Clipping and power supply sag will affect sound quality
A 2 VRMS power level corresponds to a 2.8 V peak amplifier output. Nearing this level causes clipping and distortion. For higher frequencies, where excursion is not a problem, this is the main issue affecting sound quality.
It is worse with power supply 'sag' because of low battery voltage or high current drain. Since sag is often caused by the audio amplifier itself, this is hard to solve. Some systems lower the amplifier gain when the supply sags - however these cannot typically react fast enough for dynamic signals and still distort the peaks.
Boosting the supply level with integrated DC/DC converters can reduce amplifier clipping by adding headroom, but system designers must be careful not to damage the speaker by over-excursion. The voltage boost can also increase peak battery current and cause the voltage level of a partially discharged battery to drop low enough to cause a system reset, resulting in a dropped call or audio glitch.
Safe operating range for micro speakers without protection
All these parameters can delimit an area of safe operation (Figure 2). A temperature line limits the power amplifier to avoid the worst-case self-heating temperature, and a frequency line removes frequencies below resonance to prevent over-excursion.
Figure 2. Safe operating area for a speaker.
Allowing for changes in the acoustic environment and ambient temperature ensures safe operation, but with only a modest sound output.
For a typical system, limiting the power input to the speaker to 2 VRMS (2.8 Vpeak) and adding a 1 kHz high-pass filter will create a system that remains inside these limits. When used with an 8 ohm speaker it will result in less than 0.5 W (0.9 W peak) output.
Improving the volume output
Systems should always operate near peak output. Because audio signals are dynamic, they only rarely use the amplifier's peak output voltage. Compressing the signal's dynamic range (Figure 3) increases the apparent volume without changing the peak levels (which is why commercials sound louder than the rest of the TV or radio broadcast).
These dynamic compressors work by adding gain to the quiet parts of the music, and quickly reducing it at peaks (the 'attack time'). The attack time is typically very fast (50 µs) and the corresponding decay time over which the gain is increased is typically much slower (5 seconds).
This approach again brings risks. Peak audio signals near the resonant frequency can see very large gain within the attack time. This increases the potential for over excursion and damaged speakers.
Figure 3. Sound sample before and after compression.
The output volume can also be increased by filtering out the resonant frequency. By removing the frequencies near resonance more power can be applied to the remaining signal. That drives more sound from the speaker, but the missing frequencies degrade sound quality.
The filter can be improved and narrowed by using models to predict the behavior of the resonant frequency and speaker temperature. However, any mismatch between the model and the real world can be catastrophic. A blocked speaker port, for example, changes the resonant frequency, with the filter then not protecting the speaker from damage.
Predictive models in these feed-forward systems can also calculate the speaker excursion based on the input signal. That can allow some frequencies below resonance back into the signal which improves sound quality, but it also compounds the risk, because high power signals can be delivered to the speaker where it is vulnerable to damage.
The feedback solution
Eliminating the differences between such complex models and the real world requires feedback. Feedback systems use real-world measurements to update the internal models that predict speaker behavior, and allow the system to produce more sound safely.
Key is to directly monitor the voltage and current to the speaker. This is not as easy as it sounds, since most portable audio systems use class-D amplifiers to reduce power consumption. The sample must therefore be taken after the signal is converted back to analog, which means using an external sense resistor after a power filter. This resistor lowers the system efficiency, because it consumes some of the output power.
Alternatively, more advanced current-sensing systems can be synchronized with the amplifier switching. This approach can provide more accurate results for small systems that don't use power filters on the amplifier output. This solution can be fully integrated inside the amplifier, reducing output pins.
The first step in a feedback system is to measure the speaker voice coil temperature. Because coil impedance rises linearly with temperature, an accurate current measurement can provide a stable and accurate temperature measurement. This can accurately protect against thermal speaker overload as long as manufacturing variations are properly accounted for during production.
The next major step in protecting the speaker comes from controlling the excursion directly. Basic feed-forward systems can measure temperature to estimate the speaker resonance. More advanced systems use current sensing to accurately measure the impedance across all frequencies. The impedance spectra generate an adaptive model which can accurately predict the speaker excursion.
With direct information on excursion, a system can always drive low frequencies into the speaker without damage. If the speaker port becomes blocked, the resonant frequency changes and the system will adjust the signal to prevent damage.
The excursion information can also be used to optimize the output from the speaker, rather than optimizing a fixed electrical level. Here, the speaker can always use the maximum possible excursion for the desired signal. That also improves the sound quality by avoiding clearly audible distortion caused when the speaker moves beyond its limits.
Feedback can also use information from the DC/DC converter to optimize sound quality and system performance. Monitoring the current and voltage at the DC/DC converter can detect supply sag and adjust the peak output accordingly. This can ensure that the audio signal is never clipped, and sound quality (along with system performance more generally) will not degrade as the battery discharges.
Additional feedback points can further improve sound quality and system performance while also avoiding the risks of using higher supply voltages. This brings a huge performance improvement in SPL, sound quality and speaker reliability.
A feedback-based solution gives several key advantages by automatically adapting to changes in acoustic and thermal environments. A full solution, however, must use a combination of techniques.
Adaptive excursion control is needed to ensure that the speaker membrane excursion never exceeds its rated limit. Real-time temperature protection is needed to directly measure voice-coil temperature to prevent thermal damage.
A design must prevent clipping even with sagging supply voltage, and bandwidth extension must increase the low frequency response well below speaker resonance. And an intelligent DC-to-DC converter is needed to maximize audio headroom even at low battery voltages.