Clearance drivers are a great way to learn about crossover design. But they present a problem, they disappear. I wanted to build a high quality system with medium priced drivers that I hoped would be around awhile. While this is still a hobby, I value my time. It seems a bit silly to spend hours refining a design and adding components to $10 drivers. If I'm going to the effort why not do it with better quality components.
I wanted a set of shielded mains that could work with a matching center, using readily available drivers. After much deliberation, I settled on the 7" Vifa MG18SK09-08 and the NEO3PDRFP from Bohlender Graebener. The MG looked attractive for it's versatility, low end and build quality. It should be useable sealed as well as vented. I chose the "S" version of the MG as the specs were the same for both shielded and standard versions.
The NEO3 looked like a lower cost way to try a shielded ribbon. The NEO has a small foot print (for a ribbon) and published specs looked attractive for crossover and response, as it theoretically could go as low as 2000 Hz in a second order crossover.
Box Design and Construction
I first modeled the MG in Unibox. After fiddling, I selected a 24 liter box tuned to 37 Hz. Predicted F3 of 41 Hz. I wanted to do a tower design but given the baffle width required for the driver, I wound up with a shallow box. After much calculation and trial/error I came up with a slot loaded design that gave me a 0.75" X 7.5" X 10.25" port with the net volume required (1500 cu in), adequate depth, and low port air speeds.
I confirmed the Unibox design in LspCad. I then laid out and built the boxes, recessing the MG and surface mounting the Neo. The MG has a thick cast flange that's 5/8" wide. I now have a rabbeting bit that will go to 3/4" so I used it. You may wish to pre-cut the rabbets (before the thru hole) if you want to use the full flange width.
In the box picture you can see the additional 3/8" panel I added behind the speaker cutouts. The Neo requires a rectangular cutout with not much MDF between the screws and corners. Since it's only about 1/2" deep, I left the support panel intact and drilled two small holes to accommodate the tweeter wires. At the top and bottom are pads of imitation Black Hole from PE. The layout shows the top edge of the Neo. Since the cutout is rectangular, it's probably easier to use the gasket that comes with the Neo as a template for marking the cut. I recommend you verify the layout with the drivers before cutting any wood.
Final outside dimensions on my box are 9" W X 8" D X 37" H. About 1500 cu. in. net volume. Because of vertical dispersion, you may consider a different box shape with stands. (See Listening Impressions section) Keep the width at 9" and the driver locations shown.
Since a couple of people asked, I use a very simple jig for biscuit joining. Use a flat surface for edge cuts, and a simple jig for face cuts. My Dewalt will put the slot in the exact center of 3/4" mdf. Can't miss for square, flush joints. Biscuit Jig
I set up my testing gear out in the yard. The microphone was set up at 1 meter, on the line between the MG top and Neo bottom. I reasoned that the Neo was really a line source -- not a point source. So using the tweeter axis would be difficult. I did measures of each driver, and the two in parallel. I used a 33 uF cap on the Neo to protect it. Impedance measures were taken with drivers in the box but with out the cap. I also did a nearfield measure of the MG and tried to merge it with farfield for design.
My suspicions about the point vs line source were confirmed when I tried to reconcile the "combined" measure with the two individual measures. No realistic location of the drivers in LspCad allowed me to reconcile the two curves exactly. Based on physical measurements I took my best guess and began design work. I use merged response only for baffle step confirmation. While the farfield measure looks ragged, it matches up fairly close with LspCad's prediction for in room response using only the baffle and ground effects. I've gotten used to using the farfield so I can compensate for the raggedness by eye. For the Zircon, I used a target curve that includes the BBC dip with a slightly depressed high end.
|MG 1 Meter Measurement||NEO 1 Meter Measurement|
|MG Impedance||NEO Impedance|
|Predicted in Box vs Actual|
Rather than trying to merge nearfield and farfield measurements for the woofer and desgining a flat response, I tried a different tack with the Zircon. I constructed a target curve that included baffle step and the BBC dip. I used LspCad's box model with no room interactions to determine the predicted baffle step slope and used it to detemine the target curve.
I tried several "standard" configurations of filters and chose the best one. I built it out of parts on hand and measured it again. I did both on and off axis response to get a better feel for the Neo. It's horizontal off axis performance was better than I expected. However, there was a broad peak centered at 3800 Hz. Off axis measures reduced this peak, but never eliminated it. I traced the issue to baffle edge diffraction by extending the baffle with cardboard. It was clear that I would have to address this peak in the crossover. With the almost purely resistive impedance of the Neo, I wound up with a series notch filter, and a parallel CLR impedance compensation network.
