Speaker design considerations

Again, this is such a vast and sometimes emotional subject to cover that a mere scratch of the surface is offered here. With speakers there are two important don'ts; never touch the cone assembly, even fleetingly, and never bring large speaker magnets together. The best audible results are usually obtained when there is no discernible cone movement.

Although it is easier to construct a perhaps boring rectangular box if resource is limited, it is what is done with the box that can produce the wow factor. One recalls a rather nice Bose system that mounted two speakers on the front of the cabinet and eight on the back. This acknowledged the fact that most perceived sound is reflected. To give arguably better sonic qualities, speakers could be set against the longest wall in a room to improve the bass response, usually at a distance of about a ¼ of the wall's length from the corner. These speakers sat at the same distance apart, but about a metre away from the wall, the eight rear speakers facing it.

The finish of a speaker is important if, say, sales are a consideration. To prevent warping, both sides of the wood used in a cabinet must be sealed or treated - when the construction of a box is finished a couple of coats of primer both inside and outside the box won't go amiss. This should be done after removing all debris and wood particles with a vacuum cleaner, since these can damage a speaker. Touring cabinets can be over-painted (usually matt-black) easily, although aluminium primer has been used. Those intended for outside use may have asphalt roofing felt and bitumen coverings (which not only waterproof but help reduce cabinet resonances), whereas an expensive hardwood (the dust from which can produce allergic reactions in some) may require oiling or waxing.

Some very striking, even dramatic, finishes have used the same techniques used for finishing guitar bodies, ie; spraying or graining to give a colour scheme or pattern, then spraying with some 20 coats of lacquer (allowing six weeks to cure) and then buffing to a glass finish. One subtle scheme, used on car bodies as well, used a black undercoat with alternating coats of silver and gold glitter loaded varnish which, when finished properly, gave a remarkable 'depth'. Such care and attention will then produce an object that is desirable to many for years to come. However, with a small 'production run', say for domestic use, such lengths may be considered too complex, or unnecessary, and impatience can then give rise to a regretable and unattractive appearance. Cut plywood panels can be easily made to produce curves, for large scoops for example, if steamed for half an hour before fitting, although some practice will be needed to get a first-rate fit first-time.

Most of the worst damage to speakers is caused by over-zealous/ignorant users, moisture and rodents. To help overcome the latter, fit metal grilles over all openings, diaphragms and ports.

For a beginner, ambitious projects are best avoided. A worthy bass-reflex design intended for inexpensive drive units is shown below, which can be used for surround-sound or multimedia.

The port exiting from the enclosure's top, the coil can be wound on a non-metallic former 28mm long by 28mm diameter. Wind 7 layers of 1.5mm dia enamelled copper wire. Use 100V caps and 10W resistors. Since exposed cones are finger magnets, grilles can be glued to speaker's rims. Offsetting the tweeter might improve.

A simple, effective, but larger, 2-way reflex design used a Soundlab 8" (20cm) bi-cone bass unit (8LUX, res 40Hz) in a 31 litre box (int dimens: 28cm wide x 49.5cm high x 24cm deep, 14mm teak veneered chipboard) with a port (5.5cm dia x 3.5cm long) below the woofer. A dome tweeter with a similar crossover was added and a piece of furniture foam helped to damp the box.

A 'serious party' 3-way speaker used 12" 300W Electrovoice bass units (30Hz) to replace 50W versions in 70 litre closed boxes (Q=0.7) which proved to be excellent for reggae and orchestral level classical music. These are still used today (>30 years on) as a very acceptable reference.

Compact three-way speakers appear to offer much in the way of performance versus size for domestic circumstances. One notable example, likened in it's performance to one of the (then) top-of-the-range KEF loudspeakers and rated very highly, was Armstrong's 602 intended for the respected 600 Series. This was designed by Bill Perkiss, who joined the company in the mid 1970's after leaving Goodmans. The design was originally intended to be an infinite baffle design, but ended up as a well-damped three-way reflex which is easily reproduced.

    

A similar format was used in the design (below) from the early '80s whose cabinet employed butt-jointing, although reinforcement can be added. The dome mid-range unit did not require a separate enclosure within the cabinet. During construction, this was compared favourably with the Yamaha NS1000M, Gale GS401A, KEF 105 II and Popular Hi-fi Boxers, each of which had their particular strong points.


