Hot Stuff: Loudspeaker Voice-Coil Temperatures
To generate sufficient force, the wire in a moving-coil driver must be many meters long and so is wound into a coil, archaically termed the voice-coil. To keep this coil from being too large and heavy, the wire must be thin. So even when it is formed of copper—the most electrically conductive of all metals, bar silver—it has appreciable resistance. In a drive-unit of nominally 8 ohms impedance, the resistance of the voice-coil is typically around 6 ohms.
On the one hand, this is a good thing: it ensures that a substantially resistive impedance is presented to the amplifier, albeit with appreciable reactive components. These arise from the speaker's physical properties and are converted into inductive and capacitive elements within the input impedance by the electromechanical transformer action of the voice-coil. But there is also a downside. The inherent inefficiency of direct-radiating moving-coil loudspeakers (typically only 1–2%) means that considerable electrical power is dissipated in the coil resistance. Just as with the incandescent light bulb, little of the input energy is put to good use; the rest appears as heat.
This heating of the voice-coil has various ramifications. In extremis, it can destroy a drive-unit, although that rarely occurs with modern hi-fi loudspeakers, which typically are constructed to withstand voice-coil temperatures in excess of 200°C (390°F). Although every speaker manufacturer can regale you with stories of customers who nonetheless manage to achieve it, thermal destruction of a driver is something few of us will ever suffer—although a few reading this will probably confess to having fried a tweeter at some time in their lives, as a result of driving a speaker to party levels with an inadequate amplifier.
Voice-coil heating can also induce partial demagnetization of the driver's magnet if it reaches too high a temperature as a result of the heat radiated and conducted to it. Usually, this effect is small and reversible—when the magnet cools, it returns to full strength—although it can be permanent in speakers with an inappropriate choice of neodymium magnet material.
Subtler effects on the materials from which the driver is constructed also result from voice-coil heating, and are a factor in the change of sound quality that occurs as a loudspeaker "breaks in." But potentially the largest effect on sound quality—one that doesn't recede with continued use—occurs as a result of temperature-induced changes in the voice-coil's resistance. As the voice-coil heats up, its resistance increases. This has various ramifications, one of which is that the driver's sensitivity decreases, compounding any decrease resulting from the effect of heat on the magnet. As a result, moving-coil drive-units are inherently prone to thermal compression (also known as power compression). The harder you drive them, the less responsive they become.
Copper, the most widely used voice-coil conductor, has a temperature coefficient of resistance of approximately 0.4% per degree C. So if we assume that a copper voice-coil has a normalized resistance of 1 at an ambient temperature of 20°C (68°F), its resistance will vary with temperature, as illustrated in the blue trace of fig.1. Note that by 200°C the resistance has increased by a factor of 1.73, and the driver's sensitivity will have decreased by a similar factor (–4.7dB). Fig.1 also includes plots for the less commonly used voice-coil materials aluminum (red trace) and silver (purple), which behave very similarly to copper. Although special alloys of significantly lower temperature coefficient of resistance are available, such as nichrome and manganin, these also have much higher resistivity, which unfortunately makes them unsuitable for use in voice-coils.
Fig.1 Relative resistance vs temperature for copper (blue trace), aluminum (red), and silver (purple).
Increased voice-coil resistance has effects other than reduced sensitivity. The driver's electrical damping decreases, which can substantially alter the speaker's bass alignment, making it more boomy in character when the voice-coil is hot. Crossover alignment is also affected in passive speakers, as a result of changes in the filter networks' terminating impedance. So in addition to the compression effect, there are also alterations in the speaker's transient performance and frequency response as its voice-coil temperature fluctuates in response to varying signal level.
These undesirable effects can be reduced by a variety of means, both internal and external to the drive-unit. Within the driver itself, any design feature that speeds removal of heat from the voice-coil will reduce its operating temperature and thereby the change in its resistance. Conduction, convection, and radiation losses all play a part in this and can be maximized by various means, including the use of a large-diameter voice-coil, a thick top plate, a metal diaphragm, ferrofluid, pole-piece venting, and even cooling fins on the magnet structure.
Externally to the driver, it is possible either to monitor or to simulate voice-coil temperature and dynamically adjust the input gain to counter the thermal-compression effect. With digital signal processing (DSP), it is feasible to correct for temperature-induced changes in frequency response and transient response as well, and to provide the thermal protection that becomes advisable when the driver's inherent self-limiting behavior has been circumvented.
But it can be argued that the DSP approach is a sledgehammer to crack a nut, and that a much simpler means exists to circumvent thermal compression—one that has the further benefit of also reducing some nonlinear distortions in the driver. This involves abandoning voltage drive and adopting current drive instead.
Conventional audio power amplifiers act as voltage sources: they have low output impedance compared to their load impedance, so that their output voltage is largely unaffected by changes in load impedance with frequency. This is an important attribute, as most loudspeakers have an impedance that varies considerably across the audioband, typically by a factor of 5 or more, yet they expect to receive a flat voltage vs frequency response at their input terminals.