ToneTubby 12" AlNiCo Drivers On An Open Baffle Plus Beyma CP380M With AH!1000 Conical Horn

This project started out here as an homage to vintage drivers, using Jensen A12 12" field coil drivers plus Altec 802-8D 1" compression drivers, but I ran into some reliability issues with the Jensens -- I'll suggest that the previous project is worth perusing for the background to the current endeavor. So, I put the old school components away and picked up some current production drivers -- namely, ToneTubby 12" AlNiCo and Beyma CP380M 1" compression drivers. Lynn Olson has blessed the ToneTubbys here and what feedback there is suggests that the Beyma is a pretty nice driver. As well, I sprung for a bunch of 1" drivers to measure, B&C DE250, Radian 475PB, BMS 4550, Altec 802-8D, plus the Beyma, and from a purely measurement standpoint it seemed to me that the CP380M stood out as the most appropriate/ least complicated driver for the AH!1000 horn crossing over somewhere between 1khz to 1.5khz -- I didn't listen to any of the drivers before selecting and don't have any listening impressions. So, out with old and in with the new, and I stuck with the AH!1000 conical horns.

So, where to get started? I needed to do some reading, so I started with Joseph D'Appolito's "Testing Loudspeakers" and started a thread at diyAudio, Geddes et al on Measuring Loudspeakers. Both proved to be very helpful in getting a fundamental idea of what needed to be done in terms of measurements. I would go so far as to say that having read the D'Appolito's book is a necessary condition for taking reputable, repeatable measurements. For anyone at all serious about this stuff, in the strongest terms possible I recommend picking up D'Appolito's tome as well as Dickason's "Loudspeaker Cookbook" -- they will disabuse you of any notion that a proper crossover can be either 'off the shelf' or 'textbook'. For software, I chose the ARTA suite for measuring the drivers and SpeakerWorkshop for modeling the crossover network. I got a Behringer ECM800 for measurement; calibration is a necessity for these microphones. I sent mine to Kim Girardin (email me for Kim's contact info) and it turned out to be off by about 6.6db at 11khz. For a mic preamp, I got a Rolls MP13. Made two different jigs, one for measuring frequency response & one for measuring impedence, pretty much as per recommended by ARTA. Bought a bunch of cables and a mic stand. There's not a single solid-state amplifier in the house, so I put together a LM3886 chip amp for frequency response measurement duties; the headphone output on my M-Audio USB Audiophile was sufficient for the impedence measurements. Lastly, I put together a lazy susan to facilitate taking on- and off-axis measurements. It turned out to be obscenely easy to make. I got a lazy susan bearing from Woodcraft and used some scrap wood I had laying around.

With that, I got down to brass tacks. Or, to be more accurate, started, then stopped. Started again, then stopped again, tore everything down and stuck it all up in the attic for a month, then pulled everything back down and started again... In some ways, this all became more like a job and less like a hobby. But, after trying this or that, failing, then trying again, reading some more, and speaking with people who know a thing or two, I began to get the hang of things. And, finally, here we are. These speakers aren't the bee's knees or anything like that, and I expect others could do much better with less. But, I like them a lot and I feel like I did a pretty admirable job.

When working on the crossover, I had a couple goals in mind. One was attempting to keep the complexity to a minimum, another was that I was hoping for a fairly uneventful impedence curve. But probably most prominent was attempting to get a smooth transition in directivity from the midbass to the horn driver, and consequently a fairly agreeable power response. The logical outcome of this is that the on-axis response wouldn't be of of any more importance than any of the off-axis responses, though I did pay particular attention to the 10 & 15deg off-axis response as that is approximately where I planned on sitting in relation to the speakers. To that end, I took on- and off-axis measurements for both the Beyma & horn and upper TT12, from 0 to 35degrees in 5deg increments; the 40 and 45deg measurements were gobbledygook, so I left them out. At 1.5 meters in-room, I could get a clean 5ms impulse so I'm pretty comfortable with my measurements down to 500hz.

Heading into this, I expected that I would end up with a crossover somewhere between 1 and 1.5khz -- 'the books' suggest that the 12" driver dispersion would be at 90deg or less by 1.2khz and according to Hornresp the relatively small horn would start have a widening directivity pattern under 2khz. And if you look at the graphs linked to below, that turns out to be approximately correct. In the end, a 1.2khz crossover between the upper TT12 and the Beyma worked out best. The lower TT12 is used as a 'helper' woofer to compensate for the baffle rolloff, so a first-order crossover appeared to be in order. Once the whole speaker was together, I ended taking some ungated measurements in order to verify that the lower TT12 was doing what it should.

