Speaker designs

This is such a vast and sometimes emotional subject to cover that a mere scratch of the surface is offered here.

It is recommended that the first subject of attention, in respect of implementing any sound system, no matter how big or small, must be the speakers. This is, quite simply, because these are the devices that make the sound that you listen to. Even 'professional' cabinets lack bracing, more due to economics than design and, as a result, some enclosures' cabinets resonate more than their speakers do.

Taste is paramount, find what suits you. One person may need no more than a single cone, another may need eight or more. Somebody tone-deaf may spend thousands on a system, simply to impress. His neighbour may have acreage intended to 'rock the joint', literally.

A recent web-search for speakers came up with a quite small pair for $6k with optional metal or wood side-panels for an additional $2k. As with all fashion, many things are available for high prices to those who may wish to impress, this does not however mean that the spender can (by divine right) recognise quality. It is ventured that a not dissimilar, if not superior, arrangement could be contrived for a tenth of the above mentioned cost. At one respected retailers' I was offered my own choice of makes and sizes for the repair and test area. A top-line prestigious pair were offered, but all agreed that another pair at a ¼ of the cost were superior and these were used instead. If cost is of no consequence then types like the (15k5 per pair) BeoLab 5 might satisfy.

When comparing several speakers listen to each for the same amount of time, preferably in the room and position where they will be used with a clear and unobscured sound path to the listener. Personal preference places the tweeters above the listeners' ear level, giving a more natural 'panorama'. The loudest may impress but may not be the best choice. Although it might be efficient in the mid-range which most humans are sensitive too, it may lack top and bottom-range performance.

Any large surface near a speaker reinforces the bass and can alter the mid and treble responses. A difference in sonic quality can be noticed if the speaker is moved even a foot away from a wall instead of being against it. The difference can be more marked if the speaker is at the angle between the wall and floor, or more so if right in a corner. Bass can be increased by mounting the speaker on its' side with the bass unit nearest the adjacent wall. A mid-'70s article (Radio-Electronics Dec '75 p51-53, Bookshelf Speakers Part 3) gave good advice on 'auditioning' bookshelf speakers, JPGs of these can be supplied.

If the speakers are incorrectly phased, there will be a reduction in bass and a 'hole' in the sound stage. If there is any doubt in this context (unmarked terminals, cable, etc), place both speakers together and, either playing music in mono with a high bass content or turning the treble down, swap the wires around on one speaker. Which ever position yields the greater bass is the correct one. If the speaker units are unmarked and unmounted, apply a 1½V battery to the terminals and whichever result moves the cone out (away from the magnet) mark the terminal wired to the + terminal on the battery as the + on the speaker.

Peaks and troughs in the frequency response caused by the dimensions and shape of the listening room, and the objects in it, can add colourations too, frequency 'holes' of some 15dB or more being measurable from room to room. For example, a rectangular room of 4.2 x 3.4 x 2.5m has damped resonances at about 40, 50 and 70Hz. A gap between a chimney breast and a near wall can produce an additional room resonance of 500Hz. Changing the listening position or even the drawing of curtains can have a distinct effect also. A test with several ESLs showed their tendency to 'honk' in a room with a tiled floor, but not in a room with thick carpet, reflections being considerably reduced. To illustrate this, below is a reproduction of a page from Bruel and Kjaer application note 13-101 showing the frequency responses of 5 different speakers (down) in 3 different living rooms (across).

If a frequency response curve is made over some minutes, rather than seconds, sharp dips and peaks of 20 to 30dB only a few Hertz apart are then recorded, eg;

Given the above, one can immediately question the fashion some audiophiles adhere to for utilising no equalisation whatsoever, when, with professionals, this is one of the first tools sought after initial levels are set. The bold presumption that the audio chain and listening conditions will naturally be perfect is obviously flawed. Similarly, modifications to a crossover filter may be attempting to remove a deficiency that is not resident in the speaker design, but instead is the result of a reflection or 'hole' in the listening room.

Some older notes relating to spectral qualities are copied below.

