Harsh and noisy environments are commonplace for cell phones, personal digital assistants (PDAs), and other portable communications equipment. This fact has led to the development of new audio power amplifiers (PAs). These PAs offer fully differential architectures with good radio-frequency (RF), common-mode, and power-supply ripple rejection. This article will examine the architectures of the single-ended, typical bridge-tied-load and fully differential audio amplifiers. It also will look at the effects of noise on power supply and RF rectification.
Three main types of audio-power-amplifier architectures are used in this industry: single-ended, typical bridge-tied-load, and fully differential amplifiers. Single-ended (SE) audio power amplifiers tend to be the simplest of all of the architectures. In cell phones, however, they aren't commonly used to drive speakers for applications like polyphonic ringtones or hands-free mode. Typically, SE amplifiers are used to drive headphones for listening to music in MP3 format or for gaming audio (FIG. 1).
In the typical single-supply, single-ended configuration, an output coupling capacitor (COUT) is required to block the DC bias at the amplifier's output. This prevents DC currents in the load. The output coupling capacitor and load impedance form a high-pass filter, which is governed by the following equation:
where RL represents speaker impedance.
From a performance standpoint, the main disadvantage is that the typical small load impedances—in this case between 4-Ω and 8-Ω speakers—drive the low-frequency corner frequency (FC) higher. Large values of COUT are required to pass low frequencies into the speaker. Consider a case in which the speaker load is 8 Ω. If one were to use a COUT of 68 µF, any frequencies less than 292 Hz would be attenuated.
To eliminate the output capacitor (COUT) with a single-ended amplifier, a split supply rail is necessary. This solution isn't good for the wireless environment. It would require cell-phone designers to add a DC-to-DC converter for the negative rail, thereby raising the cost and size of the solution. Furthermore, SE amplifiers are prone to "pop" when they are turned on, turned off, placed in shutdown, or taken out of shutdown. This unwanted noise occurs when there is a certain change of voltage (voltage pulse) across the speaker. It is related to the rise time, fall time, and width of the voltage pulse.
Most humans react to sounds from 20 Hz to 20 kHz. So if the pulse length is less than 50 ms, the ear won't be able to respond. At this point, the frequency will be greater than 20 kHz and no "pop" will be heard. If the pulse's rise time is greater than 50 ms, the ear will once again not be able to hear a "pop." (The frequency will be lower than 20 Hz.) The famous "pop" noise can be heard when the pulse width is greater than 20 ms. Here, the rise time of the pulse is less than 50 ms. Because the single-ended amplifier can only make a pulse if it is turned off immediately, the amplifier must ramp up in >50 ms. This speed is too slow for most smart-phone applications.
With a single-ended single supply, the "pop" also occurs because the output DC blocking capacitor holds charge. When a change occurs at the amplifier output, that voltage—combined with the voltage that's already on the capacitor—will be placed across the speaker. The result is a "pop."
Finally, delivering power to the load is a key concern when talking about audio amplifiers. When using an SE amplifier with a single supply, one end of the speaker is attached to the amplifier's output via an output capacitor. The other end is attached to ground. As a result, the potential across the speaker can only be between VDD and ground. Use the equation for output power to a load:
The maximum peak-to-peak output voltage is the supply voltage. Assuming a sine-wave output, the maximum RMS output voltage is:
The maximum theoretical output power is:
Later, it will be shown that bridge-tied-load (BTL) and fully differential amplifiers can output four times the power of an SE amplifier from the same supply and load impedance.
Today's cell-phone and portable communication devices use a common type of audio-amplifier architecture: the single-ended input with a BTL output configuration (FIG. 2). The BTL amplifier comprises two single-ended amplifiers that drive both ends of the load. The first amplifier (A) sets the gain while the second amplifier (B) acts as a unity-gain inverter. The gain of this BTL amplifier is defined as:
Due to the unity-gain inverting amplifier (B), the gain is double. One of the main benefits of this differential-drive configuration is the power to the load. With the differential drive to the speaker, one side will slew down while the other side is slewing up and vice versa. This characteristic, in effect, doubles the load's voltage swing compared to a ground-referenced load. Because there is effectively twice the voltage swing across the load, the output-power equation becomes:
The maximum theoretical output power with BTL is:
When compared to the single-supply, single-ended audio power amplifier, this doubling of voltage across the speaker results in a quadrupling of the output power from the same supply rail and load impedance.
