A class-D puts out a pulse-width modulated (PWM) signal instead of the linear signal that is typical in class-AB amplifiers. The PWM signal contains the audio signal and the PWM switching frequency plus harmonics. The class-D audio amplifier is much more efficient than a class-AB amplifier because the output MOSFETs switch from very high impedance to very low impedance, operating nanoseconds in the active region. With this technique very little power is lost in the output stage. Furthermore, the inductive element of the LC filter or speaker stores energy from cycle to cycle and ensures that switching power is not lost in the speaker.Introduction
Although class-D amplifiers have been around for a while, many people still do not understand the basics of how a class-D amplifier works and what makes it very efficient. This article explains how the pulse-width modulated (PWM) signal is created and how you hear the audio frequencies and not the switching frequencies of the PWM waveform. This article goes into detail on how outputting a PWM waveform is much more efficient than outputting a linear waveform, and why some class-D amplifiers require LC filters while others do not.
B>How does a class-D output signal (PWM) contain an audio signal?
The TPA3001D1 block diagram (shown in Figure 1) helps explain how the PWM signal is formed. First, the analog input class-D uses a pre-amplifier to gain the input audio signal and ensure a differential signal. Next, the integrator stage low-pass filters the audio signal for anti-aliasing and stability. The audio signal is then compared to a triangle wave to create a pulse-width modulated (PWM) signal. The gate-drive circuitry uses the PWM to drive the output FETs, which creates a high current PWM signal on the output.
Figure 1. TPA3001D1 block diagram.
Figure 2 shows how a typical PWM signal is formed from the comparator block in Figure 1. The audio input is compared to the 250-kHz triangle wave. When the audio input voltage is greater than the 250-kHz triangle wave voltage, the non-inverting comparator output is high, and when the 250-kHz triangle wave is greater than the audio signal, the non-inverting comparator output is low. The inverting comparator output is low when the non-inverting comparator output is high and high when the non-inverting comparator output is low. The average PWM non-inverting output voltage, VOUT+(avg) is the duty cycle times the supply voltage, where D is the duty cycle, or “on” time, t(on) / total period, T.
VOUT+(avg) = D * Vcc (1)
D = t(on) / T (2)
The duty cycle of the inverting output, VOUT-, is the 1 — the duty cycle of VOUT+. If the input is at mid-supply, the duty cycle of VOUT- and VOUT+ is 0.5.
VOUT-(avg) = (1-D) * Vcc (3)
Figure 2: Comparator's inputs and PWM outputs of a typical class-D amplifier.
The TPA3001D1 and TPA3002D2 use the filter-free modulation scheme that is used in the TPA2005D1. With this modulation scheme, the positive output, VOUT+, is the same as the typical class-D PWM, but the negative output, VOUT-, is not just the inverse of VOUT+. In this case there are two comparators and the positive integrator output is compared to the triangle wave to create VOUT+'s PWM, and the negative output of the integrator is compared with the triangle wave to create VOUT-'s PWM. Figure 3 shows the Comparator inputs and PWM outputs for the filter-free modulation scheme, where the audio signal is assumed to be a dc voltage since the audio signal contains much lower frequencies than the 250 kHz triangle wave. Figure 3 also shows the differential output voltage.
Figure 3: TPA3001D1 and TPA3002D2 inputs outputs and PWM.
Figure 4 shows the TPA3001D1 PWM outputs with a 20 kHz audio input signal. Notice how the duty cycle increases as input voltage increases.
Figure 4: Scope plot showing input signal, output before filter, and output after filter (sine wave and PWM).
An audio signal in the PWM waveform is much easier to see in the frequency domain. The PWM signal is made up of the input frequency, the switching frequency and the harmonics of the switching frequency plus side bands. Figure 5 shows amplitude versus frequency of the input, PWM output and filtered output. Figure 5 shows how the audio signal is extracted from the PWM by low pass filtering. The filtered output has the 1 kHz sine wave frequency component plus any of the 1 kHz harmonics that show up in the audio band as distortion plus any remaining ripple voltage from the switching frequency. The speaker cannot reproduce the switching frequency and its harmonics, and even if it could, the ear could not hear it. A listener would not be able to tell the difference between the filtered and unfiltered PWM signals shown in Figure 5 if they were both sent directly to the speaker.
Figure 5: Amplitude versus frequency plot showing input signal, output before filter, and output after filter.
*This is article is in two parts. Look for part 2, which examines efficiency and filtering in detail, next week.
How efficient is a class-D amplifier and how do you calculate efficiency?
