|Elliott Sound Products||Loudspeaker Cabinet Design|
One of the most popular pastimes in the DIY audio world is building loudspeaker systems. A web search will reveal literally thousands of different designs, a great many of which are (at least superficially) quite similar. It's highly likely that most of these will sound different from the others, although it's almost guaranteed that the designer will claim that his/ hers is 'better' in some way. We can be fairly certain that some of the published designs will sound very good, and others awful.
This is reflected in commercial offerings as well. There aren't many 'real' audio brands that will be truly awful, but there will be differences. This is often despite the fact that many will show frequency response (both on and off axis) to be very similar, with many sharing one or more of the same speaker drivers as used by other manufacturers. It can be very difficult to work out just why (and how) two apparently near identical designs can sound different.
Loudspeakers are the most subjective component in any audio system. Amplifiers and preamps are routinely so close to being a 'straight wire with gain' that measurements can be difficult. While CD and SACD players (as well as DVD players) definitely sound different from vinyl, blind testing the preamp-amp combinations will most commonly result in a 'null' outcome - it's usually not possible to correctly identify amp 'A' from amp 'B' with a statistically significant result. This is often not the case with loudspeakers.
There are several things that can change the sound of a loudspeaker system, even when using driver components that are identical to another system. Some of these differences will be due to the way a system has been 'voiced' - a term that means adjusting the response so the system sounds balanced and 'right' from the designer's perspective. Very few loudspeakers have a truly flat frequency response, and the way the system interacts with the listening room also changes the sound.
Few hobbyists today would argue that enclosure panels need to be rigid and acoustically 'dead'. How you get there depends on the philosophy of the designer. At one stage, hollow, sand filled panels were popular. These are certainly likely to be acoustically dead, but are difficult to make. It may or may not be possible to refill the panels after the sand has settled or the panels have expanded as the sand compacts and tries to force the panels apart.
Concrete has been used, sometimes with tiny pellets of expanded polystyrene foam to reduce the mass (so the box can actually be moved), sometimes only the baffle may be concrete. Different types of plywood are used (and no, birch ply (for example) should not sound different from some other tree species). If one box sounds different from another (identical other than material), then the material is not damped properly. Once something is acoustically dead, it doesn't matter what it's made from - dead is dead. Differences are due to resonance(s), meaning that one or more panels are not dead at all.
The cabinet shape can make a difference, even if the enclosed volume is exactly the same. While the panels may be acoustically dead, the air space within is not. The enclosed volume should never have two internal dimensions the same (such as top to bottom and front to back) as that will usually reinforce standing waves at certain frequencies. The air within the box can be made somewhat acoustically dead by adding damping material - fibreglass 'wool', or any number of proprietary filling materials that are designed to absorb the sound inside the enclosure. You'll find claims (well, perhaps not quite) that only virgin yak's wool should be used, because man made fibres 'sound bad'.
When these materials are added loosely, the effect is to make the enclosure acoustically larger. If packed in tightly, the enclosed volume is smaller. Both of these change the way the loudspeaker driver reacts with the enclosed volume, mainly at or near the speaker's resonant frequency. In many (but I suspect by no means all) commercial designs, it's expected that the driver interaction with the filled volume will be modelled and measured, and the filling adjusted to get the right amount of absorption, while minimising internal reflections to the point where they can 'do no harm'. Even this term will be variable - some will claim that -40dB is ok, others may insist on at least -60dB, while others might be content with -20dB.
Some insist that any enclosure is bad, and the speaker drivers should be free, allowed to show their naughty bits to the world should anyone peek around the back. Open baffles create a dipole effect - the sound will be (generally) equally loud directly in front or behind the speakers, with (theoretically) zero output from the sides and top. This won't be the case, but again, should the side response be -20dB? -40dB? More? Less? This is almost impossible to answer.
Such systems interact with the walls, floor and ceiling of the listening space very differently from a 'conventional' enclosure. Positioning will usually be fairly critical, but there are many who are firmly convinced that this is a better way to build a speaker. There are (of course) others who claim exactly the opposite, that enclosures are essential and that the open baffle idea is flawed.
Many people design cabinets with the deliberate aim of avoiding all parallel surfaces. This prevents (or helps to prevent) standing waves from developing within the enclosure, and is generally a good idea. However, it's not easy to do without dedicated machinery that can cut precise odd angles so that it all fits together. In some cases, you may find that adding an internal baffle at an angle within the enclose space will work, and if it's well perforated (to ensure that the total internal volume is available to the rear of the speaker cone) it may be enough to prevent major standing waves. Acoustic damping material is still needed, no matter how irregular the interior volume. The idea in most cases is to absorb the rear radiation from the speaker completely, because any sound that re-emerges through the cone will not be in phase (or in time) with the original.
The 'acoustic labyrinth' type of speaker is a (fairly serious) extension of this principle, with the length of the 'tunnel' often used to create a transmission line to reinforce bass frequencies. These cabinets used to be very popular amongst DIY constructors, but seem to have fallen from favour over the last decade or so. Part of the reason is that they are difficult and expensive to build, and the results may be rather disappointing after you've gone to all that trouble.
This article is not about specific designs. You won't find any cabinets, dimensions, crossover circuits or anything else that is available in countless books, magazine articles or websites here. What you will find is general guidelines, many based on 'ancient' knowledge, and others that are more-or-less common sense. The idea is to provide some basic information that can be used in the design of any cabinet, regardless of the drivers used.
In general, the guidelines are intended for domestic hi-fi applications, not commercial, public address or sound reinforcement systems. These have many other constraints, in particular weight and cost. When building your own systems, these are generally secondary, and the extra cost of adding an extra brace or more damping material is small compared to the overall cost of the project.
There is only a little about specific materials that could/ should be used for cabinet construction. See Section 10 for a bit more on this topic. Some people loathe MDF but love plywood (whether exotic or otherwise), and others are exactly opposite. Some materials may be difficult to get (or very expensive) in many places. One recommendation I will make is to avoid 'chipboard'. While it is still a popular material for some applications, it's generally not robust enough for a speaker cabinet. Veneered chipboard is somewhat better, but the material's structure is such that it's not easy to make a rigid box, and radiused edges expose the coarse grain structure which is time-consuming to fill to get a good surface finish.
Some cabinet shapes can be fabricated using fibreglass, but that requires a mould that is used to form the cabinet shapes. Unless you are experienced in the use of fibreglass (or carbon fibre), it's hard to recommend for hobbyist enclosures. The glass fibres and resins used are potentially dangerous without a proper face mask to prevent inhaling the fumes and/ or glass fibres. Fibreglass panels can also be quite flexible, which allows the panels to radiate sound as they flex, and bracing can be difficult to change if it's moulded into the structure. Attaching anything inside from the outside surface is generally impossible because the outer surface is usually the final finish, and external fastenings can't be concealed.
I suggest that prospective builders look at Project 181, an easy to build accelerometer intended for measuring the movement of speaker enclosure panels. It's highly recommended, because without an accelerometer you have no idea how much the panels are flexing or their resonant frequencies. Lacking this, you may end up with an enclosure that just doesn't sound 'right', even when you think you've done everything correctly. Panel resonance is always difficult to assess unless you have a way to measure it, and then take new measurements to see if the issue is fixed or not.
I do not suggest or recommend commercial software used to design speaker enclosures, with the one exception of the free program WinISD (you can find it on the Net). There are countless programs that either do (or purport to do) complete designs, based on the drivers you are using. These omissions are not because the software doesn't work, but simply because I operate as an independent individual, and I do not make specific recommendations for anything, other than components used in project articles.
Other than a few general hints here and there, I also won't be discussing general woodworking methods, choice of adhesives or finishes, the use of power tools or anything else that's well catered for all over the Net. It goes without saying that you need an area where you can generate copious amounts of sawdust, and another area (completely free of sawdust) if you plan on any high quality paint finishes. For example, classic 'piano black' tends to look a bit tatty if it has dust particles embedded all over it. Various power tools are essential, although simple enclosures can be made using only hand tools for the truly masochistic constructor .
In all cases, there will be loudspeaker driver parameters that are different from those claimed by the manufacturer, and in some cases the necessary data (in particular the Thiele-Small parameters) that are either quite wrong or missing altogether. If this is the case, I suggest that you read the article Measuring Loudspeaker Parameters, as this requires no specialised equipment and gives good results. Fairly obviously, this also extends to recommendations for particular vents or passive radiators.
This article also (deliberately) avoids making any recommendations for drivers. There are so many, and they often have a very short manufacturing period. The driver that one person loves may well be hated by others (often for obscure and illogical reasons for both 'love' and 'hate'). There's also the issue of availability - there's no point recommending a particular driver that's only available in one country, because no-one else will be able to get it easily (if at all). I will suggest that you avoid drivers that show sharp discontinuities on the impedance curve. I've run tests on a few such drivers, and the impedance discontinuity usually corresponds to a response anomaly which can be such that it simply cannot be ignored (nor equalised!).
