Colin Davies, Applications Manager Europe, Diodes Inc.
The semiconductor industry continues to deliver greater levels of integration, therefore electronic products will continue to shrink in size but grow in performance. The user experience is constantly evolving as a result, and the physical size of a device no longer determines its impact. This paradigm is enabled by improved audio and visual interfaces, in all classes of product.
Adding high fidelity audio to small and portable devices as a form of user interface has many benefits. Instead of simple monotonic alerts, a device could provide clear instructions spoken in the local language, creating an ‘eyes free’ form of user experience in consumer, industrial and medical applications. The advent of digital assistants like Amazon Echo and Google Home means people are already becoming comfortable interacting with technology using natural language.
As digital assistants use deep learning algorithms running in the cloud to understand instructions and provide local audio feedback, applications that make use of this technology need little more than a microphone and speaker. The challenge then becomes adding high quality audio capability to ever-smaller devices.
Audio amplifiers represent a specific subset of regular amplifier design, because of the propensity for noise and distortion to find its way to the human ear. As all amplifiers work fundamentally in the analog domain they are inherently linear devices, providing gain with minimal distortion. Through biasing, the operating characteristics of the transistors used to configure an amplifier can be varied, giving rise to three main classes of linear amplifier; Class A, Class B and Class AB.
Each of these classes has its relative benefits and drawbacks, allowing engineers to balance fidelity against efficiency. For example, a Class A amplifier uses just one transistor, biased to be on even when the input signal is effectively zero, which results in good linearity but poor efficiency. Class B amplifiers employ two transistors which conduct for half-cycles, so only one transistor is conducting at any time. While this delivers improved efficiency it suffers from distortion at the point where when one transistor turns off and the other turns on; so-called crossover distortion. However, by moving the bias slightly, this crossover distortion can be significantly reduced, giving rise to the Class AB amplifier. Figure 1 shows a comparison of the three types of linear amplifier.
Click image to enlarge
Figure 1: The three main classes of linear amplifier
Although a Class AB amplifier configuration can deliver good efficiency (in the region of 80%) it still experiences losses and in a bid to overcome this the non-linear — or switching — amplifier was conceived. Rather than being partially on or off most of the time, the transistors in a switching amplifier are either fully on or fully off. This is much more digital in nature and in many ways the operation is closely related to a switching power supply. A switching amplifier can deliver theoretical efficiencies of 100% and in practice upwards of 90%, with reasonable output power. These properties make switching amplifiers particularly attractive to developers of small or portable (battery operated) devices.
However, because it is a switching topology, it isn’t without its own challenges. Rather than a smooth amplified version of the input, the output is essentially a Pulse Width Modulated (PWM) square wave, which has to be converted back to an analog waveform that accurately represents the input signal.
The most widely adopted switching amplifier topology for audio applications is the Class D amplifier. The basic method of generating the PWM representation of the input is to use a comparator to switch the output on and off based on the amplitude difference between the input signal and a reference waveform. When the input is greater than the reference the output is high, when the reference is greater than the input the output is low, as shown in Figure 2.
Click image to enlarge
Figure 2: Class D amplifiers use PWM to represent the input signal
The output stage can be as simple as two FETs configured in a complimentary configuration, such that when one is on the other is off (normally achieved using one PNP and one NPN transistor). This is known as a Half-Bridge and, again, has resonance with switched mode power supply design. Perhaps not surprisingly, then, it is also common to use a Full-Bridge configuration, which features two Half-Bridge stages that allows for the path through the load to alternate, producing bidirectional current flow. This topology is normally referred to as a Bridge-Tied Load (BTL).
Due to having a DC offset of half VDD and a 50% duty cycle, current flows through the loud even under no input conditions with the Half-Bridge output stage, but with a Full-Bridge output stage the offset is present on both sides of the load. This effectively reduces the quiescent current to zero and also removes the need for a blocking capacitor when using a single supply, but perhaps more significantly it increases the output swing by a factor of 2.
Of course, because a Full-Bridge output stage uses twice as many FETs it can have additional system costs, such as switching losses, and this can effectively limit the output power. But for many audio applications this is becoming less of a problem thanks in part to the improvements in speaker technology.
Traditionally, reconstructing the waveform from the PWM output requires a low-pass filter at the output, to filter out the high frequency element of the PWM signal and allow the audio component to pass through. This would normally dictate the use of inductors large enough to handle the peak output power, which equates to greater PCB area and higher BOM costs.
This could preclude the use of Class D amplifiers in small and portable devices but it actually becomes a benefit for audio applications. Speakers are, in effect, inductive loads, so the speaker’s inherent inductance can form part of the low-pass LC filter needed to reconstruct the waveform. Using the speaker’s own resistance and inductance as part of the low-pass output filter can significantly reduce the BOM and minimise PCB space.
In addition, because speakers have a finite frequency response, filtering out high frequency elements of the PWM signal actually becomes less important. The fact that the human hearing range is also limited to the low kHz also provides a convenient ‘natural filter’ for high frequency elements.
Careful design of the BTL output stage allows for the speaker to act as a low pass filter for the PWM signal, although the amplifier design does need to specifically focus on this application area. Direct connection to a speaker can result in a current flowing through the speaker at all times, potentially causing damage to the speaker or, if a bias is present, limiting the dynamic range. To avoid this the BTL output stage needs to ensure no current flows through the load under quiescent conditions.
Of course, the human ear is also very good at detecting anomalies and so audio fidelity becomes more important. To achieve the best possible result, Class D audio amplifiers employ integrated feedback to deliver optimum THD+N (Total Harmonic Distortion plus Noise) and PSRR (Power Supply Rejection Ratio); important parameters for design engineers to evaluate when selecting an audio amplifier based on a Class D topology.
As users become more comfortable with audible feedback and control, the applications for small audio amplifiers continues to grow. As well as cellular phones and media players, automated voice responses could be used in medical or industrial equipment, as well as consumer devices and home appliances.
The availability of highly integrated Class D amplifiers will help drive this market, where the output power requirements will be relatively modest. This is where devices like the PAM8014 from Diodes are positioned. Packaged in a wafer level BGA measuring just 1.2mm by 1.2mm, this filterless Class D amplifier offers an output power of up to 3.2W with almost no external components required (as shown in Figure 3).
Click image to enlarge
Figure 3: Typical implementation of the PAM8014 filterless Class D amplifier form Diodes
As well as implementing a BTL output stage, it also features integrated short circuit protection and thermal shutdown, with a shutdown current of less than 1μA. Furthermore, the Total Harmonic Distortion (THD) at 3.2W output is 10% into a 4Ω load when operating from a 5V supply, while the THD+N drops to just 0.14% at an output power of 0.5W. Efficiency is as high as 93%, while the gain is set by the ratio of the input and feedback resistances, as shown in Figure 4.
Click image to enlarge
Figure 4: Block diagram of the PAM8014 filterless Class D amplifier from Diodes
The device includes additional measures to minimise the effect of turn-on and turn-off transients that can manifest as ‘pops’ and ‘clicks’, by disabling the amplifier until the internal reference voltage reaches half the supply voltage. Thermal protection comes in the form of a non-latching shutdown mode that is triggered if the internal die temperature exceeds 150°C, the same mode is entered if a short circuit is detected on an output pin or the supply voltage drops below 2.0V. Shutdown mode can also be entered manually by applying a logic low on the EN pin.
Class D amplifiers offer the ability to add an audible output to even the smallest device. Highly integrated solutions that combine filterless implementation with high performance, good EMI and excellent PSRR will enable a new generation of smart devices.