An Open-Loop Class-D Audio Amplifier with Increased Low-Distortion Output Power and PVT-Insensitive EMI Reduction

Paper 8-6

An Open-Loop Class-D Audio Amplifier with Increased

Low-Distortion Output Power and PVT-Insensitive EMI Reduction

Shih-Hsiung Chien

, Li-Te Wu

, Ssu-Ying Chen

, Ren-Dau Jan

, Min-Yung Shih

, Ching-Tzung Lin

and

Tai-Haur Kuo

1

National Cheng Kung University, Tainan, Taiwan,

2

NeoEnergy Microelectronics, Inc., Hsinchu, Taiwan

Outline



Background and Motivation



System Overview



Proposed Techniques



Adaptive-Coefficient Delta-Sigma Modulator



PVT-Insensitive Low-EMI Control Method



Measurement Results



Conclusions

Digital-Input Audio Amplifier



Class-AB amplifier with DAC



Class-D amplifier with DAC



Class-D amplifier with digital PWM (DPWM) gen.

Class-AB Amp.

Digital Input

Speaker DAC Load

Class-D Amp.

Low-Pass Filter Digital

Input

Speaker DAC Load

DPWM Class-D

Amp.

Low-Pass Filter Digital

Input

Speaker Load

Class-D Amplifier with DPWM



Pros



High efficiency



No need of high-resolution DAC



Cons



Distortion from class-D amp.  Degraded THD+N



Need of L-C low-pass filter for EMI suppression

DPWM Class-D

Amp.

Low-Pass Filter Digital

Input

Speaker Load

Closed-Loop vs. Open-Loop



Closed-loop architecture



Open-loop architecture  adopted in this work

 Lower complexity

 Easier design porting to advanced processes

 Smaller chip area

DSM

PCM-to-PWM Converter

(gain=1)

Analog Loop Filter Interpo-

lator (gain=1) Digital

Input

Power Stage Feedback Path (feedback factor = β ⁾

clock

DPWM

Low-Pass Filter

Speaker Load

DSM

PCM-to-PWM Converter

(gain=1)

Power Stage Interpo-

lator (gain=1) Digital

Input clock

Low-Pass Filter

Speaker Load

DPWM

THD+N vs. Output Power



Distortion and noise sources



Constant noise



Power stage distortion



Clip distortion



Low-distortion P

= max. P

with THD+N<1%

 Dominated by clip distortion due to DSM instability

T HD + N (%)

Output Power, P

(W) 1%

Low-Distortion P

DSM Instability in Open-Loop



When DSM input is large

 DSM’s quantizer overload

 clipping at DSM

 clip distortion at amp. output

 decreased low-distortion output power

Clipping Error DSM

PCM-to-PWM Converter (gain 1/k=1)

Power Stage Interpo-

lator (gain k=1) Digital

Input

DSMOUT clock PWMOUT

Power Output

Input mag.

(dBFS)

SNDR(dB)

0

SNDR @ DSMOUT

SNDR(dB)

Input mag. 0

(dBFS)

SNDR @ PWMOUT

1%

Output power (W)

THD+N(%)

THD+N vs. output power

DSM Instability in Open-Loop



To reduce DSM input: interpolator’s gain



To increase gain after DSM: PCM-to-PWM’s gain

 Clipping at PWM

 DSM instability can NOT be prevented by scaling k

Clipping Error DSM

PCM-to-PWM Converter (gain 1/k >1)

Power Stage Interpo-

lator (gain k<1) Digital

Input

DSMOUT clock PWM_OUT

Power Output

Input mag.

(dBFS)

SNDR(dB)

0

SNDR @ DSMOUT

SNDR(dB)

Input mag. 0

(dBFS)

SNDR @ PWM_OUT

1%

Output power (W)

THD+N(%)

THD+N vs. output power

Common-Mode EMI Reduction



Conventional BD modulation



Common-Mode Free BD (CMFBD) modulation [1]

[1] P. Siniscalchi and R. Hester, “A 20W/channel class-D amplifier with significantly reduced common-mode radiated emissions,” IEEE ISSCC 2009.

VDD/2 V_DD

0 VDD

0 V_DD

0

VCM OUT

OUT

VDD/2 V_DD

0 V_DD/2 VDD

0 VDD/2

VCM

OUT

OUT

Targets of This Work



Increase low-distortion output power for open-loop class-D amplifiers without



increasing supply voltage



increasing off-chip components



sacrificing THD+N at small output power



Reduce common-mode EMI without



using expensive L-C filters



PVT-sensitive issue

System Overview



Block diagram of this work



Two selectable modes



BD-Modulation Mode

Delta-Sigma Modulator

(DSM)

PCM-to- PWM Converter

Power Stage Interpo-

lator Digital

Input

OUTP Low- Pass Filter OUTN

Dual-Mode Output Stage

SEL Proposed ACDSM

x[n] y[n]

