Driver compliment:
TWEETERS: Eminence APT-200 - 90-degree dispersion horn loaded compression driver.
MIDRANGES: B&C 8PE21 - 8-inch midrange
WOOFERS: Eminence Kappa-Lite 3015LF - 15-inch woofer
PASSIVE RADIATORS: Dayton DVC385-PR - 15-inch passive radiator
TOTAL DRIVER COST: approximately $1320

Electronic Equipment:
CROSSOVER: Behringer DCX-2496
AMPLIFIERS: Hafler XL-280 (tweeters), Yamaha M-65 (midranges), Carver PM-600 (woofers).
PRE-AMPLFIER: Adcom GFP-555

Frequency measurements were performed using a system consisting of a laptop computer, Speaker Workshop software, an M-Audio Firewire 410 D/A+A/D interface and a Behringer ECM8000 microphone. In-room frequency response measurements, carried out in a large, Lshaped room measuring approximately 30'x20' with a vaulted ceiling, were used to adjust the crossover settings. No calibration for microphone or amplifier frequency response was performed. Typically, multiple distances and data acquisition settings were used when taking measurements in an attempt to identify room interactions in the signal. For all frequency responses shown, 1/3-octave smoothing was applied to the raw frequency data. No input level calibration was done so only relative SPL amplitudes are shown in the FR plots.

Tweeter:
Some of the design goals for this system include high-output and large dynamic range. While I would have preferred a dome tweeter, it's not easy to find one with sensitivity greater than 95 db/W. I'm somewhat skeptical of ribbons and it seems that most good quality ribbons are ridiculously expensive. In the end, I decided to experiment with the Eminence APT-200, a midpriced horn tweeter for professional systems.

There is very little information available about the frequency response of the APT driver or the pre-mounted APT-200 driver/lens combo except some dubious info from Eminence. I mounted the tweeter in the 24-inch wide open-baffle with the protruding lobes of the horn on each side. The frequency response I measured is shown in the graph below. The blue trace indicates the onaxis response and the red trace the response 30-degrees off axis.

The frequency response is definitely not flat and droops with increasing frequency, especially above about 7kHz. There is little output above about 17kHz although, since this is at or past the upper edge of hearing for most people, this is not of large concern. On the positive side of things, the off-axis response is nearly identical to the on-axis response. This might mean that, if the FR irregularities can be ameliorated, the driver might still work well.

Frequency Measurements:

Midrange:
I made an attempt to measure the B&C driver on-axis response in the free-field at about 0.6m distance. I didn't mount it in a baffle, since I didn't have one that would be adequate for that purpose. Take the following free-field response with a grain of salt as it was done in my living room and doesn't use any fancy measurement techniques.

When I mount this in my open-baffle, I get something much nicer (see plot below)! It now measures just about what the manufacturer advertises, with a smooth frequency response between about 400Hz and 4kHz and a well behaved roll-off in the breakup region. The response is about ±2dB except maybe for the dip at 850 Hz, which might be an artifact. This is really quite good for a pro driver. Remember, the advertised sensitivity of this driver is 98dB/W@1m!

Woofers and Passive Radiators:
Several questions were floated to the DIY community via the Parts Express and Madisound forums regarding the low frequency system design. Could an isobaric design be used as high as 300Hz-500Hz? Is it important for frequencies as high as 300 Hz to be reproduced in stereo or could two mono sources be used instead (e.g. there is a woofer on each side however the signal fed to them is mixed to mono). In other words, will switching from stereo to mono at frequencies below about 300Hz be audible?

The isobaric setup was of interest to me because the box size can be reduced by one-half. I can deal with 4-5 cubic feet but 8-10 cubic feet is just way too big unless it is built into an attic or something. After some careful study with UniBox I decided that I could get reasonable output to a little below 30Hz with the KappaLite 3015LF drivers using dual passive radiators. I also decided to use a parallel wiring connection to increase the voltage sensitivity. After searching around for a way to add the correct amount of mass, I discovered that I could buy cast lead 'pucks' having holes in the center that weighed 1.25 lbs (567 grams) each from McMaster-Carr (part number 90385K22). These were firmly attached to the PR using the handy, built-in 1/4-20 threaded rod that sticks out of the back of the cone for that purpose.

Unibox predicts a gently sloping response before the final roll-off occurs around 27Hz. The predicted frequency response for a 2.83V input and the cone excursions at high input power (51.5V or 250W to each driver) are shown in the graphs below.