Many manufacturers test speaker designs in specially constructed anechoic chambers. Some design speakers intended to accommodate 'the average living rooms'' deficiencies. Both approaches can, however, have no bearing on reality given the variables extant. For an excellent discussion of reflections see Jim Lesurfs' "Imagine a Room Full of Mirrors".

As can be seen, the listening area itself will have a greater impact on the quality of sound produced, rather than the quality of the equipment used. If a dealer then has a listening room, never feel pressured to make a 'preferred' choice, by pointing out the differences beween the dealers' listening area and your own. After all, the salesman, whose main interest is commission, will never have been there, neither will he be able to produce certification that his room is anechoic to international standards or that his hearing is perfect too. Some designs are more suited to classical music, others to electronic, but whatever the choice a good speaker will reproduce all types of music neutrally.

As with all things, over-rating is the best way to avoid breakdowns caused by under-rating. Performance will improve and components exposed to proportionately less stress will invariably last longer and sound better, especially at elevated temperatures. However, a manufacturers' data can be confusing. A good example to take are power output ratings. Systems intended for automotive installation often quote specs that verge on the ridiculous, if not incredible. One item can quote some 420 watts PMPO, which sounds impressive; but this translates as 12 watts RMS, which is quite ordinary. Watts PMPO appear to be variable in that another item quotes only 80W PMPO for some 18 watts RMS. Similarly, a small speaker is described as having a 'sound output of 2 watt', a ludicrous figure, when obviously what is being refered to is the driving amplifiers' power output (LM380 IC), the 100% efficient speaker having yet to be seen. If a supplier is coy about giving a continuous rms rating for their products, don't buy them.

An average speaker will have an efficiency of between 0.25% and 2.5%, the remainder of the input energy being converted to heat. This may not sound much, but it must be borne in mind that (universally) a speakers' sound output is expressed in dB with a 1 watt input, which can be loud. Just 3W can have neighbours rushing to complain. To move the same amount of air, a small speaker will have to have a greater voice-coil displacement than a larger one. The voice-coil of a modern bass unit may be able to move some 3-5mm before leaving the uniform field of the ring magnet. However, beyond a displacement of about ±2mm the spider supporting the cone or diaphragm will tend to decelerate it. It is because of this that many woofers produce distortion at even medium input powers.

Electrical power handling may conflict with mechanical specs, for example, a reputable 25cm unit may produce a maximum distortionless output of 103dB at 16W (60Hz) when rated at 110W, whilst a 30W 13cm unit of similar quality will produce 87dB with only 2W. Having decided on a design, ensure that the speaker, whatever its' size, will never be pushed beyond 50% of its' rating, or less being safer. Experience suggested that a ratio of 10:1 between the stated or claimed continuous rms power ratings of speaker and amplifier was realistic.

In June 2007 a 'nice' and expensive (564) small speaker used 2 off Seas MidBass L11RC/P aluminium cone drivers and a 25TFFN/G 25mm soft dome tweeter in a MTM configuration giving a frequency response of 50Hz to 20kHz (-6dB points) with a quoted power handling of 100Wrms maximum. A nominal impedance of 4 ohms was offered with minimum impedances of 3.2 ohms at 220Hz and less than 4 ohms at 140 to 520Hz (rather low from the authors' point of view). Sensitivity was quoted at 87dB with a 2.83Vrms input, or 2.5Wrms at minimum ohms. Experience suggests that the sensitivity refers to a maximum undistorted output and yet the maximum power handling is a figure 40 times higher than this. The lower figure is realistic, the upper is not.

More recently, a disappointed user sought an explanation for the demise of his pair of T1 Audio T1-124 12" automotive bass units. The label on the magnets gave two parameters '1200W' and '4R'. To dissipate 1200W into 4R requires 69.3V at 17.3A (consider the heat output of a comparable load like an electric heater, kettle, cooker ring or washing machine motor). Even if the figure quoted is peak-to-peak this implies a 300Wrms capacity. The enamelled Cu wire from which the voice coil was wound was measured at 0.5mm dia (25swg, fuse rating 15A). If it was 24swg (0.559mm dia) the fuse rating is 17A. Frankly, given the voice-coils' construction and size (only 54mm dia by 21mm long, 4 layers), a reasonable maximum at a tenth of this figure (30Wrms continuous, comparable say to a soldering iron) is considered realistic, if the speaker is to survive any reasonable length of service. With care a decent speakers' life can easily exceed 30 years, rather than days. Apart from the very optimistic power rating, prior to fitting the speakers had been put into a vice to bend the basket rims so that they would fit more neatly the bodywork they were being mounted on. Clearly no thought had been given to how such a distortion could alter the centring of the voice coils, one of which did show signs of abrasion.