Given two drivers that are otherwise identical (or sufficiently close over the required frequency range), the one with higher efficiency is usually the better choice. However, this is not an absolute position, as there can be other things that influence your final decision. This may simply come down to appearance - a driver that sounds great but looks ugly usually doesn't rank highly, unless it's hidden beneath grille cloth. Not everyone like grilles, so appearance can be important. For many people, appearance is a major factor, and disguising an 'unappealing' driver's basket-front can be an expensive and difficult undertaking.
There are many different types of enclosures, and it's not possible to cover them all in any detail. Of those listed, they are shown (more or less) in order of complexity, from the simplest to the most challenging to build. Some are very common (simple sealed and vented enclosures for example), with others used primarily by hobbyists and a few 'boutique' manufacturers. While most of the drawings are shown with a single driver, in the majority of cases there will be at least one other (a tweeter), and in some cases there will be a secondary enclosure containing a midrange driver.
No 'esoteric' enclosures are covered here. It's assumed that anyone who wishes to undertake something that quite out of the ordinary will have the necessary skills to ensure that everything is done correctly. This isn't always the case of course, so anyone who does want to make cylindrical or spherical enclosures (or anything else with a 'weird' shape) should still find many of the suggestions helpful.
Remember that the volume occupied by the speaker driver(s) needs to be added to the total volume calculated, and if a port is used, the volume of that must be included as well. The same applies to bracing materials - they all occupy space in the enclosure and need to be accounted for. You may find that you need to add extra bracing once the enclosure is (almost) finished, so a bit of extra volume can be added just in case. You can usually change the internal volume by a small amount without it having a serious impact on performance, and remember that the listening room will have much greater effects on overall sound quality than any small miscalculation of internal volume.
Speaker parameters are not absolute numbers, and in some cases they can be way off. It's always wise to measure the Thiele/ Small parameters yourself. There's an ESP article on this topic - see Measuring Loudspeaker Parameters for all the details. There are many other articles on the Net that describe speaker parameter measurements, so use the one whith which you are most comfortable.
The open baffle or dipole speaker is favoured by some, most notably the late Siegfried Linkwitz. An open baffle (or open-backed box) was used from the earliest days of amplified sound, and is by far the easiest to build. Ideally, the baffle should be large compared to wavelength (the 'infinite' baffle), but this is very difficult to achieve at low frequencies. So, while they are easy to build, they are not so easy to design (or even produce) in sizes that suit low frequencies. One wavelength at 100Hz is already 3.43 metres, so the size rapidly gets out of hand.
Figure 2.1 - Dipole 'Enclosure' ('Infinite Baffle'/ Open Backed)
For higher frequencies, it can be argued that dispensing with the box prevents internal reflections. This is quite true, but of course the rear radiation is introduced into the room, which has its own reflections, most of which are completely unpredictable and can be a lot harder to deal with than an enclosure's internal reflections. Open backed speakers are very common for guitar amplifiers, where the open back provides a stage sound that most guitarists prefer. An open backed box can be likened to a flat baffle that's been 'folded' to reduce its size. Of course, this also protects the rear of the speaker from damage in transit - especially important for guitar systems.
The sealed enclosure is very common, and can work very well if the internal volume is calculated to match the speaker's characteristics. The Thiele-Small parameters of the driver will show that optimum performance requires an enclosure of just the right size. If it's too small there will be a pronounced bass peak, followed by a sharp rolloff at 12dB/ octave. Of anything that would qualify as an 'enclosure', this is the simplest.
Figure 2.2 - Sealed Enclosure
Rather than being radiated into the room, the sound from the rear of the speaker cone is absorbed, using proprietary fibre mats, felt, carpet, fibreglass, or a combination of these materials. Ideally, no rear radiation will be reflected back through the cone, something that becomes critical at midrange and higher frequencies. Bass can be very good (often with equalisation), but this requires drivers with a larger than normal maximum excursion (Xmax). Sealed cabinets are common for instrument amplifiers (guitar, bass, keyboards).
This is probably the most common enclosure in use today. It was used in very early speaker systems, but it was basically a 'trial-and-error' design until the loudspeaker parameters were properly quantified by Neville Thiele and Richard Small. This allowed mathematical calculation of the enclosure and port sizes, and it was then possible to design a system, build it, and have it perform as expected. Many of the early 'tuned' boxes were what's now commonly referred to as 'boom boxes', because they had excessive and often 'one note' bass. Countless programs have been written to allow users to design an enclosure, based on the Thiele-Small parameters. This has removed much of the guesswork, but by themselves, the programs are (mostly) unable to provide a complete design. Most provide the necessary internal volume and port (vent) diameter and length, but further 'tweaking' is nearly always needed.
Figure 2.3 - Bass Reflex Enclosure
In these enclosures, the rear radiation is utilised to boost the bass response below the loudspeaker driver's resonant frequency. The combination of the enclosure volume and the vent length and diameter form a Helmholtz resonator, which (when done properly) reinforces the low frequency response without creating excessive bass and/or poor transient response. It's important to understand that the Thiele-Small parameters are 'small signal', meaning that the performance is not necessarily the same at high power levels. Only the bass region is affected by a bass reflex enclosure, and mid to high frequencies still need to be absorbed within the enclosure.
A variation on the 'traditional' bass reflex enclosure uses a passive radiator. This is pretty much a loudspeaker with no magnet or voicecoil, and it's generally tuned for a resonant frequency somewhat below that of the woofer. Some have weights that can be added or removed to tune the resonant frequency of the radiator. These have some advantages over a port, in that there is no possibility of 'chuffing' or other noises that a ported enclosure can create if the air velocity is too high.
Figure 2.4 - Passive Radiator Enclosure
Fairly obviously, a passive radiator takes up more space on the baffle than a port, but some people prefer them for a variety of reasons. This is a configuration that seems to be somewhat 'seasonable', gaining or losing favour for no apparent reason. There used to be many passive radiators on the market, but they appear to be less common than they once were.
An aperiodic enclosure is (kind of) halfway between a sealed and vented box. The vent is deliberately restricted, so it's either a leaky sealed box, or a 'constricted' bass reflex. There's quite a bit of information on the Net, but not all of it is useful, and design equations are hard to come by.
Figure 2.5 - Aperiodic Enclosure
The above is one of many different ways that an aperiodic enclosure can be configured. This isn't a technique that's widely known, and it's also not one I've experimented with. Many claims are made, and there are many variations - in some cases, just a small hole or a series of narrow slots is used, with appropriate damping material covering the openings. There appears to be little consensus from designers, so the technique is somewhat experimental. It's claimed that with an appropriate aperiodic 'vent' that the enclosure is made to seem much larger than it really is, and it's not uncommon to see aperiodic enclosures that appear much too small for the driver used. As I said, I've not tried this approach, but may do so when time (and motivation) permit.
Isobaric speakers are not particularly common, and are only ever used for the bass region. The benefit is that the required cabinet size is halved compared to a single driver, allowing a more compact system. The disadvantage is that the efficiency is also halved, because the same power is fed to the two drivers, but output level is not increased. Although the drivers are shown 'nested', with the front driver partially inside the rear driver, they can also be mounted face-to-face. The enclosed volume between the drivers must be small to ensure optimum coupling.
Figure 2.6 - Isobaric Enclosure
Isobaric enclosures can be used with or without a vent, depending on the desired outcome. Most speaker design software can accommodate isobaric configurations, but the mechanical details can be awkward to produce. There are some commercial isobaric enclosures, but they aren't especially common in the market. This is a good design to use if the driver you wish to use requires a box that's larger than you can accept, but no isobaric enclosure should normally be operated above around 300Hz or so. The cost, weight and relative inefficiency of isobaric enclosures limits their usefulness for commercial systems.
These are probably one of the most challenging to build, but can produce a great deal of bass over a relatively narrow bandwidth. They are used only for bass, as the dimensions are not suitable for higher frequencies. As the name suggests, these enclosures are an acoustic analogue of an electrical bandpass filter. They can have very high efficiency, but the enclosure is sensitive to variations of driver parameters. If a driver fails, it must be replaced by one with near identical parameters, or the response will not be as expected.
Figure 2.7 - Fourth Order Bandpass Enclosure
Although a fourth order box is shown, the sixth order enclosure is also used. These have an additional vent between the speaker's rear enclosure and the front resonant chamber. Bandwidth is usually fairly narrow, so they cannot reproduce a wide range of frequencies. Fourth order systems are fairly common for large sound reinforcement applications, where it's (apparently) more important to create a vast amount of noise than to consider fidelity. This isn't always true of course, but it does seem to be the case in many of the systems used for very large audiences. Some care is necessary to ensure that the effect isn't 'one note bass', where the bandwidth is so narrow that they sound as if only one note is audible (many automotive installations suffer the same problem).
These are without doubt the hardest to design, and even small variations from the 'ideal' can cause serious response anomalies. Because of the acoustic filter, some people will say that this enclosure type is responsible for 'day late' bass - there is often a significant delay from the application of a signal before the resonance is stimulated sufficiently to produce output. The delay is usually somewhat less than a full day, but you get the idea . This configuration can be extended to eighth order, but this is less common (and has a very narrow bandwidth).
