These speakers were build for a good friend of mine, who of various reasons is unable to either design or build speakers himself. I'm not too happy with the carpentry part myself, so I decided to find a nice looking finished box I could use. After having heard my own speakers, he also wanted to have some kind of Focal tweeter and at least one Seas Excel mid/woofer. The total project cost should not exceed 7500 NOK (about 1000+ US$). The last thing limited me to using only one mid/woofer. Available finished enclosures also pointed in this direction.
The sonic goal was to build a speaker as true to the input signal as possible; a true monitor. To achieve this, the frequency response must be very flat, the phase response and dispersion well behaved and the distorsion low, especially in the midrange. Those are not impossible goals. What is impossible within the previously discussed limitations is subsonic bass extension, realistic dynamics at low frequencies and the ability to play very loud. A single 17 cm mid/woofer simply can't do that.
I chose to use a box made for the German company Intertechnik for their «MS 3 mk II» speaker. This box is build of 22 mm MDF and painted white. It has precut holes for a 17 cm mid/woofer, a 104 mm diameter tweeter, reflex port and terminal cup. The net volume is about 15 liters. It also has slightly cut-off edges which, in theory at least, reduces diffraction problems. It also comes with a detachable grille.
The selected box was actually quite good, so I did very few mods on it. Of obvious reasons, I cut the tweeter hole larger to fit the big Focal tweeter. Further I covered all inner surfaces, including the plastic reflex tube with bituminous damping sheets to reduce cabinet wall resonances. This also doubled the total weight. I did not use the terminal cup, as I put a plastic box on the rear side to house the crossover and put a Speakon 4-pole connector in the plastic box. The terminal hole were covered first by a piece of 16 mm MDF.
The box volume was acoustically damped with fiberglass. All surfaces was covered with one 5 cm layer, sufficient to kill all standing wave modes and reflections within the box.
We did not want too much compromise regarding drivers, so we selected (IMHO) the best mid/woofer available today; the Seas W17EX002. It features a magnesium cone, of which Seas is the only company to provide. The driver is now slightly changed since it was introduced in 1995. This change resulted in a much reduced breakup resonance peak at 5 kHz, and it is also been said that the distorsion is even lower. The distorsion is low! The frequency response is also smoother than on the original version. Please visit Seas' home-page on the net for further details on this Excellent driver.
I measured the driver in the selected box, and this is what it looked like throught the Techron TEF-20:
This is the smoothest midrange response I have ever measured! It sounds incredible too.
The T/S parameters of the W17EX002 works well with a 15 liter box. I set the tuning frequency to 36 Hz. This results in a smooth low-frequency rolloff, and a -3 dB point around 40 Hz when placed in a listening room.
The tweeter I selected was the Focal TC120TD, unfortunately a now discontinued model from Focal. (The TC120TDX should be the same, exept that it has a phase diffusor). It features a titanium-oxide inverted dome, a very powerful ceramic magnet system, and most important; no hornloading (most tweeters are slightly hornloaded!). Even though the driver is not hornloaded, the sensitivity is quite outstanding. It does not drop down above 10 kHz either like almost every other tweeter, even 30 degrees off-axis. I've measured the distorsion as well. That did not look as good, but quite well compared to other tweeters. The frequency response is not at all as smooth as the Seas Excel, but not really bad. The Focal tweeter sounds very fresh and clean, and matches the sound quality of the Seas Excel very well.
The crossover is the brain of any speaker, the most important and most difficult part. The best drivers are useless without a crossover to feed them the right signals. To obtain near-perfect results there are so many things that must be considered; frequency response, phase response, impedance, phase correlation between the drivers, overall level, crossover slope symmetry, driver directivity, lobing patterns, off-axis response.....I could go on forever. Let me just say that all these things have been considered thoroughly when designing this speaker, but I shall only discuss a few of them here.
Before I started to design the crossover, I measured the impedance of both drivers, and imported the data along with frequency reponse measurements into CALSOD. Below 3-400 Hz I discarded the measurement data for frequency response, and relied purely on theory, as my measurements are not reliable below 400 Hz because of room interference. The rest of the curves shown here are from CALSOD simulations, but I've checked with the TEF to make sure they are representative.
Now then, what is the right theory to use below 3-400 Hz? All computer programs use the Thiele/Small model for low frequency simulations, a model which prescribes a resistive acoustic loading. These programs also relies strictly on half-space acoustic loading. In the real world, with real life speakers and listening rooms, pure resistive half-space acoustic loading does not exist. Of this reason, no box-calculators tells us how the speaker eventually will perform in our listening room.
Now some words about acoustic loading. A single point source radiates its sound energy omnidirectional, 360 degrees. This is called free-field acoustic loading; no abstacles to change the radiation pattern. When we put this source on a baffle, the sound energy is not radiated omnidirectional, but in a 180 degrees sphere. The baffle prevents the sound from radiating behind the baffle. This leads to a 6 dB increase in sound pressure. However, the baffle's acoustic loading is dependant on it's size. For frequencies which have wavelengths more than 1/4 of the baffle's size, the baffle can no longer reflect the sound wave, and will not provide acoustic loading. The loading will gradually change from half-space to free-field, and the radiation will change from 180 degrees to omnidirectional.
