Adaptive Demodulation Techniques for Next

Generation Software Defined Radios

U.S. Army RDECOM Communication-Electronics RD&E Center

Fort Monmouth, NJ 07703, USA

Contents

IntroductionModulation classification overviewResearch on commercial applicationsChallenges

Modulation Classifier

From: http://www.ottawa.drdc-rddc.gc.ca

………………………

………………

signal

Center Frequency Estimation

Demodulated signal

Modulation Recognition

What is Automatic Modulation Classification ?

BW Estimation SNR

Estimation

4

0

5 67 8 9

1 2 3

Phase Estimation

Filter

Statistical Estimation

IF Demodulation Filter

LO

A/DDemodulation Demodulation

Automatic Classification

Channel Equalizer

Channel EstimationSymbol Rate

Estimation

A non-cooperative communication technique which uses statistical methods to estimate the signal modulation types

Analog Digital

PSK/QAM Preprocessing

FSK/MSK Preprocessing

Estimation Confidence

Rating Failure

Modulation Scheme Classification Confidence Modulation Parameters

Analog Preprocessing

Classification Decision

Unknown Type

Coarse Modulation Estimation

FSK/MSK Modulation Estimation

PSK/QAM Modulation Estimation

Analog Modulation Estimation

Preprocessed IF

PSK/QAM Feature

Extraction

FSK/MSK Feature

Extraction

Analog Feature

Extraction

SNR Estimation

Templates Building

Modulation ClassificationA non-cooperative communication technique which uses statistical methods to estimate the modulation type of a unknown signal

SDR Applications (1)

4 05 678 91 2 34

05 678 91 2 3

4 05 678 91 2 3

4

0

5 67 8 9

1 2 3

4

0

5 67 8 9

1 2 3

Overcome channel fading Monitor communication spectrumRemove co-channel interferences

SDR Applications (2)Deep Space Communication

4 05 678 91 2 3

4

0

5 67 8 9

1 2 3

4

0

5 67 8 9

1 2 3

Reduce the scheduling and configuration burdens of communications

Modulation Classification Overview

Feature Extraction: Amplitude, Differential Phase, and Frequency

I

/

( )2

( )2

+

tan-1

Amplitude

IF

sqrt

timing circuit

Q

tΔΔ2φ

TΔΔφ

Freq.

Delta phase

BPFbaud rate detector

hist

ogra

m

CW PSK2 PSK4 PSK8 ASK2 FSK2

templates

reco

gniti

on tr

ee

STD

•Input: IF•Feather: Amplitude, phase, diff phase, frequency•Statistics: histogram, STD•Classifier: max correlation, decision tree•Reference: Liedtke 1984

corre

latio

n

Higher-order Transform of Constellations

The 4th order constellations

V29-8 and V29-16 constellations

QPSK

PSK8 PSK16

QAM16 QAM64

QAM32 QAM4-12 QAM16-16

QAM44-20c2

0 c2

1 c4

0 c4

1 c4

2 c6

0 c6

1 c6

2 c6

3 c8

0 c8

1 c8

2 c8

3 c8

4

Power-law:

Moment:

Cumulant:

Higher-order Statistical Features

∏ ∑= = ⎭

⎬⎫

⎩⎨⎧

=K

k

M

jk

ij

iKi

i

rpM

rHG1 1

)( )(1)|(

∑=

=K

k

krm1

440 )(

Mkr

4th order transformation of QPSK

4th order transformation of QAM16

)(4 kr

4th order dominant

points

1st order2nd

orde

r

2ndor

der

Q

I

2204040 3mmC −=

Cumulants vs. SNRs

Cyclic Spectral Analysis

∫−−

∞→+=

2/

2/

2),(1lim)( **

T

T

atjxxT

axx

dtettRT

R πττ

Theoretical spectrum correlation magnitude Gardner and Spooner 1992

•Input: IF•Features: cycle frequencies•Reference: Menguc, 2004

∫∞

∞−

−= ττ τπ deRfS fjaxx

axx

2)()( **

{ })()(),( ** ττ +=+ txtxEttR

xx

Decision Templates

Baseband

Cycle Freq

Cyclic autocorrelation

Spectrum correlation density

Time varying autocorrelation

Feature Classification: Maximum Likelihood (ALRT)

))(|( krHl QPSK

))(|( krHl BPSK

))(|( 8 krHl PSK

MAXBaseband Modulation Scheme

unknown

8PSK

QPSK

BPSK

PDFBPSK

PDFQPSK

PDF8PSK

∏=

K

k 1

(.)

∏=

K

k 1

(.)

∏=

K

k 1

(.)

• Input: baseband• Feature: Complex envelop• Classifier: maximum likelihood• References: Polydoros and

Kim 1995

Templates

Feature Classification: Histogram Correlation

∏ ∑= = ⎭

⎬⎫

⎩⎨⎧

=K

k

M

jk

ij

iKi

i

rpM

rHG1 1

)( )(1)|(

⎪⎭

⎪⎬⎫

⎪⎩

⎪⎨⎧ −−= 2

2)(

2)(

2

)(exp

21)(

σπσ

jbrrp

ik

ki

jALRT

HIST

∑=

iM

jk

ij

i

rpM 1

)( )(1

Quantize

∑=

iM

jk

ij

i

rpM 1

)( )(1

)()(k

ij rp

)( iqD

)( iqp

)( iqp

∑∑∑∑== Ω∈=

←==Q

k

iq

iq

Q

q kk

iK

kk

iKi DprprprHL

q 1

)()(

1

)(

1

)( log)(log)(log)|(

• Input: baseband/IF• Feature: frequency/diff phase• Classifier: max correlation• References: Liedtke 1984