Finally, there's the transmission line. In theory, the idea is that the line is infinitely long, but this is a little impractical for most listening spaces . Mostly, the line is designed for ¼ wavelength at the speaker's resonant frequency, and there will be some reinforcement from the open end of the line. These are notoriously difficult to get 'just right', and the process usually involves experimenting with stuffing within the transmission line until the desired outcome is achieved. An optimally set up transmission line should reduce the resonant frequency of the driver, something that no other enclosure type can achieve.
Figure 2.8 - Transmission Line Enclosure (Shorter Than Normal)
The line shown above is much shorter than normal, only because I didn't want a huge image to show one in full. The general principles are unchanged, and it's usual practice to taper the line so it gets narrower along its length. Some constructors will insist that sheep's wool is the only material that should be used, and others will use a combination of different materials to get the desired results. It's important that the stuffing within the 'line' cannot move, disintegrate or compress over time, as it's very hard to get to once the enclosure is finished and sealed. Unlike a more traditional enclosure, the internals of the transmission line can't be accessed by removing the speaker.
Although I don't intend to provide must info about horn systems here, they have to be mentioned - if only in passing. A horn acts as an acoustic transformer, reducing the high acoustic pressure at the diaphragm (mounted at the throat) to a low pressure (at the mouth) that matches the air. Horn systems can be 10dB more efficient than direct radiators, but for low frequencies the mouth (and length) need to be very large, making them impractical for home systems. The original Klipschorn™ was one of very few 'domestic' systems that used horn loading for the full frequency range. Developed in 1946, they are large and very expensive.
Fully horn-loaded systems used to be common for sound reinforcement, and when done properly are very efficient and provide sound that is/ was (IMO) vastly superior to that obtained from modern line arrays. There are several domestic and studio monitor systems that use either a horn or a waveguide (a similar principle) for the tweeter. Waveguides are becoming very common, and can be used with a 'conventional' dome tweeter to provide a small increase in efficiency and a better controlled dispersion than a dome tweeter by itself. See Practical DIY Waveguides on the ESP site for more information.
Because the design of horns is so specialised, this is the limit of what is shown here. However, construction methodology, the need to ensure that panels are not resonant and other general comments apply to any enclosure, regardless of the type of system. Panel resonances in a folded bass horn can be particularly troublesome, due to the high pressure at the throat of the horn.
While making an enclosure with no parallel sides is possible, it's very difficult for the home constructor making only a pair of enclosures. The vast majority of speakers use conventional parallel sides, front and back, top and bottom. This can still produce a very good box, but there is one thing that can make it 'better'.
There's something known as the 'Golden Ratio', signified by the Greek letter φ (Phi). There are many claims as to its inherent advantages (including aesthetics), but it does have an important characteristic ... no side is a multiple or sub-multiple of any other, so a box using the golden ratio cannot set up single-frequency standing waves across more than two panels. The ratio is defined as ...
φ = (1 + √5 ) / 2
φ = 1.61803398875...
For example, if the baffle is 400mm high, the width (or depth) should be 247mm, with the remaining dimension being 153mm. Note that these are all inside dimensions. These dimensions are not harmonically related, so there is less chance of reinforcement of particular frequencies or overtones. In reality, it probably doesn't make a great deal of difference one way or another, and it's just as easy to build a box using the 'golden ratio' that sounds bad as any other box shape (excluding a perfect cube with the driver smack in the centre of one face of course ).
The ratio can also be described as 0.618 : 1 : 1.618. Which side you choose for the baffle is largely irrelevant, but ideally it would be the narrowest side (so for the example above, the baffle would be 400mm high by 153mm wide (internal). However, this does limit the size of speaker that can be mounted on the baffle - typically to no more than 150mm (6"). If the enclosure has a sub-enclosure (for a midrange driver for example), the problem gets a bit harder. There are probably far more commercial speaker boxes that don't use the golden ratio than there are that do, so to some extent it's always going to be a moot point.
Figure 3.1 - Golden Ratio And √2 Ratio
As always, the room dimensions will have a far more profound effect on the sound, and bracing, internal damping and sound absorbing materials are just as essential as with any other enclosure shape. It's expected that very few rooms will adhere to the golden ratio, and using it for a loudspeaker doesn't guarantee anything. Overall construction methodology, with particular emphasis on bracing, can give excellent results provided some care is taken to ensure that the panels are different sizes, and that bracing is not symmetrical. Braces should always be off-centre on a panel so that the two 'sub-panels' have different resonant frequencies, but this can be hard to achieve while maintaining reasonably simple construction techniques.
While there will likely be some people who insist that the golden ratio always makes boxes sound 'better', it's not a 'magic bullet', and if it results in an inconvenient box size then by all means feel free to deviate. Another ratio that again isn't 'magic' but can work well is √2 (1.414213562...), which is also an irrational number, as is π (Pi - 3.141592654...). √2 is useful and can provide a 'better' aspect ratio than φ (in particular, you can get a wider baffle assuming the box is deeper than it is wide), but you will usually stay out of trouble provided dimensions are not direct multiples (or sub-multiples) of each other. If at all possible, try to use irrational multipliers, rather than 'simple' ratios such as 1.5 (etc.). The drawing above shows the two ratios superimposed so you can see the difference easily.
There are quite a few things that people do for appearance, that usually cause a speaker system to be less 'perfect' than the builder may have hoped. One of these is placing the drivers in a neat row, exactly in the centre of the baffle. While this means there's no 'left' or 'right' speaker (they are interchangeable), it also means that diffraction effects are magnified. Diffraction happens when a sound wave reaches a discontinuity. This is commonly the edge of the cabinet, but it includes adjacent speaker drivers as well.
When the drivers are equidistant from each edge of the cabinet, the diffraction effect is magnified. It was shown many years ago by Harry Olson that a circular baffle with a driver in the centre is the worst possible arrangement. A square baffle (speaker driver centred) is almost as bad, and the best results are obtained when the driver is mounted on a sphere. For more conventional systems, all drivers should be a different distance from each edge of a rectangular baffle. Ideally, the edges will be well rounded - not quite to the extent of producing a partially spherical baffle perhaps, but lovely square edges should be avoided.
In some cases, a diffusing or absorptive material around the driver can help, but to be effective at lower frequencies it needs to be unrealistically thick. It's not difficult to ensure that all drivers are a different distance from the edge of the cabinet though, and you only need to be concerned with midrange and treble - bass is more-or-less omnidirectional, because the diameter of the driver is small compared to wavelength.
The ideal loudspeaker would have equal dispersion at all frequencies, so that sound reflected from walls, floor and ceiling would have the same spectral energy as the direct sound. This is easier said than done, although there are a few speakers that do manage to come close. This is something that some designers strive for, while others ignore it almost completely. Even dispersion does have some major benefits of course, especially if you listen (or are forced to listen) off-axis of the system. The so-called 'sweet spot' needs to be wide enough so that everyone listening hears the same (hopefully) well balanced sound. This is achieved in only a few designs, and for the high frequencies it generally means using a well designed horn or waveguide. It's harder at the lower end of the treble range (around 2-3kHz) because in most systems, this is provided by the midrange (or mid-bass) driver, which will have a diameter that's a significant fraction of the wavelength. Some midrange drivers use a 'phase plug', which is intended to provide more even coverage at higher frequencies than a similar driver without one.
For 'bookshelf' speakers or any enclosure that will be placed against (or near) a wall or other large surface, a rear-facing tuning port is ill advised, because it won't be able to radiate into 'free space'. Likewise, placing a vent right next to the tweeter isn't sensible either. I was unable to locate any definitive papers on this topic on the Net, but it doesn't seem wise to create relatively high velocity, low frequency air movement close to the high frequency driver. The air movement is likely to cause some degree of high frequency modulation, which may be similar to so-called Doppler distortion.
Deep bass reproduction ideally needs a fairly large diameter driver, or high (sometimes unrealistic) linear excursion. When a single driver has to cover from bass all the way up to the tweeter's crossover frequency, there are inevitable compromises. Bass needs a larger driver than midrange, and once the diameter of the driver is 'significant' compared to wavelength, the off-axis response suffers. Ideally, the driver used for midrange shouldn't exceed around 125mm (5"), but if it has to handle bass as well that's somewhat on the small side of the ideal.
This isn't to say that a 125mm driver can't produce good bass - some are surprisingly good. However, one also needs to ensure that the excursion remains within the linear range at all times! That means a fairly large XMAX or comparatively low listening levels, otherwise there will likely be excessive intermodulation distortion. Expecting response below around 40-50Hz with small drivers is unrealistic, because their radiating area is too small. Multiple drivers can work, and will ideally be configured as a '2.5-way' system, where two drivers are in parallel for low frequencies, but the driver farthest from the tweeter has a rolled-off top end. The D'Appolito (invented by Joseph D-Appolito, aka MTM - midrange-tweeter-midrange) arrangement is preferred by some, but it may cause issues when the listener is not in-line with the tweeters. It's always important to keep the distance between the midrange and tweeter as small as possible to avoid phasing errors in the vertical plane (sometimes referred to (by me at least) as the 'sit-down, stand-up' effect, where the 'tone' of the speaker changes when you sit or stand).