The midrange and tweeter are offset from the centre of the cabinet so that the pair are mirror imaged on the two speakers. This improves stereo imaging, mainly by reducing edge diffraction which is a major cause of poor stereo. The worst case for diffraction is when the tweeter is mounted equidistant from the three nearest cabinet edges. Re-radiation will then take place and this will cause a discontinuity in the frequency response. By offsetting the unit the diffraction is smeared and reduced to insignificant levels. Chamfered corners at the front of the cabinet help reduce acoustic reflections which naturally occur at sharp boundaries.

Below are the V3's impedance, frequency and polar plots.

Note how the impedance of the speaker (upper trace) is frequency dependent and differs considerably from the previous example. Lively loads like isobariks, ESLs, ribbon types, etc, can easily dip below 1 ohm, then causing catastrophic output stage failures. In amplifier tests 'simulated loads' may be adopted, that for an ESL is shown below. These may not accurately reflect reality, or connecting media. For example, a real ESL can measure 1.8 ohms at 20 kHz to over 60 ohms at 150 Hz.


Impedance and phase plots of a representative sample of different speakers showing variations between models.


As a general rule, the power dissipation of an amplifier driving a 60° reactive load (usually considered to be a worst case loudspeaker load) will be roughly that of the same amplifier driving the resistive part of that load. For example, a loudspeaker may at some frequency have an impedance of 8R and a phase angle of 60°. The real part of this load will then be 4R, and the amplifier power dissipation will roughly follow the curve of power dissipation with a 4R load.


An important point to make is that care should be taken to seal and make airtight all joints, seams and gaps (with the obvious exception of ports) since these can make quite disconcerting and very unmusical noises. Corner joint battens (glued and screwed) stiffen a structure considerably and can help reduce the likelihood of leaky seams (these can be located with talcum powder or chalk). Since speakers, at some point, might have to be removed for replacement or testing, the use of, say, self-tapping screws to mount them is not recommended, especially if the enclosure material is chipboard, which can be quite porous to air if not sealed. Ideally, use bolts with T-nuts.

For larger sizes, a very solid box is a prequisite, so don't underestimate the need for bracing to reduce box resonances. After all, why waste wanted acoustic energy in rattling a box, and it's contents, to their later possible detriment? The panel around the speaker mounting hole can be doubled in thickness. However, front to back braces with additional lengths to increase strength and rigidity across long surfaces help as well. The normally wasted circular cut-outs made for speakers have been quartered and used for this. Don't neglect to include such items in the volume calculations. Effort made in these areas will definitely be repaid by improved performance, and having built a box remember that one might be reluctant to disassemble it. Bear this in mind when considering access for servicing. A sub-bass enclosure (closed box, driver 10"(25cm)+, Q=0.7) might look like this (face down).

Note the provision at the rear of the speaker for a separate power supply, amplifier and filter. This reduces extensive disassembly for repairs and can help balance a light structure if the speaker is heavy. If an enclosure or stand requires 'ballasting', loose sand or lead shot can burst seams and joints, especially when moving them. Fixed sheets of lead and even lead-loaded concrete can be used, but don't over-estimate the strength of the overall structure. Taking the time to find the optimum position for a small quantity of ballast will avoid the damage caused by indiscriminate assumptions.

Enclosure linings, often 2" fibre-glass insulation wadding, are intended to reduce reflections and standing waves. They do not decrease the effective volume of the enclosure and neither do they increase it. Polyester fibre (for sleeping bags and duvets) can be used, which does not break up as some types of fibre-glass can. The individual panel frequencies can be more evenly distributed - this can be achieved by using nonrectangular cabinet walls but this can result in a design that some might consider too complex.

A number of good books are available, two whose designs have been built are given below.

"Designing, building and testing your own speaker system" by David B Weems, ISBN 0-8306-8372-7 (0-8306-3374-X pbk). A simple but effective approach, the larger scale tapered-pipe speaker is recommended (below). Larger units using 8" bass units with a 'presence' cone gave a good account, four or more of which and a LF bandpass enclosure make for a very respectable domestic arrangement.

Note how the central partition will stiffen the structure. Use corner joint battens as well (not shown).

"High power loudspeaker enclosure design and construction" (Eminence, ISBN 0-9518252-1-6) offers some excellent advice. All you'll need to know, scoops, w-bins and bandpass bass units are covered.

