Speaker Design: An overview of we're trying to achieve           

I don't think it's enough to simply say that a speaker should reproduce an electrical signal as faithfully as possible. For other equipment such as amplifiers this a reasonable aim. But the sound of speakers and their interaction with room acoustics is already factored into a recording to some degree. After all, the music was tracked, mixed and mastered on speakers of some kind. At each step in the recording process the the inaccuracies of the monitor speakers are being compensated for.

A successful speaker design has anticipated the present and future needs of its users. The commercial, artistic, technical and aesthetic needs. Of course not everyone's needs are the same; and thus there is room for many speaker designs to claim the status of being the best.

Speaker design encompasses so many competing principles and different disciplines that 'best' usually means a good set of compromises.

It's often stated that there is no such thing as a 'classical' speaker or 'rock & roll' speaker. A great loudspeaker will suit all types of music. Each genre is typically recorded and mixed on particular types of speakers. In order to hear similar sound to the musicians and recording engineers it makes some sense to listen to the reproduction on the same speakers in a similar acoustic environment.

In the case of 70's & 80's rock & roll that probably means JBL's. In the case of classical that might mean B&W's. In the case of some hip hop that might mean a ghetto blaster!

Art has always been led by technology. The introduction of photography changed the way artists viewed the world. The invention of the piano at the turn of the 18th century changed the direction of music in Europe. The use of sequencers and samplers precipitated contemporary dance music. The technology influences or sometimes invents the art form.

If a form of music is dependant for its character upon boomy bass, then that must be respected when designing loudspeakers to suit that style of music.

The trick in speaker design is to devise a speaker that represents as many different genres of music as well as possible, in addition to the usual constraints forced upon the designer, such as cost, size and appearance. 

So what are some of the attributes designers and listeners need to consider when evaluating or designing loudspeakers, and what is their significance?

The perception of human hearing is loosely defined as 0dB. Typical very quiet listening rooms might have ambient noise levels of 20dBA. The threshold of discomfort starts for many people with music playing at an average level of around 100dB with brief peaks extending up to perhaps 120dB. So it seems we need to consider a dynamic range of about 100dB and a maximum average SPL of 100dB. Not surprisingly, this sits well with 16-bit digital audio's maximum dynamic range of 96dB.

However, while I don't advocate listening at peak levels approaching 120dB, real musical instruments at close range can produce levels much louder than this for short periods. Either the recording of a close miked snare drum needs to be limited to fit into the available dynamic range or else the entire signal chain including the speakers needs to be capable of more. The opportunity now exists to break through the 96dB limitation of 16-bit audio with 24bits of resolution. This is starting to appear in recording studios where bits are often lost during audio processing. But most recordings still end up being mastered at 16bits.

So until better recordings come along, 100dB maximum average SPL at the listening position with a peak capability of 120dB seems a good, although tough, target to aim for. Whilst this sounds like a lot more than typical speaker systems are capable of, it must be remembered that speakers are limited by heat dissipation and displacement and can typically handle much higher power than their name-plate ratings during transients. Whether they can do so with a reasonable degree of linearity is another question.

The driving amplifier also needs to be able to provide enough power without too much clipping at the required SPL's taking into the speaker's electro-mechanical conversion efficiency. This is a significant problem in its own right. - Very high continuous output power, transient power, bi-amping, efficient speakers, reducing the listening distance are all methods that can be employed to ensure sufficient SPL.

A well designed driver will simply go louder as the electrical input is increased. Seems obvious but being electro-mechanical devices, speakers are non-linear when pushed hard. Power compression issues show up under severe useage, which is why it's a phenomenom that has mostly been explored by the professional speaker manufacturers. However even for domestic use, loud musical passages if played at high volume can push drivers into regions of operation that result in power compression.

Power compression is minimised through efficient thermal design. Vented magnetic assemblies and large diameter voice coils are common techniques employed.

The axial amplitude response of a speaker is the best known benchmark for comparing speakers in a 'technical way'. More commonly called 'frequency response'. In fact the frequency response should also include the phase response to be a full description. The axial response is measured at one single fixed point in front of speaker, often in line with the tweeter, but not always.

