Paper 8-6 An Open-Loop Class-D Audio Amplifier wih Increased Low-Disorion Oupu Power and PVT-Insensiive EMI Reducion Shih-Hsiung Chien 1, Li-Te Wu 2, Ssu-Ying Chen 2, Ren-Dau Jan 2, Min-Yung Shih 2, Ching-Tzung Lin 2 and Tai-Haur Kuo 1 1 Naional Cheng Kung Universiy, Tainan, Taiwan, 2 NeoEnergy Microelecronics, Inc., Hsinchu, Taiwan
Ouline Background and Moivaion Sysem Overview Proposed Techniques Adapive-Coefficien Dela-Sigma Modulaor PVT-Insensiive Low-EMI Conrol Mehod Measuremen Resuls Conclusions 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 2 of 32
Digial-Inpu Audio Amplifier Class-AB amplifier wih DAC Digial Inpu DAC Class-AB Amp. Speaker Load Class-D amplifier wih DAC Digial Inpu DAC Class-D Amp. Low-Pass Filer Speaker Load Class-D amplifier wih digial PWM (DPWM) gen. Digial Inpu DPWM Class-D Amp. Low-Pass Filer Speaker Load 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 3 of 32
Class-D Amplifier wih DPWM Pros High efficiency No need of high-resoluion DAC Cons Disorion from class-d amp. Degraded THD+N Need of L-C low-pass filer for EMI suppression Digial Inpu DPWM Class-D Amp. Low-Pass Filer Speaker Load 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 4 of 32
Closed-Loop vs. Open-Loop Closed-loop archiecure DPWM Feedback Pah (feedback facor = β ) Digial Inpu Inerpolaor (gain=1) DSM PCM-o-PWM Converer (gain=1) clock Analog Loop Filer Power Sage Low-Pass Filer Speaker Load Open-loop archiecure adoped in his work DPWM Digial Inpu Inerpolaor (gain=1) DSM PCM-o-PWM Converer (gain=1) clock Power Sage Low-Pass Filer Speaker Load Lower complexiy Smaller chip area Easier design poring o advanced processes 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 5 of 32
THD+N vs. Oupu Power Disorion and noise sources Consan noise Power sage disorion Clip disorion 1% THD+N (%) Oupu Power, P OUT (W) Low-Disorion P OUT Low-disorion P OUT = max. P OUT wih THD+N<1% Dominaed by clip disorion due o DSM insabiliy 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 6 of 32
DSM Insabiliy in Open-Loop When DSM inpu is large DSM s quanizer overload clipping a DSM OUT clip disorion a amp. oupu decreased low-disorion oupu power Digial Inpu Inerpolaor (gain k=1) DSM Clipping Error DSM OUT PCM-o-PWM Converer (gain 1/k=1) clock PWM OUT Power Sage Power Oupu SNDR(dB) SNDR @ DSM OUT Inpu mag. (dbfs) 0 SNDR(dB) SNDR @ PWM OUT THD+N vs. oupu power 1% Inpu mag. (dbfs) 0 THD+N(%) Oupu power (W) 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 7 of 32
DSM Insabiliy in Open-Loop To reduce DSM inpu: inerpolaor s gain To increase gain afer DSM: PCM-o-PWM s gain Clipping a PWM OUT DSM insabiliy can NOT be prevened by scaling k Clipping Error Digial Inpu Inerpolaor (gain k<1) DSM DSM OUT PCM-o-PWM Converer (gain 1/k >1) clock PWM OUT Power Sage Power Oupu SNDR(dB) SNDR @ DSM OUT Inpu mag. (dbfs) 0 SNDR(dB) SNDR @ PWM OUT THD+N vs. oupu power 1% Inpu mag. (dbfs) 0 THD+N(%) Oupu power (W) 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 8 of 32
Common-Mode EMI Reducion Convenional BD modulaion VCM V DD 0 V DD 0 V DD V DD /2 0 Common-Mode Free BD (CMFBD) modulaion [1] VCM V DD V DD /2 0 V DD V DD /2 0 V DD /2 [1] P. Siniscalchi and R. Heser, A 20W/channel class-d amplifier wih significanly reduced common-mode radiaed emissions, IEEE ISSCC 2009. 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 9 of 32
