Class D amplifier

A Class-D amplifier , Anglicised and in general usage Class D amplifier (engl. Class-D amplifier ), and switching amplifiers , digital amplifier or digital amplifier called an electronic is amplifier , which primarily as a power amplifier ( power amplifier ) is used . It is characteristic that an analog or digital audio signal is brought into a sequence of pulses by means of a suitable method, for example pulse width modulation (PWM). As a result, the output stage can be operated in switching mode, whereby the switching elements (practically always transistors ) are either maximally conductive or maximally insulating and therefore only know two states. In contrast to the intermediate states of linear operation used in conventional Class A, Class B or Class AB amplifiers , these two working areas show only little power loss . The PWM signal then describes both the frequency resolution of the time axis (corresponding to the sampling rate in digital audio technology) and the dynamic resolution of the level (corresponding to the quantization or bit depth of the audio signal). A reconstruction filter ( low pass ) behind the power stage generates a continuous voltage curve that corresponds to the input signal. Another method is pulse rate modulation .

Class D results from the continuation of the amplifier modes marked with letters . The common term digital output stage or digital amplifier actually describes the special case of the class D amplifier, also known colloquially as “fully digital”, and gives the false impression that a PWM amplifier would amplify using digital technology or only amplify a digital input signal. In fact, the generated pulse width signal is an analog signal with basically infinite resolution on the time axis. Due to the square-wave voltage, it is similar to a digital signal precisely defined by the clock rate in the width of the square-wave pulses. The development of an amplifier for digital input signals up to the control of the required power switching transistors represents a challenge for analog circuit developers.

Class D amplifier types

Class D amplifier with pulse width modulation, MOSFETs and analog input

Block diagram of a half-bridge PWM output stage

An amplifier with pulse width modulation (PWM) and analog control is described below. There are various other analog and digital methods or refinements, which, however, have in common that a signal with only two voltage states is created which corresponds to the input signal on average over time. Examples are pulse density modulation ( pulse frequency modulation ), delta-sigma modulation or sliding mode control . In contrast to the Class AB push-pull output stage , which can amplify a signal analog to the original signal until it reaches the maximum voltage available from the power supply, the PWM output stage works according to a different principle: A symmetrically operating triangle generator oscillates at a typical frequency from approx. 250  kHz (corresponding to a frequency resolution of 96 kHz sample rate ) up to a few MHz, the level of which is compared by a comparator with the level of the input signal to be amplified. In order to correctly map a 20 kHz audio signal sampled at 44.1 kHz, the switching frequency (or working frequency) of the class D amplifier must be at least 100 kHz, as at least five switching cycles can describe the 20 kHz wave. Due to the structure as a comparator, the circuit changes the analog audio signal into a square wave , as can be seen in the block diagram. If the triangular signal is larger than the audio signal, the output jumps to "high", if it is smaller, it jumps to "low". The maximum pulse width is smaller than the cycle time of the working frequency and can therefore never be switched on or off (high or low) for longer than one clock cycle. The audio signal is now in the pulse duty factor of the PWM signal. The mean value is thus roughly proportional to the mean value of the audio signal. This PWM signal is fed to the output stage, in which the actual amplification takes place, consisting of two power transistors in switching mode for one positive and one negative half-wave each.

A PWM Class D amplifier is built either as a half bridge with two transistors and symmetrical supply voltage (positive and negative supply voltage to ground), or in the case of a simple voltage supply with four transistors as a full bridge, which uses two half-bridge stages for load sharing, each half of the Supply current to the output. However, due to the higher switching losses, a full bridge amplifier has an overall efficiency that is up to 10% lower. The transistors of a half bridge basically switch with each clock cycle (one transistor for the switching state high, the other for low). To describe the audio level - (minus infinity), the transistors switch on and off evenly, the ratio of low and high states of the pulse width signal is then 50%. In contrast, in order to describe an audio signal with 100 Hz at full deflection, one transistor is always switched on for the maximum time and the other transistor is switched on for the minimum possible time, for example to represent the peak of a positive half-wave extending over several clock cycles of the operating frequency. In order to exclude a short circuit by switching both transistors at the same time, a mandatory time delay, the so-called deadtime, is inserted between the switching cycles . This delay results in losses in both the frequency and quantization resolution and the resulting corruption of the signal in an increased distortion factor (THD) of the amplifier. For this reason, an attempt is made to keep the deadtime as short as possible, whereby an effective deadtime (including 10% of the rise and fall times of the transistors) of 10 ns is used for the MOSFETs usually used in audio . ${\displaystyle \infty }$

The MOSFET IRF6645, for example, has a rise time of 5.0 ns (rise time) and a fall time of 5.1 ns, for example the IRFI4212H-117P has a rise time of 8.3 ns and a fall time of 4.3 ns.