Research on Commercial Applications

Maintain a constent BER by varying modulation schemesModulation schemes: QPSK, 16QAM, and 64QAMData frame based modulation recognitionA pilot symbol is used in forward channelReference: Jain, P.; Buehrer, R.M, “Implementation of adaptive modulation on the Sunrise software radio,” The proceedings of the 45th Midwest Symposium on Circuits and Systems, Volume: 3 , 4-7 Aug 2002. Pages:III-405 - III-408

Research on Adaptive Modulation Based on SDR - Cooperative

Slow Flat Fading

ChannelsTransmitter

Feed Back Channel

ReceiverDATA DATA

Pilot Pilot

Pilot Data

Environment limitation and restrictionElimination of the signal overhead informationAttractive for packet data services

Why Applying Non-cooperative Demodulation

Data output

Modulation Recognition

Preprocessing

Choose Demodulator

Modulation Recognition

Air-interface PreprocessingRF

IF

Delayer Demod

Deference Between Military and Commercial Applications

. .SIGINT SDR .

Real time classification demodulationSNR low high Candidates unlimited limited QoS friend / foe packet lossPulse shape unknown knownBandwidth unknown knownBaud rate unknown knownBlindness more less

Assume:Equally likelySymmetrical QAM/PSK

Purpose:Reduce the processing time

Issues:Low utilization of available informationMay need longer data length for randomness

Reduced form constellation

Q

I

Q

I

Constellation for QPSK

Research of Nolan et al. Reduce Form Constellation

Automatically recognize BPSK, QPSK, 8PSK, pi/4QPSK, 16QAM, FSK, MSK, GMSK, AM, FM, CW, and SSB using decision tree for spectrum, variance, and baud detection analysis.

IF

Data output

Transmitter

Adaptive Receiver – Ishii et al.

OSC

RF Modulation Estimation

Demodulation

•Spectrum

•Envelope

•Baud

•Amplitude

•Phase

BPF

Decision tree

Thresholds

Automatically recognize BPSK, QPSK, 8PSK, and 16QAM using amplitude and differential phase variances. Channel gain estimation is discussed (2000).Automatically recognize BPSK, QPSK, and 8PSK using differential phase and maximum likelihood test.

IF Data outDemodulation

OSC

BPFRF

Maximum Likelihood Modulation Estimation

Noise Variance Estimation

Blind Modulation Estimation Umebayshi et al.

Transmitter

Research of Menguc and Jondral (1)Air Interface Identification for SDR

Identify (Verify?)TDMA-GMSKOFDM-PSK/QAMCDMA-QPSK

Base band

Data output

Transmitter

RF

Likelihood Test

Calculate Cyclic

Autocorrelation

•Determine Number of Interfaces

•Estimate Carrier and BW

Preprocessing

Demodulation

Recognize Air Interfaces

Threshold

Research of Menguc and Jondral (2) Magnitude Plot of the Cyclic Autocorrelation

Estimations

GMSK OFDM CDMA

* O. Menguc, “Air interface identification for software radio systems,” Ph.D. Dissertation, University of Fridericiana Karlsruhe, Nov. 30, 2004.

IssuesProcessing speedNeed universal front end

Research of Simon and Divsalar (1)Data Format classification for SDR

∑∑−

=

−

=⎟⎟⎠

⎞⎜⎜⎝

⎛<⎟⎟

⎠

⎞⎜⎜⎝

⎛ 1

0,

0

1

0,

0

22coshln22coshlnbb K

nManchesterk

K

nNRZk r

NPr

NP

large;2ln||small;2/

{)cosh(ln2

xxxx

x−

≅

•Discriminate NRZ and Manchester code

•Reduce complexity

•Extend to non-coherent case

Ln cosh(x) vs x

Use two curves approximate ln cosh(x) in order to simplify the ML computation

small

Research of Simon and Divsalar (2)Reduced Complexity ML Implementation

SNR Estimation

BasebandData

Other Research Results on Modulation Classification Based on SDR

Gu et al. “Channelized receiver platform of SDR based on FPGAs, Proceedings of The 5th IEEE International conference on ASIC, Vol.2, Oct. 2003, pp.840-843.Yang, “An enhanced SOFM method for automatic recognition and identification of digital modulations,” Proceedings of the 2nd

IEEE International Workshop on Electronic Design, Test and Applications (DELTA’04), Jan. 2004, pp.174-179.Ko et al. “Modulation type classification Method using wavelet transform for adaptive demodulator,” Proceedings of 2004 International Symposium on Intelligent Signal processing and Communication System, Vol.46, Oct. 1995, pp.211-222.Hooftand Darwish, “A reconfigurable software digital radio architecture for electronic signal interception, identification,communication and jamming,” COTS Joural, April 2002, pp.31-35.

Normalization

RF signal

Pulse Timing and Matched Filtering

Local Oscillator

Band Pass Filter

Quantization and Address Mapping

Noise Power Estimation

Look Up Table

Decision Process

Confidence Check

Estimated Modulation Scheme

Carrier and Carrier Phase Tracking

Real-time Data

Output

Choose Demodulator

Delayer Demodulation Process

Adaptive modulation is not only an important information warfare practice but also an effective tool to maximize the data capacity and minimize the transmission error in SDR applications.Automated modulation classification is a solution in handling the non-cooperative communication problem for SDR.Blind estimation of modulation parameters such as center frequency offset, carrier phase, pulse shape, symbol rate, and bandwidth is critical to the robustness of modulation classification.A good modulation classifier should be able to identify modulation scheme fast and robust.

Summary

Future Work

Faster estimatorShorter data lengthLower SNRBetter channel estimationBetter QoS