We also need to look at what 'significant' means in terms of wavelength.
In general, the diameter of any loudspeaker driver should ideally be less than ½ wavelength at the highest frequency of interest, but that can be extended at the expense of dispersion. The driver's cone diameter should always be smaller than 1 wavelength. Wavelength is determined by ...
λ = c / f Where c is the velocity of sound (nominally 343m/s at 20°C), λ is wavelength, and f is frequency
From the above, it's apparent that smaller drivers are ideal for the midrange. A 65mm cone (nominally a 90mm driver) will have almost perfect directivity up to 2kHz, and is generally acceptable up to 3-4kHz. Some drivers include a phase plug which is intended to improve the directivity at higher frequencies. Some can be effective, others not so good - it depends on whether the manufacturer has included it solely for aesthetics or performance (the latter is more expensive, because it requires many tests to get it right). While it's common for people to use 150mm mid-bass drivers with a tweeter, it's hard to get a tweeter that can cross over at a low enough frequency to prevent poor off-axis response.
A very common crossover frequency is 3kHz, at which frequency a complete wavelength is only 114mm. The midrange cone should ideally be no more than half that (57mm) but simple reality dictates that it will almost always be larger. A 100mm (4") driver is a reasonable compromise, with the cone being pretty close to the optimum diameter. This almost always means that the system will be a 3-way, since a 100mm speaker isn't going to be very useful for bass. In general, 3-way systems can perform very well, and there's rarely any need to exceed that - other than adding a subwoofer of course. While technically that makes the system 4-way, the sub is usually mono, so only one is used in most systems.
In some cases, it may be possible to use a waveguide to load the tweeter and allow operation to a lower frequency, but these can be difficult to design and build for the hobbyist constructor. The secondary advantage of using a waveguide is that it moves the tweeter back from the baffle, and can help to 'time align' the woofer and tweeter. Waveguides are discussed in the contributed article Practical DIY Waveguides (a three part article). Designing a waveguide that does the things you want (and none of the things you don't want is not a trivial undertaking.
Of course, the points made above are suggestions, and are not intended as 'rules'. Many very successful commercial systems use a larger mid-bass driver, and can still perform very well. There will be 'disturbances' in the off-axis response (especially with low-order crossover networks), but not everyone agrees that the polar response has to be perfect over the full frequency range. For example, if your listening room is acoustically treated to eliminate most reflections, the off-axis response is only important if you listen off-axis. Room treatment can have far a greater influence on what you hear than most people realise, and while important, that's not an area where I have significant experience, and no products (whether commercial or DIY) will be discussed.
Ultimately, while perfection is always nice to have, I don't think that any commercial loudspeaker has actually achieved 'perfection' as such. The same can be said for room treatment and (although to a far lesser extent) electronics. It's not at all difficult to design and built preamps and power amps that have distortion so far below the audible limits that they contribute little or no degradation of the sound. However, this has never stopped people from going 'one better', to the point where it can be difficult to measure any anomalies with the best equipment available.
Ideally, all loudspeaker drivers in a system will reproduce the energy of a transient simultaneously from the listener's perspective. This nearly always means that the tweeter should be set back on the baffle, or its output will be slightly ahead of the midrange driver - the sound from the tweeter will reach your ears first, closely followed by that from the midrange. The time difference may only be 75µs or so (up to around 150µs is not uncommon with larger mid-bass drivers) but that small difference can make a surprisingly large difference to the frequency response. It's often affected more off axis too, because of the relatively large area of the midrange driver.
There is a fairly extensive look at time alignment (Phase, Time and Distortion in Loudspeakers), but it's largely from a purely theoretical standpoint. In reality, people often go to great lengths to set the tweeter further back than the midrange driver to ensure that the acoustic centres of the drivers are aligned properly, but this can cause other issues - especially diffraction. If a horn or waveguide is used for the tweeter, this might be sufficient to move the acoustic centre so it's in line with the midrange, but doing so does not automatically mean the system will sound any better.
Before embarking on time alignment, you need to determine the acoustic centre of each driver. This is rarely as simple as aligning dustcaps or voicecoils (or any other part of the motor structure), and it usually varies with frequency. To get results that are useful, you must measure using the time domain (using an impulse test rather than a frequency sweep). By definition, a frequency sweep measures in the frequency domain. The impulse can be a short tone-burst, or just a single impulse generated by measurement hardware/ software or some other means of creating a repeatable pulse stimulus. In reality, you'll probably have to measure in both the time and frequency domains. If this is done carefully (and with the crossover network you plan to use in place), it should be possible to get results that will be entirely satisfactory.
Time alignment between bass and midrange drivers is generally not important, because any offset is (usually very) small compared to wavelength. Since bass frequencies are (pretty much by definition) comparatively slow, a short impulse of (say) 100µs is simply not possible, as that corresponds to a frequency of 10kHz or more. Consequently, if there's a 100µs time difference between the bass and midrange (assuming a crossover frequency of around 300Hz) it will not cause any audible variation. There most certainly is an effect, but at less than 0.03dB it pales into insignificance compared to normal speaker variations (and the room hasn't been considered yet).
In some cases, the relative alignment of drivers can be improved by adding a very short delay - perhaps digital, or using phase shift networks to achieve the same end. Again, doing so will not necessarily make anything sound 'better'. It might be different, but 'different' is not the same as 'better', although our ear-brain mechanism will often conflate the two. It's common for us to hear 'better' when the result is merely slightly 'different'.
All sorts of delay ideas are used for time alignment, but they are mostly not applicable to passive crossovers. One technique that has been proposed is an L/C (inductor/ capacitor) 'ladder' network, but this is not something to be approached lightly. The cost is likely to be considerable, and it's very difficult to get a flat response. Yes, you can obtain phase shift, but there are usually much easier ways to go about it. In an active crossover, a time delay can be created by an all-pass filter (usually several in series), but this isn't without issues either. Phase shift networks are a common solution to obtain short time delays, but the delay is not consistent - it varies with frequency. So, even if the offset is perfect at the crossover frequency, it will not remain 'perfect' over a wide frequency range. This causes ripples in the frequency response, and wide-bandwidth phase shift networks are hard to design and require many opamps.
Sometimes, designers use different crossover slopes for the midrange and tweeter to achieve the phase shift necessary for time alignment. Anyone can do this of course, but it requires a good measurement system to ensure that the results are as expected, and is usually difficult to get 'just right'.
If the baffle is sloped backwards to achieve time-alignment, you will be listening to the drivers off-axis, so their off-axis response has to be good enough to allow this without causing response errors. Some constructors (including manufacturers) have used a stepped baffle (usually with the 'step' at a 45° angle), but this means that the midrange and tweeter drivers can't be located as close to each other as they should be. It's no accident that some midrange drivers (as well as some tweeters) have flat sides or a curved profile on the tweeter surround so the two can be located as close to each other as possible. This isn't done for fun - the two sound sources need to be as close as possible to ensure minimal destructive interference (combing effects).
If the drivers are separated by a true step (i.e. 2 × 90°) then you risk creating what I like to call a 'diffraction engine'. The output from both drivers will be subjected to potentially extreme diffraction, which again will cause combing (a situation where the response varies widely depending on the listening or measuring position). Using separate enclosures stacked one above the other (with offset to 'time-align' the drivers) can have much the same effect. This can even extend to loudspeakers that cost more than a mid-priced luxury car, but there is no suggestion here that they are somehow 'no good'. This is merely an observation.
Often, attempting to ensure that everything is physically 'perfect' in terms of an impulse (time aligned) doesn't necessarily result in a system that is better than one where the drivers are mounted on the baffle in a conventional manner. Every small aberration can be measured, but often it will not be audible in situ. You may hear a difference, but again, being different doesn't necessarily mean better. Despite the claims of some, measurements are far more revealing that our hearing ever will be, as hearing evolved primarily to keep us alive ... music is wonderful to have, but we don't need it to survive in the world .
Time alignment is not necessarily essential, and there are countless well regarded commercial loudspeaker systems that don't use anything fancy to correct for minor time delays. If you're lucky, the time difference may be such that reversing the phase of the tweeter may be sufficient to ensure that there is very little disturbance in the frequency domain. The time delays involved are usually short (less than 200µs is likely to be typical). In some cases, a minor tweak to a passive crossover (shifting its nominal frequency a little for example) can achieve good results. While it's certainly possible to calculate the shift needed, it's usually simpler to do it experimentally (some might call this 'voicing' the system - a fancy name for a bit of trial and error).
While we humans can't resolve very short time delays, we will easily hear any destructive interference caused, which typically manifests itself as a notch at the frequencies where the phase is altered by the delay. Although sound will travel a mere 34mm in 100µs, its effects can still be audible. Whether the small notch or ripple is audible or not depends on the resolution of the drivers used, although the room acoustics will always have a far more significant effect overall.
