Tantalum Capacitor Benchmark in Portable Audio Applications R. Faltus, T. Zedníček AVX Czech Republic s.r.o., Dvorakova 328, 563 01 Lanskroun, Czech Republic email: [email protected] Ian Smith Wolfson Microelectronics plc, Westfield Road, Edinburgh, EH11 2QB, UK J. Zajaček, J. Šikula Department of physics, Brno University of Technology Technicka 8, 616 00 Brno, Czech Republic A B S T R A C T Designers of the latest portable audio/video equipment have a wide choice of capacitor solutions for coupling and line applications. The key design criteria includes clear noise-free filtering (high quality audio), stability with temperature, small size, light weight and reasonable cost versus performance value for consumer electronics. The typical capacitance value demand is from 100µF to 330µF for headphone coupling and 10µF for line applications. More capacitor technologies – tantalum, MLCC, NbO and aluminium can be used to meet the capacitance requirements. Tantalum capacitors are increasingly used today in such applications despite in some cases, especially the 10µF line some cheaper solutions are available. The paper, that will be prepared together with a leading audio chip manufacturer Wolfson Microelectronics plc, Scotland, will present results of a capacitor benchmark study in small portable audio applications. coupling of circuits to high input impedance like Introduction operational amplifiers inputs, low value of coupling There are several technical features required from capacitor can be sufficient for a perfect bass tone capacitors used in portable audio devices. transfer. In such cases like line inputs and outputs Noiseless audio filtering performance and high fidelity in capacitors with values from 1 µF to 10 µF are usually passing audio signals measured by total harmonic used. distortion (THD) is required. Stability of electrical Coupling of output devices like headphones or parameters over temperature and mainly over time is loudspeakers demands higher capacitance required to assure an overall sound quality for a long because their nominal impedances are in the range operating time. In most of the cases, portable audio of only 4 Ω to 32 Ω. For example headphones with devices emphasize miniaturization and it often constrains ZN = 32 Ω and considering low pass frequency fL = designers to choose smaller and lower profile capacitors. 50 Hz require capacitance C = 100 µF. Besides decoupling circuits, where capacitors are It demonstrates that the transfer ratio of the coupling largely connected in parallel to power supply lines circuit is optimal when impedance of capacitor is near to of signal processing integrated circuits, amplifiers or smaller than impedance of load. etc., we have a.c.-coupling circuits. This is where Capacitors made by different technologies can be used the capacitor directly affects quality of the audio for both line and output device coupling circuits. signal and that’s why its careful selection is Depending on the technology, they can exhibit different important. effective serial resistance (ESR) change profiles with The purpose of coupling circuits in audio devices is frequency. Together with parasitic effects like piezo effect to separate the unwanted DC voltage from the (in the case of MLCC capacitors) high value of ESR at useful AC signal which we want to let go through to low frequencies can affect sound quality of all devices. In next step of signal processing or to the output our study we focused on how different coupling capacitor device. A coupling circuit can be imagined as a technologies affect overall audio quality, measured by simple C-R differentiator with nominal capacitance noise background and total harmonic distortion plus C of coupling capacitor and input resistance R of noise (THD+N). next unit like amplifier, signal processor or output device like headphones or loudspeaker. We are