A number of enthusiasts prefer single speaker designs driven directly. The proprietor of a local B&O dealer, Jabez Gough of Gough's (no longer extant), came up with a design (1961 UK patent 912,430) using 8" drivers that generated much interest. This consisted of a cabinet with hinged lid to direct the sound (a perhaps crude precursor to acoustic lens technology?), with ports to either side of the speaker which was mounted on the top, under the lid. Importantly, perhaps acknowledging the variance in speaker spec, it was held that dimensions, including that of the seams was not critical, although the author acknowledged that a lighter cone and coil mass could improve. A suitable amplifier could be the Roger's Cadet, or the later Linsley Hood '69 class A. Larger versions seen included the Goodman's twin-axiom 10/12 speakers (bi-cone) and other 'exotics'. This author did not find the resulting sonic performance disagreeable. One critic suggested that the inclusion of a coil of 24 x 24 x 1" fibreglass below the speaker reduced a 'boxiness' quality.

Tony Gee's Solo-103 is another single-cone design that some might like. Wayne Jaeschke's Dayton D3 is a good choice for DIYers or if more bass is required Dave Tenney's Dayton 8s are worth a look.

Where it is considered critical, and possible, the author will attempt to align separate channel's multiple voice-coils in a plane equi-distant to the listener's position, a method used by a small number of manufacturers whether the speakers be mounted to radiate horizontally or vertically; see Beovox Uni-phase and 2 series, Beolab 5, Linkwitz PA applications, etc. A slanted or raked baffle is used in other designs. For example David Huckle retains this in a Troels Gravesen / Jesper Spohr design.

Although high-tech composites can appear in some enclosure designs (eg: B&W Nautilus), wood is considered to be best by many, some manufacturers emphasising the fact that their sub-bass units are made of this. Chipboard can be porous and require sealing. Denser MDF suits domestic circumstances, Marine Birch plywood suiting large PA touring cabinets. Concrete can be employed for LF designs, although thought should be given to the pattern before commiting to a cast, together with a pumice aggregate if movement is required. Innovative shapes can be made, but the need for a vibrating table, to expel air, might have to be considered.

Interesting sub-bass (20-80Hz) reinforcement can be achieved by mounting a bass unit at the end of a tube (transmission line) whose length represents a ¼ of the desired wavelength (smaller versions are sometimes seen in car audio). A relatively simple construction, the critical element is sourcing the tubes. Choose a fibrous, eg; thick cardboard, composition to reduce resonances. The rear of the speaker should be 'open' and, if placing upright, have a heavy base. Performance will vary with the type and size of driver (preferably large), and to the extent and amount of damping material (polyester, wool, fibreglass, etc) placed in the tube. Ideally, the inner diameter of the tube should match the diameter of the speaker's diaphragm. However, the general arrangement of a system using 18"/45cm (600W, 33Hz) speakers (tube = 24"/61cm dia, 12.6'/3.84m length - 22Hz) that performed well is shown below, although smaller designs using low-resonance 8"/203mm bass units and 6'/1.83m tubes (40Hz) can impress. Notwithstanding architectural resonances, if desired or if space is at a premium this type can be slung horizontally from a ceiling in which case a heavy base (lead-loaded concrete) can be dispensed with. If one is used, channels for core ventilation and wiring can be cast in this.

The larger scale approach is considered to give far better performance than smaller sub-woofer units. A 20 litre unit intended for a living room using two 6½"/165mm speakers can, with filtering and a 2"/50mm port, offer a 'wide' bandwidth of, say, 25-110Hz and a 'narrow' band of 18-65Hz (-6dB). However, even when driven by a 100W amplifier, the response peaking at between 25 and 35Hz will, for those who have heard better, result in a 'boominess' which will not do the justice that a larger arrangement will.

Significant improvements in existing closed box systems have been seen when a second box of half the existing speaker's internal volume is made to match the existing one, which it then sits beneath. Two ports are then calculated and fitted, one connecting the two cavities and an external one on the new box.

Power distribution varies across the spectrums covered by the various drivers used. 'Normal' play-back will give the following ratios;

In disco use, frequencies below 150Hz and above 5kHz may be subject to excessive equalisation thus requiring greater power handling capacity.

With crossovers, air-cored coils may be bulkier but have a better ability to pass transients than ferrite-cored coils, which can momentarily saturate on high power peaks. Keep them well-spaced to prevent any flux linkage between them. A 'Loudspeaker Passive Crossover Choke Coil Calculator' can be found here and a 'Guide to producing inductors and other wound electric component's here. To reduce losses, aim for a DC resistance of less than ½ ohm. A wire diameter of at least 1.5mm is recommended.

