CORE Metadata, citation and similar papers at core.ac.uk

Provided by Universiteit Twente Repository

288

A Power Efficient Audio Combining Switching and Linear Techniques

R.A.R.vaader Zee’, A.J.M. van Tuijl’ ’MESA Research Insuute, University of Twente, Enschede, the Netherlands ’Philips , Nijmegen, the Netherlands R.A. R.vanderZee@el. utwente.nl [email protected]. corn

Abstract

Integrated Class D audio ampI13ers are vely power +Vdd eficient, but require an external filter which prevents further integration. Also due to thisfilter, large feedback Modulator factors are hard to realise, so that the load influences the - andtransfer characteristics. The amplifier presented in this paper consists of a switching part that contains a much simpler filter, and a linear part that ensures a low distortion and flat f’equency response. A 30W version was realised. The switching part of the amplifier was integrated in a BCD process. Together with a linear part and with a louakpeaker as load, it has a flat f’equency response +/- 0.3dB, a -VSS 1 dissipation that is up to 5 times lower than a traditional Figure 1. Typical class D audio amplifier class AB audio amplifier, and a distortion of <0.02% overpower andf’equency range.

1. Introduction During the last few years, there has been a growing market demand for audio with a high output power and a large number of channels. This is a result of the increasing dynamic range of audio equipmentand the movement towardsmultichannel sound systems tm 1mO (surround sound). At thesame time, however, the fM dimensions have become smaller. Mini sets, car radios, Figure 2. Loudspeaker impedance and PC multimediaequipment have only littlespace available. This requires integrated audio amplifiers with shielding are necessary. Another problem with switching a small number of external components and a low power amplifiers is the distortionand the influence of the load. dissipation.Traditional class AB amplifiers, which are Figure 2 shows the impedance of a loudspeaker. The still largely used, are not suited tomeet these demands. class D filter, however, is designedfor areal and Of the existinghigh efficiency audio amplifiers[ 1-71, constant load impedance. The result of connecting the class D amplifiers [3-61 have the highest efficiency. A loudspeaker is shown in Figure 3. The flat line is the typicalclass D amplifier consists of a modulator that simulated transfer of an ideal classD filter followed by a converts an analogue or digital audio signal into a high fourth order Butterworth filterwith a comer frequency of frequencyPulse Width Modulated(PWM) signal, 30kHz, loaded with the specified load impedance of 42. followed by a half bridge power switch (Figure 1). The The other line showswhat happens when the loudspeaker low frequency audio signal is reconstructed by means of is connected. The transfer deviates several dB’sfrom the anLC-filter. For sufficient suppression of the carrier flat line. This will colour the sound impression. Applying frequency,typically a fourth order filter is necessary. feedbackbefore the filter [3,4] doesnot solve this There are some disadvantages to this principle. Due to problem. Feedback after the filteris much more difficult, theswitching action, Electro Magnetic Interference and high feedback factors can not be realised[5,6]. (EMI) can be a problem; acarehl design and good 289

I7 the main load current. L1does not have to be very linear, 18 because AB determines the output voltage. IS 14 3. Dimensioning 13 iii 12 3.1. Current ripple and coil = 11 10 The choice of the threshold currentIh and the coil LI

0 depends on two important parameters. The first is the

8 switching frequency.

7 100 lwo 1ow

f [Hz1 Figure 3. Simulated class D transfer with resistor and loudspeaker load

Even when these problems are overcome, the filter can not beintegrated. A reduction of thecomplexity (number of components) of this filter is desirable. This paper describes the designand realisation of an amplifier t- that has a switching amplifier and a linear amplifier in Figure 5. Typical currents parallel [7,8]. This way, thenumber ofexternal componentsis reduced, and the transfer is much less Figure 5 shows the typical currents in the amplifier, dependent on the load or the filter characteristics. from which theswitching frequency for any V, and dVJdt canbe calculated: (refer to Figure 4 for the 2. Thecircuit principle meaning of the symbols) The circuit principleis shown in Figure 4.