discussing the input resistance R instead of input Measuring appliance impedance Z for purposes of simplification and For resistance R is equal to modulus of impedance Z. WM8960_6158_QFN32_EV1_REV2 (Ref. 1) with the Then low pass frequency is: WM8960 chip were used. The WM8960 (Ref. 2) is low all measurements, evaluation kit power, high quality stereo codec designed specially for Equation 1: fL = 1 / (2πRC) portable digital applications produced by Wolfson Microelectronics plc. Configuration of bypassing internal Designers choosing the right value of capacitor for the AD and DA converters was chosen, so only the coupling circuit consider the input impedance of the amplifiers and mixers of WM8960 internal structure were following circuit versus low pass frequency (see Eq. 1), in the signal path (Fig. 1). which is governed by the human ear ability to hear bass tones and it’s usually in the range of 20 to 50 Hz. For 2 Noise background of WM8960 EV1M kit with various output coupling capacitors -11 10 C1 standard tantalum 220µF/10V C2 low ESR tantalum 220µF/10V C3 2 x parallely polymer tantalum 100µF/6.3V ® C4 standard Oxicap (NbO) 220µF/6.3V ® C5 low ESR Oxicap (NbO) 220µF/6.3V C6 aluminium electrolytic 220µF/16V C7 2 x parallely MLCC (X5R) 100µF/4V -12 -1 Su [V Hz ] 10 2 f -3 -13 10 -2 f -14 10 Figure 1: WM8960 codec internal configuration -15 10 10 100 1000 10000 f [Hz] Measurements were realized on the right channel (HPR Figure 2: Results of background noise measurement output) and the parallel left channel was shorted at the input of the developer’s kit to minimise the possibility of The resulting spectra of background noise is seen in crosstalk. Used outputs are dedicated for headphones Figure 2, where the output capacitors are as follows: and we took advantage of onboard load of 16Ω resistor C1 – standard tantalum 220 µF / 10 V; max. which was connected at the output of coupling capacitor. ESR = 0.5Ω @ 100 kHz The kit was supplied from external stabilized +5V power C2 – low ESR tantalum 220 µF / 10 V; max. supply. The configuration (see Fig. 1) and setting of gain ESR = 50mΩ @ 100 kHz was loaded from PC using configuration software and C3 – 2 x parallel polymer tantalum 100 µF / 6.3 V; total USB interface. max. ESR = 35mΩ @ 100 kHz,(see Ref. 6) ® C4 – standard OxiCap (NbO) 220 µF / 6.3 V; max. ESR = 0.4Ω @ 100 kHz, (see Ref. 5) Noise background C5 – low ESR OxiCap® (NbO) 220 µF / 6.3 V; max. First we measured the frequency spectrum of the ESR = 45mΩ @ 100 kHz background noise of the evaluation kit configured as C6 – aluminium electrolytic 220 µF / 16 V, (see Ref. 7) described above and then compared the behaviour for C7 – 2 x parallel MLCC (X5R dielectric) 100 µF / 4 V different output coupling capacitors. An input coupling capacitor was fixed by use of a standard tantalum The next step was to observe the noise spectrum of a capacitor 4.7 µF/ 10 V and the input of the measured configuration with an output MLCC capacitor which was channel was shorted. The gain of the internal input PGA actuated by an electrodynamic acoustic exciter. In amplifier of the WM8960 was set to 10.5 dB while other practice, a separate board with the MLCC was amplifiers in the signal way stayed in default setting 0 dB. mechanically fixed to the exciter which was driven by The HPR output was connected by coaxial cable to amplified white noise signal of waveform generator measuring amplifier 3S Sedlak AM22 (Ref. 3) with Agilent 33220A. configuration: 60 dB amplifier – 0.003 Hz to 30 kHz passing filter – 10 dB amplifier. The output was connected to a sampling card Advantech PCI 1716L (Ref. 4). Data was processed by special FFT with resulting variable frequency step and total amplification of 80.5 dB. 3 Noise background of WM8960 kit with MLCC output coupling capacitor R = 16Ω and it was connected to digital THD+N meter NTI Minilyzer ML1. -11 10 C7 2 x parallely