The pulse-width-controlled square-wave signal, which now superimposes an infinitely high frequency on the audio signal at its switching edges, is then freed from the higher frequency components outside the audio spectrum by means of a low-pass filter and sent to the loudspeakers. The high-frequency switching operation results in increased interference signals in the area of ​​the PWM frequency or their harmonics , which are preferably emitted through the loudspeaker lines and require increased interference suppression measures to avoid radio interference and to comply with EMC regulations . To avoid phase shifts due to the capacitive and inductive elements of the output filter as well as radio interference suppression, filterless modulation methods such as frequency spreading are also used , whereby the interference is spread over a larger frequency range. Class D amplifiers implemented in this way are available as so-called spread spectrum class D amplifiers and, depending on the output power and length of the connection to the loudspeaker, do not require a low-pass filter at the output.

Quantization resolution

While the frequency resolution can be refined by increasing the working frequency, the dynamic resolution is dependent on the working frequency, the effective deadtime and the input signal level. The dynamic resolution between two successive points of intersection of the input signal and the triangular voltage is analog or infinitely fine, whereby the distance or the width of the pulse width signal would have to change with every changed amplitude of the audio signal, but these distances or widths can never be finer or finer. be shorter than the specified, effective deadtime. This is statistically noticeable especially with very high (positive or negative) values ​​of the input signal to be sampled, i.e. especially with the level peaks of an audio signal at full level, which intersect with the level peaks of the triangular voltage at very short time intervals. As soon as an intersection falls within the deadtime, the switching edge only occurs after the end of the deadtime, which means that the corresponding dynamic value cannot be displayed. Since the switching edge is delayed by the entire deadtime in the worst case and a completely different amplitude value can already be present at the input at this point in time, the resolution in the worst case can generally be limited by the deadtime. If the switching edges of the PWM signal are longer than the deadtime, the following switching edge is again subject to the analog and infinitely fine resolution. Based on a switching frequency of 100 kHz (corresponding to a digital frequency sampling with 44.1 kHz sample rate) with a cycle time of 10 µs (10,000 ns) and an effective dead time of 10 ns, a generally simplified one is therefore in the worst case or in all other cases , relative resolution of 1: 1000 limited to one clock cycle is achieved (for the entire modulation range with positive and negative half-wave together), with = 1024 thus corresponding to a 10-bit digital resolution. This theoretical calculation of the resolution allows for a simplification on the one hand any headroom as well as further rise and fall times of the transistors, which are lengthened by thermal changes, inductive or capacitive components in the signal curve or due to load and thus worsen the resolution, and on the other hand an analog or infinitely fine resolution between two half or quarter periods of a cycle or the individual intersections of the input signal and triangular voltage. If the class D amplifier is operated at an operating frequency of around 250 kHz with a cycle time of 4 µs (4,000 ns) (with a frequency resolution of 96 kHz sampling rate), the relative dynamic resolution related to a clock cycle is again at an effective deadtime of 10 ns, at 1: 400, with = 256 corresponding to 8 bit digital resolution. With a switching frequency of 1 MHz, corresponding to a cycle time of 1 µs (1,000 ns), the relative resolution of 1: 100 remains within one clock cycle, with = 64, corresponding to a resolution of 6 bits. For comparison, the 16-bit dynamic resolution in the CD standard is 1: 65536 (with = 65536). ${\displaystyle 2^{10}}$ ${\displaystyle 2^{8}}$ ${\displaystyle 2^{6}}$ ${\displaystyle 2^{16}}$

Class D driver ICs

In addition to the classic circuit structure with MOSFETs, there are also integrated class D amplifiers, for example the Si824x with an analog input section, half-bridge circuit , 120 W power per channel at a PWM frequency of up to 8 MHz (corresponding to a cycle time of 125 ns), one programmable, minimum deadtime of 0.4 ns, a rise time of 20 ns and a fall time of 12 ns. If the effective deadtime is defined again as deadtime plus 10% of the rise and fall time, then with this integrated circuit a minimum effective deadtime of 3.6 ns or - in the worst case - a quantization resolution of 1:34 is achieved.

