Notes
20 March, 2006: Doh! Just noticed that I never gave the full model number for the tweeter. It's
the T25CF001. DO NOT use any of the ther Excel tweeters, they won't work.
15 March, 2004: Added TurboCAD files for the crossover.
4 Jan, 2004: Changed series resistor in the tweeter network from 2 Ohms to 4 Ohms.
This is the resistor value I have been using, but I assumed most listeners would prefer
the more aggressive sound of the 2 Ohm. User feedback has told me that the 4 Ohm resistor is
preferred.
6 Sept, 2003: Added Usher 2nd order 2-way (scroll down).
2 April, 2003: New, bigger boxes, new pictures, and measurements. The I enlarged the boxes after looking at the size of the XO's.
Those 14ga copper foil coils are HUGE! The new box is 8.5"x14.25"x10.5" (WxHxD).
5 Jan , 2003: I have completely reworked the crossover of this speaker. It is now
a second-order at about 2kHz. This was a complete revelation for the Scan Speak
driver. Going second-order forced me to deal with the two break-up modes in the audible range.
It also allowed near-perfect phase matching and linearity over a very wide frequency range.
The cost is increased component count. There are now nearly twice as many
components as the old 4th-order XO. However, the rewards outweigh the cost.
To put it in the simplest possible terms, the clarity
and balance on this speaker are simply amazing. The 18W/8545K is a truly a diamond in the rough.
Left in its natural state, it's just another rock. But a few well-placed cuts, and it's
brilliantly beautiful.
I would have absolutely no problems putting this speaker up against anything selling under $3,000.
I would have absolutely no problems putting this speaker up against anything selling under $3,000.
Box Design
The first step was to measure the 18W/8545K's to see what the real specs are.
They came to
I like to make my boxes as narrow and short as practical to reduce their visual impact. The 18W/8545K is about 7" wide, so this led me to an 8.5" wide box. I needed 14.25" height to fit the drivers in, which left 10.5" depth to get the 16L volume (all dimensions are external). The test box is 0.75" MDF without any bracing.
My final boxes are not braced, but lined as much as possible with Black Hole 5. I consider this optional, but when you've already spent this kind of money, why not go a little further.
The woofer is centered on the box's centerline 8.5" from the top. The tweeter is off-center by 0.75" (to reduce edge diffraction) and 2.5" from the top of the box. Of course, the boxes are laid out in mirrored pairs.
- Fs = 34.8Hz
- Re = 5.7 Ohms
- Qms = 2.96
- Qes = 0.44
- Qts = 0.38
- Vas = 34.8L
I like to make my boxes as narrow and short as practical to reduce their visual impact. The 18W/8545K is about 7" wide, so this led me to an 8.5" wide box. I needed 14.25" height to fit the drivers in, which left 10.5" depth to get the 16L volume (all dimensions are external). The test box is 0.75" MDF without any bracing.
My final boxes are not braced, but lined as much as possible with Black Hole 5. I consider this optional, but when you've already spent this kind of money, why not go a little further.
The woofer is centered on the box's centerline 8.5" from the top. The tweeter is off-center by 0.75" (to reduce edge diffraction) and 2.5" from the top of the box. Of course, the boxes are laid out in mirrored pairs.
Crossover Design
The Scan Speak 18W/8545K
Here is the Scan Speak 18W/8545K in all its glory. The break-up modes are quite obvious. The one at 850Hz adds a shrillness to the midrange. The peak at 2700Hz would be nearly cut off with a 4th-order crossover, but is only a few dB down with a second order XO. Also notice the high-order roll-off. With 4th-order, this is to the designer's advantage. But with a 2nd-order, it will cause phase-matching issues in the XO region, which can make the speaker sound boxy. Both of the break-ups and the slope must be dealt with. The fall-off below 600Hz is baffle-step loss.
I don't think you'll find very many people who will argue about the need to trap out the resonance at 850Hz. But why add the expense of a second trap? This next image shows the XO without the second trap, overlaid on an ideal 2nd-order response. You can see the upper break-up mode is quite prominent. Not only that, but when combined with the steep roll off, make the filter response become 3rd-order acoustic. This will cause integration problems with a 2nd-order filter on the tweeter.
Here is the result of adding the trap. Perfect 2nd-order response to 25dB down.