Figure 6.1 - 145µs Displacement, Phase Shift Network Vs. Polarity Reversal
As an example of the topics discussed above, a 24dB/ octave Linkwitz-Riley crossover was simulated. The crossover frequency is 3kHz (2.83kHz to be exact), and a three stage phase shift network was compared to reversing the polarity of the tweeter. For what it's worth, this is almost identical to the arrangement my speakers use, and the larger than normal offset is because I use a ribbon tweeter. The phase shift network gives the response shown in red, and the green trace is the result when the polarity of the tweeter is reversed. It's pretty obvious that reversing the phase of the ribbon tweeter gives a significantly better response than the phase shift network.
A phase shift network used as a delay is optimal when the dips are of equal amplitude (peaks are more audible and are nearly always unwelcome), and that's the case here. The phase network was staggered, using different value caps to spread the delay over a wider range. A two stage phase-shift network was worse than the three stage staggered version, and no phase network could compare to the simple phase reversal. Time alignment is (or can be) very tricky, and sometimes the least obvious method gives the best result.
It's worth noting that locating the acoustic centre is not a simple process. I set up an experiment in my workshop, and it's fair to say that the results were inconclusive at best. I used a 25mm dome tweeter and a 100mm mid-bass driver, wired in parallel. The pair was pulsed by discharging a 33µF capacitor into the pair, and the tweeter was moved from having the magnets in-line (both on the bench top) to having the two mounting surfaces in line. The total distance was about 40mm, and while there were differences, they were not pronounced. Part of the problem is that the mid-bass is slow compared to the tweeter, so there was no possibility of seeing separate impulses. The best response was obtained with the rear of the magnets in-line, and the impulse response is shown below.
Figure 6.2 - Magnets Aligned
The above trace as obtained with the magnets aligned (roughly aligning the acoustic centres for the drivers used). This is a 'better' response, but without performing a frequency scan it's hard to be certain. I have a pair of almost identical drivers in a small box that I use as my secondary workshop monitor, and (predictably) their mounting surfaces are on the same plane. This box was (many years ago) designed by the late Richard Priddle, and was well regarded at the time.
Figure 6.3 - Mounting Surfaces Aligned
The above looks pretty much ok, and the impulse is reproduced fairly accurately. However, the positive and negative peaks are a bit lower than they should be, and there's a small 'ripple' in the second positive peak. I couldn't hear the difference between this and the first plot shown above, but the microphone picked it up easily. The time delay from the mid-bass is 117µs, equivalent to a distance of 40mm.
You can calculate the time delay and/ or distance travelled with the following formulae ...
c = d / t
d = t × c
t = d / c
c = Velocity of sound (nominally 343m/s), d = distance in metres, t = time in seconds
Ultimately, and despite the offset that usually exists with most drivers, the effects are never as drastic as a simulation might indicate. Simulations work in the electrical domain, where it's possible to get almost infinitely deep notches if drivers are 180° out of phase at some frequency. Acoustically, this doesn't happen. While there is every chance that you will get a notch due to phasing of adjacent drivers, what's important is whether it's audible or not. Remember that the response of every driver you look at is never flat, but can vary by up to ±5dB in many cases. This is particularly true at higher frequencies, and depends on many factors. Cone drivers of 100mm diameter or more can have some fairly serious variations above 1kHz, and these variations are exacerbated off-axis. Mostly, the 'disturbances' caused by non-aligned acoustic centres will be less than those from the driver, so it may be a moot point.
It's a fairly easy matter to run a simulator to see the effects of any time misalignment, but they operate in the electrical domain. For example, you can use an 'ideal' transmission line, which lets you set the characteristic impedance and delay time to anything you like. The results of an electrical simulation are always extremely pessimistic, because the electrical domain (and the simulator) are close to exact, and completely fail to account for the same signals mixing in the air, rather than electrically. The differences are so significant that you'll nearly always get an answer that's not only pessimistic, but often quite wrong. Simulator packages are designed for circuit simulation, and the results do not apply to the acoustic response, other than by accident. That's not to say that such experiments are useless, but you need to be aware of the differences between electrical and acoustical summing.
Ideally, your speaker will be a point source, so that all frequencies emanate from the same place in space. Tannoy has (for a long time) made speakers that are as close to a real point source as you're likely to find. Their coaxial speakers have a horn-loaded tweeter that's concentric with the woofer/ mid-bass, and it uses the main cone as part of the horn. Tannoy is not alone - there are several other makers of dual-concentric drivers. While this can work well, it's not recommended if the 'main' driver has significant excursion, as that will change the horn parameters. There are also other concentric drivers that use a sectoral horn mounted to the centre polepiece of the main driver, so cone excursions will have little effect. Celestion makes a coaxial driver that uses only one magnet, but has separate voicecoils for the HF and LF sections. While they claim that they are phase coherent, this may or may not be the case in reality. Seas has a similar driver (L12RE/XFC), but I don't have any details other than what's shown on the website.
Others (especially car speakers) claim to be 'concentric', but mount a small tweeter in front of the main driver, either on a sub-frame or an extension of the woofer's centre pole. While these are likely a good choice for a car, few people will find them to be satisfactory for hi-fi. Tweeter diffraction is likely to be fairly extreme, and due to the small tweeter they usually have to be crossed over at an unrealistically high frequency. This doesn't mean that good results can't be obtained, but they will rarely compete well against separate drivers selected for their performance.
These coaxial designs are not universally loved (many hate them with a passion), but they remain the closest to a true point source as you are likely to find. The main point is that all loudspeaker drivers are a compromise, and coaxial/ concentric designs are no different. Ultimately the driver selection comes down to cost, and the designer deciding what they can or cannot live with. Audio is very personal, and what works depends on what you prefer listening to. If you find that a single wide-range, high-efficiency driver suits the music you like, then that's what you'll probably use. There are several wide-range drivers that are commonly used in DIY projects, and they tend to be used predominantly by those who imagine that the key to 'good sound' is simplicity. This may be true in some cases, and if this approach is aligned with your wants/ needs, then you have to be prepared to spend serious money for anything 'decent'.
One of the issues with wide range drivers is that the cone area is large compared to wavelength at high frequencies, so they often have a very small 'sweet spot', and might not sound so good off axis. They also tend to be rather expensive, and like many of the different arrangements mentioned here, they aren't something I've worked with. My primary work in audio is on the electronics rather than speakers, and there are so many different speakers on the market that it would be impossible (and impossibly expensive) to test even a small percentage of them.
The simple fact is that most commercial loudspeakers use individual drivers for mid-bass and treble (plus 'super treble' if you think you can hear above 20kHz). Many of these receive rave reviews (is there any other kind?), and all speakers I've built (both for hi-fi and sound reinforcement) have used individual drivers. Some were disappointing and didn't last, others are anything but, and are still in use. I've not tried to build a true point source speaker, and the only coaxial driver I have is sub-par in most respects, and hasn't been used for anything other than a few experiments.
Of course, this does not mean that coaxial drivers shouldn't be used. If you find one that suits your needs and sounds good, then you get the benefit of well controlled dispersion, very little lobing, and a true point-source - at least for the mids and highs. Large coaxial drivers (e.g. 300mm (12") or greater) become a compromise, and the horn tweeter isn't to everyone's liking in any size. With many of the smaller drivers, the 'horn' is more of a waveguide than a true horn, potentially minimising the oft-complained of 'horn sound'. Choosing drivers that have a tweeter suspended in front of the main driver might work for you, but I know of no commercial loudspeakers that use that arrangement. It's not uncommon for in-wall, ceiling and car speakers to use this approach, but these are (usually) not regarded as 'hi-fi'.
As most readers will be very aware, I recommend active crossovers wherever possible. This means that each loudspeaker driver has its own amplifier, and in the DIY world this is not especially difficult or expensive to do. Passive crossovers (using capacitors, inductors and resistors) take the full-range signal from the power amp, and divide the frequency range so that each driver gets only those frequencies it can handle. There is no doubt whatsoever that a very well designed and executed passive network can sound very good indeed, but unless every precaution is taken there will be interactions that may make excellent drivers sound dreadful.
There are several articles that cover passive crossover design, and these should be read through thoroughly so you know what to expect. Simple (e.g. 6dB/ octave, preferably series) passive crossovers can work much better than you may expect, but they are limited to relatively low power use. Because the slope is so gradual, it's easy to get excessive tweeter power at frequencies below the crossover point, and where the tweeter is least able to cope with the dissipation and/ or excursion. For example, a 3.1kHz, 6dB/ octave series crossover has reduced the tweeter voltage by only 17dB at 310Hz. To put that into perspective, if you have a 50W/ 8Ω system, the power at 310Hz is over 500mW - that might not sound like much, but it's probably more than the tweeter was designed to handle at that frequency. First order (6dB/ octave) crossovers can be used if you have well behaved drivers, and don't intend to use amplifiers of more than around 30W or so. If you plan to use a 6dB/ octave network, a series configuration is preferred (see Series Vs. Parallel Crossovers).