MLCC (X5R) 100µF/4V C7 MLCC capacitor actuated by acoustic vibrations Distortion for different output capacitors with fixed input standard tantalum capacitor C11 -40 -12 C1 standard tantalum 220uF/10V C2 low ESR tantalum 220uF/10V C3 2 x parallely polymer tantalum 100uF/6.3V C4 standard Oxicap (NbO) 220uF/6.3V C5 low ESR Oxicap (NbO) 220uF/6.3V C6 aluminium electrolytic 220uF/16V C7 2 x parallely MLCC (X5R) 100uF/4V C7 MLCC actuated by acoustic vibrations 2 -1 Su [V Hz ] 10 -45 -13 THD+N [dB] 10 -14 10 -50 -55 -60 -15 10 10 100 1000 10000 f [Hz] -65 10 100 1000 10000 100000 f [Hz] Figure 3: Noise spectrum with MLCC actuated by Figure 4: Harmonic distortion for different output coupling acoustic vibrations capacitors Figure 3 shows the original background noise spectrum A piezoelectric effect of ceramic capacitors was observed with C7, MLCC output capacitor without any vibration and measured when the electrodynamic acoustic exciter source marked as “C7” and the second spectrum is for was mechanically fixed to the board with output coupling the same configuration but under acoustic vibrations as MLCC capacitor C7. The exciter the same like above described above. The exciter had parameters Z = 8Ω, was driven by amplified white noise from waveform Pmax = 0.25 W and it was driven by P = 3 mW in this case. generator Agilent 33220A and its measured input power was P = 0.18 W. Influence of output coupling capacitor technology Piezoelectric over acoustic quality effect influence over THD+N and comparison of different technologies are seen in the Overall acoustic quality of the evaluation kit configuration Figure 4. was measured by THD+N for fixed input coupling The drop of all characteristic curves between 5 kHz and capacitor C11 a standard tantalum 4.7 µF/ 10 V and 20 kHz (Fig. 4, Fig. 5) is caused by the brickwall filter different output coupling capacitors, as the background inside of THD analyser. The filter removes the third noise measurement above. harmonic component of the signal and so the measured All internal amplifiers of the WM8960 stayed in default THD+N is lower. gain setting of 0 dB. Harmonic signal from a waveform generator Agilent 33220A within the range of 10 Hz to 20 kHz was connected to the right channel of the evaluation kit input while the input of left channel stayed shorted to minimise crosstalk. The level of the harmonic signal was Up-p = 200 mV. The appropriate board output was loaded by resistor 4 Influence of input coupling capacitor technology over acoustic quality Different output capacitors as well as different input capacitor’s influence on THD+N was measured together with piezoelectric effect of MLCC (See the Fig. 5). For all the measurements, a fixed output coupling capacitor was used – C5 low ESR OxiCap (NbO), which exhibited the best results in the previous noise and THD+N measurements. Distortion for different input capacitors with fixed output low ESR Oxicap capacitor C5 -40 C12 polymer tantalum 4.7uF/10V C13 standard Oxicap (NbO) 4.7uF/10V C14 2 x serially standard tantalum microchip 10uF/10V C15 2 x serially consumer tantalum microchip 10uF/4V C16 2 x serially polymer tantalum microchip 10uF/10V C17 MLCC (X5R dielectric) 4.7uF/10V C18 aluminium electrolytic 4.7uF/63V C11 standard tantalum 4.7uF/10V C17 MLCC actuated by acoustic vibrations THD+N [dB] -45 -50 -55 -60 -65 10 100 1000 10000 100000 f [Hz] Figure 5: Harmonic distortion for different input coupling capacitors Frequency dependencies of THD+N are for following input coupling capacitors: C11 – standard tantalum 4.7 µF / 10 V; max. ESR = 9Ω @ 100 kHz C12 – polymer tantalum 4.7 µF / 10 V; max. ESR = 0.5Ω @ 100 kHz ® C13 – standard Oxicap (NbO) 4.7 µF / 10 V; max. ESR = 3.1Ω @ 100 kHz C14 – 2x serially standard tantalum microchip 10 µF / 10 V; total max. ESR = 15Ω @ 100 Khz C15 - 2x serially consumer tantalum microchip 10 µF / 4 V; total