A linear amplifier supplies a set amount of supply current for the desired output voltage. The supply current equals the output current in a bridge-tied-load (BTL) class-AB amplifier. A class-D amplifier is a sampled system and delivers a set amount of power into the load for a given period. The class-D amplifier works at delivering the same amount of power from the power supply to the load by putting out a pulse-width modulated (PWM) signal and using the decoupling capacitor and the output filter inductor or inductance of the speaker (for filter-free modulation) as energy storage elements. The PWM signal makes the output voltage switch between the supply rails, yielding very little voltage drop across the output transistors. In contrast, the class-AB output FETs spend most of the time in the active region between the supply rails causing much dissipation and low efficiency.
An ideal class-D amplifier would have 100 percent efficiency because it attempts to supply an equal amount of power from the supply to the load. The ideal MOSFETs of the class-D amplifier would have zero drain-to-source resistance in the “on” state,- rDS(on),- and infinite drain-to-source resistance in the “off” state, -rDS(off). Unfortunately, all MOSFETs have a nonzero rDS(on) and a finite rDS(off). The power loss due to rDS(on) and rDS(off) is conduction loss. A voltage divider is formed by rDS(on), rDS(off) and the output load or speaker, RL. The value of rDS(off) is large enough that it can be ignored when calculating efficiency. A voltage divider using rDS(on) and RL is shown in Figure 6.
The efficiency equation, which is defined as the ratio of output power to power supplied, is shown in Equation 5. The filter inductor or inductance of the speaker (for filter-free modulation) keeps the high frequency switching currents low such that the current taken here is the current in the audio band. The switching current loss is taken into account in the quiescent loss discussed below. The current through rDS(on) is equal to the current through the load, which causes output power to drop out of Equation 5, making efficiency due to conduction loss independent of output power. The efficiency due to conduction loss is shown in Equation 7.
Efficiency = POUT / PSUP (5)
Efficiency (CONDUCTION) = iL^2 * RL / iL^2 * (2rDS(on) + RL) (6)
Efficiency (CONDUCTION) = RL / (2rDS(on) + RL) (7)
Equation 7 can be used as a first order approximation to calculate the effect that rDS(on) has on efficiency. For an rDS(on) of 0.1 ohms and a load resistance, RL, of 4 ohms, the efficiency is 95 percent. If the rDS(on) is raised to 0.3 ohms, the efficiency drops to 87 percent.
The amplifier's bias currents, gate charge, and switching currents also dissipate power. To calculate the efficiency due to two or more losses, PSUP in Equation 5 needs to be broken down in terms of output power and power dissipated.
Efficiency = POUT / PSUP = POUT / (POUT + PD1 + PD2 + PD3 …) (9)
The amplifier's bias currents, gate charge, and switching current losses can be considered independent of output power, since the conduction losses dominate at maximum output power, and can be summed into a quiescent loss, PQ. The quiescent loss is calculated by measuring the supply current when the device is active with no input signal (with the filter and load that will be used in production), and multiplying that current by the supply voltage to get the quiescent loss.
PQ = IDD(q) * VCC (10)
To use the efficiency equation (9), the power dissipated in conduction loss must be extracted from Equation 7. Solve Equations 7 and 9 for power dissipated from conduction loss, PD(CONDUCTION). The result is shown in Equation 12.
Efficiency (CONDUCTION) = RL / (2rDS(on) + RL) = POUT / (POUT + PD(CONDUCTION)) (11)
PD (CONDUCTION) = POUT * 2rDS(on) / RL (12)
Insert the dissipation losses from Equations 10 and 12 into Equation 9 for the class-D efficiency.
Efficiency = POUT / POUT + (POUT * 2rDS(on) / RL) + PQ (13)
The quiescent losses dominate at low output power levels and the conduction losses dominate at high power levels.
A class-D amplifier has much higher efficiency than a class-AB amplifier. The higher efficiency means lower power dissipated, which allows a 12V class-D amplifier to be used without a heat sink while a heat sink is required for an equivalent class-AB amplifier. Figure 7 shows the measured power dissipated of a stereo class-D amplifier, TPA3002D2, compared to the power dissipated of an ideal stereo class-AB amplifier. At 10 W of output power, the TPA3002D2 dissipates only 3.7 W into 4 ohms while the class-AB equivalent dissipates 14 W!
Figure 7: Output power versus power dissipated for a TPA3002D2 compared to an equivalent class-AB amplifier.
Why do some class-D amplifiers require filters while others do not?
