Second Try
Second verse, same as the first, but with a sane(er) crossover point (i.e., another educational exercise).
The same satellites built using:
A pair of 0.25 cubic foot enclosures from Parts Express for about $120.
A pair of Tang Band W5-704D mid-woofers, about $25 each from Parts Express.
A pair of ND20FB-4 tweeters, about $8.00 each from Parts Express.
Passive crossovers, designed using in-box measurements with SounEasy.
The speaker terminals and crossover parts were about $30 each from Parts Express.
All drivers were flush-mounted.
This time with an Fc of 3800 Hz. Still too high, but much more usable. I just didn't have the experience manipulating crossover filter elements to get the crossover point where I wanted it without pushing the phase response of the drivers apart.
I'm attaching what I originally wrote about the project at the bottom of the page. It's surprising how I see the project in hindsight, looking back on it 1 1/2 years later. When I wrote my notes, I was fired up. My first modest effort was making sound! (Not great sound, but usable.) I felt great about it. Now, I don't think it would take me more than a few days to decide to change it. The results were still problematic (some lobing still present), and the sound was actually very bright.
The notes I wrote on the project are generally correct, I wouldn't change my driver mounting methods and basic construction techniques (even the hideously ugly, free-form, glued to the walls XO). If you look, the inductors, as disorganized they appear, are mounted as far apart as I could place them and oriented to minimize electromagnetic coupling. All connections in the speakers (all wire connections, and, of course, the XOs) are soldered for longevity. No push-on or screw terminals anywhere (the lug in the picture below was cut off for final assembly, the XO input wires were soldered directly to the binding posts). The 45° relief on the back side of the baffle is required to let the sound generated by the back surface of the cone pass into the enclosure. Without the relief, the spider shelf will almost completely block the path to the enclosure.
However,
I designed the crossovers with no baffle step compensation thinking the roll-off in frequency response below 600 Hz wouldn't occur because the speakers are placed next to a TV and near the walls (see the Third Try pictures). This was a big boo-boo. When I rebuilt them the third time, I designed the XOs for 5 dB of baffle-step compensation and it worked great.
I noticed from the frequency response plots I generated with SoundEasy that I was not tailing my tweeter response correctly. I should have followed the natural roll-off of the tweeter, instead of trying to follow the response of my test amp with an ADC/DAC converter pair in its preamp stage. This screws up the phase response of the tweeter used in the model (it reduces the accuracy of the Hilbert Transform used to extract the phase response), and the predicted frequency response of the XO model was not accurate. Second boo-boo.
Also, I was not using a calibration file for my Behringer ECM8000 measurement mic. Third big boo-boo. My mic has a big hump in its frequency response (about +6 dB centered @ 11 kHz, starting @ 8 kHz and extending to about 14 kHz). I tried to ignore this, but at the time I thought it was the tweeter's actual response and I know it influenced my XO design. Send your measurement mic to Kim Girardin (he'll measure its response to 45 kHz), use a test amp with decent frequency response (less than ± 0.5 dB, from 20 Hz to 50 kHz), and you can see how to correctly tail the tweeter's response to 50 kHz.
I apologize for not really finishing this design with the W5-704 and ND20s. It's really not worth saving. I wouldn't hesitate to use the W5-704 in the Dayton box, but I wouldn't use the ND20, Id use the ND28 and a crossover point no lower than 2.5 kHz. This kills the lobing problems (important for a near field monitor), and puts more of the treble load on the tweeter, where it belongs. The baffle layout is OK, but if I did it again, I'd try mirrored pairs, with the tweeters mounted at about one o-clock, just as close to the woofers as in the picture above.
Jay R. Taylor
March, 2010