array either the bottom or top would need to be moved away from the wall. I don't know, I'm tired and burned out for the day.

http://faculty.tru.ca/rtaylor/publications/dipole_cbt_implementation.pdf

[bradley.s]

I don't know how to test impulse response. I'll have to learn how to do it.