max. ESR = 15Ω @ 100 Khz C16 - 2x serially polymer tantalum microchip 10 µF / 10 V; total max. ESR = 4Ω @ 100 Khz C17 – MLCC (X5R dielectric) 4.7 µF / 10 V C18 – aluminium electrolytic 4.7 µF / 63 5 Summary Review table Technology of the capacitor Background noise transmission standard tantalum low ESR tantalum polymer tantalum standard OxiCap® (NbO) low ESR OxiCap® (NbO) standard tantalum microchip High CV tantalum microchip polymer tantalum microchip aluminium electrolytic MLCC 0 + 0 0 ++ n/a n/a n/a ++ 0 Explanation: ++ very good, n/a + good, Output capacitor influence over THD+N 0 + 0 ++ n/a n/a n/a 0 - 0 neutral, Input capacitor influence over THD+N ++ n/a 0 n/a ++ ++ 0 ++ + Insensitivity to mechanical vibrations ++ ++ ++ ++ ++ ++ ++ ++ ++ - - not good capacitance range not suitable THD+N total harmonic distortion + noise - Low ESR capacitors pass less noise to the harmonic component of the measured signal. output than standard ESR parts; the exception is polymer tantalum capacitors which exhibit similar behaviour like standard parts. - Conclusions and Recommendations Low ESR output capacitors exhibited lower Special attention should be paid to selection of overall THD+N, specially low ESR OxiCap® output capacitor technology due to the sound quality of (NbO). - audio device is more sensitive to output coupling In the position of input coupling capacitor, capacitor than input capacitor. standard polymer exhibited higher THD+N; all - other technologies were comparable. - using MLCC is very sensitive to mechanical vibration - output capacitor performance was inferior to the performance in both input a and OxiCap ® Low ESR OxiCap® technology was the best in were rated the second best, closely after low good ESR OxiCap®. The selection between tantalum output and applications. - and performance. Low ESR tantalum capacitors tantalum and OxiCap® capacitors. showed tantalum lower frequencies. THD+N and noise background. The MLCC capacitors ESR technology that has worse result especially at and its piezo effect has negative influence over Aluminium low technology capacitors, except tantalum polymer in both positions of input and output capacitor - The best audio performance can be achieved by OxiCap ® depends on application requirements like mounting space and operating There is a certain drop of THD+N characteristic voltage. curves between 5 kHz and 20 kHz (Fig. 4, - Fig. 5). This is caused by the brickwall filter MLCC can not be recommended for output coupling capacitor. Special attention should be inside of THD analyser, which is removing third paid to the risk of piezo effect that can be 6 - induced for example by vibrations of the PCB as 4] Datasheet of Advantech PCI 1716L measuring PC verified during this test. card http://taiwan.advantech.com.tw/unzipfunc/Unzip/1- Aluminium capacitors performed well in the M2QHV/PCI-1716L_DS.pdf tests, however a special care should be paid to their limited reliability, capacitance drop with time and lead-free process compliance. This is 5] T.Zednicek, B.Vrana et col.,”Tantalum and Niobium beyond scope of this paper. For Reference Technology Roadmap”, http://www.avx.com/docs/techinfo/tantniob.pdf see 7. 6] T.Zednicek, „Tantalum Polymer and Niobium Oxide References 1] Manual and software of capacitors“, Wolfson http://www.avx.com/docs/techinfo/newtant.pdf WM8960_6158_QFN32_EV1_REV2 evaluation kit, 7] T.Zednicek, „AVX OxiCap™ Outperforms Aluminum http://www.wolfsonmicro.com/products/WM8960EV1 Electrolytic Capacitors in Consumer Applications“, 2] Datasheet and product flyer of Wolfson WM8960 http://www.avx.com/docs/techinfo/oxi_vs_al.pdf digital audio codec chip, http://www.wolfsonmicro.com/products/WM8960/¨ 8] Application note „A.C. Coupling Capacitor Selection” http://www.wolfsonmicro.com/uploads/documents/en/WA 3] Producer of measuring N0176.pdf devices, http://www.3ssedlak.cz/en/index_en.html 7