It was reported that these speakers had failed simultaneously although this is unlikely. Both had failed open-circuit so one had probably failed first and (probably) the volume then increased to overcome any perceived deficiencies until the other gave up the ghost. Having worked with large bass arrays (and the amateurs who use them, especially in respect of reggae systems) some owners have refused to accept that up to half of the drivers have failed and require replacement. If a sufficient sound output is available from the surviving units, then what appears to be of utmost importance is the visual presentation of a large number of bass drivers whether they are actually functional or not.

One study (elektor, 6/86) concluded:

High efficiencies are only possible with large effective diaphragm areas.
Large cone areas result in lower distortion than small diaphragms.
The efficiency cannot be improved by more than 6dB however much the electrical input is increased. The main reason for this is that, particularly at low frequencies, the mechanical power handling capacity becomes the limiting factor.
The electrical power handling capacity, because of modern construction methods and improved speech coils, has become one of the least important parameters of a loudspeaker system.

If using an oscillator (sine wave) to test speakers always remember to test at low power levels. 1W should suffice, as 2W or more at high frequencies can destroy tweeters. Test out-of-doors on a flat hard surface with the microphone on the floor at a distance of 1 metre with the speaker being measured as close to the floor as possible (this method reduces erroneous measurements from reflections). When the frequency is adjusted to a point where the output falls to half that of the reference, this can be noted as a -3dB point. Drop-out or absence of a signal may mean that the tone is out of human, or the equipments', range and that the amplifier, speaker or say a crossover component is getting very hot instead. High frequency testing can result in troublesome radiated harmonics. For this reason, square waves are not recommended. Pulse (amplitude-modulated) techniques can be used to test a speakers' dynamic performance, but again, care must be taken to eliminate the measurement of reflections. A useful 'Audio Oscillator with Tone Burst' was described by J.T.Tiernan in the October '82 Wireless World.

Filtered noise can be used, however, speech and music can be perfectly adequate sources. Setting the speaker where it will be heard and any tone controls and filters to flat, adjust speaker inputs (usually the tweeters') for as natural speech as is possible at a level commensurate with those talking actually being in the listening area. Music should confirm the initial assessment, bar any changes to bass emphasis.

Controls (for, say, the tweeter and mid-range units) can be brought to the speakers' front panel to ease adjustment, although decent L-pads can cost more than a driver. When adjusting a passive crossover design, cables from the units can be brought outside the enclosure, say via a hole intended for the connectors, provided this is made airtight during testing to ease access and work on the crossover.

Ideally, there should be nothing other than the speaker cable between a speaker and amplifier. However, occasions arise when 'low-loss' cables can cause problems by adding capacitance. With a low-impedance load offering a lively reactance, like isobariks, it can be seen why. Jim Lesurf, a designer with Armstrong, has an excellent site where he mentions 'The Society for Cruelty to Amplifiers'.

Most amplifiers have an output relay that trips in the event of a DC failure. If a speaker is valuable, will be inaccessible, can be expected to be abused or is intended to be part of an array, the fitting of an internal cut-out makes eminent economic sense. Set this to trip at a sensible level below the maximum handling power. An LED across the cut-outs' relay placed on the speakers' front will quickly identify 'losses'. A large system, thus protected, shuts down comfortably as the volume approaches criticality, and automatically restores as the volume is reduced, a boon if the inexperienced have access to the controls. In fact, to prevent speaker overloads, a suitable bulb can be put in series with a speaker, the bulb being cheaper to replace. As the load is driven harder, the bulbs' filaments' impedance rises, then limiting the current through the load.

Such an approach is useful when faced with unknown or dubious loads and is often used in professional systems.

