1. 2015 IEEE Int. Conf. on Communications
Robust Transceiver to Combat Periodic Impulsive Noise In Narrowband Power-line Communications
Jing Lin1, Tarkesh Pande2, Il Han Kim2, Anuj Batra2, Brian L. Evans3
1 Qualcomm Inc.
2 Texas Instruments Inc.
3 The University of Texas at Austin
9 June /2015

2. Outline Introduction Noise in PLC and previous work Contribution
Modulation Diversity
Noise Estimation using sparse Bayesian techniques
Results

3. Periodically varying and spectrally shaped noise
Wideband impulses
Narrowband interferences
Sub-channel SNR is highly frequency-selective and time-varying

4. Previous vs. Proposed Transmitter Methods
Throughput Reduction
Channel/Noise Info at Transmitter
Previous
Adaptive modulation
[Nieman13] ✗
Full
Concatenated error correction coding (PLC standards) ✔
None
Proposed
Time-frequency modulation diversity
Partial

5. Modulation Diversity (I)
Sub-channels
SNR s1 s2 s3 s4 s5 s6 s7 s8 s9
s10
s11
s12
s13
s14
s15 X
Symbols b1 b2 b3 b4 b5 b6 b7 b8 b9
b10
b11
b12
b13
b14
b15 ✔ X
Bits
Data rate = 1 bit / channel use
[Schober03]

6. Example: Hochwald/Sweldens Code
Map N bits to a length-N codeword consisting of PSK symbols
Special case: PSK repetition code
Constellation mappings are optimized for channel statistics
000
110
001
010
011
100
101
111
Optimal length-3 code in Rayleigh fading channel
[Hochwald00]

7. Proposed Time-Frequency Mapping (Transmitter)
Allocate components of a codeword to time-frequency slots
Require partial noise information
Narrowband interference width
Burst duration
Subcarriers
Time-domain noise … … … …
OFDM symbols

8. Diversity Demodulation (Receiver)
Combine signals received from N sub-channels
Estimated sub-channel
Diversity Demodulator
Received signal
Log-likelihood ratio (LLR)
Estimated noise power

9. Noise Power Estimation (II)
Offline estimation
Utilize silent intervals between transmissions
Semi-online estimation
Between transmissions: Estimate start/end instances of all stationary intervals
In transmissions: Estimate noise power spectrums
Time
Offline
Semi-online
Transmission
Workload at the noise power estimator
Low
Med
High

10. Proposed Semi-Online Estimation
Measure noise using cyclic prefix
Formulate a compressed sensing problem
(where )
Collect multiple measurements in the same stationary interval
Cyclic Prefix
OFDM symbol
Noise
NBI
AWGN + -

11. Proposed Semi-Online Estimation (Cont.)
Apply sparse Bayesian learning algorithm
IG - Inverse Gamma dist.; IW - Inverse Wishart dist.
EM - Expectation maximization
Prior [Zhang11]
EM Updates
Row sparsity
Temporal correlation
Diversity Receiver
Slicing Error Estimation
Hyper-prior

12. System Parameters Parameters Reference System TFMD System
Sampling Frequency
400kHz
FFT Size
256
CP Length 30
Data Subcarriers
23:58 (36 tones)
Convolutional Code
Rate ½ K=7
Reed-Solomon Code
235/251
N/A
Interleaver Size (Bits)
4032 (packet) 36
Packet Size (Bytes)
235
Data Rate (kbps)
23.5 25
ΔT symbols (TFMD) 4
ΔK Tones (Nd =2/Nd=3)
18/12
Noise: Model LPTV with three regions [70% 29% 1%]
Channel: Flat

13. Results Coherent mode gains : > 6dB Non-Coherent mode gains: 8dB

14. Conclusions
Modulation diversity as an effective method for improving performance in PLC channels
Developed a methodology for estimating noise variance
Exploits the cyclic prefix
Uses Sparse Bayesian learning techniques