AUDAX speakers home page

KLS9
Noel Keywood:
"Up to now, our loudspeakers have been a little different from the norm but this design, KLS9, deliberately uses a conventional format: it's a two-way floorstander. There are two reasons for this. Everyone wants to know how to design a loudspeaker and a two-way is an easy place to start. KLS9 is something of a design exercise. Secondly, by using high-quality drive units and not skimping on the cabinet size-wise, it is relatively easy to produce a budget design that is easy to build but offers superb performance."

CHOOSING THE DRIVE UNITS
In a two-way 'speaker, the bass unit must cover a wide frequency range, from the lower limit set by the box, all the way up to meet the tweeter, at around 3kHz. Because this covers not only the bass region but the midband too, it is called a bass/midrange driver. It's asking a lot of any unit to work over such a range and in practice, most bass/midrange drivers get a bit rough in the crossover region, from 2kHz-4kHz. Using a small cone minimizes this this problem, but the drive unit then has to work hard moving air to produce bass. Bass distortion rises and low frequency power handling deteriorates.

To achieve reasonable power handling as well as deeper bass with more punch, an 8in cone is needed. This is the most popular choice amongst manufacturers and it was my choice for KLS9. To maintain quality above 2kHz, I chose to use a High Definition Aerogel (HDA) cone driver from Audax, the HM210ZO. This is one of their most advanced designs.

HDA is a composite material comprising Kevlar and Carbon fibres, aligned along the polymer chain of an acrylic gel. The cones are actually baked hard, although they retain a peculiar slight stickiness on their front surface. HDA is extremely light, yet rigid, with optimised internal damping. Audax fit the unit with an edge-wound voice coil of flat copper wire, mounted on a heat-resistant Kapton former, and the whole assembly is supported by a rigid, cast Mazak chassis, not a cheap pressed steel affair.

Modern synthetic cones are more consistent in their properties than paper types, giving a clean sound even at high volume. Paper 'breaks up' badly, sounding sharp and coarse. By the way, the instrurnent that puts most harmonics into the break-up region of a bass/midrange cone is the violin, which is why violin sounds so different between 'speakers.

The tweeter I chose to use was the well tried and tested Audax TWO25MO. This uses a fabric dome for good sound quality (I would not use a metal dome by the way; they are too hard and clattery). It comes down far enough to meet the bass/midrange unit at around 3kHz and has sufficient sensitivity to match. It is also very reasonably priced.

THE CABINET
Not unsurprisingly, when the Thiele-Small parameters of the HM210Z0 are plugged into a computer programme to calculate box volume, the optimum figure turns out to be the size of a conventional floorstander. That's because nowadays, drive unit manufactures tailor parameters to produce a drive unit optimized for a real-life cabinet. The HM210Z0 suits a medium - to - large ported cabinet.

From experience we have found it best to go for a high-pass response that produces a little bass peaking around the 30Hz-60Hz region. This gives real weight to bass, which most listeners prefer. With the HM210Z0, using a box volume pass box of 2.6cu ft (0.075 cubic metres), bass reaches down to 36Hz (-3dB), which is low. You will find that our cabinet is 0.072 cubic metres volume in fact, but internal sound absorbent makes the cabinet appear bigger as far as the drive unit is concerned.

Generally speaking, the larger the cabinet the better - up to a limit. If you make the cabinet for this driver any larger than 2.6cu ft in volume, it will starts to peak up in the bass, sounding boomy and one-note. What I have done is to choose the highest volume figure possible, consistent with flat response, meaning +1dB of peaking at 40Hz. This gives low box Q, good damping and a relatively clean step response, suggesting solid sounding bass with weight but little overhang. Practice bore out theory; KLS9 delivers very deep, tight bass – it can really thunder. Conveniently, for this volume of cabinet, dimensions can be set to give a large but not monstrous floorstander, measuring 93cms high, 25cms across the front panel and 38cms deep.

THE PANELS
However, cabinet volume can be reduced, by up to 50%, for those who might want a smaller speaker or lighter bass. The best approach here is to knock up a pair of experimental cabinets of 2.6cu ft volume and then slowly reduce the volume with bricks until you get the bass quality you are after. The bass drivers can be removed to get the bricks in, or the rear panels if they are screwed in place.

