Karl F. Nieman † , Marcel Nassar ‡ , Jing Lin † , and Brian L. Evans †

Karl F. Nieman†, Marcel Nassar‡, Jing Lin†, and Brian L. Evans†

Pacific Grove, CANovember 6, 2013

FPGA Implementation of a Message-Passing OFDM

Receiver for Impulsive Noise Channels

IEEE Asilomar Conference on Signals, Systems, and Computers

†Wireless Communications and Networks Group, The University of Texas at Austin, Austin, TX‡Mobile Solutions Lab, Samsung Information Systems America, San Diego, CA

2

Smart Grid CommunicationsLocal utility

MV-LV transformer

Smart meters

Data concentrator

Home area data networksconnect appliances, EV charger and smart meter via powerline or wireless links

Smart meter communicationsbetween smart meters and data concentrator via powerline or wireless links

Communication backhaulcarries traffic between concentrator and utility on wired or wireless links

Low voltage (LV)< 1 kV

Medium Voltage (MV)1 kV – 33 kV

Background | Impulsive Noise Mitigation | Mapping to Hardware | Implementation

3

Impulsive Noise in 3-200 kHz PLC Band Outdoor medium-voltage line

(St. Louis, MO)

Cyclostationary noise becomes asynchronous after interleaving

Indoor low-voltage line (UT Campus)

= 1 MHz

Interleave

Background | Impulsive Noise Mitigation | Mapping to Hardware | Implementation

Impulsive noise can be 40 dB above background noise

4

Impulsive Noise in OFDM Systems

FFT spreads received impulsive noise across all FFT bins– SNR of each FFT bin is decreased– Receiver communication performance degrades

IFFT Filter + FFTEqualize

r and detectorVector

of symbolamplitudes(complex) Channel

Receiver

𝐬 𝐲

Gaussian () +

ImpulsiveNoise ()

Background | Impulsive Noise Mitigation | Mapping to Hardware | Implementation

Impulsive Noise Mitigation (Denoising)

• FFT bins (tones)– Transmitter null tones have zero power– Received null tones contain noise

• Impulsive noise estimation– Exploit sparse structure of null tones– is over complete dictionary– is sparse vector– is complex Gaussian ()

5

IFFT Filter + + FFTEqualize

r and detectorImpulsive

noise estimatio

n

Gaussian () +

ImpulsiveNoise ()

Vectorof symbolamplitudes(complex)

+-

Channel

Receiver

Ω is set of null tones (i.e. ) is DFT matrix

𝐬 𝐲

||

¿

+¿

Conventional OFDM systemAdded in our system

Background | Impulsive Noise Mitigation | Mapping to Hardware | Implementation

6

Impulsive Noise Mitigation Techniques

MethodLow SNR

High SNR

Non-Parametric

?

Computational

ComplexityNulling/Clipping[Tse12]

Low

Thresholded Least Squares/MMSE[Cai08]

Med

Sparse Bayesian Learning[Lin13]

High(matrix

inversion)l1-norm minimization[Cai08]

High

Approximate Message Passing (AMP) [Nas13]

Med

com

pres

sive

sens

ing

• Compressive sensing approaches are used for low SNR• AMP provides best performance vs. complexity

tradeoff

Background | Impulsive Noise Mitigation | Mapping to Hardware | Implementation

7

• M = null tones• N = FFT size• Iterate

– Time-frequencyprojections

• Mostly scalar arithmetic and data• Parallelizable for hardware

implementation– FFT/IFFT, exponential, vector multiplies, divisions

Approximate Message Passing (AMP)1.