Adaptive Coefficient

Set[n]

OUT_P OUT_N

bead C bead OUTP L

OUT_N

L C



Low-EMI Mode [1]

DSMA

DSM_B 20

0

-40

-80 -20

G ai n (dB )

0.1 1 10 100 192

Frequency (kHz)

Trade-Off in DSM Design



Two DSM Designs



DSM

: high in-band noise suppression



DSM

: full-scale stable input range



NTF plot

Root-locus plot

-1 -0.5 0 0.5 1

Real axis

Im agi nar y ax is

unit circle

DSM_A DSMB

1 0.5

-0.5 0

-1

Proposed ACDSM



Adaptive-Coefficient Delta-Sigma Modulator (ACDSM)



Small x[n]  coef. with high in-band noise suppression



Large x[n]  coef. with full-scale stable input range

g1[n]·H

Quantizer

g2[n]·H g3[n]·H g4[n]·H g5[n]·H

y[n]

x[n]

a1[n] a2[n] a3[n] a4[n] a5[n]

b1[n]

b2[n]

b₃[n]

b4[n]

Direct Coefficient Switching



Coefficient is switched between



Small x[n]  Set

(high in-band noise suppression)



Large x[n]  Set

(full-scale stable input range)

 large internal transient spike  DSM unstable

Delta-Sigma Modulator (DSM)

Set

y[n]

x[n]

Set

Set

= [g

,...g

,a

,...a

,b

,...b

]

Set

= [g

,...g

,a

,...a

,b

,...b

]

ACDSM Algorithm



Linear-interpolated coefficient changing

 operating coefficient set is changed with small Set

 internal transient spike is reduced

N Sets Delta-Sigma Modulator (DSM)

Set

Set_Δ

y[n]

x[n]

Set

Set

Set

Proposed ACDSM Algorithm

Dynamic Range (DR) Plots



The ACDSM simultaneously achieves



a wide stable input range



high in-band noise suppression

100

80

60

-1 -0.5 0

DSMA

DSM_B ACDSM

120

-40 -30 -20 -10 0

Input Magnitude (dBFS)

SNDR (dB)

100

80

60 50

CMFBD Realization



Previous low-EMI control method [1]

[1] P. Siniscalchi and R. Hester, “A 20W/channel class-D amplifier with significantly reduced common-mode radiated emissions,” IEEE ISSCC 2009.

turn-on:

M2, M3

turn-on:

M5, M6

turn-on:

M1, M4

S1 S2

S0

M1

M₅ M₆

M2

M₃

M4

OUT_P OUT_N

V_DD

V_G1

V_G4

VG5 VG6

Speaker Load

Previous Low-EMI Control (1/3)



In state S

turn-on:

M₂, M₃ turn-on:

M₅, M₆ turn-on:

M₁, M₄

S₁ S₂

S₀

V

, V

t

t

t

OUT

OUT

S₀ M₁

M5 M6

M2

M3

M4

OUT_P OUT_N

V_DD

V_G1

VG4

V_G5 V_G6

Speaker Load

V

,

V

Previous Low-EMI Control (2/3)



In transition from S

into S

turn-on:

M2, M3

turn-on:

M5, M6

turn-on:

M1, M4

S₁ S₂

S0

t

t

t S₀

Speaker Load M1

M₅ M₆

M₂

M₃

M₄

OUTP OUTN

V_DD

VG1

V_G4

VG5 VG6

V

, V

OUT

OUT

V

,

V

Previous Low-EMI Control (3/3)



In state S

turn-on:

M₂, M₃ turn-on:

M₅, M₆ turn-on:

M₁, M₄

S₁ S₂

S₀

t

t

t S₀ S₁

Speaker Load M₁

M5 M6

M2

M3

M4

OUT_P OUT_N

V_DD

V_G1

VG4

V_G5 V_G6

V

, V

OUT

OUT

V

,

V

PVT Variation Effect (1/2)



Significant shoot-through current

t

t

t Speaker Load

M1

M₅ M₆

M₂

M₃

M₄

OUTP OUTN

V_DD

VG1

V_G4

VG5 VG6

Shoot-through

turn-on:

M2, M3

turn-on:

M5, M6

turn-on:

M1, M4

S₁ S₂

S₀

V

, V

OUT

OUT

V

,

V

PVT Variation Effect (2/2)



Additional output voltage transition

t

t

t Speaker Load

M₁

M5 M6

M2

M3

M4

OUT_P OUT_N

V_DD

V_G1

VG4

V_G5 V_G6

Additional transition

turn-on:

M₂, M₃ turn-on:

M₅, M₆ turn-on:

M₁, M₄

S1 S2

S₀

V

, V

OUT

OUT

V

,

V

CMFBD Realization



Proposed low-EMI control method

M₁

M5 M6

M₂

M₃

M₄

OUTP OUTN

V_DD

V_G1

V_G4

V_G5 V_G6

Speaker Load

turn-on:

M₂, M₃, M₆ turn-on:

M₆ turn-on:

M₅, M₆ turn-on:

M₅ turn-on:

M₁, M₄, M₅

S_C S_D S_E

S_A S_B

Proposed Low-EMI Control (1/4)



In state S

M1

M₅ M₆

M₂

M₃

M₄

OUTP OUTN

V_DD

VG1

V_G4

VG5 VG6

Speaker Load

turn-on:

M₂, M₃, M₆ turn-on:

M₆ turn-on:

M₅, M₆ turn-on:

M₅ turn-on:

M₁, M₄, M₅

S_C S_D SE

SA S_B

t

t

t S_A

t

V

OUT

OUT

V

V

,

V

Proposed Low-EMI Control (2/4)



In transition from S

into S

t

t

t S_A

t M1

M₅ M₆

M₂

M₃

M₄

OUTP OUTN

V_DD

VG1

V_G4

VG5 VG6

Speaker Load

turn-on:

M₂, M₃, M₆ turn-on:

M₆ turn-on:

M₅, M₆ turn-on:

M₅ turn-on:

M₁, M₄, M₅

S_C S_D SE

SA S_B

V

OUT

OUT

V

V

,

V

Proposed Low-EMI Control (3/4)



In state S

t

t

t S_A

t M1

M₅ M₆

M₂

M₃

M₄

OUTP OUTN

V_DD

VG1

V_G4

VG5 VG6

Speaker Load

turn-on:

M₂, M₃, M₆ turn-on:

M₆ turn-on:

M₅, M₆ turn-on:

M₅ turn-on:

M₁, M₄, M₅

S_C S_D SE

SA S_B

S_B

V

OUT

OUT

V

V

,

V

Proposed Low-EMI Control (4/4)



In state S

t

t

t S_A

t

S_C M1

M₅ M₆

M₂

M₃

M₄

OUT_P OUT_N

VDD

V_G1

V_G4

VG5 VG6

Speaker Load

turn-on:

M₂, M₃, M₆ turn-on:

M₆ turn-on:

M₅, M₆ turn-on:

M₅ turn-on:

M₁, M₄, M₅

S_C S_D S_E

S_A SB

S_B

V

OUT

OUT

V

V

,

V

Chip Micrograph

2 .45 mm

1.5 mm

DS M M

of L

M

of R

M

of L

M

of L

M

of

L

M

of R

M

of R

M

of

R

G at e D ri v e r

Digital Audio Processor

0.2 mm

0.3 mm

THD+N vs. Output Power



Measurement condition: 24-V

, 8- Ω, BD modulation



30-W low-distortion output power

 20% increase by ACDSM

0.1 0.5 1 5 10 20 30 Output power (W)

1.5

THD+N (%)

1 0.5

0.1 0.05

DSM

DSM

ACDSM

1.5 1

0.5 0.2 0.1

20 25 30

increased by 5W

EMI Measurement



Conducted EMI



Radiated EMI

0.15 0.5 1 5 10 20 30 Frequency (MHz)

60

40

20

Level (dBμV/m)

low-EMI mode BD-modulation mode

Frequency (MHz) 60

20

0.15 0.5 1 5 10 20 30

8 dBμV/m

Level (dBμV/m)

30 64 98 132 166 200 360 520 680 840 1000 Frequency (MHz)

50

30

10

BD-modulation mode

low-EMI mode

FCC class-B standard

24 dBμV/m

31 of 32

© 2014 IEEE IEEE Custom Integrated Circuits Conference 2014

Comparison

( )

Power Output

Normalized ₂

L DD

OUT

R η V

P

⋅

= ⋅

(1)

g q y ( ) Chip Area

(mm

)

Process

3.74 (stereo) 0.18 μm BCDMOS

23.9 (5.1-ch) 0.35 μm HVCMOS

0.76 (stereo) 65nm CMOS

Supply Voltage V

(V)

Nominal Load R

( Ω) Peak Efficiency η (%) Output Power P

(W)

@ 1%THD+N

Normalized Output Power

@ 1%THD+N

DSM Max. Stable Input (dBFS)

EMI Reduction

(dB μV/m)

This Work 24

8 90 30 1.03 +0.2 8 (conducted)

24 (radiated)

(3)

JSSC 2012 [3]

18 8 88 13 0.83

-1.2 -

JSSC 2010 [4]

3 8 88 0.4 0.92

-0.7 -

Conclusion



A 30-W open-loop class-D amplifier is implemented for a 24-V supply voltage and 8- Ω load



The ACDSM simultaneously achieves



high in-band noise suppression



wide stable input range

 20% low-distortion P

increase



The proposed low-EMI control method

 PVT-insensitive common-mode EMI reduction