Another point to consider is the bypass capacitor (CBYPASS). This capacitor is the most critical one in the circuit. It serves several important functions. First, CBYPASS determines the rate at which the amplifier starts up. If the amplifier ramps up slowly, it can reduce the "pop" noise. The CBYPASS and high-impedance resistor-divider network, which generate the mid-rail, result in an RC time constant. As mentioned previously, no "pop" will be heard if this time constant is greater than 50 ms.
The second function of CBYPASS is to reduce the noise that's produced by the power supply. This noise is caused by coupling into the output drive signal. It is derived from the mid-rail generation circuit that's internal to the amplifier. The noise appears as the degraded power-supply rejection ratio (PSRR). In a system with a noisy supply, it can affect THD + N.
Compared to SE audio amplifiers, the advantage of this type of architecture is the amount of output power from the same supply rail. In addition, the output DC blocking capacitor can be removed. After all, the DC offsets are now canceled by both sides of the speaker biased around VDD/2. Now, the low-frequency performance is only limited to the input network and the speaker response.
This type of configuration also has a clear disadvantage, however. If any noise is coupled into the single-ended input, it will still be present on the output and multiplied by the amplifier gain. Because amplifier B has no feedback to the input, any high-frequency noise coupled onto the outputs also will result in clicking and buzzing. This effect is called RF rectification.
FULLY DIFFERENTIAL AMPLIFIERS
A new type of audio power-amplifier architecture is currently in production with many cell phones, PDAs, smart phones, and new wireless devices. It is the fully differential audio amplifier shown in Figure 3. The fully differential amplifier gain is defined as:
Fully differential amplifiers have differential inputs and outputs. These PAs consist of differential and common-mode feedback. The differential feedback ensures that the amplifier puts out a differential voltage, which is equal to the differential input times the gain. The external gain-setting resistors act as the feedback loop.
Common-mode feedback ensures that common-mode voltage at the output is biased around VDD/2, regardless of the common-mode voltage at the input. This feedback is internal to the device. Using a voltage divider and capacitor, it creates a steady mid-supply voltage. To ensure that one output doesn't clip before another output, the output is biased at VDD/2.
Fully differential amplifiers share all of the advantages that the BTL amplifiers have over the SE amplifiers. They also have three main advantages over typical BTL amplifiers. First, the input coupling capacitors are no longer required. When using fully differential amplifiers, inputs can be biased at voltages other than mid-supply. The amplifiers that are being used must have a good common-mode rejection ratio (CMRR). The amplifier's inputs can be biased from 0.5 V to VDD − 0.8 V for the TPA6203A1 and TPA2010D1. If the inputs are biased outside the input common-mode range, however, input-coupling capacitors should be used.
Secondly, the mid-supply bypass capacitor, CBYPASS, is no longer required. Any shift in the mid-supply affects both the positive and negative channels equally. As a result, it cancels at the differential output. Removing the bypass capacitor slightly worsens the PSRR. Yet that ratio may be acceptable when an additional external component is eliminated. The fully differential amplifier's last main advantage is that it boasts better RF immunity. For this strong point, it can thank its good CMRR and fully differential architecture.
To get the load output power, use the same calculation that was used for the BTL amplifier. This amplifier also is fully differential. Keep in mind that one side of the speaker is slewing up as the other side is slewing down and vice versa. Again, this aspect doubles the voltage swing on the load compared to a ground-referenced load. The maximum theoretical output power with BTL is:
As with the BTL amplifier, this doubling of voltage across the speaker results in a quadrupling of the output power from the same supply rail and load impedance. Compared to the previous amplifiers, this type of architecture's biggest advantage relates to noise immunity.
The three main sources of noise for the audio power amplifier are:
- Power-supply noise
- Noise coupled on inputs
- Noise coupled on outputs
Usually, variations on the voltage supply will lead to small, incorrect variations on the amplifier output. PSRR is the ability to reject these effects. Typically, it is expressed in decibels. For the TPA6203A1 fully differential audio power amplifier, for example, the PSRR value is specified as −87 dB from 217 Hz to 2 kHz at 3.6 V. Using the standard equation for PSRR, the output voltage can be calculated:
For a 500-mV variation on the power-supply rail, the change in differential output voltage is 22 µV.
