Paul Kittinger emailed me a number of months back asking if I collaborate with him on a design for the Midwest Audiofest, sponsored by Parts Express. Paul and I have known each other going back over ten years, if my memory serves me correctly, when he came to my home for a speaker DIY. Paul has since moved to Ohio, and I’m here in California.
Paul had a concept, and asked if I would design the crossover. He wanted to design an ML TL based on the RS225 and the RS28F. I had done a project similar, my RS22528A, a number of years ago and was quite familiar with the strengths and weakness of these units. I’d always been impressed with these drivers, and my testing of the new versions (RS225-8 unshielded and the RS28F) showed outstanding performance, regardless of price. As much as I liked my original design, there were a couple of improvements that I thought could be done, but I just didn’t have a reason to build a completely new design.
And along came Paul with his RS22528FMLTL. Of course, he came up with a better name, the duet.
Highly stylized version in black if you build it that way
An 8″ driver offers notably improved low end performance over the typical 7″ midwoofer based two way, and an ML-TL alignment allows outstanding bass performance and with control of midrange reflections. However, 8″ drivers need to be crossed quite a bit lower than a similar 7″ driver, ideally in the 1.3-1.8k range, which limits the number of tweeter choices significantly. My testing of the RS28F shows it to be up to the task, handling a crossover in the ~1.3k-1.4k range without difficulty.
Paul emailed me the tower dimensions, and I built a mock up test box. I used a sealed PE box and mated it to a baffle of the exact dimensions of Paul’s tower. For crossover design purposes, this would be sufficient. I measured the drivers on the baffle outdoors, then designed a suitable crossover.
A PDF with Paul’s writeup can be found at the link below. The design turned out quite well, so much so that it was awarded first place in its category at the Midwest Audiofest.
More on the crossover design for the curious.
There are quite a few ways of designing a crossover network. I use Praxis, but no reason you couldn’t use SE, omnimic, ARTA or any number of the modern PC based software measurement systems.
Step one was to use a measurement point 2-3 meters out, and use lspCAD to target an eighth order LR curve in the region of the crossover. I try to get as high as possible and measure outdoors. You can see my test box in the picture below. The baffle is identical to Paul’s Duet, but the back side isn’t exactly the same. It should be very close to the correct FR and adequate for crossover design. I can usually get usable window of ~7ms after subtracting time of flight.
The topology is a notched fourth order curve. The idea behind this topology is to start with a fourth order LR set of curves, then add an additional reactive component to generate a notch in the stop band. The notch has the effect of having a very steep slope. Then you just adjust the crossover Fc so that the steep slopes overlap and now you have a bit of a hybrid crossover that is overall fourth order, but in the region around Fc, approximately eighth order. Probably the easiest way to understand this crossover is through a series of graphs.
First, consider just a fourth order LR.
Then, add the reactive component and you end up with notches in the stopband. As noted by the blue comment, you have to use the optimizer to adjust the region of Fc to target an eighth order topology.
The beauty of this approach is that it completely suppresses the rather large and broad breakup of the raw RS225. Below you’ll see a set of raw curves for the two drivers. The region starting at 3k and peaking at around 6-7k needs to be suppressed aggressively.
Alternatively, a more typical fourth order LR with a notch could be used. The relative merits of this crossover have been discussed previously elsewhere, so I won’t go into it here. See my write up for the RS22528A, do a google search for old forum posts on the PE and HTguide boards.
Step two is to build the actual physical crossover and remeasure. Occasionally, the resultant FR curve is spot on. I target flat initially. If it’s not flat, then my job at this point is to figure out why not. I’ll typically go back to lspCAD and look at the topology, and adjust individual component values to engineer corrections. Then I’ll make component changes and remeasure over and over, one by one until I get the flattest possible curve.
Usually flat sounds pretty good. In this case though, when I emailed the preliminary crossover to Paul, he felt it was too forward. This leads to that age old, endless debate, “what is the right target for a speaker’s FR curve?” In general, I target flat, but occasionally flat doesn’t sound right. I had Paul mail me a tweeter and the crossover, just to verify that the tweeter was not the problem. I measured the FR of the crossover, as well as the distortion performance of Paul’s RS28F. Both of these measured fine. Although I can’t prove it conclusively, I suspect that the low crossover causes the tweeter power response to bloom quite a bit lower than we are used to. That is, the tweeter’s very wide power response causes quite a bit of off axis energy in the range of 1.3-3k compared with a more traditional crossover at 2-3k. It’s just not what we’re used to listening to.
So, it was time to engineer a dip. The easy part of this is that placing an amp side resistor on the tweeter leg attenuates the tweeter curve. Unlike a tweeter side resistor, placing the resistor on the amp side causes relatively more attenuation as you go down in frequency. This turns out to be perfect to engineer a modest dip which is most prominent immediately after Fc. In essence, a single resistor causes what is effectively a “BBC dip.” We can argue about the BBC dip, and what it really is, another time. I use the term because most folks have an idea of what it is.
If the dip bothers you, or you think it might bother you, just build the crossover with the resistor, and then use a switch to short out the resistor. When shorted, the response will be nominally flat, but leaving the switch open will leave the dip in. So you can have it both ways.