Zaph|Audio - SB Acoustics 2-Way : Final Crossovers

Crossovers - Which ones?

Let's cut to the chase. The crossovers shown here were auditioned for quite some time and are the results of a lot of subtle or not-so-subtle tweaks, both in CAD and by ear, that at the end were put back through CAD to arrive at the final values. None of these have the absolute values that the optimizer produced. It's often not easy to realize values that they spit out and it was important to provide values readily obtained, since certain ones are somewhat critical. And to be complete, these are the values that appear to be stocked by the the large U.S. distributer, Madisound. The values I would prefer to use are available directly from Solen for the most part with more options as to component values, but I suspect that many will not purchase directly from Solen. Madisound carries a sub-set of the Solen components as well as some of their own products outside of that list. Key to this is the availability of a couple of laminated core inductors that are suggested due to their much lower resistance, thus with less insertion loss that can be significant.

The fundamental differences between these crossovers relate to tradeoffs accepted in each and they all do have those. My primary goal, aside from good sound of course, was to minimize the cost of the crossover. As is often said in discussions, it makes no sense to buy less expensive drivers, then try to hammer them into submission with an elaborate (read expensive) crossover. The main tradeoffs that distinguish the three are presented below in table form.

One quick note, you can make quick visual comparisons of the summed/power responses in the page that covers the off-axis.

A Few Details

The design graphs below show only the on-axis responses. The optimization actually used a combination of On-Axis and 5 degrees up, down left and right, a 5-point averaged window relative to the tweeter axis and all at 1m. This did not include diffraction variations nor driver directionality, only changes related to the center-to-center spacing coupled with the crossover. Another page shows the predicted response for various off-axis points, again related only to the crossover. Driver directionality and diffraction are very much more complicated to include and require significantly more time if it is be measured and/or modeled. The diffraction analysis page shows various angles reviewed in the decision on where to place the drivers. The page covering the off-axis and power response related to the crossover itself helps to show more of what influences the in-room system response, but even that is not the full story. Some of the aspects of the power response are detailed in the page covering off-axis response.

This was actually a last-minute addition. The breakup of the woofer will be an issue for some who may otherwise be interested in it. This crossover pushes the breakup down more than 35db on-axis. At the same time the tweeter highpass provides more than enough protection with the low Fc. Note carefully that the drivers are NOT connected with the same polarity. This may seem odd, especially since it was targeted as a Linkwitz-Riley 8th order that would normally be connected with the same polarity. The individual drivers were optimized, then summed. As it turns out, the driver offset and summed response optimization yielded a quasi-LR8. That is, the woofer lowpass is rather close to that target, but the tweeter highpass was allowed to vary with the results shown.

.

This crossover was the original intended response. That is, the layout was chosen so that the diffraction signature integrated well with the tweeter highpass. It succeeded. There was no trap needed. Even though this crossover has a low Fs, the tweeter isn't stressed too much due to the higher order rolloff. Notice that the only real difference in tweeter output between this one and the 2K LR4 version is in the range of 1-3kHz. In fact, the tweeter rolloff increases at the low end in comparison, so the difference is almost a wash between the two as far as the tweeter is concerned.

There is one peculiarity. This is an approximate LR4. The phase target was not met. The phase can't be worked to get a perfect L-R response. Either the tweeter is extended too low or the woofer breakup rises too high. Auditioning showed this to be a very good crossover nevertheless. The biggest drawback is complexity and cost.

The second crossover is a good balance for sensitivity, tweeter extension, dispersion and component count. The trap for the tweeter Fs is necessary to control the peaking due to interaction with the crossover if an "ideal" LR4 target is to be achieved.

This alternate version of the previous crossover came about when I was about to assemble the impedance compensation circuit for the tweeter. I had left it out and measured the response for verification and noticed the peak in response from the tweeter. When I looked at the assembled components and considered the extra cost of the trap, I decided to try it without one. The original response was a very close LR4 highpass. Removing the trap and cutting the first cap in half from 20 to 10µF left a nearly perfect response to the -20db point where it turned down more, as we can see here. However, the summed response on-axis, off-axis and the power response changed very little. It sounds essentially the same, so I'm putting them both up since some might still prefer the nearly "ideal" highpass shown above.

Three components removed, one cap cut in half, seems like a reasonable tradeoff to me. That's how I've left the crossover that I have assembled to use permanently. Well, almost, but we'll cover another small tweak for those who care less about the absolute aesthetics. That came about when measuring and testing at the end and involves some additional diffraction control as an option.

This crossover originally had the least number of components that still sounded good. In fact, I think that I'd have a hard time distinguishing it from the other ones in a double-blind test. The differences are probably most readily discerned at high volumes related to the tweeter demand, if then, and might be noticeable by the slight bass sensitivity drop. The tweeter Fs peaks, but is actually meeting the target response down to 1K. The rolloff below 1K is higher order and much like the 2000Hz alterate one. This alters the phase response and is partially responsible for the phase departure from target below 2KHz. Keep in mind, though, that this is only for the 1m on tweeter-axis point.

I've left this one in for those who prefer a higher crossover point (less stress on the tweeter) and to show how quickly the breakup of a woofer can become more prominant. You can see how an increase of less than 200Hz in the crossover Fs affects the breakup output.

You might expect there to be a problem with the woofer breakup not suppressed to at least 20db down, but it doesn't make its presence obvious. I suspect that it's due to the cone material and the harmonic distortion profile. Hard cones always make a breakup on-axis such as this one more audible, at least in my experience. Since the SB17 cone is a more damped material with breakup more broad, it's not nearly as noticeable. I've been pleasantly surprised that it's not the problem that I had assumed it would be. Suppress the breakup significantly and the phase takes a turn for the worse (sorry for the bad pun). The effectively relaxed slope (not using a trap on the breakup) helps to bring the phase of the two drivers closer to the target in the crossover area. It's all about tradeoffs.

The tweeter peaking at 1kHz close to its Fs could be suppressed further using a trap roughly the same as that used in the second crossover, but its probably past the point of diminishing returns. It does affect the phase of the tweeter in the crossover region as well, so it's not a good cost/performance consideration.

Another consideration is the air-core inductor in the trap. This is a drawback because the insertion loss (resistance) is part of the reason for the slight loss in sensitivity below about 400Hz. I have not found any quality laminated-core inductors of less than 1.0mH so far, so I had to use a 14ga. air-core as a reasonable cost alternative. Even so, you can see the slight drop in sensitivity in the lower region of the woofer.

David L. Ralph © 2009