Targes of This Work Increase low-disorion oupu power for open-loop class-d amplifiers wihou increasing supply volage increasing off-chip componens sacrificing THD+N a small oupu power Reduce common-mode EMI wihou using expensive L-C filers PVT-sensiive issue 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 10 of 32
Sysem Overview Block diagram of his work Digial Inpu x[n] Proposed ACDSM Dela-Sigma Modulaor (DSM) Adapive Coefficien Se[n] y[n] Dual-Mode Oupu Sage PCM-o- PWM Converer SEL Power Sage Inerpolaor Low- Pass Filer Two selecable modes BD-Modulaion Mode L C L Low-EMI Mode [1] bead C bead 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 11 of 32
Trade-Off in DSM Design Two DSM Designs DSM A : high in-band noise suppression DSM B : full-scale sable inpu range NTF plo Roo-locus plo Gain (db) 20 0-20 -40-60 -80 DSM A DSM B Imaginary axis 1 0.5 0-0.5-1 uni circle DSM A DSM B 0.1 1 10 100 192 Frequency (khz) -1-0.5 0 0.5 1 Real axis 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 12 of 32
Proposed ACDSM Adapive-Coefficien Dela-Sigma Modulaor (ACDSM) Small x[n] coef. wih high in-band noise suppression Large x[n] coef. wih full-scale sable inpu range b 1 [n] b 3 [n] b 2 [n] b 4 [n] x[n] g 1 [n] H g 2 [n] H g 3 [n] H g 4 [n] H g 5 [n] H a 1 [n] a 2 [n] a 3 [n] a 4 [n] a 5 [n] Quanizer y[n] 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 13 of 32
Direc Coefficien Swiching Coefficien is swiched beween Small x[n] Se A (high in-band noise suppression) Large x[n] Se B (full-scale sable inpu range) x[n] Dela-Sigma Modulaor (DSM) y[n] Se B Se A Se A = [g 1A,...g 5A,a 1A,...a 5A,b 1A,...b 4A ] Se B = [g 1B,...g 5B,a 1B,...a 5B,b 1B,...b 4B ] large inernal ransien spike DSM unsable 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 14 of 32
ACDSM Algorihm Linear-inerpolaed coefficien changing operaing coefficien se is changed wih small Se Δ inernal ransien spike is reduced x[n] Dela-Sigma Modulaor (DSM) y[n] Se Δ Se Δ Se Δ Se Δ Se B Se 1 Se N Se A N Ses Proposed ACDSM Algorihm 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 15 of 32
Dynamic Range (DR) Plos The ACDSM simulaneously achieves a wide sable inpu range high in-band noise suppression 120 100 100 80 SNDR (db) 80 60-1 -0.5 0 60 50 DSM A DSM B ACDSM -40-30 -20-10 0 Inpu Magniude (dbfs) 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 16 of 32
CMFBD Realizaion Previous low-emi conrol mehod [1] V DD V G5 M 1 M 5 M 6 M 3 V G1 M 2 Speaker Load M 4 S 0 M 1, M 4 S 1 S 2 M 5, M 6 M 2, M 3 [1] P. Siniscalchi and R. Heser, A 20W/channel class-d amplifier wih significanly reduced common-mode radiaed emissions, IEEE ISSCC 2009. 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 17 of 32
Previous Low-EMI Conrol (1/3) In sae S 0 V DD V G1 V G5 M 1 M 5 M 6 M 3 V G1, M 2 Speaker Load M 4 V G5, S 0 S 1 S 2 M 1, M 4 M 5, M 6 M 2, M 3 S 0 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 18 of 32
Previous Low-EMI Conrol (2/3) In ransiion from S 0 ino S 1 V DD V G5 M 3 V G1, V G1 M 1 M 5 M 6 M 2 Speaker Load M 4 V G5, S 0 S 1 S 2 M 1, M 4 M 5, M 6 M 2, M 3 S 0 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 19 of 32
Previous Low-EMI Conrol (3/3) In sae S 1 V DD M 1 V G5 M 5 M 6 V G1, M 3 V G1 M 2 Speaker Load M 4 V G5, S 0 S 1 S 2 M 1, M 4 M 5, M 6 M 2, M 3 S 0 S 1 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 20 of 32
PVT Variaion Effec (1/2) Significan shoo-hrough curren V G1 Shoo-hrough M 1 V G5 M 3 M 5 M 6 V G1, M 2 V DD Speaker Load M 4 V G5, S 0 S 1 S 2 M 1, M 4 M 5, M 6 M 2, M 3 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 21 of 32