Digital Class D amplifiers

With switching amplifiers it is possible to carry out most functions digitally. The input signal is then usually a pulse-code-modulated signal that is converted into a control signal for the output stage by a signal processor or a specialized digital modulator circuit. In addition to the pulse width modulation already described, the delta-sigma modulation is used here. Because of the quantization errors in the output stage signal caused by digital processing , methods for noise shaping are used. Only when controlling the output stage the digital domain is left - therefore provides a digital amplifier, in principle a " performance - digital-to-analog converter " is.

advantages

Power consumption of various (ideal) push-pull power amplifiers

Efficiency of various (ideal) push-pull power amplifiers

Class-D amplifiers are characterized by more economical consumption and lower waste heat, both in mains and battery operation. Since it is therefore possible to use smaller heat sinks, and in the case of integrated circuits, it is even possible to dispense with heat sinks, the result is a more compact design. The theoretically ideal class D amplifier has a power-independent efficiency of 100%. The efficiency of ideal (push-pull) analog amplifiers, on the other hand, is between 78.5% (Class-B) and 50% (Class-A) at full modulation, but in the partial load range (with Class-B linear with the output voltage, with Class-A quadratic with the output voltage) continues to drop and accordingly generates a multiple of the output power in waste heat. Real class D amplifiers have efficiencies of 85 to 94% at full modulation, whereby efficiencies of over 60% are possible even in the low load range at 1% of the maximum output power.

disadvantage

In the case of class D amplifiers that are not fed back, in which any interference signals in the output are not counteracted, load-related fluctuations in the supply voltage due to overshoots and undershoots of the half bridges cause distortion of the output signal, since the supply voltage affects both the amplifier circuit and the load supplied in itself. If the supply voltage is not completely smoothed (due to an inferior power supply unit or hum loops ), there is also an audible 50 Hz hum when idling without an input signal, especially in low- priced active monitors , which together with the background noise, depending on the input gain setting, is often enough can be heard one meter away. A negative feedback is only possible to a limited extent in order to avoid instabilities, since with any inductive and capacitive filter components at the output of the amplifier, frequency-dependent phase shifts occur, which also impair the locatability of the stereo image due to the associated transit time differences (corresponding to a passive crossover ) . Furthermore, the high-frequency signal of the modulator and the square-wave voltage generated on the power side can lead to interference in other assemblies both within the amplifier and via the loudspeaker lines if this is not carefully decoupled and shielded . Compared to other amplifier types, class D amplifiers have an increased phase noise or a limited signal-to-noise ratio due to the deadtime in the switching cycle with the described, statistically distributed dynamic resolution losses and, in particular, at very loud levels, an increased THD value in integrated circuits such as the MAX9709 or TAS5630B at the performance limit with a THD> 10%. Also, due to the high internal resistance of the switching transistors and the inductance of the output filtering, the class D amplifier does not have the attenuation factor of a class AB amplifier and is therefore more sensitive to inductive feedback from the speaker's voice coil.

scope of application

Class D amplifiers are used as audio amplifiers mainly in public address systems with high performance and high energy efficiency, as modulation amplifiers in amplitude-modulated radio transmitters and for wireless power supply of medical implants. A wide field of application can also be found as hi-fi amplifiers and in active boxes in consumer and home recording areas, as well as wherever high efficiency is important with low power , e.g. B. in power amplifiers for headphones in battery-operated devices, mobile phones and MP3 players . Due to the combination of low bandwidth requirements and the increased power requirement compared to higher frequencies, they are also used in amplifiers for subwoofers . In full bridge circuits, Class D amplifiers are also found in pure power electronics in switched-mode power supplies, inverters and frequency converters.

Class D hybrid amplifier

A combination of the energy efficiency of the class D amplifiers with the linearity of the class AB amplifiers is achieved by entangling the two systems. A possible circuit design sees z. For example, the load circuit of the class AB amplifier is supplied by the filtered output signal of the class D amplifier, with both amplifiers being controlled by the input signal. While virtually all the disadvantages of the class D amplifier are decoupled from the loudspeaker or the load, the class AB amplifier itself only contributes a relatively small proportion of the total output of the hybrid amplifier. The class D amplifier alone (with an efficiency of 90%) would have a power loss of 10 watts at 100 W output power, whereas a class AB amplifier alone (with an unfavorable efficiency of 50%) would have a full 100 W power loss. The hybrid amplifier described here has an efficiency of around 80% and a power loss of 30 W.

