Improving on a 2300 dollar speaker (Should you really have to!)

[pjpoes]

I have had this JmLabs Electra CC900 center channel for some time that I was using with my DIY mains. I decided to take it apart, make a schematic for the crossover, and take some measurements. First thing of course is that, I needed a baseline. Then my goal is to tinker a little and see if I can get a better response. I wasn't originally going to post this little project as I thought it probably was not of much interest to most of you. However, when I saw the response I decided it may have some academic value after all.

I have always felt that this center had a somewhat chesty, hands over mouth sort of sound to it. Dialog was typically drawn to the center, and it had a surprisingly bland sound, given that its focal and all. I bought this used from a shop going out of business, but I believe its new retail was something like 2300 dollars. I was a little shocked at how bad this speaker turned out to be for its price. The cabient is pretty good, void free plywood for the front and rear baffles, and MDF for the top, bottom, and sides. The sides and front baffle are roughly 1" thick, the rest is 3/4" or so. There is one central brace made of plywood. The crossover is a typical 4th order topology with an LCR conjugate network on the midbass drivers. The midbass drivers are placed physically forward of the tweeter to help with phase alignment, but are still not at an equal zero point (I am estimating roughly half an inch on the z-axis as compared with the tweeter). The crossover uses all 100v polypropalene capacitors and 20 gauge aircore and steel laminate inductors. Air core for the tweeter, laminate for the woofers.

All of that isn't bad, not overbuilt, but not as bad as I have seen. Where the problem came is with the measurements I took for baseline. At first I thought maybe it was something I was doing wrong, maybe the drivers need more space to come together, maybe its a bad angle, strange room interaction, who knows. I took probably close to 100 different measurements using various techniques, and they all showed roughly the same behavior, a sizable 10db suckout in the upper midrange lower treble. I took measurements of just the drivers and made up frd and zma files so I could model this in a crossover simulator, and the reason for this suckout became very clear. The tweeter is wired in reverse phase of the midbass drivers and the particular transfer functions chosen cause there to be a suckout in that region. With the phase normal you have the typical null at the crossover point but with better summing on either side. I placed an order for the parts today, so hopefully during the week I can rebuild it, but my plan is a similar topology (So I can reuse the circuit board), but with values that allow proper summing with the drivers wired in phase. I'm hoping this will also improve the step response as well. I will include the graph of the overall response, and then when I get the parts I will show schematics and new response.