Initialization 2. Output Linear 3. Output Non-Linear

5. Input Non-Linear 4. Input Linear

𝜏 𝑖𝑝 (𝑡 )=∑

𝑗𝜏 𝑗𝑥 (𝑡 )

𝜏 𝑖𝑠 (𝑡 )= 1

𝛾𝐵+𝜏𝑝

𝑖 (𝑡 )=𝜏 𝑖𝑠 (𝑡 ) ( 𝑦 𝑖−𝑖 (𝑡 ) )

𝜏 𝑗𝑟 (𝑡 )= 𝑁

𝑀𝜏𝑠 (𝑡 )

𝑗 (0 )=0

𝜌 𝑗=𝜂 𝑗

1+𝜂 𝑗 𝑗 (𝑡+1 )=

𝛾 𝐼

𝛾 𝐼+𝜏 𝑗𝑟 (𝑡 )

𝜌 𝑗𝑟 𝑗 (𝑡 )

Background | Impulsive Noise Mitigation | Mapping to Hardware | Implementation

8

Synchronous Dataflow (SDF) Model• Targeted architecture for real-time streaming

performance:– Xilinx Virtex V field programmable gate arrays (FPGAs)– Embedded x86 computers running real-time OS (Phar Lap ETS)

• SDF model of OFDM receiver with AMP noise mitigation:

• Periodic schedule is

Background | Impulsive Noise Mitigation | Mapping to Hardware | Implementation

(6950𝐎 ) (278𝐀 ) (278𝐁 )𝐂𝐃𝐄𝐅𝐇𝐈

Task

Processing

O Input samples from ADC

A Resampling FIR filters

B Time and Freq. Offset Correction

C FFT + Index Active and Null Subcarriers

D AMP Noise Estimation

E FFT + Index Active Subcarriers

FSubtract Noise Estimate, De-Interleave Reference Symbols

H Zero-Forcing Equalization

I Equalize and Detect

Mapping AMP to Fixed-Point• Variables sized using MATLAB Fixed-Point Toolbox• Most variables sized within 16-bit wordlengths

9

sizing for using graphical tool

Background | Impulsive Noise Mitigation | Mapping to Hardware | Implementation

10

Graphical High-Level FPGA SynthesisNational Instruments Communication System Design Tools

– LabVIEW DSP Design Module– LabVIEW FPGA– LabVIEW Real-Time

Background | Impulsive Noise Mitigation | Mapping to Hardware | Implementation

2. Output Linear

𝜏 𝑖𝑝 (𝑡 )=∑

𝑗𝜏 𝑗𝑥 (𝑡 )

𝑖 (𝑡 )=𝐼ΩFFT ( (𝑡 ) )−𝜏 𝑖𝑝 (𝑡 ) 𝑖 (𝑡−1 )

Step 2 of AMP

DSP diagram replaces

thousands of lines of VHDL

code

AMP-Enhanced OFDM Testbed

11

RT controller

LabVIEW RT

data symbol generation

FlexRIO FPGA Module 1 (G3TX)

LabVIEW DSP Design Module

data and reference symbol

interleave Ref. symbol LUT

zero padding

(null tones)

generatecomplex

conjugate pair

256 IFFT w/ 22 CP insertion

NI 5781

16-bit DAC

RT controller

LabVIEW RT

BER/SNR calculation w/ and w/o AMP

FlexRIO FPGA Module 2 (G3RX)

LabVIEW DSP Design Module

NI 5781

14-bit ADCsample

rate conversion

time and frequency

offset correction

256 FFT w/ 22 CP removal,

noise injection

FlexRIO FPGA Module 3 (AMPEQ)

LabVIEW DSP Design Module

null tone and active

tone separation

channel estimation/

ZFequalization

AMP noise estimate

Subtract noise

estimate from active

tones

data and reference

symbol de- interleave

Host Computer

LabVIEW

sample rate

conversion

256 FFT, tone select

testbench control/data visualization

diffe

rent

ial M

CX p

air

TX Chassis RX Chassis1 × PXIe-10821 × PXIe-81331 × PXIe-7965R1 × NI-5781 FAM

differential MCX pair(quadrature component = 0)

1 × PXIe-10821 × PXIe-81332 × PXIe-7965R1 × NI-5781 FAM

Background | Impulsive Noise Mitigation | Mapping to Hardware | Implementation

2-mode Gaussian Mixture noise injected here: ~

Results• System implemented using G3-PLC signaling structure

MHz, (real-valued), active tones• Receiver w/ AMP was mapped across two FPGAs

– ‘G3RX’ – Downsampling, IFFT, time/frequency offset correction– ‘AMPEQ’ – AMP algorithm, equalization, and detection

12

Utilization

Trans.