Those who persistently overdrive amplifiers into low-impedance loads are usually blissfully ignorant of the cost of repairs, when required, and will invariably claim burnt-out equipment is inadequate, rather than themselves. Similarly, a speaker coverages' tendency to 'beam' or narrow like a spotlight when overdriven will be missed by those who simply like loud noise.

Passive crossovers are viable for systems up to 500W. By their nature, these introduce losses and component failures can be ignored, fault often been sought elsewhere presumably because they are hidden away. The Beovox M100 had 10 caps, 6 coils and 9 resistors in its' crossover, apart from the five speakers and protection circuit, and some of these could get very warm! Such a speaker would be expected to give an adequate sound output with an power input of about 5W, or less.

It can clearly be seen from the above how a small investment in a handful of common components can be an insurance against greater loss.

Biamplification, where the spectrum is split between a woofer and tweeter, usually between 500 - 1k6Hz, has the advantages of allowing smaller amplifiers (often in-built) to produce a given sound pressure level whilst reducing distortion from another part of the spectrum being overdriven.

Some integrated active systems can be complex (and expensive) to build, especially if compact labyrinth or transmission line designs are involved, but these can offer a flexibility. An instructive project, by Giesberts and Baggen (elektor, Jul/Aug '93, pp8-12), offers Linkwitz correction. Some newer designs can mix class AB and D amplifiers, some very compact models employing pure class D. With such an approach, the need to ventilate heat-sinking is negated, thus placing less constraint on the cabinets' design.

Where active crossovers are used in PA systems, care can be taken to ensure that speakers are not run outside their design bandwidth, and any distortion, losses or costs inherent in passive crossover components are avoided. Signal conditioning can then be achieved with smaller, but more accurate and higher quality, devices. Protect HF drivers from switch-on transients with a decent high-voltage bi-polar capacitor in series with each; formula = 40,000/fc µF, where fc is the crossover frequency. Some piezo designs require a 47R, 3W in series to limit high frequency resonance. Best slopes are considered to be 12dB for bass/midrange and 18dB for tweeters/horns or, if desired, higher slopes of 24dB can be used instead. 'Filler-drivers' for cross-over points can be used although their subjective value may be questioned given the extra outlay required in respect of the extra speakers, amplifiers and filters required.

Some custom layouts have demanded load limiting with heavy-duty resistors in series with the speakers (passive crossovers), common arrangements involving multiples of 10R, 50W (20W in free air) resistors. At no time has this arrangement ever been noted by listeners. Speaker systems thus 'ballasted' appear to have greater longevity.

Some may disagree, but experience has shown that when faced with building an array with a quantity of assorted speakers (even of industrial proportion), aiming for a higher load impedance appears to result in a more pleasing sound. Failures are also notably fewer. In practice, there appears to be no listener so far met who can determine a speaker systems' impedance by its' performance. Low-impedance speakers can introduce problems where previously there were none, their reactance adding to the capacitive effects of 'low-loss' cables. Similarly, passive crossovers feeding low impedance units will require larger value, and thus more expensive, components. For class A, AB, B or H outputs use 8-15 ohm loads, thus reducing THD, incidence of clipping and increasing damping factor. For low-voltage class C or D, use 4 ohms, although some (usually higher voltage) designs will perform better with 8.

Leslie speakers, often used in conjunction with upright organs where the speaker is rotated in its' cabinet, can fail because of the bearings used. These can be mercury-wetted and over time the mercury will particulate, ie; break-up into small particles, giving a high-impedance characteristic. If the bearings are removed and subjected to a sharp physical shock along the axis of rotation, the mercury will then resume its' former liquid and low-impedance state.

An older speaker can go intermittent if say a soldered joint on the cone has 'dried'. A delicate touch is needed here or the cone will be easily ruined. Intermittancy on warm up could be due to a break in the voice-coil, contact being remade on cooling.

Motional feedback speakers (where voltage induced across a coil wound on the voice-coil former is fed into the feedback loop of the driving power amplifier) appeared in some active designs but did not prove popular, presumably due to the expense of providing the extra coil. To overcome this, some designs derived a signal by mounting an electret microphone near (1cm) the speaker cone, then mixing this (positive feedback) with the incoming bass signal.