THE CROSSOVER
Analysis of the HM210Z0's frequency response by our B&K measuring microphone and Hewlett Packard 356 IA FFT spectrum analyser showed it rolls off above 3kHz and, as expected from an optimally damped and consistent synthetic material, doesn't break up badly at high frequencies. It looked suitable for a slow-ish roll-off Butterworth (or thereabouts, damping-wise) second-order filter response, but in the end I found a first-order (-6B/octave) low-pass filter was sufficient, as it should be with a good driver. The 1.2mH inductor has a ferrite core, which allows a low DCR of 0.65Ohms to be achieved, maximising sensitivity and leaving driver/box Q unaffected.

The high-pass section for the tweeter was engineered to make it flat in response terms and a good phase match with the bass/midrange unit. The first-order low-pass filter combines with the HM210Z0 in 1 complex fashion, because the driver unit itself rolls off above 3kHz. Also, the drive units are horizontally displaced by 30mm, which amounts to 90 degrees of phase shift at 3kHz. Experiment showed that best phase matching was obtained with a first-order filter feeding the tweeter, making the crossover network very simple. I usually expect to use second or third-order networks for the tweeter, but the slow roll-off of the HM210Z0 allows the tweeter to be rolled in high up the audio band, keeping low frequency energy out, and with this arrangement the phase matching proved excellent.
First-order filters don't provide much flexibility in the all-important response tweaking department but in this design it didn't matter. Frequency response was as I prefer it, with a slow downward trend, no crossover suckout or dip around 3kHz at all, and no phase errors in this region to put in an off-axis suckout or dip. Keeping the midrange flat ensures good detailing and vocal projection without having to resort to any artificiality, like brightness.

Talking of which, a tweeter should never be allowed to peak. The TWO25MO's response is tailored by the series high-pass capacitor so it exhibits A)no peaking and B)it rolls off slowly toward 20kHz. This ensures there's no sting in the treble output to allow treble to rise toward 20kHz, giving painfully incisive treble.

In the end, KLS9 turned out to need a minimal crossover, much simpler than any of our earlier designs. That makes it even easier to build, of course and less expensive too.

I expected fairly benign load characteristics as a result of this simplicity and wasn't surprised to see a gently undulating impedance curve, as our analysis shows. This means there is little reactance in the load, always a good thing, since energy storage is kept to minimum. The curve sits high up, meaning overall impedance is high, because of the HM210Z0's intrinsic 6.5Oms DC resistance, which sets minimum impedance.

Overall impedance measured 12Ohms, meaning the speaker draws little current, so it will not stress amplifiers. When impedance is high , sensitivity usually suffers. In this case, it turned out to be a respectable 88dB SPL (sound pressure level) from one nominal watt of input (2.84V), so amplifiers of 30-80 watts are suitable, according to room size.

BUILDING THE KLS9
The cabinets should be built from 18mm MDF according to our cutting plan. The most important parameter is cabinet volume, but it is not so crucial that demensions must be exact. Glue the cabinet with Evode Resin W or similar. It can be screwed as well, if desired - use coarse thread screws made for MDF. Note that we put in a bracing panel, with a cut-out in it, two-thirds of the way down. This is to brace the side panels and avoid boxiness.

The Assembly Sequence:
1. Glue top, bottom and side panels together, then screw/pin on the front panel to hold them in shape whilst the glue dries.
2. Glue and screw the internal bracing panel into place.
3. Cut driver cut-outs on front panel.
4. Glue damping pads of heavy, natural fibre carpet felt onto every internal panel. An alternative is the bituminous felt used to damp car panels.
5. Cut terminal panel and port cut-outs on rear panel. Fix port into hole and glue in place.
6. Screw and glue rear panel into place.
The cabinet can be accessed internally through the bass/midrange cut-out on the front and the terminal panel cut-out at rear. Internal damping wool can be put in through the front opening.
7. Solder wires onto drive units and lead them out through the terminal panel cut-out when screwing drivers into place.
*Note polarity – the narrow pin on both drive units is positive (+, marked red on woofer). Also, the tweeter has reversed phase, so connect +ve from the crossover to the tweeter's - ve (wide) terminal. *
8. Solder crossover components onto terminal panel. Connect drivers to crossover then screw the terminal panel into place.
We placed the terminal panel right behind the bass/midrange unit so we could fix it using nuts and bolts, if need be. A 7mm recess had to be routed out for the chassis front face, leaving I 7mm for screw thread. We were worried they might not hold after repeated use, but they proved OK in practice.
CABINET PROPORTIONS
The cabinet was proportioned so that a point midway between the bass/midrange and the tweeter would be 2ft 6in from the floor - ear-height in a normal seat.