Thesecond important parameter is thepower P bandwidth of D. The slew rate of IL~should always be largerthan the slew rateof ILs. When weassume a sinusoidal input signal Vo=aVssin(2?rf t), the maximum frequency that can be amplified without distortion is:

Figure 4. Circuit principle When D is not able to provide the full load current, A switching part (D), consisting of the two switches AB could supply the rest. This is the approach in [7], SW1,2and the coil Ll, is connected to the output. The resulting in a very low switching frequency. In that case, other coil and the two capacitors of the standard class D however, AB must be able to supply the full load current filter, however, are replaced by an (integrateable) class for high frequency audio signals. In the design presented ABamplifier (AB), alsoconnected to the output. AB here, we want D to provide the full load current for all controlsthe output voltage. Theinput signal to D is audio frequencies, so that AB can be small. derived from AB'S output current. The output current of Practical values for a realisation are: a = 0.85, Vs = AB is measured by A. The system is self-oscillating: if 1SV, RLs = 40, and I& = 1OOmA (raising the quiescent switch SWl isclosed, and SW2 is open,the current power by 0.9W). According to equation (2), for a power through L1 increases linearly with time, and flows right bandwidth of 2OkH2, L1,should be 20pH. With equation into AB. This is measured by A, and when IAB exceeds a (I), it can be calculated that the resulting free running certain (small) value, SWI is opened, and SW2 is closed. switching frequency at Uo=O would be 2.25 MHz. Since Then, the current through L1decreases, etc. Because IAB this leads to unacceptable switching losses in the power oscillates between two small threshold currents (+/- Ih), switches, L1 is chosen 80pH. With thisvalue, the the power dissipation in AB is small, while D delivers amplifier can deliver a full power sinusoid up to 5kHZ, 290 decreasing to 20% of full power at 20kHz. This makes theamplifier a suitable candidate for Transient InterModulation distortion (TIM). However, audio signalshave a limitedhigh frequency contents [lo]. Computer simulations with a behavioural model of the compl Handshake amplifierrarely showed any distortion ofthis type. Lonic Preliminary listening tests did not reveal any audible & drivers differences either. comp2 SR FF 3.2. AB’S output impedance AB’Soutput impedance mustbe low to achieve a -vsw good rejection of the switching frequency. At least,it Figure 7. Comparators and power switches shouldbemuch smaller than the minimum load impedance. When AB is realised as an amplifier with a The amplifiers are inverters with a feedback resistor. first order roll-off loop gain, the output impedance seems 6 stages offer a total gain of more than 10,000. An SR inductive. This virtual output inductance (0.1SZ@ 1MHz flip-flop creates the output such that the comparators is approximately 15nH) can form a resonant circuit with behave like one comparator with hysteresis. A chip photo a capacitive load, making the self oscillating loop unstable. An output inductor of >>15nH (a small air coil) in series with the load stabilises the system. Calling the virtualoutput inductance LAB, the switching frequency attenuation is LJLAB, which is in our case 74dB.

4. Realisation The modular structure of the amplifier is reflected in the experimental realisation. The linear part is at present still external, and builtwith a commercially available Figure 8. Chip photo of sensing and comparator power OPAMP. Itmust source and sink 1OOmA. The section. switchingpart of the amplifier was realised in two modules in a BCD process. This comparator module is followed by thepower switches, realised as DMOS . A bootstrap capacitor provides the upper gate voltage, and a control LJ circuit avoids common codduction of the two transistors ,I by means of a handshake procedure. 5. Measurements A 30W version of the amplifier was measured. The maximum efficiency is 85%. This is slightly lower than in a normal class D amplifier (90% is possible [5]) due to the extra dissipation of AB (quiescent power and current 1 ripple dissipation). The dissipation for audio signals up comp2 to 15W is shown in Figure 9. Above 15W, audio signals are heavily distorted through clipping. These -Vs c’ measurementswere done with the IEC-268 testsignal Figure 6. Output current sensing circuit that has the frequency- and amplitude characteristics of audio, and gives a good prediction of amplifier efficiency Figure 6 shows the circuit diagram ofthe output for audio signals [ 113. As a comparison the dissipation of current sensing circuit. The output current ofAB is a standard (arbitrary) class AB amplifier is displayed. measured by sensing its supply lines. The measuring The loadhardly influences the performance of the resistors are 0.1 SZ, and also integrated. Two copies of the amplifier, as is shown in Figure 10. Figure 10 displays output current are made, each of which receives a the transfer characteristics for both a resistorand a opposite offset by means of I*. In Figure 7, they are loudspeaker connected to the output. The decrease at followed by two comparators. Regenerative comparators high audio frequencies is the result of the output offer the lowest power-delay products [9], but since no inductor, but it is still a major improvement over Figure external clock signal is available, a multistage amplifier 3, which has the same scale. design is chosen. 29 1