With high-cost (and very directional) ESLs or top-line digital speakers, avoid those anxious or keen to reduce reflections in rooms with capacious windows or 'minimalist' designs (all hard surfaces can add very significant reflections), optimum subjective listening positions being, generally, severely limited by perception which can usually be described as intense, at best. Some systems, with in-built automatic digital signal processors such as the BeoLab 5, can even self-adjust, to a degree. The Visaton DS4s, using two different DSPs and (for stereo) 8 amplifiers, needed a sound engineer to fully adapt to the listening room. Equipment such as this may be 'nice' and may impress but is incapable of altering a rooms' acoustics alone. Reflections, reverberations and even the consequences of moving the furniture can upset some sensibilities. Similarly, disappointment will arise when it is determined that such equipment is neither child or drunk-proof, exposed speaker cones being especially irresistible. Rather like valves. Since a number of owners find repairs unaffordable, a decent insurance policy (all-risks) is advised.

Some speaker loads, like ESLs and isobariks, can be difficult for amplifiers to run and disappointment can arise when expenditure does not match expectation. On occasion, a mismatch can result in one part of the audio chain deciding to 'eat' the other! To avoid this, research exotic types that exhibit lively loads and obtain assurances that equipment is matched to a required situation, and volume levels, before signing on the dotted line. Haden Boardman offers some good advice on ESLs which, as he notes when their operation is not understood, can be destroyed by amplifiers' exceeding 15W. The important thing with ESLs is not the drive power, but the amps' ability to withstand short-circuits, and often it is suggested that a high power 2R5 resistor should be wired in series with them to maintain stability. The same can be said about isobariks.


A deep bass needs a large room to obtain the desired movement of air or wavelength. The lowest frequency obtainable in any room is relative to its' maximum dimension. For this, at 20Hz, you need a space of 17m (56')+, across which you can generate the desired movement of air. At about 300Hz and below, bass performance will generally be decided by the form of the listening room and its' contents, including the listener! Placement of sound sources will figure largely, even more so than the quality of the equipment used, although this in itself cannot be underestimated. Altec and Electrovoice made sturdy bass units, but Eminence did well too.

When a bass speaker starts to 'fart', this is not 'true' bass. It is the sound of overdriven cones hitting the limit of their travel and indicates that permanent damage is imminent, if not already extant (spider crashing on top plate, voice coil bottoming on back plate, voice coil coming out of gap above core or physical limitation of cone, etc). True bass you feel. It makes you melt.

Personal preference favours large diameter bass speakers, say 18" 400W, in big boxes, such as scoops or w-bins, although high-power 12" units with resonances of, say, 22Hz or less are definitely worth a look. The acoustic impedance experienced by a speaker cone reduces as the inverse square of the frequency. A bigger cone is more efficient and to achieve an enclosure that matches the resonance, this too must be large in order to reproduce the lowest octave faithfully, rather than making a 'boomy' special effect. A high power driver will also be less likely to exhibit excursion and power handling limitations.

Bass reflex (ported) cabinets are physically smaller than sealed boxes and therefore are more economically viable for high-power systems. Sealed enclosures, using the conventional air suspension principle are not really suited to high-power applications and are more suited to domestic situations. The intensity may be increased when mounted in a bass reflex (increasing the effective diaphragm area) or horn enclosure (increasing the radiation impedance). Transmission lines are considered best, but can take up inordinate amounts of room if not 'folded' which can detract from a sub-bass units' performance.

Domestic arrangements cannot always cope with the industrial bass approach where a sealed enclosure can approach the dimensions of the listening room. However, there has been an increase in interest in smaller bandpass designs, which some consider a dark art. Many constructors find a 10" unit sufficient for, say, a living room, such a size requiring a smaller enclosure. However, a recent demonstration, using a class D driven 15" 350W speaker with an internal volume of only 29 litres, showed how far this art has come. Speakers can be driven outside their quoted frequency ranges, albeit with reduced power rating and Linkwitz correction (see also below) can overcome the adverse effects that a smaller enclosure has on low frequency performance. The 350W sub-bass unit above (100mm dia voice coil, res 35Hz) would sensibly be driven at about 30W max. Older amplifier designs have been modified for experimental purposes. Alternatively, newer more efficient designs can be purpose-built. Interesting and reasonably compact sub-bass (20-80Hz) reinforcement can be achieved by mounting a bass unit at the end of a tube, utilising the same resonance effect as a church organ.