The front was deliberately kept narrow, for best imaging. Wide-fronted loudspeakers image messily, so beware if you are of thinking of altering this dimension. You will notice that the cabinets are 'handled' by offsetting the tweeter, so there are right and left-hand speakers. This places the tweeter asymmetrically, dispersing phase and diffraction effects, and it reduces the path length to the cabinets inside face, pushing effects up-band. Bear this in mind when cutting the front panels, since they must be different. The tweeter sits closest to the inside panel of the speaker, which viewed from the front as a stereo pair.

We used a router to inset the chassis of the bass/midrange unit, making for a flush baffle in order to minimise destructive reflections. The tweeter has been mounted over the edge of the bass/midrange chassis, to get it as close as possible; it is one wavelength away at 3kHz. I did not want this distance to be exceeded, since I rate good imaging very highly. For this reason, all the cabinet front edges were routed to have a smooth radius on them, to avoid edge-diffraction. This makes the cabinet difficult to veneer, so we applied Unibond as a sealant and then eggshell black paint.

The crossover is so simple it can be soldered directly onto the input terminal panel. Glue the crossover components on with a hot-melt glue gun or, better, drill the panel and bolt on two three-way tag strips. The capacitor, inductor and resistor can then be soldered on. Keep the inductor and capacitor apart and it mutual right angles to minimise inductive coupling ('cos a capacitor is a coil of foil with inductance).

We suggest you use Solen audio quality polypropylene capacitors. The inductor is ferrite cored for lowest DCR, measuring 0.65Ohms. These properties must be replicated, or performance will change. The resistor will not get hot; make it 5W-9W. We used a high temperature wirewound, but really this resistor is best made up front IW carbon film types, because of their neutral sound. The terminal panel we show is a bi-wave type and the speaker is bi-wireable of course.

Because the panel is positioned directly behind the bass/midrange unit, a pad of carpet felt should be placed over it to eliminate reflections from the rear of the bass/midrange driver. The reflex port itself is not too critical. To keep distortion down, it must be made fairly large. Our port was made 4cms in diameter and 4cms deep. You might like to try experimenting with these dimensions. The port is placed close to the floor on the rear baffle. Since it works low, around 35Hz, it is close enough to the main driver to avoid phase problems, so positioning is not too critical.

Ideally, the base should have floor spikes fitted. The cabinet side panels could well be braced further by a 1in dowel or similar, because this speaker produces a lot of bass energy.

And finally, to the matter of internal damping. Regular readers will know we recommend long-haired wool. It must be teased out and should be supported internally on threads to prevent settling over time. The bracing panel is useful for this purpose. Put an open gauze over it and wool on top, as well as in the bottom chamber. Don't over-stuff the speaker or it will sound lifeless and the port will not work properly. Reflex cabinets must not be too heavily damped for this reason. The amount of damping used is a matter for experiment, being sufficient to suppress cabinet boom but not enough to create a 'dead' sound.

Most people like to finish their cabinets with a good veneer. To retain beveled edges, I suggest hardwood or stained softwood 1/4 rounds are routed into the edges. Veneer up to these edgings. Bear in mind that all around the tweeter, up to 12in away, there should be no protuberances, especially ridges. The surface acoustic wave it produces must not be obstructed, nor should it be caused to diffract off sharp surface discontinuities. Attention to detail here, together with good drive units, narrow front baffle and accurate phase matching between drivers yields very sharp, clean-edged images, plus well embodied cymbals made of solid metal, rather than the sort of mellifluous representation that arises as a result of phase errors and anomalies.