paint on these guys is an epoxy which has to be used within about 5
hours after mixing. It also has a hefty price tag of $85 a quart,
It's a transparent epoxy, with the appearance of a candy apple lacquer seen on custom cars. We got the idea after seeing a semi at a truck stop.
These speakers were designed and built in late 2008.
The woofer is a Peerless 220WR P831709 bought in April, 1992 for $93 each. At that time, a pair of tens and twelves were also purchased. I still have the twelves, unused.
The midrange is a modified Radio Shack 5 inch midrange with a whizzer cone. It cost about $18. Initially, it sounded bad, schreechy and a Focal 5K013L was considered but here was an opportunity to do some experimenting. The back of the cone was coated with about 0.0625" (1/16th inch) RTV, an uncured rubber in a tube which sets after several hours. The screechiness disappeared. In fact, they sounded much like the Focals.
The tweeter is a very old Wharfedale dating to the sixties/seventies. It has a purple mylar dome with a rather wide annulus for a tweeter. The main reason for using it was that it's purple. As luck would have it, it sounds very smooth. The crossover frequencies are around 400hz and 4kHz. I have this thing for decades over octaves. It has been said that 3 octaves is good for a mid section, the wider the band, the better, as long as it's within the range of the driver. The decade goes a little further but not too far. For instance, 400 to 3200 is 3 octaves and 400 to 4000 is a decade.
Crossovers. Ahhh, I could write a book on them.
Way back in the early sixties, I learned that a filter works best when designed to match the load impedance at the frequency at which the filter is to operate. In 1981, I bought a book titled BUILDING HI-FI SPEAKER SYSTEMS by M.D.Hull of Phillips Corp. It substantiated what I learned in the sixties as all their crossover formulae use the speaker impedance at the crossover frequency rather than making an assumption of 4, 8 or 16 ohms. This was quite evident in the midrange part of a three way system. The midrange needs a bandpass filter and unlike all other crossover designs I've seen that use the standard nominal 4, 8 or 16 ohm impedances, the actual impedance at the crossover frequency is used. For instance, consider a midrange operating in a band from 400 hz to 4000 hz. A typical 8 ohm 5 inch driver can have an impedance of 8.5 ohms at 400 hz and 20 ohms at 4 kHz. If 8 ohms is assumed throughout, the filter will require 49.8 uF and 0.32 mHy in series with the driver. This is a first order Butterworth, 6dB/octave.
Now, if we use the component values thus calculated above, the crossover frequencies shift to 376.2 hz and 9952 hz. While this is close at the low end, the high end isn't even in the ball park. Also, most 5 inch units are hard pressed to be linear to 10 kHz; most roll off around 5kHz to 6kHz at roughly 6dB/octave due to voice coil inductance and mass loading of the cone & coil assembly. A device, called a ZOBEL can be used to flatten the impedance of a loudspeaker. It's a series resistor and capacitor wired in parallel with the driver. It adds another reactive element to the system, already plagued with such things and has another effect, insertion loss. If one is going to design a passive network for any speaker, the impedance of that speaker should be measured with the speaker in the box in which it is to operate as the air load will affect the impedance. The same applies to designing the zobel.
Passive networks can create many unexpected results. Insertion loss, phase shift, slight loss of transient response. Using very low resistance coils will reduce insertion loss but they can be extremely expensive. Most, if not all of the problems of passive networks can be reduced or eliminated by multi amping the system. The biggest problem here is cost. A three way will require 3 stereo amps and an electronic crossover, none of which come cheap.
The center system is described here. BRIGGS
The flanking black d'Appolito systems are partially described here. BLACK
|Response of the above speakers outside at a distance of 12 feet.|
|The next graph shows
the response, again outside but at 3, 6 and 12 feet. The high
frequency bands rises at 1.5kHz, 3.5kHz and 5.5kHz
to 9.0kHz are unexplained at this time. Theories are as
follows. Keep in mind that the power to the speaker was held at 0.01
watt (0.3v into 8 ohms @ 1.0kHz. OK. If it
were ambient noise, it's safe to assume it was constant as it wouldn't
show in all 3 sweeps considering also that each sweep is separated in
time by about 2 minutes, to move and measure the mic distance, run into
the house and initiate the sweep. In contradiction, ambient
noise as above described/assumed would appear to decrease with respect
to the speaker's output since an increase in distance would
decrease the SPL from the speaker. So, the conclusion is that it
results from interference in the radiation patterns of the midrange and
tweeter and/or ground reflections. The block wall of the house and
fence are about 30 feet away on the sides of the speaker but only about
10 feet behind the mic. The speaker position was constant; the mic
was moved. At 12 feet from the speaker, it was closest to the rear
fence, about 10 feet. The lateral distances remained
constant. At any rate, the listening position in the room in which
they will be heard is about 14ft.