Despite the number of parts, the crossover is pretty straightforward (I'm not sure I would go so far as to label it 'minimal'). Generally, I used zobels and the conjugate in order to get the drivers to behave a bit more nicely, and make an easier job of choosing the crossover parts. Horses for courses, and I suspect others with more/ different experience than I could make do without them. The lower TT12 is 1st-order electrical at 270hz. Without the zobel the cone breakup at ~2.5khz is barely attenuated, so in order to get away with just a coil the rising impedence needs to go. In fact, the values I chose for the zobel actually lead to a very slight decline in the impedence above 1khz, so rolloff is a bit higher than 6db/octave. For the upper TT12, the components form a 2nd-order electrical filter at ~1.2khz. Again, without the zobel, I'd be looking at a third-order network and the added complexity that accompanies it -- in particular, the impact to phase response. For the horn driver, the conjugate smooths the impedence peaking around 2.2khz and also dampens some off-axis energy at 2.7khz. The horn & driver is offset by 1.8125" and connected out of phase in order to line up phase reponse of the individual drivers at the intended crossover point at ~10deg off-axis. I came up with the offset distance by finding the relative acoustic phase centers of the two drivers, fiddling with the horizontal offset in SpeakerWorkshop to come up with a reasonable offset, then adjusting the horn location so that the absolute acoustic phase centers are separated by the prescribed amount. I also tried a small inductor, between .05 & .2mH and with & without various resistors in parallel, in series with the treble driver, but in the end I preferred the sound without any sort of >10khz dampening. The resulting impedence curve for the speaker isn't too bad; I wish it were a bit more monotonous and had a more neglible peak at 2.2khz but I'm not bothered enough to put a conjugate across the entire network. It does dip down to 3.9Ω, so I have the speakers hooked up to the 4Ω tap on my amplifiers.

I simulated the crossover in SpeakerWorkshop, then tweaked the component values by listening. The component values don't correspond to any hypothetical filters (Butterworth, Linkwitz-Riley, etc.), though I did use them as a starting point. I simply adjusted the values to get a reasonable frequency & phase response in SpeakerWorkshop then further fine-tuned them while listening. All the jpgs I've posted are simulations done with the final component values. All of the frequency reponses use 1/24 octave smoothing, as ARTA appears to require some smoothing in order to export .FRD files that include the minimum phase response. Lastly, I don't have any graphs of the final speaker, as there's just no way to get decent, gated measurements of >=10ms at a minimum of 3 meters in my living room -- any and all validation of the final product is derived exclusively from listening.

So what did I blow off in terms of the design? I gave no thought to vertical lobing. Well, I thought about it and decided that adequately addressing it was overreaching for now and that I should blow it off for the time being. Something to address in the future as I tend to think it's an issue with the substantial vertical separation between the upper TT12 and horn... Also, I need to work on the sub-120hz region. I have some large-diameter prosound high-Q woofers right now which are okay, but I just got a quad of OB15 from AE Speakers so sooner rather than later, I hope. And, I'll need to come up with some sort of active filtering. I have a design already, but we'll see how it goes.

[20 Apr 09] From a technical design perspective, I've finished this project -- there is, however, still a desperate need for some prettying up. First, since my previous update, I've fiddled with the components that form the L-pad and constant-directivity equalization (Rs, Rp, and Cs in the above schematic). Rs remains as before, but Rp has changed from 4.3Ω to 4Ω -- a pretty minor change that brings down the horn's level by a couple tenth's of a dB. Also, Cs has gone from 2.7μF to 1.0μF. Over a period of several months, the resonance/ break-up at ~16kHz was getting to me a bit (or at least that's what I propose it was), so I backed off on the CD eq. I consider this entirely a matter of preference and wouldn't be suprised if many or even most would prefer a higher value here. The changes are minor enough, I haven't bothered re-posting any of the measurements or the crossover schematic.

Second, I finished up the (sub)woofers. I'm using the OB15 from Acoustic Elegance; the driver isn't listed on their website currently, but it's the same price as the IB15 and can be had for the asking. I posted the Theile/Small #'s at AudioCircle; you can find them here. Here's a couple of jpegs of how I have them setup: pic1, pic2, and pic3. The drivers are wired in parallel and driven by a Crown K1 amplifier.

I tried my hand at passive filtering using R's and C's between the main amplifier's speaker out and the sub amp's line in, but the insertion loss was way, WAY too high (as one might expect). So, I ended up putting together a (sub)woofer controller using opamps. I made do with OPA2134s, but I would suspect that there are any numbers of components that would work just fine in this application. Here is a schematic for the signal portion and there is the board layout. It's volume control -> open baffle compensation -> 2nd-order lowpass -> 2nd-order highpass at ~28Hz with variable Q -> phase correction. I'm not much with solid state, so I got most of the IP elsewhere. Rod Elliot at ESP has a good three-parter on working with opamps here, here, and here. The OB compensation is from Linkwitz here. The variable Q high-pass I saw at John Pomann's page here. I didn't feel like doing the bipolar power supply myself, so I got a +15/-15 kit plus transformer from Twisted Pear Audio here. The lowpass is variable via the DIP8 plug-in; Marchand does it this way on at least one of its products. The highpass is set at ~28Hz but the Q of the filter is variable between 0.8 and 4.0, allowing a bit of fine-tuning. The phase correction is pretty standard stuff. The active filter 'in' comes from the main amp's binding posts. I'm pretty happy with how the OB15 have worked out. It took me about 30 minutes once I got out my measuring gear to dial in the (sub)woofers. Seriously, I have no idea how people get anything done without a sytem to measure what's what.

