Spring 2010

DIY Open Baffle With Widerange Drivers
A.K.A. The third ugliest amazing speaker you'll ever see.
Article By Jeff Poth

Difficulty Level

  This is the best speaker I've built at the time of this writing (that is mostly January, 2010). The speaker is almost two years old but the woofer... much older. Older than some of you reading this! Let us dive right in: The drivers are the JBL 123A (or, preferred, 123A-3/2213) and Fostex FF85k (preferred FF85KeN).

OK, Out Of The Pool!
Dirty trick, eh? We'll do some backstory. I have to mess with my writing format to keep you fickle Internet clicky types involved! About the time of this speaker's concept, I had just moved cross-country. All my tools and hi-fi gear were living with my wife, 3,000 miles away, and I needed a speaker, both as a project to keep me busy (and hence out of trouble!) and as a way to listen to music for a few months until my beautiful wife and daughters joined me in semitropical paradise. It had to be easy to build, as I had a very limited number of items to work with.

The Woofer
At and about this time, I was reading Troels Gravesen's site when I came across his pages relating to the L100. For those of you who don't know, the JBL L100 is one of the top selling speakers of all time, and was insanely popular in the 70s and 80s. A number of versions exist, up through the L100T3, which is a much superior design to the original L-100, but also is a totally different speaker. The original L-100 was a set of superb drivers in a very poor overall design. What caught my eye was the 123A 12-inch woofer. Troels had done some measurements on a low-diffraction baffle, and got impressive smooth, extended bandwidth from this driver. As I didn't have my tools, I needed to keep the project simple, and a smooth, extended bandwidth tends to allow a simple crossover which can be tuned much more easily. So began the hunt for 123A. I got a pair each of 123A and 2213, the 123A-3 equivalent. JBL often had both pro and consumer part numbers for the same driver in these days.

This woofer has a "Lansaplas" coated, ribbed, curvilinear cone, a high-ish Qes, and a low Fs. It also has a high quality AlNiCo magnet structure and JBL build quality. The combination of a curvilinear cone with low inductance and a lot of damping material not only gives it a fairly well damped, extended top end, but also a well extended low end- in the right box.

These drivers are available; the L-100 was the top selling speaker in history, in its day. You'll just have to hunt for a nice pair. A known good pair can be usually had for $200, you can find cheaper if you really look. $200 is a perfectly fair price for the level of quality these drivers have, both sonic and build. If you come up with L100s for the woofers, please try not to destroy good examples of the speaker- if the cabs are in good shape, you can offer them for resale, whole, and find another set of woofers. Also, be aware that if the cabs are NOT in good shape, the LE-5 midrange and tweeters are quality drivers too.

The 2213
The 2213 is a modest upgrade with a slightly improved motor design. The frame is nicer on 123A however, so if you're looks conscious of the back end... 123A. Sonically, 2213 is just a little cleaner and "bigger" sounding.

Note: These AlNiCo drivers have the potential for being de-magnetized. This leads to a higher Qts and less efficiency. You will need to be able to test these to confirm. I used a Dayton woofer Tester 3 for this task. Proper values should be approximately Fs: 25Hz Le: 0.6mH Qts .5

The Midtweeter
I wanted to do a simple crossover for this project, as mentioned above. This requires an extremely robust tweeter. The Fostex FF85k was the obvious choice and available from Madisound. I had previously used the FF225k, which is a widerange driver (not really full range, but covers 100 Hz to 10 kHz very nicely). Its little brother is a very slick 3" extended treble Fostex's cool UDR surround technology with a copper sleeve over the pole. This is a 3" 'fullrange' that can be thought of, for the purposes of this project, as a tweeter that can play down into the midrange cleanly. The awful sound of so many 1" dome speakers in your typical "hi-fi" 6.5" midwoofer and 1" dome tweeter speakers is due to the 1" running out of steam. This driver has about nine times the surface area, and thus needs to move much less. For the extreme highs, it has a direct coupled aluminum dome dustcap, which acts much like a more common whizzer cone construction. This trick has been done since the 1950s and JBL's legendary LE8T fullrange (amongst others, E130 being particularly popular amongst guitar players).

For those of you who want a significant upgrade, the fully modified FF85KeN from planet_10 hifi is a big improvement, with better clarity, dynamics, and a more complete disappearing act. This is one amazing driver, taking the FF85k a significant notch above it's already extremely impressive sound quality. More information on the modified driver at this link. 

