Six years ago, digital amplifier technology was new, and manufacturers wishing to adopt it faced three major obstacles. First, the signal path changed architectures. Digital amps accept a pulse-code modulation input and output a high-voltage, pulse-width modulation output. That required a switch from an analog-centric data path to an all-digital data path. Second, power is delivered to the speaker through a MOSFET H-bridge, replacing the linear, Class AB amp. The third hurdle, and the toughest one to clear, was that the power supply typically (but not always) transitioned from a loosely regulated linear supply to a switched-mode power supply (SMPS).
A digital amp applies a 30- to 40-volt signal from a power supply directly across a speaker and reverses it at a frequency of several hundred kilohertz. Sound is re-created by changing the duty cycle (pulse width) of this voltage. The voltage is applied across the speaker with four MOSFETs in an H-bridge configuration. The transistors are either fully on or fully off, so there is limited thermal loss. Digital amps typically have no feedback and apply the power supply voltage directly to the speaker, so the power supply typically needs better regulation than a system with feedback.
Here are some design guidelines aimed at avoiding common problems in designing with an SMPS and digital amplifier. Following these recommendations will cut development time.
Do
• Use a switched-mode power supply. Compared with a linear supply, an SMPS is smaller and lighter, and has a better cost-to-watt and -volume ratio. The trade-offs are increased electromagnetic interference, higher complexity and a different load-handling curve. Those trade-offs can be overcome, but they require new techniques.
• Pay attention to system layout and to the higher EMI of SMPS switching. The high voltage requires that certain design rules be followed and that regulatory-agency safety approvals be obtained.
• Examine the short-term and long-term power. An audio signal has a high crest factor, meaning the peaks may be high, but the average power remains far below the peak. Worst case, the average audio power at full power will be approximately 1/8 full power. For example, in a 5.1-channel home theater system capable of 100 watts/channel, the average power needed will be 600 W/8 = 75 W. SMPS efficiency is approximately 80 percent, so if the power supply can provide 100 W, it should work well. The Federal Trade Commission has a requirement that after a one-hour warmup period where all channels are driven at 1/8 power, two channels must be driven at full power for 5 minutes.
• Use an off-the-shelf SMPS, if possible. An SMPS is usually used in digital TVs, DVD receivers and DVD players. Because of the high volume, the cost has come down drastically in the past few years.
• Reduce the source impedance of the SMPS. One difference between digital and linear amps is that digital amplifiers have an open-loop architecture. The source impedance of the SMPS has a directly proportional relationship to the total harmonic distortion of a digital amp. The best way to address this is to place the SMPS as close as possible to the digital amplifier board, and to use wide power supply traces and low-gauge power wires.
Don't
• Assume the digital amplifier is the source of EMI. In almost every customer engagement that reported an EMI issue to Texas Instruments, the problem was traced not to the digital amplifier, but to the SMPS. EMI from a digital amplifier is minimal if the reference layouts are followed, especially for the high-current signal paths.
• Forget the differences in the overload handling. SMPSes and linear power supplies have different overload-handling methods. The transformer in a linear supply has an impedance. As the current increases in the transformer windings, the resulting IR drop causes the output voltage to decrease and the transformer to heat up. A thermal sensor is usually included in the transformer to prevent a permanent failure. An SMPS has a different protection mechanism. The voltage drop is much more gradual as the current load increases. However, once the current or thermal limit is reached, shutdown occurs almost immediately. This is not a problem, as long as the overload point is properly specified.
• Overspecify the power supply regulation. Because digital amps are open loop, they have a lower power supply rejection ratio (PSRR) than linear amps. An SMPS with less than 5 percent regulation works well for most designs. PSRR is defined as 20 x Log(Vout(f)/Vinjected(f)), when no audio signal is present. Because the output of TI's digital amplifier is muted when there is no input audio signal, one could claim to have infinite PSRR. But a 1-kHz audio signal is typically injected, and the output power of the injected signal is measured. The result is seen above for TI's digital amps, showing about 60 dB of PSRR, especially at the low frequencies, where it is most important.
By Kevin Belnap (k-belnap@ ti.com), digital home audio product-marketing engineer, Texas Instruments Inc.
See related chart