The full-power cone excursion graphs show that even at 250W the travel is well within the maximum limits (±17mm for the Eminence driver and ±22mm for the passive radiators) and below the Xmax of the KappaLite drivers (9.5mm). The SPL is almost 120dB at 200 Hz at this power setting! I did not perform frequency response measurements on either the raw drivers or the finished cabinet so hopefully I am not too far off target. While testing for leaks, I fed a 17Hz sine wave to the bass cabinets and the resulting output was pretty weird. There was enough sound power at that frequency to modulate my voice and I suspect that I may have achieved an even lower tuning than what is shown in these graphs.

Cabinet Construction
To save time and effort, a minimum of effort was put in to esthetics and woodcutting. Construction is largely out of one-half-inch thick MDF boards with bracing from 2x2 furring strips in the bass box. The midrange and tweeter are mounted in an open-back cabinet. I just slapped some sides (12"x48" particle board shelving) on to a 24"x48" piece of MDF into which I had cut holes for the drivers. After studying Siegfried Linkwitz's design models for the Phoenix, I knew that I needed at rather wide baffle. Not much effort went in to its construction, since I only intended to investigate the acoustic signature of open-back driver loading. I must admit it is uncolored and the sound field seems to be more evenly located throughout the room. I'm not sure how to make this type of setup attractive and reach 300Hz with this driver so I may just use ol' reliable - a stuffing damped, closed box loading - in the final design.

Since I wanted to use the woofer section as high in frequency as 300Hz, I felt I needed to have the outer woofer of the isobaric configuration facing outwards. I designed and constructed a small chamber between the front panel and a second panel located about 8 inches behind the front panel that is bounded and supported by two vertical braces situated about 15-inches apart (see pic, below).

this area will become
   the isobaric chamber

vertical brace

side chamber, connected to
the rear compartment

The vertical braces define the sides of the isobaric chamber and separate it from two small side chambers that are connected to the rear chamber via holes in the panel. The two woofers of the isobaric configuration are mounted back to back and are offset so that the magnets are next to each other. The KappaLite drivers use neodymium magnets which are relatively small in diameter, allowing the driver offset to by only about 7-8 inches and allowing them to barely fit in to 24-inches. Below is a view into the rear chamber after the rear firing woofer and the passive radiators have been installed. Visible are the lead weights that have been attached to the threaded rods of the passive radiators using wing nuts and the openings to the two small side chambers in the front.

lead 'puck' attached to PR
threaded rod with wing nut

connection to 'side chambers'
that flank the isobaric chamber

View of rear chamber showing rear firing woofer and passive radiators.

System Frequency Response
I ended up using crossover frequencies of about 3.5kHz between the midrange and tweeter (with a slight overlap) and 250Hz between the midrange and woofers with 24db/octave Butterworth filters. I tried to keep midrange frequencies out of the woofers as much as possible but, because the B&C midrange driver frequency response starts to roll off below about 350Hz, a 3dB boost is used to flatten the overall response around the crossover frequency.

After repeated tweaking and measuring to smooth the response, I was able to obtain a frequency response for the open-back panel that is shown in the graph below. This is about ±2.5dB from 400Hz-15kHz which is pretty flat although not perfect. With these drivers it is difficult to get the FR to be much flatter and I am satisfied with the sound. I didn't measure the response of the woofer section but just matched the gain by ear.

Future Directions:
I can definitely say that I am happy with the concept prototype. The next step is to decide how to rearrange the woofer cabinets and integrate the midrange in the final revision so that the result is a little more visually appealing.

I have decided to switch out the Eminence APT tweeters for a slot compression tweeter from Beyam (CP21F). I am very happy with the B&C midrange. The Eminence KappaLite drivers are also good and have a sufficiently smooth frequency response to be used as high as 500Hz (or above). I just didn't like how the lower midrange sounded through these drivers (although there was absolutely no damping/stuffing material in the cabinet) so I'm considering a 4-way with two 10-inch drivers working between 90Hz and 450Hz. I will put the KappaLites in a smaller (less than 4 cubic feet) sub-enclosure. Along with the mass-tuned passive radiators. this will cover 27Hz-90Hz and be capable of a 120dB SPL/1m over that frequency range.

Lastly, if I didn't have the versatility of the Behringer crossover at my disposal this project would have been a challenge, especially with the APT tweeter. I can 100% reccommened this tool to anyone who is interested in using active crossovers.