Passive crossovers should ideally be at least 12dB/ octave, but to get them to work well, impedance compensation is essential for both the mid-bass and tweeter. This makes the crossover network fairly complex, and if good quality parts are used it will be expensive. Higher order passive networks can be used, but anything above 18dB/ octave (3rd order) becomes a very costly undertaking. As the filter order is increased, so too is the need for accurate component values and impedance compensation. Even a small variation of impedance across the crossover region can have serious effects on the accuracy of the network. Likewise, the tolerance of the parts used in higher order networks become more critical, and even a small variation of voicecoil resistance (due to the power dissipated) can have serious effects on the network's performance.
Unfortunately, even getting everything 'right' doesn't always mean that the speakers will sound any good. A very useful tool for optimising the crossover frequency is the Project 148 State Variable Crossover, which was designed with this very application in mind. I've been using variable frequency electronic crossovers for many years (nearly 40 at the time of writing!) to find the 'sweet spot' between drivers. While crossover frequencies are often dictated by the driver parameters, sometimes you need to go a little outside of the recommended parameters unless you have drivers that are specifically designed to work well together. This is regrettably uncommon, even when the drivers are made by the same company.
This is one of the factors that has led some people to believe that crossover design is a 'black art', whose intricacies are known only to a select few. This is not the case at all, but there's a lot more to it than buying a generic crossover network from a hobbyist supplier, wiring it to the loudspeakers and considering the job completed. Unless the drivers have impedance compensation, the results can be mediocre at best, but rarely 'horrible' unless you do something seriously wrong.
Many 'modern' systems are using DSP, with a fully digital signal processing chain. Unfortunately, this involves some serious processing power, and is not without its problems either. Several people who have bought the Project 09 analogue Linkwitz-Riley crossover board have done so after deciding that the analogue to digital and digital to analogue converters (along with the DSP itself) created too many 'artefacts' for their liking, and resorted to returning to an analogue solution. As far as I'm aware, no-one who changed back to analogue has decided that the DSP is 'better'. While it makes it easy to add delay (and optionally equalisation) if necessary, the process can degrade the signal unless the very best DSP chips are used (along with high quality ADCs and DACs).
This doesn't mean that they are no good - some are very, very good indeed, but probably not if you only pay a few hundred dollars for the complete setup. The digital process also has very limited headroom, since most run with only a 5V supply, so the absolute maximum signal level is generally below 2V RMS. Both the bit rate (aka sampling frequency) and bit depth (the number of bits available for processing) are important. When complex filters (often with equalisation) are performed, the system has to use at least 24 bits or low-level detail may be lost as it passes through the processing chain.
Designing loudspeaker systems is not a 'black art', but it is full of traps for the unwary. Probably one of the most common mistakes (see note below) is to align drivers down the centre of the baffle, which has the advantage that you end up with speakers that can be swapped - there is no 'left' or 'right' speaker. The diffraction effects aren't always readily audible, but they will exist and can make getting a flat response very difficult (within the abilities of the drivers used of course). If you're making a set of 'utility' speakers then it doesn't matter, because no-one expects them to be perfect. On the other hand, if you shell out several hundred dollars for good speakers, then it's worth the effort of making separate baffles for each enclosure. Mostly, they will be mirror images, so the extra effort needn't be that great.
Note: It's highly debatable as to whether aligning drivers down the centre of the baffle is a 'mistake' or not. There are many well regarded speakers that do just that, and they don't seem to have engendered the wrath of reviewers for doing so. My preference has always been to offset the drivers to ensure that no dimension from the tweeter (or midrange) to the edge/ top of the baffle is the same, as that minimises diffraction problems. A popular (and comparatively recent) technique is to use a waveguide for the tweeter, making diffraction effects (almost) a non-issue.
There is considerable difference of opinion as to whether the tweeters in an asymmetrical baffle should be on the 'inside' (closer together by maybe 100mm or so) or the 'outside'. My preference has always been for the inside, but it's something that you need to try for yourself and decide which way sounds better. It may be that neither is actually better, just slightly different. There seem to be as many opinions on this as there are people writing about it, so it's up to the builder/ listener to decide.
One thing that is important is to ensure that the tweeter is directly above the midrange (not offset). This can often be at odds with keeping drivers at different distances from cabinet edges, especially with very narrow enclosures. If there is an offset, you will get uneven dispersion around the crossover region, with the radiation pattern tilted [ 5 ]. With a system using an MTM (mid-tweeter-mid) arrangement, you might find that a small offset improves directionality in the preferred direction, but this means that left and right speakers cannot be interchanged, and extensive testing will be necessary to ensure that dispersion is properly controlled at all frequencies.
It's generally agreed that the tweeter should be at ear level while sitting in your preferred position. For speakers that aren't tall enough, stands should be considered mandatory. Many households can't readily accommodate 'true' floor-standing speakers, as most tend to be rather imposing. While smaller cabinets are often described as 'bookshelf' designs, actually locating them on a bookshelf is usually a bad idea - especially those with a rear vent which would be obstructed, reducing bass output. It's also likely to be difficult to use sufficient toe-in (pointing the boxes towards the listening position). A lack of toe-in can often result in a 'hole-in-the-middle' sound, where the central position (which is supposed to be the prime listening position) has a pronounced response dip, and often some odd phasing issues. There are a few people who prefer 'straight-out' speakers, and some even prefer toe-out (speakers splayed), but this rarely (if ever) improves the sound stage or imaging.
Figure 9.1 - Baffle Cross-Section
In general, the baffle should not be recessed to allow for a grille cloth or protective cover. It's tempting to do so, but it can cause considerable disturbances due to diffraction. The speaker drivers should be recessed into the baffle though, so there are no discontinuities across the face of the baffle itself. For small 'near-field' speakers (as might be used with a computer for example) it probably doesn't matter, but I have been able to measure a small 'glitch' in the response of speakers that are surface mounted. Minimising discontinuities is almost always beneficial, but some will always remain because of the way most speakers are manufactured. Just the surround and cone of a midrange driver can create small but measurable response anomalies, but reality tells us that there's no much that can be done to avoid this.
Rounding (or chamfering) the edges of the baffle (and in general, the 'rounder' the better) minimises edge diffraction, but there will always be limitations due to the material's thickness and the rounding bits that we have available. While taking things to extremes (such as a cylindrical enclosure) can reduce edge diffraction to the minimum possible, that's not always practical and the drivers always need a flat mounting surface anyway. Such enclosures can (and have) been made, with some constructors going to the extreme of using spheres. It may be the optimum shape acoustically, but it's rarely practical (and a sphere is a cow to build unless it's made from fibreglass or similar).
Some speaker enclosures have tapered, angled tops and upper sides [ 8 ], to keep the baffle area around the tweeter as small as possible. In a few cases, the baffle is trapezoidal, with the tapered sections extending from the top to the bottom of the enclosure (or a significant part thereof). These are usually difficult to build, but if done properly can give very good results. This is taking the idea of 'rounding' the edges/ corners to extremes, but it can produce a good result if done properly.
With a double-thickness baffle, it will be stiffer and less resonant if the two layers are different materials. For example, plywood may be preferred for the outer surface, and MDF is then ideal for the second layer. Because the two materials are so different, resonance is minimised. If a slightly flexible adhesive is used between the two, they will be decoupled to some extent, which reduces the Q of any resonance that does exist. Tee-nuts or other metal threaded inserts can be placed between the two layers so the cutout for the woofer/ midrange drivers can be radiused on the inside. This reduces internal diffraction.
There are innumerable materials that can be used for loudspeaker enclosures. Many commercial systems (especially small PA powered boxes) use ABS or a similar thermoplastic material. While the original setup cost is very high, enclosures can be produced rapidly and for relatively low cost. Most of these boxes include the horn flare (or waveguide) for the high frequency driver, as well as appropriate cutouts for the crossover or amplifier module. What they lack in cost is made up for by the appearance, and usually little or no finishing is required, other than removing any adhesive residue after the two halves are glued together.
However, while they are cheap to build and usually look quite good, the plastic almost invariably lacks rigidity, despite the curved surfaces. While these are fairly strong, they are anything but stiff, and I've not come across one that could be called rigid (regardless of the definition that the maker may use for the term). Bracing is difficult, and while most do have ribs moulded into the interior, they are nowhere near strong enough to prevent panel resonances. Fibreglass (with or without carbon fibre) is very strong, but is also very difficult to repair if the box is damaged.
There are many plastic composites available, but few (if any) are suitable for home construction. The requirement for a mould means that it's not economical for building just a couple of enclosures, and thermo-set plastics require an autoclave or large oven to cure the resin. This is clearly not practical for home construction for the vast majority of hobbyists.
A favourite for many is plywood. It's a very good material, and offers high stiffness for its weight, but it is usually also very poorly damped. If a panel resonates, it will usually do so with some vigour, and good bracing practices are essential (see next section). There are many constructors who think that MDF is 'no good', but that may be due to poor construction techniques, or even because they are used to the panel resonance(s) produced by plywood. MDF has been the material of choice for many major manufacturers for some time, and it's now possible to powder coat MDF given the proper (and very expensive) equipment.
