Zaph|Audio - SB Acoustics 2-Way : Crossover Considerations

Designing the Crossover - Some Issues to Consider

Now that the box design is out of the way, including the layout of the drivers on the baffle, and the driver in-box measurements are complete, let's move on to the make-or-break part of the design, the crossover. While bad drivers cannot be made to sound great, it is true that great drivers can easily be made to sound bad or mediocre and the crossover is the key. We'll look at a number of the factors that go into the decisions made. That said, keep in mind that there is no best or perfect crossover.

Woofer Issues

Several considerations for a woofer include low end extension, sensitivity in the midband, upper end breakup, directivity and power handling. These are probably the more important considerations. The distortion profile is also important, but that should be considered prior to selecting the driver. In this case, the choice was fixed. Fortunately this driver has good distortion characteristics. This means that we can consider lowering the system sensitivity to provide a bit more low end extension without as much concern for displacement concerns as we would with some other drivers. The low end extension is largely dictated by the box highpass that was covered in the section on the box design. For the rest, let's look briefly again at the raw in-box response.

Two curves are shown because of the difficulty of measuring the low end of a driver other than a tweeter in anything other than a true anechoic chamber. I set the overall system sensitivity through auditioning various levels while working on the crossover. The measured low end response was not and not even measured close-mic until I was well into the crossover work. It wasn't until I decided to describe the design in detail that I noticed the discrepency in my original raw driver model. You can see the difference in the two curves below.

Two other characteristics stand out. One is the low sensitivity at the low end. This is after having adjusted about 1.5db from the 4-pi response that gets reinforced by the room response. The other one is the breakup above 4kHz. Both are quite a challenge. Add to this the fact that the driver is a 4-ohm unit, so some options in a crossover can become problematic. As you'll see, there are ways around some of this. Tradeoffs have to be made, but doing so appropriately can minimize some of the issues. The impact of this discrepency can be seen in the two curves below.

The blue curve is what was originally considered to be the final system response, at least with the thought that the low end was accurate. It had been based upon the overlay of the baffle step response seen previously. The red curve is the result of the analysis detailed on the 2-pi to 4-pi page. This is more in line with the perception of the system. A small bump at the low end of a 2-way gives more apparent bass. The bump shown in the red curve came about through reduction of the overall system sensitivity due to perceived lack of bass.

Tweeter Issues

The tweeter has a separate set of issues. One is overall sensitivity, usually less of an issue with a tweeter and not one in this case. A second issue is the low end extension. This limits how low we can cross it without distortion becoming an issue. The SB tweeter has a reasonably good distortion profile and the manufacturer claims that it can be used down to 1500Hz. As we'll see, one crossover takes advantage of this. As mentioned on the baffle layout page, the baffle layout was selected to blend with the tweeter low end response to provide for better integration when designing the crossover. We'll see how this was justified in the 1500Hz crossover, the original target.

The plan was to position the tweeter so that the diffraction signature coupled with the tweeter raw response would combine to provide a broad, low-Q peak for the diffraction step. This makes the crossover work a bit easier because passive crossovers with transfer functions that correct high Q peaks/dips are limited and can require inordinately large component values. In other words, a broad peak is easier to fix than a narrow one, the latter at times being impossible to do passively due to insertion loss (resistance) and/or unrealistic component values.

The other issue is the peak in response around 15kHz. This is maintained to some degree off-axis, it's not an internal reflection issue as demonstrated in the tweeter tweaks page, so the top end was rolled off a bit to balance this out. This will roll off the range from 15-20kHz. I found it necessary, but others may prefer to leave the peak. The change in the crossover is simple, adjust the value of an inductor or remove it altogether. The system may sound rather sharp, however.

Components and Tolerance

I'll bet you were thinking that this meant component tolerance. You're partially correct. The values used in a crossover aren't necessarily etched in stone and component production tolerances won't have a huge impact. Some changes and substitutions of different values can be made with little if any audible impact. We'll examine a few that you may want to consider if you think that some small changes might be warranted or if you don't have the exact values. The final component values were selected to be exactly those available as single units as much as possible. The one area that is very sensitive is the resonant trap in the woofer circuit, the tweeter Fs resonant trap and the zobel in the tweeter when any of those are used. If you can't measure changes made, I strongly suggest using the recommended component values.

David L. Ralph © 2009