Rec. AMP+Eq

FPGA 1 2 3total slices

32.6% 64.0%

94.2%

slice reg. 15.8% 39.3%

59.0%

slice LUTs 17.6% 42.4%

71.4%

DSP48s 2.0% 7.3% 27.3%blockRAMs

7.8% 18.4%

29.1%

Received QPSK constellation at equalizer output

conventional receiver

with AMP

Resource Utilization

Background | Impulsive Noise Mitigation | Mapping to Hardware | Implementation

Bit-Error-Rate Measurements

13Background | Impulsive Noise Mitigation | Mapping to Hardware | Implementation

unco

ded

bit-e

rror-r

ate

(BER

)

signal-to-noise ratio (SNR)

8 dB for 30 dBimpulsive

noise4 dB for 20 dB

impulsive noise

No loss (or gain) in non-

impulsive (AWGN) noise

14

Conclusions• Approximate Message Passing Framework allows

– Impulsive noise mitigation at low and high SNR– Conversion of matrix operations to scalar and vector

operations– Parallelization and efficient mapping to hardware

• Up to 8 dB impulsive noise mitigation achieved using– Fixed-point data and arithmetic– Streaming G3-PLC rates

• LabVIEW project and FPGA bitfiles available here:– http://users.ece.utexas.edu/~bevans/papers/2013/fpgaReceiver/index.

html

Background | Impulsive Noise Mitigation | Mapping to Hardware | Implementation

References[Cai08] – G. Caire; T. Y. Al-Naffouri; A. K. Narayanan, "Impulse noise cancellation in

OFDM: an application of compressed sensing," Information Theory, 2008. ISIT 2008. IEEE International Symposium on , 2008.

[Tse12] – D-F. Tseng; Y. S. Han; W. H. Mow; L-C. Chang; A.J.H. Vinck, "Robust Clipping for OFDM Transmissions over Memoryless Impulsive Noise Channels," Communications Letters, IEEE , vol.16, no.7, 2012.

[Lin13] – J. Lin; M. Nassar; B. L. Evans, "Impulsive Noise Mitigation in Powerline Communications Using Sparse Bayesian Learning," Selected Areas in Communications, IEEE Journal on , vol.31, no.7, 2013.

[Nas13] – M. Nassar; P. Schniter; B. L. Evans, "A factor graph approach to joint OFDM channel estimation and decoding in impulsive noise environments," IEEE Trans. on Signal Processing, accepted for publication, 2013.

[Max11] – Maxim and ERDF, "Open Standard for Smart Grid Implementation," 2011.

15

16

Questions?

17

Backup Slides

18

Powerline Communications (PLC)

• Uses orthogonal frequency-division multiplexing (OFDM)

• Communication challenges– Channel distortions– Non-Gaussian impulsive noise

Categories Band Bit Rates Coverage Enables Standards

Narrowband 3-500 kHz

up to 800 kbps

Multi-kilometer

Smart meter communication

• (ITU) PRIME, G3• ITU-T G.hnem• IEEE P1901.2

Broadband 1.8-250 MHz

up to200 Mbps <1500 m Home area

data networks•HomePlug•ITU-T G.hn•IEEE P1901

Background | Impulsive Noise Mitigation | Mapping to Hardware | Implementation

19

AMPEQ.lvdsp(first half)

Background | System Design and Implementation | Demo | Conclusion

(second half)

Approximate Message Passing (AMP)

20

= number of null tones

= FFT size

• Reconstruct time-domainnoise from frequency-domain null tones

• Iterate until convergence

• Algorithm consists of:• Mostly scalar arithmetic• FFT/IFFTs• Exponential

• Targeted at G3-PLC signaling structureBackground | Impulsive Noise Mitigation | Mapping to Hardware |

Implementation