All in all, I'm quite pleased, both with how the woofers are working out as well as with how the speakers in toto take care of business. I think it will be a good, long time before I take on another speaker project.

[29 May 09] So, claims about being 'done' were slightly premature. I ended up tweaking the crossover a bit, adding a notch filter to bring down some excess energy centered at around 4kHz. The network is now a bit more complicated, but overall I'm liking it better with the notch filter than without. I'm not certain the changes would be to everyone's taste; the presentation is definitely a bit warmer now, but without any loss of detail or articulation. For me, the modification brings the timbre more in line with what I recall hearing from the OMA AC1 -- having gotten to hear those speakers a couple of times now, they really have become a 'reference'.

Latest Frequency Response, from On-Axis to 35-Degrees Off-Axis

[22 Jan 10] While I do like what the AE OB15s give me, I wanted to see if I could put something together that was more-or-less fullrange and would be amenable to being driven by the type of amps I like to build: namely, low-powered, no-feedback/ low-dampening DHT amps. So, here's what came out of it: a REALLY large open baffle 2.5way (my wife's comment was "That's a whole lot of speaker"). Treble- and mid-drivers are as before, but the .5 helper woofer is now 4x Eminence Kappa-12A. It works out to be pretty high-efficiency, I'm guessing ~98dB/1W since that's about what the ToneTubby is & and there isn't much to drag it down, and an easy load. I think the impedence generally looks pretty good -- relatively smooth impedence & benign electrical phase angles -- and should be a good match for the amplifiers I have here in-house. It sounds that way, anyways; listening impressions thus far have been generally positive. Here is the resulting impedence and electrical phase curves. Despite the planned loss of low-end extension versus the previous setup (45Hz vs. 30Hz, respectively), one reason for trying out fullrange speakers, rather than sticking with my previous arrangement of main baffles plus SS-driven helper (sub)woofer baffles, is that the previous setup, to me, wasn't particularly well-suited to the endeavor of building amplifiers that I find interesting, for a couple of reasons: the main panels didn't have an extended-enough frequency response on the low-end (only ~120Hz) and I didn't like having to use the 4Ω tap on the output solely to compensate for a dip in impedence ~200Hz.

Over there are the specs for the Eminence Kappa-12A drivers I'm using. The Qts is heretically low for open baffle usage, but the Theile/Small numbers suggest that sensitivity for four of them should be >109dB/2.83Vac, leaving room to tailor the response so that at Fs sensitivity would still be ~95dB/2.83Vac and a decent match for the ToneTubby. I used just a series resistor to attenuate the midband while providing a good deal less attenuation around the resonant frequency of the woofers. Simulations I ran suggested that the latest go-around should work reasonably well. I would've liked to use 15" drivers, but I couldn't fit four of them on a baffle and keep the mid- and treble-drivers at a reasonable height. Here is the crossover I'm using with this iteration. The .5 is now first-order electrical at 120Hz. The mid- remains as before, but the treble- section of the network has been tweaked to roll-off some of the diaphragm breakup at 16kHz; I like it quite a bit better with this.

Unfortunately, my hopes for an f3 of around 45Hz were too optimistic. I pulled out the measurement gear, ran the frequency measurements, and the speakers turn out to be strong down to the low-50s then roll off very quickly thereafter. A little shy of expectations, fatally so. While there isn't a great deal of bass, what there is certainly keeps up and is pleasantly tight and dynamic. Part of the reason for choosing the Kappa-12A is that they have no need for dampening from the amp -- the book value for qts is .27. Then, I can use all the series attenuation needed without worrying about loose, flabby bass. They do integrate quite well with the mids, and they are definitely more taut than the OB15 I have. But, that could also be because that the Eminence are a bit more rolled-off than the AE drivers, lending to the sense of lean-ness in the bass.

Now if only the response didn't drop off a cliff under 45Hz or so. I'm a little suprised by how much I miss that last octave, especially considering that the OB15 only covered down to 30Hz. I could get one of those GR-Research servo subs, but with two huge (they're 36"Wx39"H, by the way) panels should I really have to? A task for another day, I suppose. One accidental positive to the experiment is that with such a larger baffle than I had previously been using, the ToneTubby now reaches <200hz. Consequently, I'm crossing over the .5 much lower, more than an octave lower. This is as I had planned, but I had not expected that with just one TT now covering 200Hz and up, the mids are quite a bit more clear. Not that I was bothered by the heretofore unnoticed congestion, but there is now a much greater sense of clarity -- a really nice bonus and future revisions will endeavor to keep this intact. As well, while the bass isn't all there, what there is of it I really appreciate; there's a impact and dynamic to it that I wasn't getting with the OB15s. I'm planning further research down the path of using lower-Qts drivers on open baffle, perhaps with a dedicated variable output impedence SS amplifier. I think that the chapter on fullrange open baffles with solely passive crossovers is completed; I just can't see how one can get 40Hz @ 98dB/1W without a literal wall of drivers.