The filter-free modulation scheme was developed to greatly reduce or eliminate the output filter. The filter-free modulation scheme minimizes switching current, which allows a very lossy inductor or even a speaker to be used as the storage element in place of an LC filter and still allows the amplifier to be very efficient.
The traditional class-D modulation scheme has a differential output where each output is 180 degrees out of phase and changes from ground to the supply voltage, VCC. Therefore, the differential pre-filtered output varies between positive and negative VCC, where filtered 50 percent duty cycle yields zero volts across the load. The typical class-D modulation scheme with voltage and current waveforms is shown in Figure 8. Note that even at an average of 0 volts across the load (50 percent duty cycle), the peak output current is high, causing filter loss, thus increasing supply current. An LC filter is required with the traditional modulation scheme so the high switching current is recirculated in the LC filter instead of being dissipated in the speaker.
Figure 8: Traditional class-D modulation scheme's output voltage and current waveforms into an inductive load with no input.
The filter-free modulation scheme, as shown in Figure 9, has each output switching from ground to the supply voltage. However, VOUT+ and VOUT- are now in phase with each other with no input. The duty cycle of VOUT+ is greater than 50 percent and VOUT- is less than 50 percent for positive voltages. The duty cycle of VOUT+ is less than 50 percent and VOUT- is greater than 50 percent for negative voltages. The voltage across the load sits at zero volts throughout most of the switching period, greatly reducing the switching current, which reduces I2R losses in the filter and/or speaker. The low switching loss allows the speaker to act as the storage element while still keeping the amplifier very efficient.
Figure 9: Filter-free class-D modulation scheme's output voltage and current waveforms into an inductive load.
Although the switching frequency components are not filtered out, the speaker has high impedance at the switching frequency so little power is lost in the speaker. The speaker also cannot reproduce the switching frequencies and even if it could, the human ear cannot hear frequencies higher than approximately 20 kHz.
The 5V filter-free class-D audio amplifiers like the TPA2005D1 can be used without an output filter if the traces from amplifier to speaker are short. The TPA2005D1 passed FCC and CE radiated emissions with no shielding with speaker wires 10 cm long or less. Wireless handsets and PDAs are great applications for class-D without a filter. The higher voltage filter-free class-D amplifiers like the TPA3001D1 and TPA3002D2 require a ferrite bead filter in all applications.
A ferrite bead filter can often be used if the design is failing radiated emissions without an LC filter, and the frequency sensitive circuit is greater than 1 MHz. This is a great option for circuits that just have to pass FCC and CE because they only test radiated emissions greater than 30 MHz, and a ferrite bead filter is better than an LC filter at attenuating the high frequencies >30MHz. If choosing a ferrite bead, choose one with high impedance at high frequencies, but very low impedance at low frequencies.
An LC output filter must be used if there are low frequency (< 1 MHz) EMI sensitive circuits and/or there are long leads from amplifier to speaker. Figure 10a and Figure 110b show typical ferrite bead and LC output filters.
Figure 10a: Typical ferrite chip bead filter (Chip bead example: NEC/Tobin: N2012ZPS121).
Figure 10b: Typical LC output filter, cutoff frequency of 27 kHz for a speaker resistance of 8 ohms.
A class-D audio amplifier creates a pulse-width modulated, PWM, signal by comparing the input audio waveform with a triangle wave. The class-D amplifier outputs the PWM through an inductive element, the filter inductor for traditional class-D or the voice-coil of the speaker for filter-free class-D. The class-D amplifier is more efficient than a class-AB amplifier because a class-D amplifier takes the required output power from the supply instead of taking the required current from the supply and dissipating the left-over power in the output transistors. A stereo class-AB amplifier dissipates 14 W when outputting 10 W from a 12 V supply with a 4 ohm load, while the TPA3002D2 dissipates only 3.7 W under the same conditions. The TPA3001D1 and TPA3002D2 use a modulation scheme that allows them to use a ferrite bead filter instead of a full LC filter.
1 TPA2000D2 2-W Filterless Stereo Class-D Audio Power Amplifier Datasheet, Texas Instruments, Inc., March 2000, publications number SLOS291D.
2 TPA2005D1 1.1-W Mono Filter-Free Class-D Audio Power Amplifier Datasheet, Texas Instruments, Inc., July 2002, publications number SLOS369B.
3 TPA3001D1 20-W Mono Class-D Audio Power Amplifier Datasheet, Texas Instruments, Inc., December 2002, publications number SLOS398.
4 TPA3002D2 9-W Stereo Class-D Audio Power Amplifier with DC Volume Control Datasheet, Texas Instruments, Inc., December 2002, publications number SLOS402.