Sub-sonic performance, though desirable (cinemas may have resource and room to devote to kilowatts at 7 or even 3Hz!), may have unforeseen consequences. Thus, if experimenting in this range, caution is advised in respect of high power levels affecting people and objects, once their natural resonance is hit. Special precautions may then become necessary for tone arms, ornaments, picture frames, furniture, windows and doors. Humans have a variety of resonances depending on the density of different body parts and some people may be intolerant and become uncomfortable. Architectural resonances, for example, will alarm neighbours and can result in frantic calls to the emergency services! A paper written in respect of 'brown noise' suggests that a frequency of 22.25Hz will have consequences for the human colon, although in practice this has not been noticed. A frequency of about 20Hz can 'rattle' eye-balls and some 7Hz has even been associated with the perception of 'supernatural' experiences. Other reported side-effects include recklessness, euphoria, low efficiency, dizziness and nausea. Most music rarely goes below 40Hz, although film soundtracks can, and do, especially 'blockbusters'. A lower roll-off above 15Hz then seems sensible, although many find 30Hz sufficient. Run bass units in positions that 'throw' across the length of a room, even above a crowd. Avoid corners if you have neighbours.

The 'Low-cost Horn Loudspeaker System' by Toneburst (WW, May 1970) contradicted this advice and used concrete.

Speaker cables

Extensive tests ('blind' and otherwise) were conducted using identical lengths and runs of 2-core 79 x 0.2mm Cu 18A (0.58p per m) and 3 x 42 x 0.12mm silver-plated OFC 25A (7.65p per m) speaker cables with a variety of amplifiers (solid-state and thermionic) and speakers. No discernible improvement was heard or measurable using the (>13x) more expensive cable. In fact, in some instances instabilities arose, particularly with isobariks, creating problems where there were previously none, the capacitance of the cable adding to that of an already reactive low-impedance load.

Low-loss cables can cause problems (ringing), particularly with square waves and other transients, if an output inductor is fitted to an amplifiers' output, this being intended to counteract the capacitance in a reactive load making it appear more resistive to the amplifier. In such circumstances, replacing the inductor with a high wattage 0R22 resistor can improve the transient response.


Ultrasonic transducers can produce tight beams of acoustic energy and, if amplitude modulated, this can be useful, say, in communicating discreetly over distance. A horse race in the UK was 'nobbled' by transducers mounted in a pair of binoculars, the horse targeted thus convinced it had a swarm in its' ears. The binocular case contained the power supply. A US company produces industrial duty types suited for the military or law enforcement.

Back in the '70s, a welding set whose spark was modulated by an audio signal suggested a new approach to spherical acoustic radiation patterns, although definition, especially that of bass, was severely limited. A similar principle was demonstrated by Magnats' (Wagner & Wagner 1980) Corona Plasma loudspeaker - a 27MHz oscillator of 20W was amplitude modulated by an input which then was used to generate a high voltage developing 15,000°C at a needle point giving an output of 3kHz to 200kHz at 114dB in all directions.

Tip: if speakers are to be moved/transported, shorting the speaker input will result in the voice coil being 'held' by the magnets' field, thus helping to protect it from shock. Remember to remove this link before use. A simple arrangement can use a relay to protect a combos' speakers, for example, which should ideally have an individual power rating far greater than that of the mains transformers', if single or paralleled.

Thought has been given to a tilt-switch sensor for such a circuit, to protect a falling speaker before it hits the deck, combos being particularly prone. However, none have been tested.

speaker design considerations

more of a similar bent

Contact me at paulkemble@hotmail.com
especially if you want additional content to this page
or if you find any links that don't work. Don't forget
to add the page title or URL. Take care!