Manufacturers give measurement results as taken from a distance of 3.1 feet or 1 meter and under anechoic conditions. I know of no one who listens to speakers under those conditions. (computer speakers excepted and NOT under anechoic conditions)
I've seen published by some manufacturers, pdf files of $400 midrange units that exhibit a 3 to 4 dB bump in the 950hz to 2.5khz range.
The following curves were measured with an input of 0.3v which is 0.01125w into 8 ohms. The average SPL is about 62 dB at 12 feet. This will go to about 82 dB with an input of 2.83v, or 1 watt.
At 3 feet, it equates to about 94 dB. Realistically, I'd say around 91 dB, based on the approximation made regarding the 62 dB average.
Here, we have two sets of curves, the lower triad being lowered by -20 dB for visual clarification. The lower set is also smoothed to half an octave, a typical practice used by many manufacturers to make the response curve look better. The bump in the lower resister is due to ground plane reflection; it becomes less prominent as the distance of the mic is increased.
Keep in mind, these curves were obtained outside with the nearest reflecting surface other than the ground being 25 feet away, namely the house and the block wall. Also, there is no ceiling, approximating anechoic conditions. This is in no way reflective of what one would measure in a room.
Some construction piccies
|Waiting for the primer to dry. A spray gun was used and this had to be done on a windless day, not too difficult to get in Arizona.|
|The front grilles.
The quarter round has yet to be installed at the perimiter.
If memory serves me well, these panels are 3/8 inch thick
the right channel unit with the grille removed. The sqwaker is in
it's own chamber; the tweeter is closed in the rear.
The crossovers for each system are slightly different as the driver impedances were measured in their respective enclosures and rarely are two driver impedance curves identical. Coils were trimmed to less than 1% accuracy. The capacitors were adjusted to within 1% of calculated values using mylars of less then 0.3 uF, some as low as 0.04 by adding the appropriate values in parallel with the larger one, also a mylar or polypropylene. All coils are air core; those in the woofer section are 14 or 16 AWG while the rest are 18 AWG.
Inductances and capacitances were measured with LMS (described below) whereby the value can be determined at any frequency which does affect the measurement. Caps were selected that had the lowest ESR, which, in all honesty, is moot compared to the resistance of the speaker wire. But, I figured, what the heck.
Don't pay any attention to those two critters on top; they get into everything.
|The numbers as
calculated by Bass Box Pro, version 6. Years ago, I used a
computer program I wrote in BASIC to do all this but speaker
measurements had to be made manually. I now use LMS, Loudspeaker
Management System by LinearX which simplifies things a lot at the cost
of the LMS hardware. But, I wanted a new toy.
The LMS is very close to my manual measurements, checked later upon the acquisition of a Fluke 8808A bench meter.
|Normalized amplitude response from
BassBoxPro-6. This is an ideal response assuming a linear
elasticity of the annulus, which doesn't happen. It usually loses
linearity at about 60% of its excursion in either direction but is still
well within its maximum excursion limit.
The voice coil height usually determines Xmax which ism usually half the difference between the voice coil length and the thickness of the top plate. However, usually before this, the annulus and/or spider will become nonlinear. If it were made looser to maintain linearity, the voice coil would move beyond the magnetic field resulting in severe harmonic distortion as well as heating the coil faster, resulting in higher resistance (impedance) and lower power transfer. Magnetic field saturation is also imminent.
Then there's eddy currents in the coil, he's always there but at lower levels, he's kind of incognito. Add to this, back emf which increases with excursion, all of which the amplifier has to resist. At high levels, it's similar to driving a car with one foot on the brake.
Back to the loudspeaker main page