Those intimidated by the thought of winding half a kilo of copper wire on to a toroid might like to consider using the secondary windings of mains transformers (isolate primary connections) instead. Transformers with open-circuit primaries can thus be re-used.

Capacitors should have high voltage ratings to prevent possible failure under high power drive. 50V ratings are often seen, but 100V is obviously better. Useful types can be those intended for mains motors and flourescent lighting. Such components can reduce costs and give performance indistinguishable from 'audiophile' components, combined with high voltage ratings.

Choosing a speaker

Let's say a domestic system envisaged requires good bass drivers. Two ranges are considered; one intended for hi-fi and the other for high power use. Plotting some of the specs can help the selection process, eg;

The first graph shows how low frequency performance is proportional to size, as is efficiency, shown by the second graph. Taking the 8" units as an example it can be seen how output can vary in one size, but different models, by as much as 13dB and by 17dB across the whole range selected. Most speaker's frequency ranges are quoted at -6dB. Looking at the frequency responses, the two largest speakers tend to peak after about 1kHz, suggesting that their use be restricted to 500Hz and below.

Taking these considerations into account four choices are made;

6½"; low power, but a useful resonance for it's size. A good choice, perhaps, for the tapered-pipe speaker above.
8"; more economical than 225W version.
10"; good combination of power, resonance and cost, although lowest impedance must be noted.
15"; lowest resonance, ample rating, reasonable cost.

Limitations of cone travel will usually occur long before power ratings are exceeded. As a consequence, working maximum outputs of driving amplifiers can be a tenth of those stated, or even less.

Once mechanical and economic considerations have been made Thiele-Small parameters can then be used to determine an optimum enclosure. If aiming for an optimum bass performance when building a number of identical enclosures, design and build for the lowest resonance (largest volume) given by the speaker's specification tolerance, then reduce volume by adding polystyrene blocks or tiles and/or cut ports to length to suit the individual driver fitted. In a multi-way system where three or more units cover different parts of the spectrum, ensure each speaker has it's own compartment to prevent, say, a mid-range unit being modulated by a bass unit's output. Tweeter's diaphragms are usually mechanically isolated and need not be so mounted.

Tolerances of ±10%, whether mechanical or electrical, are often considered acceptable (resonant frequencies can be quoted at some ±15-30%, the constructor can measure these but some might find the procedure too complicated for small production runs). Avoid mounting a speaker in the dead centre of a baffle since this will result in standing waves and pronounced peaks and troughs in the frequency response. Aim for an assymetrical baffle layout. The accepted ratio of height/width/depth is 2.3/1.6/1 although other shapes are permissible.

The speaker cone alignment of some even respectable makes of large bass drivers when supplied as new can be questionable to say the least. This can be checked, preferably as soon as possible after purchase, by driving the unit with a low frequency whilst listening for any 'scraping'.

Isobariks vs Linkwitz

The search for the greatest possible bass output from the smallest possible enclosure is frequently driven by the fact that, say in a three-way system, the bass unit will often consume twice the power of the other two channels put together. Both approaches have their limitations.

Given the aesthetic inelegance of the isobarik clam-shell approach and it's extreme vulnerability of exposed speaker assemblies (as with any design employing exposed baskets or cones, particularly from inquisitive fingers and other accidents), attempting to force two tightly coupled, but invariably unequal, speakers to act as a single unit with one's most prized amplifier is tempting fate, and as other's experiences have shown, invariably does.

With fcs quoted at ±15-30%, attempts can be made to match units, but results can disappoint, unless one has access, say, to a manufacturer's entire production run. The load presented to the amplifier can be very lively especially with low-loss cables which can add capacitance. Ideally then, if only one unit blows (usually the front one), the pair should be replaced. Then there can be a power amplifier to repair... For this type of load (as with ESLs!), current-limited paralleled output pairs and heavy-duty diodes clamping the output to the power rails are recommended, ie: assume intermittent short-cicuits and/or back EMF, which quite a few amplifiers simply will not/cannot tolerate.

The Linkwitz correction, applying a bass boost between a low-pass filter and power amplifier, does not rely on the uncertainties involved in mechanical interactions and their possible consequences at high energy levels. Some constructors, however, can be tempted to opt for a smaller driver which consequently wears out sooner (electrical power handling conflicting with mechanical specs / optimism conflicting with the real world).