Particle board (aka chipboard) is one material that was once used extensively by low-cost manufacturers and hobbyists. It is sub-optimal in almost all respects, and in particular the structural integrity is found wanting, so corner braces are almost always needed to the box doesn't disintegrate during handling. Particle board can be obtained with high-grade real wood veneer, and while this makes the finished item look better, it's still fairly low-strength. If a veneered finish is used, this pretty much eliminates the possibility of using radiused edges, so edge diffraction will be greater than desirable. In general, particle board has very little to commend it, veneered or not. Attaching drivers and connection panels is irksome, because the screw holes will become useless after only a few insertion/ removal attempts. The use of Tee nuts or similar is essential, and even they should be glued in place or they may fall out during assembly (or disassembly should changes be needed).
Many manufacturers are now using advanced composites, which let them create any shape relatively easily. Cellulose reinforced resins and 'exotic' plastic resins are common, but the requirement for moulds to create the finished shape (and an autoclave to cure the resin) means that these are generally not suited to the DIY approach. It's not impossible of course, but even getting the materials in small quantities may prove difficult. 'Traditional' materials will almost always be the best choice for DIY, because an enclosure can be built using only basic hand and power tools.
Finishing is another matter entirely, and that is not covered here. The requirements for proper spray booths and an extreme dust-free environment are essential for the classic 'piano black' finish, or any other high gloss finish. Less labour (and equipment) intensive finishes are more common in the DIY sector, although there are no doubt some who will be able to achieve very high quality surfaces with a well equipped workshop. Ultimately, the final finish depends on what you can achieve within your budget and with the tools you have to hand.
These are the most critical parts of any enclosure. Bracing is necessary to increase the stiffness of the panels, and it has the side benefit of forcing any resonances to higher frequencies. High frequencies have less acoustical power, and are easily absorbed by the damping material used - provided that it selected for its effectiveness. Fibreglass (as used for home insulation) is very good, but it shouldn't be used in vented enclosures because tiny glass fibres may be ejected from the port, and these should not form part of the atmosphere of the listening room.
As noted earlier, braces should be asymmetrical, and not simply pieces of timber placed neatly around the inside of the cabinet. The greater the asymmetry, the less chance there is of creating sub-panels with the same resonant frequency. The effectiveness of the bracing can be tested with an accelerometer (see Project 181 - an audio accelerometer for speaker box testing). This will tell you just how much a panel vibrates, and at what frequency (or frequencies). If you find a panel that appears to have too much vibration, additional bracing will be necessary to reduce it to a level you find acceptable.
Note that the P181 article also shows screen captures of vibrations measured on a test box I have in my workshop, and includes a number of ideas that you can use to create strong, non-resonant enclosure panels. It's not particularly detailed, and it was my own experiments measuring panel resonance that led to this article being written. It's a complex field, and while some resonances are 'benign', many others are quite the opposite.
Braces should be made from rectangular hardwood (e.g. 50 × 25mm/ 2 × 1 inch), and always glued firmly in position with the short edge to your panel. This provides much greater stiffness than the alternative mounting. Ideally, they will also be screwed (or nailed if you must) to ensure a firm bond while the glue sets. Other than the normal bracing that you'll often use where panels meet, the braces should ideally be at an angle (with no two angles the same), and they don't need to extend right to the corner with any bracing, as the corners are extremely stiff already.
More substantial bracing is needed for the baffle, which should ideally be double thickness to withstand the momentum of the diaphragm (for the bass driver in particular). Braces between the baffle and the rear of the enclosure also help to prevent vibration. Many construction articles show 'window frame' bracing, which can certainly work, but these are incredibly difficult to install at the odd angles that can be very helpful in ensuring that no two panels (or sub-panels) have the same resonant frequency. The left drawing shows asymmetrical bracing, where each sub-panel is a different size. The right drawing has four more-or-less identical sub-panels, and they will all resonate at roughly the same frequency. This is usually unwise, but it may be alright for smaller enclosures where the resonant frequencies are all well above the highest output from the midrange.
Figure 10.1 - Bracing, Right And (Usually) Wrong
Remember that it's the outside of your enclosure that needs to look good. The internal construction with angled braces and odd shapes is not visible, and should be designed for rigidity and performance, not appearance. Deadening materials (e.g. bitumen tiles, heavy felt or other mass-damping treatment) needs to be very well bonded to the interior of the treated panels so it cannot move, rattle or fall off. All internal wiring has to be secured properly to prevent rattling as well, because it can be very difficult to correct after the box is sealed up and you only have access via the speaker cutouts.
If the back (for example) is made removable, then I strongly recommend that it be secured with metal thread screws, with 'tee nuts' or some similar threaded metal insert. Wood screws can't be inserted and removed more than a few times before the thread cut into the timber starts to disintegrate (especially true with MDF or particle board!). This also applies to the driver mounting screws - wood screws are generally a poor way to mount the drivers. You can also use a metal bar, ring (for speaker drivers) or angle with threaded holes, provided it's well attached and doesn't vibrate or rattle. Suitable gasket material is essential to stop whistling noises as air passes through any small gaps. These gaps (if present) can also adversely affect the performance of tuned enclosures, because they represent losses that reduce the effectiveness of the tuning. It may seem counter intuitive, but metal thread screws work very well in holes tapped into hardwood, provided it really is hard! Some Australian hardwoods are so hard that they can destroy a drill bit, and these take a tapped thread very nicely indeed.
The type of acoustic damping material used is a matter of personal choice. Fibreglass is very good, but isn't suitable for vented boxes, as glass fibres may be ejected from the vent. Most suppliers stock damping materials, and to be effective they should be coarse to the touch so there is considerable friction between the fibres to absorb as much energy as possible. Foam is generally not suitable, because it a) doesn't usually work very well, and b) because it tends to disintegrate after a few years. Foam surrounds were once common for woofers, but eventually the foam gives up and the surround has to be replaced or the driver scrapped. Damping materials need to be as acoustically absorbent as possible.
There is much disagreement as to whether vented boxes should have damping material or not. My view is that it's essential, because without it there will be excessive upper bass and midrange energy bouncing around inside the cabinet. This can often be heard through the vent, so if you can hear anything that isn't at the tuned frequency coming out of a vent, the cabinet needs damping. Another technique that can be used is to set up diffusers within the box. These will be different heights and widths, and spaced at 'irrational' intervals. Damping material is still necessary, but you may find that you need less of it if you have effective diffusion. Upper bass and midrange can also be deflected with an internal angled brace, so that the energy is directed towards a well damped section of the enclosure. Bass (which has long wavelengths) will not be affected. Remember that anything approaching ¼ wavelength at any frequency can be your friend or your enemy, depending on how it's implemented.
It's almost always necessary to add braces from the baffle to the rear of the enclosure, and also between the sides. These need to be attached very firmly, because the stresses can be quite high at high power levels. While it would be 'nice' if these braces could be angled so that remaining panel resonance(s) are different frequencies, this is usually impractical for a number of reasons. Some people have used braces from the rear of the woofer (or mid-woofer), but this isn't easy to get right, and can't be used easily if the driver has a vented rear polepiece.
I suggest that you also have a look at the Small Satellite Loudspeaker System design that was described back in 2007 (it's a 3-part article). The bracing is done with aluminium 'U' section (25 × 25mm, 3mm thick), which is much stiffer than MDF and most timber. It also uses very little of the internal volume, but a very reliable method of gluing is necessary because aluminium has an annoying habit of oxidising itself (much like anodising, but thinner), and the layer of oxide can creep under the adhesive and it may 'let go' after a few years. If done well (with a proper two-part epoxy - not the 5-minute stuff) it should stay put for longer than you'll use the speakers.
Finally, you have to decide whether you will use spikes (for floor standing enclosures or at the base of the stand), or whether you will use a stand for smaller speakers. It's usually preferred that the tweeter should be at eye level when seated and listening (actually ear level, but ears and eyes are at close to the same level on most people's heads). Some people love spikes and consider that any loudspeaker no so equipped must sound dreadful, while others have the opposite view. Spikes are obviously not suitable for polished floors unless they sit in little cups, and these are also an area of controversy amongst many audiophiles (just like almost everything else in the signal chain). Use what you feel suits your system the best, and you don't need to spend a fortune - gold plating does not make spikes sound better!
The price range for spikes/ isolators is quite astonishing, with prices from AU$20 for a set of eight, up to over a thousand dollars for a set of four! There are some fairly outrageous (IMO) claims made for the expensive types, but claims and reality are usually not backed up by any science. I (naturally) will make no recommendations one way or another, and there are so many conflicting opinions that I can only suggest that you do your homework, and decide for yourself which way you want to go.
Stands will usually be selected to suit your tastes, decor and budget. Heavy stands add mass to the system making it less likely to move with woofer excursion, and it's important that the stands don't have any audible resonance. While it's unlikely that resonance will be audible (it will be hard to excite any resonant mode unless the boxes are flimsy to start with), sturdy and acoustically 'dead' stands will give some peace of mind. Some provide cable management and/ or the ability to be sand-filled to eliminate (or at least damp) resonant frequencies.