Web links

Commons : Class D Amplifiers  - collection of pictures, videos and audio files

• Jun Honda: Class D amplifier modules, audio design without headaches. Weka Fachmedien GmbH, October 18, 2012, accessed on January 2, 2017 . 
• Gerd Kucera: Chipset for the development of audio class D amplifiers. Vogel Business Media GmbH & Co.KG, January 24, 2008, accessed on January 2, 2017 . 

Individual evidence

1. ↑ Eric Gaalaas: Class D Audio Amplifiers: What, Why, and How. Analog Devices, Inc., June 2006, accessed January 2, 2017 . 
2. ^ David Jones: Comparison of feedback implementations for digital audio amplifiers. AspenCore, November 4, 2007, accessed January 2, 2017 . 
3. ↑ Hubert Reith: power supplies for class d. Retrieved January 2, 2017 . 
4. ^ Class D Amplifiers: Fundamentals of Operation and Recent Developments. Maxim Integrated Products, Inc., accessed January 2, 2017 . 
5. ^ Karl Stephan: Analog and Mixed-Signal Electronics . John Wiley & Sons, Inc., January 2015.
6. ↑ limited preview in Google Book search (section 10.12)
7. ↑ Jun Honda, Jonathan Adams: Class D Audio Amplifier Basics. (PDF; 305 kB) Infineon Technologies AG, accessed on January 2, 2017 (English). 
8. ↑ Sergio Sánchez Moreno: Class D audio amplifiers: theory and design. (PDF; 310 kB) (No longer available online.) June 2005, archived from the original on May 16, 2017 ; accessed on January 2, 2017 (English). Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice.  
9. ↑ Class D amplifiers, switching amplifiers. fairaudio HiFi Lexicon, accessed on January 2, 2017 . 
10. ↑ IRF6645 DirectFET Power MOSFET. (PDF; 252 kB) Infineon Technologies AG, August 5, 2005, accessed on January 2, 2017 (English, data sheet). 
11. ↑ IRFI4212H-117P Digital Audio MOSFET. (PDF; 254 kB) Infineon Technologies AG, August 21, 2006, accessed on January 2, 2017 (English, data sheet). 
12. ↑ Hubert Reith: Principal sources of error from class d. Retrieved on January 2, 2017 (oscillograph image of the actual switching edges). 
13. ↑ Texas Instruments enables significantly lower system costs for consumer audio products. Texas Instruments, August 24, 2009, accessed January 2, 2017 . 
14. ↑ Kristin Rinortner: Reduce the interference spectra with several switching regulators. Vogel Business Media GmbH & Co.KG, December 14, 2010, accessed on January 2, 2017 . 
15. ↑ MAX9709 25W / 50W, Filterless, Spread-Spectrum, Stereo / Mono, Class D Amplifier. (PDF; 1.3 MB) (No longer available online.) Maxim Integrated Products, Inc., 2014, archived from the original on January 14, 2012 ; accessed on January 2, 2017 (English, data sheet). Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice.  
16. ↑ Glen Ballou: Handbook for Sound Engineers . Ed .: Audio Engineering Society. Focal Press, 2015, pp. 851 ( limited preview in Google Book search).
17. ↑ Si824x Class D Audio Amplifier Reference Design. Silicon Laboratories, accessed January 2, 2017 . 
18. ↑ Si824x Class D Audio Drivers. Silicon Laboratories, accessed January 2, 2017 (English, data sheets, circuit diagrams and user manuals). 
19. ↑ Ashish Gokhale: Improving Performance and Reliability in Class D Amplifier Circuits. Hüthig GmbH, August 2, 2011, accessed on January 2, 2017 . 
20. ↑ Matt Felder, Evan Ragsdale: Class D amplifier with digital input. Hüthig GmbH, March 25, 2013, accessed on January 2, 2017 . 
21. ↑ David Jones: TAS5630B 300-W Stereo and 400-W Mono PurePath HD Analog-Input Power Stage. (PDF; 2.6 MB) Texas Instruments, November 2010, accessed on January 2, 2017 (English, white paper). 
22. ↑ limited preview in the Google book search
23. ↑ Johan Huijsing, Michiel Steyaert, Arthur van Roermund: Analog Circuit Design . Springer Science & Business Media, March 20, 2013.
24. ↑ Stefan Stark: A High-Power CMOS Class-D Amplifier for Inductive-Link Medical Transmitters. University College of London, August 2015, accessed January 2, 2017 . 
25. ↑ Glen Ballou: Handbook for Sound Engineers . Ed .: Audio Engineering Society. Focal Press, 2015, pp. 852 ( limited preview in Google Book search).