The "Final" Crossover
After several hours of listening and tweaking, I invited an acquaintance to audition them. Joseph plays professionally in the local symphony. He definitely has an educated ear, (and some high end gear). We sat and tweaked for another couple hours. He spotted some issues that we eventually modified using my ACI Jaguars as a reference.
Despite my efforts to reduce components, the presented crossover is about as minimalist as I could get. I tried to keep the impedance up a bit with out sacrificing sonic quality. The purely resistive profile of the Neo allows the use of a single padding resistor. A true Lpad will drop the upper end of the curve to around 3 ohms.
Jan 1 2003 Update
As the Zircon's break in, the Vifa has picked up considerable bottom end. Moving the towers closer to the wall (about 8" from the back of the box) definitely over emphasizes the bass. The first published crossover had full 6db of baffle step compensation. I tried crossovers using coils of 2.0, 1.7 and 1.5 mH (with appropriate modifications to the rest of the components). The revised crossover uses a 1.5 mH 18 ga perfect lay that I had on hand. It reflects about 3 db of baffle step and tones down the 2000-3500 hz range just a little. If you will be putting these anywhere close to the wall, (8-20") the crossover above is a better choice. If you choose to move them out into the middle of the room, (or really like bass) the full 6 dB crossover may be your choice.
After subsequent twiddling with half a dozen crossovers, I think the one presented below is the best compromise of soundstage, clarity, balance, imaging and impedance (5.8 ohm minimum at about 4500). If you want to voice this yourself, I suggest padding resistors of 6, 6.5, and 7 would be a reasonable selection. In it's present form, it's a hint more forward than my Jaguars, with comparable high end. My final changes were really preference not improvements.
Jan 26 2003 Update
I replaced the 18 ga coils with 1.5 mH 16 ga coils. The lower DCR brought up the high end a bit. I also used Mills resistors to do the final padding. I found 7 ohms to be just about right. The Zircons definitely get brighter as the NEO's break in. In head to head comparisons, I found definite differences between the Eagle's I used orginally and the Mills that I finally used for padding. For the NEO, I definitely recommend the Mills for padding. I tried changing out the PE 2%, "audio grade" sand casts I used on the trap networks but heard no audible effects.
|Predicted Crossover Response||Reverse Null||Predicted Impedance|
The clarity of highs from the Neos is spectacular. With the diffaction effects neutralized by the notch filter, the Zircon's present a very wide, pinpoint soundstage. Instrument presence extends well outside of the speakers. Horizontal off axis dispersion is very good - much better than I expected from the Neo's horn loaded faceplate. Vertical off axis is another matter. While the overall effect of standing (vs sitting) is not objectionable, there is a definite vertical sweet spot -- a characteristic of ribbons in general. It's clear that I must add about 4" to the height of the Zircon as I built it. (Tweeter about 38" off the ground would be ideal)
The MG is typically Vifa. Neutral sounding, with crisp authoritative bass. Very clear on piano, female vocals, guitar and percussion. It's not quite a Scan Speak, but it has plenty of slam for a 7" driver. Good off axis response. With the tuning of my box, I like the Zircon out from the wall about 18". Move away 10 feet and they simply disappear into a wall of sound.
In head to head comparisons with my ACI Jaguars, the Jags have more depth of soundstage and a bit more bottom end. (The Zircon will do low to mid 40's, the Jag's will get to the 30's). I thought the Zircon's had a bit more detail on the high end. The Zircons definitely have a "different" sound than the Jaguars. (Ribbon vs Dome) but I'm not sure how to describe it.
The Zircon's definitely reveal flaws in recordings. Tape hiss is very audible in some of my older analog mastered CD's. On well recorded material they definitely shine. On the "Trinity Sessions" CD (Cowboy Junkies), Margo Timmons vocal on "Mining for Gold" just shimmers. The harp intro to "I Don't Get It" is searing. On "Well, Well, Well", (Maria Muldaur and Mavis Staples chilling duet on Muldaur's "Fanning the Flames" CD) you can hear Staples swaying movement and breathing over Cranston Clements steel guitar. I received a boxed set of Cirque du Sollei sound tracks for Christmas from my sister. (CSWAG from her tours with Cirque). The Dralion CD is totally engrossing and brought back the experince of the live show.
So I am I ready to trade my Jaguar's for the Zircon? Not quite, but it's very close. For $150 each in components, I'd happily live with the Zircon's as my mains.