It's unfortunate that it carries so much significance because it's not only thing you hear unless you happen to sit 1m from your speakers in line with the tweeter where the measurement microphone was placed. Professional sound recordists do sometimes sit that close in a listening position called near-field monitoring.

Near-field refers to listening to speakers in a place where the direct sound of the speakers is significantly louder than the reverberant sound field. In any room the sound continues to bounce around the walls creating a reverberant sound field after the direct sound has ceased.

The further the listener sits from the speakers the less effect the direct sound has on the perceived tonal balance of the speakers and the more important the reverberant sound field becomes.

The power response of a speaker is a measure of the total sound power level a speaker creates versus frequency. The total power will give rise to a reverberant sound field once the speaker is placed in a room.

Ideally a speaker will have an approximately constant power response. This coupled with a flat on axis frequency response will give a speaker that will not change it's sound characteristic as it is rotated away from the listener.

In practice this is very difficult achieve. A direct radiator has a dispersion characteristic that narrows as frequency increases. If such a speaker has its on-axis response equalised to be flat, the power response will drop off slowly as frequency increases because the speaker is putting less total sound power into the room at high frequencies. This is bad but not catastrophic, because as mentioned in the opening paragraphs, this is what we are used to hearing with most loudspeakers.

Multi-way speakers have a  dispersion characteristic that reduces as frequency increases for one driver and then repeats the pattern as the smaller higher frequency driver takes effect, putting a dip in the power response near the crossover frequency. This a more serious deficiency as it cannot be equalised out of the system. Either the axial or off axis frequency response will suffer.

Various means can be employed to control the directivity of a speaker and so it's power response. Horns and waveguides are the best known examples. Some people don't like the sound of horns very much however, citing extra distortion and tendency for the performance to project too far into the listening space due to excessive directivity.

I consider these distortions to consist of harmonic distortion and its equivalent, intermodulation distortion.  As harmonic distortion is relatively easy to measure, it's not surprising that it is a common metric by which drivers and speaker systems are judged.

Both harmonic and intermodulation distortion are easy to hear when dealing with pure tones. It's fortunate that music is complex in nature and tends to mask many distortion products. With pure tones, distortion as low as 0.1% is audible, which is about as good as a driver ever gets.

Obviously less distortion is better. Fewer higher order harmonics is also better. Linear & non-linear distortion can be tackled in the design of transducers themselves and in the construction of speaker cabinets.

Symetrical magnetic systems, flux modulation reduction, linear suspension design, rigid cone materials, cone damping, are all areas to be optimised in a driver's design in the pursuit of minimum linear distortion. 

Massive cabinet panels, and careful tailoring of the resonance frequencies of those panels can be optimised within the design of a speaker cabinet to minimise the production of distortion due to re-radiation of sound coupled mechanically and acoustically to the cabinet panels.

I consider these distortions to be every other distortion and artifact that isn't covered by linear & non-linear distortion. Rattles, buzzing, diffraction effects, cone break-up all cause unpredictable distortion products and artifacts. They are arguably the most important factor in determining the unique sound signature of a given driver or speaker system.

Minimising these distortion products is mostly achieved by careful drivers selection along with sturdy cabinet construction. However, careful consideration of baffle diffraction effects and driver mounting can also help to prevent strange noises.

These so-called other distortions are troublesome not only because they're hard to predict, but also because they're hard to measure. Furthermore, the measurements are difficult to correlate with the perceived sonic character of the speaker. As with linear & non-linear distortion they're quite easy to hear when pure tones are used to excite the speaker.

When considering distortions from speakers, it's also worth remembering that various items in the listening area are responsible for producing all manner of rattles and noises. Windows and heating appliances are especially problematic. Fortunately the ear can separate some of these noises from the direct sound as they emanate from a spatially and temporally distinct source. As it's usually pretty easy to secure loose items, it's worth tracking down and eliminating the rattles.

A CD with a swept sine wave from about 40Hz to 200Hz or an adjustable frequency tone generator will assist in finding drivers that need their mounting screws tightened or glass panes that vibrate.

Now that we have some ideas about how speakers work in an acoustic environment, we can look at doing some design work. System design starts with the objectives at the top level and expands to show more and more detail. For example, there's no point in specifying a particular amplifier until we know the desired SPL, the speaker efficiency and the acoustics of the listening space.

to be continued....