Don't Make Me Box Your Ears!
Having found a woofer that was compelling and a lovely midtweeter to match, the enclosure design needed to be addressed. In this case, an 'unclosure' was the better choice. The 123A is a monster woofer, with prodigious low end output within its modest 7mm Xmax. A box for this driver wants to be 5 cubic feet or larger, the size of a large mini-fridge. An open baffle of the proper design, however, will play down to the 30s and 40s, with careful design. Open baffles are not known for having deep bass. This design relies upon floor reinforcement to buy some extra juice down low, without going to a much larger baffle.

The Cowpushers
Time spent playing with Xlbaffle, thanks to Thorsten Loesch for designing and Dave Dlugos for hosting, gave me a size approximation. Using acoustic theory and working with Tolvan's The Edge, I came up with the angled edge design. As it turned out, it worked out very well. The speakers are, necessarily, tilted backwards to aim the drivers towards the listening position. The amount of tilt is wholly dependent upon the listener position. You want the plane formed by the baffle tilted back slightly more than directly facing you when you look down at them from your listening seat. If you were to draw a line from the dustcap of each driver to your ears, you'd want the lines to be equal lengths. This is important, as a fairly high crossover point was utilized in this design. Adjust to taste but don't deviate much. This is also critical to getting the sound dispersed for use as background music, as if they are too far forward, the soundfield will be largely limited to the bottom half of the room.

Crossover
Crossovers are the hardest part of any speaker design, and this one was no exception. I got good results with a compensated series crossover at 500 Hz, but was still not quite happy.  After a lot of experimentation, I came up with a combination of acoustic and electric crossover that produced an exceedingly flat response from 30 Hz to 20 kHz. By exceedingly flat, I mean in room over a 60 degree window, +/- 3dB, and only deviates from this within a 90 degree window vis a vis >10 kHz rolloff. This speaker interacts very favorably with a real world room, as it's very consistent- extreme off axis angles are controlled via dipole cancellation and acoustic absorption.

The great part here is the crossover is very simple. A first order filter is used on the 3-inch, 10uF. Use a good quality cap. The 12" does not work with a first order filter via series inductance; its small inductance interferes with both efficacy of a series inductance, as well as creating phase wrapping that cannot be corrected for with electrical solutions. My solution was an acoustic one. A felt (adhesive backed F13, part number 8745K53 at McMaster) pad was implemented on a metal grille from Parts Express that created a low-pass acoustic filter. This allowed excellent phase tracking at the crossover point and an ideal textbook first order filter behavior (read: transient response is superb).

One lump remains. A 200Hz lump, to be specific. This is the result of the baffle reinforcing some frequencies, as well as a reduction in capacitor efficacy at Fs on the Fostex. This is the hardest part of the filter, fortunately, it's not that hard. You WILL want to spend money for a very high quality inductor and capacitors here. I used Solen 10 AWG aircore coils and a composite (several values in parallel) capacitor bank to get what I needed. I recommend avoiding any cored inductors here if possible, and keeping the DCR low. It's a big part, an expensive part, and a part that you don't want to skimp on. 12 AWG would be good, 14 may be sufficient, but absolutely no smaller if you're using aircore. If you're using cored, 14 is okay, but I recommend aircore. Remember, this part is the one that's passing the first few, most energy dense, octaves through it. The capacitors are also quality important, but are likely less of an issue here. Several quality poly capacitors in parallel, 100V or more will do the trick. The resistor can be a generic 10W sandcast type; it is not going to be passing significant current apart from at the notch frequency. Sandcast resistors do have some significant inductance, but this isn't really an issue here. When using these resistors for high frequency circuits take care.

Directionality
Open baffles are a very different case than "box" speakers, in a variety of ways. Some of their strengths are:

These things do not come without their compromises, however:

There is more to it than above, and the curious reader would be wise to spend some time reading about dipole behavior, most specifically the studies at Linkwitz Lab and Music & Design. A speaker hobbyist who reads and understands the content at those two sites will be fairly well-versed in many of the issues in loudspeaker design.