Figure 11 shows that the distortion witha resistor load Furthermore,the transfer ofthe amplifier is less at lkHz remains less than 0.02% up to the clipping point dependenton the connected load. This is a newstep at 30W. The distortion with a loudspeaker as load, and at towardshighly integrated power efficient audio different frequencies always remains smaller than0.02%. amplifiers. Future work will concentrate on designing the The measured residualswitching noise atthe output is linearamplifier in the same process, and mergethe 3mV.Table 1 summarisesthe main specifications of thedifferent modules on one chip. amplifier. 20

15 1 Max. outmt Dower 1 R, c=4R.THD=0.1% I 30W j 1 Maximumefficiency 1 P0,,=3OW I 85% i Quiescent power 1.8W -j Power dissipation Po=lOW, IEC268 3.5W i test signal 5 " I..,.,...I.. "..I ...... ,.I..^ ,.,., _...... ".l." ..I..._" ...... THD+N 20Hz-20kH2, 0.003- ' (filter 22Hz-22kHz) 1W-30W 0.02% I 0 Frequency response 20Hz-20kHz, R or 0.01 01 1 10 +/- 0.3dB j LS load Po Iwl Figure 9. Power dissipation for IEC-268. jWitchattenuation 1 f=500kHz,RLs=4R I 77dB _I: 7. References

[I] P. Buitendijk "A 40W integrated carradio audio amplifier". Int. Conf. Consumer Electron.,Dig. Tech. Paper, Rosemont, IL, June 5-7, 1991, pp. 174-5. [2] E. Botti, T. Mandrini, F. Stefani "A High-Efficiency 4x20W MonolithicAudio Amplifier forAutomobile Radios Using a Complementary D-MOS BCD Technology". IEEE J. of solid state circuits, Vol 31, No 12, Dec 1996, p 1895-901. [3] Z. Lai, K.M. Smedley. "A new extension of one-cycle control and its application to switchingpower amplifiers", 8, IEEE Transactions on Power ,vol. 11,Iss. 1,p.99-105. 1000 10000 100 1000 [4] Jon Hancock "A Class D Amplifier Using MOSFETs f [Hz1 with Reduced Minority Camer Lifetime", J. Audio Eng. SOC. Figure 10. Transfer with resistor and loudspeaker Vo1.39, No.9, Sept.1991, pp.650-62. [5] Niels Anderskouv, Karsten Nielsen, Michael A. E. Andersen "High Fidelity Pulse Width Amplifiers based on Novel Double Feedback Techniques", AES Preprint 4258, 100th AES convention, 1996 May 1 1- 14. Copenhagen. [6] Brian E. Attwood, "Design Parameters Important for the Optimization of Very-High-Fidelity PWM (Class D) Audio Amplifiers", J. Audio Eng. SOC.Vo1.31, No.] 1, Nov.1983, pp.842-53. [7] R.A.R. van der Zee and A.J.M. van Tuijl "A High Efficiency Low Distortion ", 103rd convention of the Audio Engineering Society, New York, Sept. 1997, preprint #460 1. [8] Peter Garde "High-efficiency low distortion parallel amplifier", United States Patent, No. 4,516,080, May 7, 1985. 01 1 10 1w [9] Jieh-Tsorng Wu and Bruce A.Wooley "AIOOMHz Po w Pipelined CMOS Comparator" IEEE J.Solid-state Circuits, Figure 11. THD+N @ 1kHz, filter 22Hz-22kHz vo1.23, pp. 1379-85, Dec. 1988. [IO] R.A. Greiner, Jeff Eggers " The Spectral Amplitude 6. Conclusions Distribution of Selected Compact Discs" Journal of the Audio Eng. SOC., vol. 37,April 1989, pp. 246-75. It is possible to use a linear amplifier to do most of [ll] R.A.R. van der Zee, A.J.M. van Tuijl "Test Signals for Measuring the Efficiency of Audio Amplifiers". 104'h the filtering of a class amplifier for only little D a extra convention of the Audio Engineering Society, Amsterdam, 16- power dissipation. This way, the external filter has less 19 May 1998. components,and does not have to be very linear.