PVT Variaion Effec (2/2) Addiional oupu volage ransiion V DD V G1 V G5 M 1 M 5 M 6 M 3 V G1, M 2 Speaker Load M 4 V G5, S 0 S 1 S 2 M 1, M 4 M 5, M 6 M 2, M 3 Addiional ransiion 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 22 of 32
CMFBD Realizaion Proposed low-emi conrol mehod V DD V G5 M 1 M 5 M 6 M 3 V G1 M 2 Speaker Load M 4 S A M 1, M 4, M 5 S B M 5 S C S D S E M 5, M 6 M 6 M 2, M 3, M 6 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 23 of 32
Proposed Low-EMI Conrol (1/4) In sae S A V DD V G1, V G5 M 1 M 5 M 6 M 3 V G1 V G5 M 2 Speaker Load M 4 S A M 1, M 4, M 5 S B M 5 S C S D S E M 5, M 6 M 6 M 2, M 3, M 6 S A 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 24 of 32
Proposed Low-EMI Conrol (2/4) In ransiion from S A ino S B V G5 V DD M 1 M 5 M 6 M 3 V G1, V G1 V G5 M 2 Speaker Load M 4 S A M 1, M 4, M 5 S B M 5 S C S D S E M 5, M 6 M 6 M 2, M 3, M 6 S A 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 25 of 32
Proposed Low-EMI Conrol (3/4) In sae S B V DD V G1, V G5 M 1 M 5 M 6 M 3 V G1 V G5 M 2 Speaker Load M 4 S A M 1, M 4, M 5 S B M 5 S C S D S E M 5, M 6 M 6 M 2, M 3, M 6 S A S B 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 26 of 32
Proposed Low-EMI Conrol (4/4) In sae S C V G5 V DD M 1 M 5 M 6 M 3 V G1, V G1 V G5 M 2 Speaker Load M 4 S A M 1, M 4, M 5 S B M 5 S C S D S E M 5, M 6 M 6 M 2, M 3, M 6 S A S B S C 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 27 of 32
Chip Micrograph 1.5 mm Digial Audio Processor DSM 0.2 mm 0.3 mm Gae Driver M 1 of L CH M 2,4 of L CH M 3 of L CH M 3 of R CH M 2,4 of R CH M 1 of R CH M5,6 of LCH M5,6 of RCH 2.45 mm 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 28 of 32
THD+N vs. Oupu Power Measuremen condiion: 24-V DD, 8-Ω, BD modulaion 30-W low-disorion oupu power 20% increase by ACDSM 1.5 THD+N (%) 1 0.5 0.1 DSM A DSM B ACDSM 1.5 1 0.5 0.2 0.1 increased by 5W 20 25 30 0.05 0.1 0.5 1 5 10 20 30 Oupu power (W) 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 29 of 32
EMI Measuremen Conduced EMI Level (dbμv/m) 60 40 20 0.15 0.5 1 5 10 20 30 Frequency (MHz) Radiaed EMI BD-modulaion mode 60 8 dbμv/m 20 low-emi mode 0.15 0.5 1 5 10 20 30 Frequency (MHz) Level (dbμv/m) 50 30 BD-modulaion mode 24 dbμv/m FCC class-b sandard 10 low-emi mode 30 64 98 132 166 200 360 520 680 840 1000 Frequency (MHz) 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 30 of 32
Comparison Supply Volage V DD (V) Nominal Load R L (Ω) Peak Efficiency η (%) Oupu Power P OUT (W) @ 1%THD+N Normalized Oupu Power (1) @ 1%THD+N DSM Max. Sable Inpu (dbfs) EMI Reducion (2) (dbμv/m) g q y ( ) Chip Area (4) (mm 2 ) (1) Process This Work 8 (conduced) 24 (radiaed) (3) 3.74 (sereo) 0.18μm BCDMOS P Normalized Oupu Power = 2 ( η V ) /(2 R ) DD 24 8 90 30 1.03 +0.2 OUT L JSSC 2012 [3] 18 8 88 13 0.83-1.2-23.9 (5.1-ch) 0.35μm HVCMOS JSSC 2010 [4] 3 8 88 0.4 0.92-0.7-0.76 (sereo) 65nm CMOS 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 31 of 32
Conclusion A 30-W open-loop class-d amplifier is implemened for a 24-V supply volage and 8-Ω load The ACDSM simulaneously achieves high in-band noise suppression wide sable inpu range 20% low-disorion P OUT increase The proposed low-emi conrol mehod PVT-insensiive common-mode EMI reducion 2014 IEEE IEEE Cusom Inegraed Circuis Conference 2014 32 of 32