When we consider loudspeakers, the baffle of the box produces half-space acoustic loading down to the frequency where one-quarter wavelength of sound equals the baffle size. A 20 cm wide baffle will start to "unload" from about 400 Hz, and the speaker's response will drop by 6 dB below 400 Hz. This phemonemon is often referred to as "the baffle step". The "step" is not absolute, but more like a 1st order function.
When we put the speaker in a room, an other acoustic effect is produced; the boundaries of the room will start to acoustically load the speaker. An "average" room loading curve is recommended by Martin Colloms in his book "High Performance Loudspeakers", and it looks like this:
When we add the baffle loading of our box to the room loading shown above, we get something like this:
This is theory. According to my own experiments and measurements, a speaker designed for true halfspace loading will exhibit an approx. 3 dB dip centered around 150-200 Hz in a room. This makes the speaker sound bright, and it may also sound bass-heavy at the same time if the low frequency extension is sufficient. An absolutely flat frequency response from 100 to some 2000 Hz is in my opinion extremely important. Almost all fundamentals in music happen in this area. Big dips in the response are a lot worse here than below 100 Hz or above 2 kHz.
The above discussed 3 dB dip at 150-200 Hz leaves us with a low frequency sensitivity of about 85dB/1W/1m with the Seas woofer. A quick calculation tells us we have to burn 7 dB (or 80% !) of the tweeter's sensitivity in the crossover to obtain a flat frequency response at 85 dB, and at least some 4-5 dB of the midrange. Do we really want that? Instead I have designed a crossover that crosses over at the selected frequency and slope, but creates a total frequency response that rises smoothly by 7 dB from 100 Hz to 2 kHz. This frequency rise is corrected by a simple passive equalizer placed somewhere in the signal chain at line-level (preferrably between the control and power amp). We now have a speaker that has quite nice sensitivity in the important midrange, and is capable of playing more than loud enough with a 50W amplifier.
I decided to cross over at 2750 Hz. I always use a 4th order Linkwitz-Riley type of filter and this case is no exception. The goal of the filter is NOT to create a true 4th order electrical function, but to create an acoustic response from the drivers that complies to the 4th order function.
For the woofer it is absolutely necessary to kill the resonance peak at 5 kHz. I do this with a L-C parallell circuit. This circuit together with a series inductor does it all; both 4th order rolloff and kills the resonance peak. This is the filter type Seas recommends for the Excel magnesium drivers.
The tweeter looked simpler to design a filter for, since it's rolloff is more relaxed that the wild Seas mid/woofer. Even so, I struggled a lot with this. I ended up with a third order filter topology.
I also designed impedance equalizing networks to flatten the total impedance of the speaker. These are shown in red colour on the crossover schematic below. These parts may be skipped, but most amplifiers benefit from a flatter impedance load.
Here's the final crossover:
The 12,5 mH inductor should as well as all the others be an air-core type. I used an inductor with 0,8 mm wire. This coil has a DCR of 3,6 ohm, but that's not a problem here since I put a 8 ohm resistor in series.
To minimize the coloration of the crossover, I used CFAC inductors (except for the big 12,5 mH in the impedance equalizer). There may be some sonic differences between different brands of polypropylene caps, but the difference between CFAC and all other inductors is really great. The quality of the CFAC really shows off through the Seas Excel mid/woofer. If you can't afford CFAC's all the way, at least use it in series with the mid/woofer.
I used Reoderstein ERO MKP1840 caps, but other polypropylene types may be used without spoiling the sound quality.
The passive equalizer that must be input in the signal chain somewhere looks like this.
For this filter to work properly, a source with an output impedance less than 100 ohm is required. The load impedance should not be less than 10 kohm. However, variations in load impedance above 10 kohm does not make any significant changes in the filter function. If your source has a high output impedance, you can subtract it's value from 1,3 kohm, and replace the 1,3 kohm resistor with this value. A 600 ohm output impedance would require a 700 ohm resistor (680 ohm is the nearest E24 standard value).
This is what CALSOD tells us about the frequency response without the required equalizer:
A good check on proper filter symmetry is to invert one of the drivers. There should now be a large symmetric dip at the crossover point. This is how it should be:
When we put in the required equalizer, we get a frequency response that is very good:
The tweeter peaks a little at the top end, but no adult person can hear it anyway, so why then bother? As we can see, the low end rolls off smoth, which works well in an ordinary room. The speaker will not become bass-heavy even with a corner placement. When judged from listening only, the speaker plays «flat» to about 40 Hz. If you want more bass, the cure is called subwoofers.
The speaker lobes upwards when placed with the woofer above the tweeter. The crossover is also designed to have its on-axis direction 5 degrees upwards. This means you should be at the same height as the woofer's centre when listening at 2 meters distance.
The impedance stays flat at 5 ohms over a wide frequency range, and the phase angle is within +/- 20 degrees:
The speaker looks very good, just take a look (click the images for a larger view):
And here's a full view of the entire hardware collection (well, at least the part he wants to show us...) of Karl Anders Ellingsen, who owns the speakers. Check that grin!
Does it sound good? Build it and find out!
Stig Erik Tangen
Copyright © 1997 Roy Viggo Pedersen