In TDMA and GSM cell phones, the most notorious voltage-supply noise comes from the RF stage switching on and off. The GSM phone switches on and off at a rate of 217 Hz. When the RF power amplifier switches on, high current is drawn from the power supply. As a result, the power supply dips by up to 500 mV. An audio amplifier with poor PSRR will cause more than 217-Hz harmonic clicking noise in the speaker.
To see the impact of a 500-mV supply-voltage drop occurring at the rate of 217 Hz, three fully differential audio power amplifiers were tested: the 3.1-W Class-AB TPA6211A1, 1.25-W Class-AB TPA6203A1, and 2.5-W Class-D TPA2010D1 (FIG. 4). The results from testing the TPA6203A1 and TPA2010D1 show that due to the fully differential amplifiers' PSRR, the variations in the supply rail have virtually no effect on the output signal. Consequently, it won't cause a 217-Hz harmonic clicking sound in the speaker.
With noise that's coupled onto the inputs of a single-ended input amplifier, the main issue is that it will be multiplied by the closed-loop gain. The unwanted noise will then appear on the amplifier's outputs. These types of amplifiers have virtually no noise immunity apart from filtering the input signal prior to the amplifier.
In contrast, the fully differential amplifier is very good at rejecting noise. This amplifier only multiplies the difference between both inputs. As a result, any common-mode interference that's coupled onto the differential-input traces will be effectively ignored by the amplifier. The best way to understand this immunity to input-coupled noise is by looking at the CMRR:
For an example of how the CMRR affects the amplifier's AC noise immunity, use the TPA6203A1 1.25-W, fully differential, Class-AB amplifier. First, take the CMRR equation above and solve it for output voltage:
The TPA6203A1 has a CMRR of −74 dB across 20 Hz to 20 kHz with a gain of 1 V/V. Assume that the common-mode noise, which is coupled onto the inputs, is 100 mV on each input. The noise transferred to the output can then be solved by the equation below:
This equation results in a 20-µV ripple on a differential-amplifier output. In the case of the single-ended input amplifier, the result would be 100 mV multiplied by the closed-loop gain.
When using a BTL output configuration, the most common noise that's heard on the speaker is the RF power amplifier switching on and off at 217 Hz. Typically, this switching can be heard as a clicking or buzzing. To see why BTL amplifiers aren't immune to noise coupling onto their outputs, look at Figure 5 and Figure 6.
During the on state, the radio-frequency power amplifier sends data to the base station. In the laboratory, testers held a GSM phone about 10 cm away from the audio amplifier. Next, they looked at the signal that was picked up on the audio amplifier's outputs. This noise looked like an RF signal gated by a square wave. The actual screenshots are shown in Figure 5.
Looking at the full bandwidth (>20 MHz), one can see that the signal is picked up on each amplifier's outputs. This has no effect, however. The speaker cannot reproduce signals at that high a frequency. When looking at the limited bandwidth (<20 MHz) in the BTL architecture, however, the inverter follower (BTL amplifier) tries to respond to the gigahertz signal. This attempt causes dips on its output (OUT−) at a rate of the gated square wave (217 Hz for GSM). The dips, in turn, result in clicking or buzzing on the speaker.
In this measurement, noise was injected on the outputs—not the inputs. When it was band-limited, OUT+ stayed relatively constant because its input, IN−, did not have noise injected onto it. OUT− had a lot of ripple because OUT+ was the input to OUT−. The inverting amplifier from OUT+ to OUT− tried to respond to the gated radio-frequency waveform. But it only responded to the low frequency. If noise also was injected on the input, OUT+ would have been a lot noisier due to poor CMRR.
The fully differential amplifier's outputs were injected with the same noise as the typical BTL amplifier. When bandwidth is limited, they show no noise because of its differential feedback to the inputs. This effect can be seen in Figure 6. Here, a GSM cell phone was placed near the outputs of a BTL and a fully differential amplifier. The resulting waveform was captured. If noise was injected on the inputs, the fully differential amplifier would have rejected the noise by its CMRR. Compared to a typical BTL amplifier, the fully differential amplifier obviously has the best immunity to RF noise.
THE BOTTOM LINE
Audio power amplifiers are prone to pick up noise from the harsh environments in portable wireless-communication devices. The typical BTL audio power amplifier has several limitations. If noise is coupled onto the amplifier's inputs, outputs, and power supply, it causes clicking and buzzing. In comparison, the fully differential amplifier excels in this type of environment. Thanks to its fully differential feedback and the ability to cancel the effects of RF rectification, it minimizes the "buzz" in cell phones.