The critical component here is "Directionality". While this speaker does not qualify as "CD", it is significantly more consistent with its dispersion pattern than many speakers, and this was one of the major design goals, within some other constraints. The nature of an open baffle speaker is that as frequency decreases, the front and rear radiation from the speaker driver is more and more able to cancel. The result is a figure 8 pattern, with the "waist" of the 8 representing the driver on the baffle, and the round portions representing forward and rearward radiation. The frequency specific behavior of this effect will depend largely upon the size and shape of the baffle. Smaller baffles tend to be more consistent with their off-axis losses (good) but lose more bass (bad). It's a balancing act.

In addition to dipole directionality, the driver itself tends to be directional. The larger a driver is, the lower in frequency it will begin to "Beam". In the case of our 12" woofer, we have a beamwidth that has significantly tapered by the 2 kHz crossover point. The FF85k begins to be directional at about 3 kHz, and is fairly consistent in its directional behavior up until about 8 kHz (for a cone driver). It is more directional in this range than might normally be expected for a 3" driver, which is to our benefit in this application. Above 8k it becomes very much a laserbeam, so for very high frequencies, you will want to sit very close to on-axis. This is not a wide sweetspot, omnidirectional speaker, but it has sufficient dispersion for a few good listening positions. You and your sweetie on the sofa should be just fine. But this leaves us with the problem of a significant change in dispersion at the 2 kHz crossover frequency.

There are several design decisions that help us accommodate this problem. One of the major causes of directionality in large drivers is the cancellation from different arrival times (leading to 180 degree out of phase signals) from the different parts of a large driver (when it's large relative to wavelength). This is part of the usage of phase plugs. Our acoustic absorption used to form the low-pass characteristic interferes with this cancellation at some angles, and may well smooth the off-axis response (at the cost of on-axis extension -- a benefit, in this unique case). This is theoretical, I have not measured it, but is very likely. Another benefit is the cone geometry of the 123a/2213. The "Curvilinear" style cone tends to have improved high-frequency dispersion characteristics over straight-walled cones. Finally, a large amount of lossy acoustic felt, as used on the woofer grille, is built up around the edges of the FF85k to create a lossy barrier and induce some directionality at and below the crossover point. This is not sufficient for full idealized matching, but helps. The enhancements of this paragraph help minimize it.

Construction
There will be less construction details in this article than in some others. The construction is exceedingly simple. I used a 1" baffle which was made of layered 0.25" and 0.75" high-quality ply, screwed and glued together. This was done for stiffness. Care was taken to avoid obstructing the flow of air away from the driver. The scalloped construction with bevels was done with hand-tools, not a router proper, so please forgive the roughness. A proper construction would have some improvements to the build quality here.

The construction pays attention to maintaining an open flow to the back of the driver. This is a valuable consideration in any speaker build, but moreso with open baffles and small drivers like the ff85k, where it would be extremely easy to obstruct the path of the backwave. It's also important to provide at least some recess for the large woofer. I chose a ¼ inset because it was what tooling allowed. Remember from the intro, that this pair was made with minimal tooling, as my tools proper had not made the move yet. A full flush mount for this driver would be preferable from a performance perspective.

You can see the primary "spine" used to brace the driver magnets (this helps reduce the energy transferred into the baffle by the driver frame) as well as stiffening the baffle itself, and providing the rear foot point which sets baffle tilt so that the baffles don't simply fall over. This particular build also has substantial bracing of the baffle, which is recommended. The method you use is up to you, the "ribs" shown here are fairly effective, and the cutouts are to prevent standing waves from forming between the bracing structures. Yes, I know they're ugly, thank you very much.

In short strokes, you can assemble these, with care, with little more than a jigsaw and a drill (as I did), but a proper toolset, including a circular and/or tablesaw, a router with a circle cutting jig, drill and bits, and other odds and ends are going to be required. The rough aspects of the design are important- the shape of the baffle, the effort towards stiffening it and reducing energy transfer into it (bracing the magnets), the crossover, the tilt of the baffle. Some things, however, you can make your own. How much absorption you want around the edges of the midtweeter for example, and whether you want to make funny looking felt pieces around the sides of the woofer, or are happier with doing a full flush-mount.

Closing Thoughts
In closing, I hope you find this to be worthwhile reading, and moreover, worthwhile building. The simplicity of the crossover and the overall performance of the package is the result of careful refinement, many iterations having been tested before arriving at the final performance.