Ideally, there will be a layer of felt, non-slip rubber or sound deadening material between the cabinet and the stand to ensure there can be no rattles at any frequency. Beware of systems that are top-heavy if you have small children - no-one wants to see their offspring crushed by a 100kg speaker box! Some have the ability to be permanently attached to the loudspeaker, while others are just intended for the enclosure to sit on the stand with no attachment. Personally, I'd avoid that, but it depends on the stand, the loudspeaker and your circumstances.
As with so much in audio, there are as many opinions as there are authors, and just because a few people agree with one idea or another does not make it reality. Something that works well in one environment doesn't necessarily mean that it's suitable for your needs, and in some cases the 'product' offered is nothing more than snake-oil, and won't achieve anything useful at all. It's up to the constructor to work out what works in the specific environment where the speakers will be used. For example, using ultra-hard (perhaps tungsten tipped) spikes on a tiled floor is probably unwise. Along similar lines, titanium spikes won't 'transform' the sound, despite the (considerable) cost - a set of three can be obtained for as little as €2,199 (about AU$3,530 at the time of writing!), although some are available at an ever-so-slightly less insane price. Personally, I completely fail to see the point of buying a set of spikes that cost more than a set of very decent drivers (but they do come in a padded carry-case). I'll let you be the judge as to whether this qualifies as snake-oil .
There really aren't many 'conclusions' that apply, because nearly everyone has differing opinions on what is 'good', 'bad' or indifferent. Ultimately, if you are building your own speakers for your use, then it only matters that you are happy with the results. There are many conflicting needs, including what you can (or cannot) deposit in your lounge room lest you incur the wrath of your 'better half'. Aesthetics always plays an important role, and if you have small children then even greater limitations may apply. Having 100kg speakers on nice stands may look great, but not if they can fall over and squish a child. You may not like using a grille (I don't), but keeping small fingers away from delicate tweeters becomes a priority.
As I've mentioned in several articles, electronics (as with almost everything else) is a compromise - the 'art' is in making the compromises in such a way that the end result isn't ruined. This is almost always easier said than done, unfortunately. No-one wants to have to re-build speaker cabinets because of some fundamental error (of judgement or construction), especially since construction usually represent a considerable effort and cost. In some cases it may be possible to 'rescue' an enclosure by adding bracing or damping materials, but if you don't get the basics right then the time, effort and materials may be wasted (and I freely admit that this has happened to me a couple of times). Consider that major manufacturers may build a number of prototypes before they get the performance they expected, but this isn't something that most hobbyists can afford. Mostly, we 'mere mortals' have to try to get it right the first time, and can't afford to generate vast amounts of scrap material in the pursuit of 'perfection'.
Speakers are without doubt the most compromised of all the components that go together for a complete hi-fi system. The individual drivers are a compromise, and not always due to cost - even very expensive drivers are still compromised by the materials and the laws of physics. When multiple drivers are used together, the compromises are simply magnified, but they are even greater if you try to get everything from a single driver. Passive crossover networks are always a major compromise because they use inductors, which are the most imperfect passive components made. Yes, you can have them wound with flat silver sheet to minimise resistance (and your bank balance), but they will always have self-capacitance that can cause issues. Not everyone likes electronic crossovers, even though there are far fewer compromises involved, and changes are easily made (both to frequency and level for each driver). However, you need a dedicated amplifier for each of the individual drivers.
The goal of this article is (if nothing else) to give you some pointers towards reducing the (sometimes significant) effects of an enclosure that is (of course) yet another compromise. There is no such thing as a 'no-compromise' speaker box - without compromise, you have nothing at all. Even if you use the very best materials, that doesn't mean that they are without flaws. The same goes for bracing, damping and deadening materials. If you get everything right, you should end up with speakers that sound good - musical and suitable for the material that you listen to, but even then you will not get perfection. Electronics can be made easily with response that is dead flat from DC to daylight (well, not quite daylight perhaps), with distortion of all types that's difficult to measure. No loudspeaker, however expensive, can come close. Then there's the room ... getting that right is a major undertaking.
To give you an idea of the time and effort that can go into building a 'nice' pair of speakers, see New Speaker Box Project - Part 1. Not that they are 'new' any more - they were built in 2001, and upgraded to ribbon tweeters about 5 years later. They are in daily use to this day, and have failed to disappoint in any way. Are they 'perfect'? Not at all, but they do sound very good with all types of music (along with video sound tracks, etc.). This is certainly not something I'd want to tackle again, especially since I'm more than 20 years older (they were made when I was in my early 50s). That doesn't mean that I won't build any more speakers, but they will be (probably considerably) smaller and less complex overall.
One particularly troubling 'claim' I saw was that "we can hear everything we can measure, but we can't measure everything we can hear". Reality is exactly the opposite. Measurement systems are accurate to fractions of a dB, and are far more revealing than our hearing. The often neglected part of our hearing is our brain, and it lies to us. If we expect to hear a difference, then there's every chance that we will hear a difference, even when there is none at all. There are several different names for this, with one being the 'experimenter expectancy effect', and it applies to everything. This is why medical tests are double blind, so neither the experimenter or 'victim' (experimentee) knows whether they have received the drug being tested or a placebo. Audio tests also need to be double-blind, although this is very difficult with loudspeakers. Some of the major manufacturers have set up very advanced systems to ensure that the test is a close to true double-blind as possible, but this isn't an option for most hobbyists.
An issue that's not discussed nearly often enough is the difference between a microphone and our ears (actually our complete hearing mechanism). A microphone is dumb - it cannot distinguish the difference between the direct sound and a reflection, which is why anechoic chambers are used by some major manufacturers. As a result, microphones in a room will rarely give a true indication of a system's response, because the direct sound and early reflections cause 'wobbles' in the output graph that don't necessarily exist. I once ran a test and was able to measure when a coffee cup was moved - naturally, this was completely inaudible, but the mic picked up the difference quite clearly. What should be equally clear is that was an anomaly - we simply do not hear such tiny differences because our brain knows they are unimportant!
Even though this article is far longer than I intended, I trust that it helps. By necessity, it's an overview - the idea was never to describe a complete system, but to provide guidelines that I've applied based on my own constructions, tests and measurements using an accelerometer, and acoustic measurements of a great many drivers over the years. Loudspeaker construction is one of the most labour intensive (and expensive) undertakings for DIY people, and anything that helps prospective constructors to get it right has to be useful. I hope I've succeeded.
Over time, our hearing will accommodate even serious response errors (this is called 'breaking in' by many snake-oil purveyors), but if the response is restored to flat (perhaps by using equalisation), it will sound completely wrong for some time - until our ear-brain combination 'breaks in' to the new response. If you have access to a decent equaliser, I offer the following challenge ...
Notch out a frequency around the middle of the frequency range (600-700Hz for example). The sound will be quite wrong for a while, but after perhaps 30 minutes you will adjust your expectations. Next, restore the response to flat, and hear the huge peak in the midrange. It will initially sound dreadful, with a 'honky' sound that quite obviously can't be right. However, keep listening for a while and that sensation goes away, and everything sounds normal again.
To say that this is confronting is putting it mildly. If you have never performed such a test it's unlikely that you'll believe it possible, which is why you must do it. Until you experience this for yourself, you are 'sucker bait' for snake oil of all kinds. People tend to think that they can remember what something sounded like 'before' and 'after', but in reality our auditory memory is limited to a few seconds! Naturally, there can be response anomalies that are so gross that we do remember them for much longer, but subtle changes are not in that category.
In closing, the hobbyist must consider that even the best speaker in the world may sound dreadful in some rooms. Even with typical furnishings, moving your head by 100mm is generally enough to affect the frequency response by as much as ±10dB [ 6 ]. This is measured by a microphone, which is completely lacking our brain's processing facilities and takes a reading at a fixed point is space. It's important to understand that we humans do not hear these extreme variations, because our ear-brain combination removes much of the interference that causes the measured to vary so wildly. However, this does not mean that we won't hear such radical variations if they are created by the sound source - the loudspeaker. However, over time we will adapt, and even seemingly outstanding differences can become the 'new normal'.
There is one thing that we don't become accustomed to (at least not to the same extent), and that's distortion. More specifically, intermodulation distortion. If great enough, this turns everything you hear to mush - there is no definition and all clarity is lost. This is why most of the 'best' loudspeakers use multiple drivers, so the frequencies are separated into individual drivers and intermodulation is reduced. However, a system using multiple drivers must be designed properly, or it may cause more harm than good. Using more than a 3-way system is unlikely to improve matters (excluding a subwoofer, which creates a 4-way system).
In addition to the above, there are a few brand names mentioned and quite a bit of 'general research' that doesn't warrant a direct reference. Before embarking on your next speaker project, I recommend that you do your own research, and ensure that you get a balanced overview - relying on one opinion (or forum thread!) is unlikely to give reliable answers.
|Copyright Notice. This article, including but not limited to all text and diagrams, is the intellectual property of Rod Elliott, and is © 2019. Reproduction or re-publication by any means whatsoever, whether electronic, mechanical or electro- mechanical, is strictly prohibited under International Copyright laws. The author (Rod Elliott) grants the reader the right to use this information for personal use only, and further allows that one (1) copy may be made for reference. Commercial use is prohibited without express written authorisation from Rod Elliott.|