The Seas T25 Excel
Here is the raw Seas tweeter. The depression above 10kHz is a measurement artifact. Again, note the prominent peak at about 1700Hz.
And after the filter is applied. As is the woofer filter, the tweeter filter contains a trap with the combined purpose of dampening a peak and forcing the response into a 2nd-order slope. Again, you can see a near-perfect match to the ideal curve.
Crossovers
Here is the final result. Note that the tweeter required a ladder delay to force it into phase alignment. With a second order filter, you don't have enough components available to work phase alignment into the filter, so sometimes you have to add it in its own circuit. The other possibility is to build a sloped baffle.As for part selection, I chose SonicCaps, copper foil coils, and Mills resistors. For coils, I suggest 14ga on the 1.5mH woofer coil, and suggest 16ga everywhere else. If you're looking for places to save money, component quality can be dropped in the traps without sacrificing much, if anything. I would use at least Dayton or Solen caps, and decent quality air-core coils. You can probably get away with 18ga coils on all the traps, since they are not really resistance-sensitive and not much current flows through them.
Here are the TurboCAD LE files for the crossover:
Low Pass
High Pass
To use them, download the free copy of TurboCAD Learning Edition. Print the files, and cut a piece of whatever (hardboard, plywood, MDF) to the sizes indicated. You will need two boards of each (one for each speaker). Next, attach the diagram to your board using 3M spray-on adhesive, and drill 1/4" holes where the black dots are (hint - stack two boards together and clamp while drilling to do both boards at once). After pulling the paper off, mount the components where indicated, using zip-ties through the holes (or over the side of the board if the component is near the edge and there are no holes shown). The holes in the corners are for mounting screws if you decide to use them. Wire the components together according to the thick black lines. Where inductors are shown as squares, they are mounted standing up.
Final Results
I was pleased with my previous 4th-order filter. I am stunned by this one.
The 18W/8545k is maligned in some circles as heavy handed and harsh. It has a reputation of
delivering excellent bass, but falling apart in the mid-range. I can report to you that,
with proper care in the XO, it is probably cable of matching the mid-range clarity of the Excel
mag cones without losing any of their famous bass.
If your looking for an excellent speaker for under $1,000, this is a great option. With budget XO components, you can bring it in for less than $700. Go for the Sonic Caps and Hyperlitz coils and you're looking at a little more than $800. It costs about $100 more per pair than the 4th-order design, and is definitely worth the added expense.
When I initially showed this design, there was some concern for crossing the Seas T25 at 2kHz 2nd-order. I admit that's a bit close to the ragged edge, but it should be distortion free up to about 95dB at average listening distances.
By the way, the scribbling on the front is the instructions for making the driver holes :).
If your looking for an excellent speaker for under $1,000, this is a great option. With budget XO components, you can bring it in for less than $700. Go for the Sonic Caps and Hyperlitz coils and you're looking at a little more than $800. It costs about $100 more per pair than the 4th-order design, and is definitely worth the added expense.
When I initially showed this design, there was some concern for crossing the Seas T25 at 2kHz 2nd-order. I admit that's a bit close to the ragged edge, but it should be distortion free up to about 95dB at average listening distances.
By the way, the scribbling on the front is the instructions for making the driver holes :).
Usher 2nd-Order 2-Way
Someone was kind enough to send me a box with the Usher drivers in them for me to
do a 2nd-order crossover. The results are shown below.
As far as the Usher drivers are concerned, they are good quality units. I think the woofer may be a bit overpriced. It lacks the refinement of the Scan Speak 18W/8545K, especially when used in low-order XO's. It seems well-suited to 4th order, but there is a nasty ridge in the waterfall at about 3kHz which is audible in a 2nd-order crossover. Using a stronger tweeter, like the Seas Excel, might reduce the problem.
All in all, I was satisfied with the Usher version, but I'd rather have the ScanSpeaks. The box is the same as the Ella.
As far as the Usher drivers are concerned, they are good quality units. I think the woofer may be a bit overpriced. It lacks the refinement of the Scan Speak 18W/8545K, especially when used in low-order XO's. It seems well-suited to 4th order, but there is a nasty ridge in the waterfall at about 3kHz which is audible in a 2nd-order crossover. Using a stronger tweeter, like the Seas Excel, might reduce the problem.
All in all, I was satisfied with the Usher version, but I'd rather have the ScanSpeaks. The box is the same as the Ella.