Fig 1. This shows the spectra of both waveforms when the Class D output is issuing silence (no audio, duty cycle = 50%) with a switching frequency of 333 kHz.
Fig 2. The blue line shows the energy present in the 30-MHz band as produced by a conventional, unmodified Class D amplifier output. The brown line displays the same Class D output, but with a ±5% modulation by a pseudorandom sequence with a 40-kHz pattern repetition frequency.
Fig 3. Another technique for reducing EMI modifies the modulation scheme so one side of a differential or bridged Class D output pair can cease switching when the audio baseband signal’s amplitude is large enough. This figure shows the one-sided modulation with a filtered differential Class D output and unfiltered positive and negative switching outputs.
Fig 4. The lack of one-sided modulation (green) and the use of one-sided modulation (blue) produce two different Class D output spectrums.
There has been rapid growth in recent years in the number of portable devices incorporating powered speakers—cell phones, MP3 players, GPS systems, laptops and notebook computers, tablet computers, gaming devices, toys, and so forth.
The Class D (or switching) type of audio amplifier is frequently chosen to drive the speakers in these applications because of its reduced heat dissipation (important in compact products) and relatively high efficiency (for increased battery life) compared to the conventional Class AB design.
One possible drawback to the switching topology of the Class D amplifier, however, is its tendency to emit electromagnetic radiation that can potentially interfere with other nearby electronic devices. External, passive filtering can moderate this interference, but it adds cost, area, and complexity to the end product. Fortunately, other internal circuit design methods can be employed to mitigate electromagnetic interference (EMI).
Edge Rate Control
To amplify audio, the output (or outputs, in a differential configuration) of a Class D amplifier alternately switches back and forth between two power supply rails (typically a positive rail and ground) at a frequency 10 times or more than the highest audio frequency to be amplified (perhaps 300 kHz or higher).
This switching pattern is modulated so a simple low-pass filter, sometimes comprising the speaker itself, can recover the audio. The switching transitions tend to be very rapid, perhaps 2 ns or less, and thus contain a significant degree of high-frequency energy.
This can lead to EMI radiation from the interconnection wires, particularly if there is no low-pass filter in the signal path, and if the lead length between the amplifier and speaker is at all significant (more than perhaps 1 cm).
One method used to mitigate this EMI radiation is a reduction in the slew rate of the amplifier output transitions. Reduction in the slew rate by a factor of 10, for example, increasing the rise and fall time from 2 ns to 20 ns, has a pronounced effect on the radiated energy produced by the Class D amplifier.
Figure 1 shows the spectra of both waveforms when the Class D output is issuing silence (no audio, duty cycle = 50%) with a switching frequency of 333 kHz. Throughout the majority of the spectrum between 30 MHz and 1 GHz, the reduction in high-frequency content is approximately 20 dB.
In systems that incorporate FM broadcast reception electronics (88 to 108 MHz) and/or cell-phone or wireless Internet circuitry (700 MHz to 2.7 GHz), this offers a substantial reduction in EMI and therefore also in the risk that system performance could be compromised. However, the EMI reduction advantages of this edge rate control (ERC) do not come without cost.
First, the efficiency advantage offered by Class D amplifiers comes primarily from having the output devices always either fully on or fully off so the instantaneous dissipated power in the output devices, P = VI, remains essentially zero at all times (unlike Class AB amplifiers in which the power device VI product is never zero). The introduction of (or increase in) a span of time at each switching transition during which V ≠ 0 while the load current I ≠ 0 causes a modest increase in on-chip dissipation and therefore a decrease in efficiency.
Second, while a non-ERC output stage is essentially simply a large inverter (perhaps including shoot-through or crowbar current mitigation), an ERC output stage includes additional circuitry that regulates the gate voltages of the pull-up and pull-down devices to properly produce the desired, controlled slew rate at the output terminals. Depending on the approach taken, this adds both die area (cost) and current consumption (reduced efficiency). All told, the efficiency penalty due to the addition of ERC may be on the order of 1% to 2%.
While ERC is an effective means of attenuating EMI that arises in the frequency range above 30 MHz, and which is also restricted by limits imposed by United States Federal Communications Commission (FCC) regulations, it does not address the fundamental carrier frequency of the Class D amplifier switching output and its related odd (square wave) harmonics that fall in the range below 30 MHz so well. The blue line on Figure 2 shows the energy present in this band as produced by a conventional, unmodified Class D amplifier output.
To reduce the height of the fundamental and overtone spikes in the Class D output spectrum, it is possible to add a small amount of frequency modulation to the amplifier’s clock circuitry—perhaps with a modulation index of ±5% or so—that does not affect the amplified audio quality.
Although there are numerous choices for the characteristics of the source of the modulating signal, one standard practice is the use of a pseudorandom pattern with a repetition rate (full pattern repeat frequency) that is above the highest expected audio signal frequency (generally 20 kHz) by a suitable margin. This prevents the generation of tones that could otherwise fall into the audio band.
The brown line on Figure 2 shows the same Class D output, but with a ±5% modulation by a pseudorandom sequence with a 40-kHz pattern repetition frequency. Approximately 10 dB across the spectral range suppresses the odd order harmonics of the fundamental clock frequency.
An additional approach that can be taken to reduce EMI is a modification to the modulation scheme that allows one side of a differential or bridged Class D output pair to cease switching when the amplitude of the audio baseband signal becomes sufficiently large.
The opposing output, which continues to switch, essentially then can accept full responsibility for modulating the output signal the remaining distance up to its full peak value. With this scheme, for a significant percentage of time depending on the audio source material, only one output is switching.
Therefore, the EMI during that time is cut in half (Figure 3 and Figure 4). This has the added advantage of reducing the fixed switching losses due to the charging and discharging of power device gates and other parasitic capacitances. It also reduces the time the outputs spend in ERC transitions that, as described above, have a modest efficiency penalty.
The disadvantage of this technique is that the overall forward gain of the amplifier is somewhat reduced. As such, there is a minor increase in total harmonic distortion (THD) and noise.
Class D amplification is commonly used in portable devices for its power efficiency advantages over conventional Class AB amplifiers. Its primary disadvantage is its inherent generation of EMI that can negatively impact nearby electronics. Yet these known and effective IC design techniques can substantially mitigate EMI concerns without the burden of additional external components.