The Center for Signal & Image Processing Georgia Institute of Technology
Constrained Clipping for PAPR Reduction in VLC Systems with Dimming Control
Kai Ying1, Zhenhua Yu2, Robert J. Baxley3 and G. Tong Zhou1 1 Georgia Institute of Technology, Atlanta, GA, USA 2 Texas Instruments , Dallas , Texas , USA 3 Bastille Networks, Atlanta, GA, USA Outline 2 The Center for Signal & Image Processing
Introduction Double‐sided Clipping Constrained Clipping Numerical Results Intensity Modulation/Direct Detection (IM/DD) 3 The Center for Signal & Image Processing
Channel model VLC: visible light DC biasing IR: infrared light
Electric signal Electric signal LED Photodiode Light intensity Dynamic‐range‐limited and real‐valued
TittTransmitter: RiReceiver : light emitting diode (LED) converts A photodetector (PD) or an image sensor the amplitude of the electric signal generates an electric signal proportional to into the intensityypg of the optical signal. the intensity of the received optical signal.
Differences from Radio Frequency (RF)
Real‐valdlued
Dynamic‐range‐limited (turn‐on/saturation)
DC biasing required PAPR Reduction 4 The Center for Signal & Image Processing
Motivation of peak‐to‐average power ratio (PAPR) reduction
Consideringgg signals with the same variance Low amplifier Low power High PAPR/peaks Low signal-to-noise- plus-distortion ratio More Clipping Distortion
Signal 1 2
1
1 W orking range
-1 -1
-2
Variance: 1 Amplifier gain: 2 PAPR: 1 PAPR Reduction 5 The Center for Signal & Image Processing
Motivation of peak‐to‐average power ratio (PAPR) reduction
Consideringgg signals with the same variance Low amplifier Low power High PAPR/peaks Low signal-to-noise- plus-distortion ratio More Clipping Distortion
Signal 2 2 2 3 W orking range
-3 -2 -2 Variance: 1 Amplifier gain:2 Amplifier gain: 1.1547 PAPR: 3 Clipping distortion Lower signal power Double‐sided Clipping 6 The Center for Signal & Image Processing
Clipping is the simplest way to reduce PAPR Motivation of clipping
Tradeoff between clipping distortion and signal power
InputAmplifier Bias-Tee Output
Unclipped Preprocessed signal OFDM signal Amax
B
Amin Double‐sided Clipping 7 The Center for Signal & Image Processing
Clipping is the simplest way to reduce PAPR Motivation of clipping
Tradeoff between clipping distortion and signal power
InputClipping Amplifier Bias-Tee Output
Clipped Preprocessed signal
OFDM signal Amax
B
Amin Double‐sided Clipping 8 The Center for Signal & Image Processing
Difference from amplitude‐limited system
Double‐sided clippin g
Clipping ratio is impacted by both gain and bias
hx() hx()
Amax Amax
B1 B
Output signal Output signal Amin Amin
0 x 0 x
Input signal Input signal
GBAA=1， (max min ) / 2 GBB=1， 1 Double‐sided Clipping 9 The Center for Signal & Image Processing
Difference from amplitude‐limited system
Double‐sided clippin g
Clipping ratio is impacted by both gain and bias
hx() hx() A max Amax
B1 B
Output signal A min Amin
0 x 0 x
Input signal
GGBA1maxmin， ()/2 A GGBB11， Double‐sided PAPR 10 The Center for Signal & Image Processing Orthogonal frequency division multiplexing (OFDM)
Time domain: Real-valued
Frequency domain: Hermitian symmetry PAPR on both sides
Upper PAPR (UPAPR): Lower PAPR (LPAPR):
Asymmetric factor
Joint CCDF of UPAPR and LPAPR Simple Clipping 11 The Center for Signal & Image Processing Clipping in time domain
Distortion in frequency domain
Error vector magnitud(de (EVM )
Clipping ratio / Normalized clipping levels
Upper side:
Lower side:
To satisfy the EVM requirement, clipping levels cannot be very low. Too conservative! Constrained Clipping 12 The Center for Signal & Image Processing Block diagram xn xn X k X k xn
Distortion Mitigation Subcarriers Sort in Largest subset satisfying ascending order
Subcarriers with Subcarriers with small distortions large distortions
Keep unchanged Mitigate distortion
EVM requirement Constrained Clipping 13 The Center for Signal & Image Processing
How to revise subcarriers with large distortions
Modifyin g fre quenc y domain s ymbols will chan ge the time domain signals. We make the change as small as possible.
Parseval’s Theorem
Minimize the change in frequency domain Q
Ek X k
E k X k X Th
X k
I R. J. Baxley, et al., Constrained clipping for crest factor reduction in OFDM, IEEE Transactions on Broadcasting, vol. 52, no. 4, pp. 570‐575, Dec. 2006. Numerical Results 14 The Center for Signal & Image Processing
PAPR reduction with various clipping levels
0 10
-1 10 F DD
-2 oint CC 10 JJ =6 dB, =5 dB upper low er =7 dB, =6 dB upper low er =8 dB, =7 dB upper low er =9 dB, =8 dB upper low er -3 =10 dB, =9 dB 10 upper low er Orignal OFDM
5 6 7 8 9 101112131415 (dB) Numerical Results 15 The Center for Signal & Image Processing
Probability that signal exceeds 9 dB in UPAPR or 8 dB in LPAPR
0 10 =1 dB} =1
-1 10 =9 dB, dB, =9 CCDF{
-2 10 6 6.5 7 7.5 8 8.5 9 9.5 10 (dB) upper The Center for Signal & Image Processing Georgia Institute of Technology
Thank you!
Q&A