PRIN 2005 WOMEN Project Research unit: Università degli Studi di Napoli Federico II 5 th Meeting of WOMEN Project Trento, February 12, 2008 Fabio Sterle

PRIN 2005 PRIN 2005 WOMEN ProjectWOMEN ProjectResearch unit: Università degli Studi di Research unit: Università degli Studi di Napoli Federico IINapoli Federico II

55thth Meeting of Meeting ofWOMEN ProjectWOMEN ProjectTrento, February 12, 2008Trento, February 12, 2008

Fabio SterleFabio Sterle

5th Meeting of WOMEN Project – Research unit of Napoli Febr. 12, 2008

Outline IQ imbalance in RF systems

Blind estimation of the receiver IQ imbalance parameters IQ imbalance in DVB-T systems

MIMO systems Multistage DF MMSE equalizer synthesis

Dynamic addressing-based routing scalable protocols for MANET Tree-based routing protocol

1. D.Mattera, F.Sterle, “ML estimation of reciever IQ imbalance parameters,” Third International Waveform Diversity & Design Conference, Pisa (Italy), 4-8 June 2007.

2. M. Lipardi, D.Mattera, F.Sterle, “MMSE Equalization in Presence of Transmitter and Receiver IQ Imbaclance,” Third International Waveform Diversity & Design Conference, Pisa (Italy), 4-8 June 2007.

3. D. Mattera, L. Paura, and F. Sterle, “MMSE WL equalizer in presence of receiver IQ imbalance,” to be published on Trans. Signal Process., April 2008.

4. M. Caleffi, G. Ferraiuolo, L. Paura, “Augmented Tree-based Routing Protocol for Scalable Ad Hoc Networks,” IEEE Internatonal Conference on Mobile Adhoc and Sensor Systems (MASS '07), Pisa (Italy), 8-11 October, 2007.

5. M. Caleffi, G. Ferraiuolo, L. Paura, “On Reliability of Dynamic Addressing Routing Protocols in Mobile Ad Hoc Networks,” Proc. of Wireless Rural and Emergency Communications Conference (WRECOM '07), Roma (Italy), October 2007.

6. M. Caleffi, G. Ferraiuolo, and L.Paura, “A Reliability-based Framework for Multi-path Routing Analysis in Mobile Ad-Hoc Networks,” to appear on International Journal of Communication Networks and Distributed Systems.


IQ imbalance in RF systemsIQ imbalance in RF systems


Direct-frequency conversion

The imperfections in the analog stage are due to low-cost fabrication technologies.

(received signal)

( ) ( )(1 )cos( 2) (1 )sin( 2)

( ) ( )(1 )sin( 2) (1 )cos( 2)c c

s s

y t r t

y t r t

( ) ( ) ( )

( ) ( )c sr t r t jr t

x t n t

useful signal noise

Low-pass equivalent modelLow-pass equivalent modelLPF

LPF

A/D

A/D

Digital S

ignal Processor

(1 )cos 2 2LOf

(1 )sin 2 2LOf

( )RFr t

1

2

amplitude imbalance

phase imbalance

*( ) ( ) ( ) ( ) ( )

cos / 2 sin / 2

cos / 2 sin / 2

c sy t y t jy t r t r t

j

j


Direct-frequency conversion

The imperfections in the analog stage are due to low-cost fabrication technologies.

(received signal)

( ) ( )(1 )cos( 2) (1 )sin( 2)

( ) ( )(1 )sin( 2) (1 )cos( 2)c c

s s

y t r t

y t r t

( ) ( ) ( )

( ) ( )c sr t r t jr t

x t n t

useful signal noise

Low-pass equivalent modelLow-pass equivalent model

y(t)= yc(t)+j ys(t) coincides with r(t) only when =0 and =0.

the receiver IQ imbalance can be compensated if and are known.

LPF

LPF

A/D

A/D

Digital S

ignal Processor

(1 )cos 2 2LOf

(1 )sin 2 2LOf

( )RFr t

1

2

amplitude imbalance

phase imbalance


IQ imbalance in multicarrier system

IFF

T

P/S

1

N

s

s

s s S/P

FF

T

DirectDown-Conversion

(IQ imbalance)y

kh

0 1

0 1

0 1

2 0 1

1 1 0

L

L

cL

L

L L

h h h

h h h

h h h

h h h h

h h h h

H

2

,

1,1 2,2 ,

1with , 0,1, , 1

c

c H

j ik

Ni k

HN N

F e i k NN

N

r H s n

H F ΛF

hF

0

The OFDM system:

The input signal of the DDC stage is:

kykr * y r ry



The OFDM system:

IFF

T

P/S

1

N

s

s

s s S/P

FF

T


(IQ imbalance)y

kh

*

*

y r r

y Fr Fr

*1 1

*2

*/ 2 * @/ 2 2

*/ 2 1 / 2 1

*/ 2 2 / 2

*2

N

N N

N N

N N

N

r r

r r

r r

r r

r r

r r

Fr r Fr r

* y r r ky ykr



The OFDM system:

IFF

T

P/S

1

N

s

s

s s S/P

FF

T


(IQ imbalance)y

kh

*

*

y r r

y Fr Fr

* y r r

*1 1

*2

*/ 2 / 2 2

*/ 2 1 / 2 1

*/ 2 2 / 2

*2

N

N N

N N

N N

N

r r

r r

r ry

r r

r r

r r

ky ykr



The OFDM system:

IFF

T

P/S

1

N

s

s

s s S/P

FF

T


(IQ imbalance)y

kh

*

*

y r r

y Fr Fr

@ @ y Λs Λ s w

* *, ,

* *, 2, 2 2

1, / 2 1k k k k k k k k

k k k k N k N k N k k

y s s w k N

y s s w

IQ distorsion or interference

* y r r ky ykr


IQ imbalance compensation

The IQ imbalance at the receiver can be compensated provided that the IQ imbalance level is known:

In fact, it is easy verified that:

1 2 2w

*

2 2 2w

* * *1 2k k kw y w y r


ML Estimation of the IQ parameters

The IQ imbalance parameters are blindly estimated according to the maximum-likelihood (ML) criterion.

Assumptions:

1) The transmitted signal is circularly symmetric.

2) The noise is zero-mean white Gaussian circularly symmetric.


MLE for Gaussian signal

The transmitted signal is modeled as a Gaussian process.

( ) , (0,1/ 2)k k k k kx x kT a jb with a b N

Likelihood function for unknown α and θ (N received samples):

2 1 21/ 2

21

2 2

1 1ˆ ˆˆ ˆˆ ˆ( , , , ) exp ( , , )2ˆˆ ˆ2 ( , , )

ˆ ˆˆ ˆˆ ˆ( , , ) (1 ) ( , )

NT

n k g n kk

g n

g n n

y y C yC

C C

2 2

1

2 2 2 2

{ } { } { } { }ˆˆ sin

{ } { } { } { }

k k k kk k k

k k k kk k k k

y y y y

y y y y

2 2 2 21 1CRLB( ) cos ( )(1 ) CRLB( ) cos ( )

4N N

Cramèr-Rao Lower Bound (CRLB)

1) low computational complexity2) robustness


Numerical results: estimation

Parameter setting:• K is the number of received samples• SNR 1/σn

2=25dB• 64-QAM• AWGN and Frequency Selective Channel (FCS)

very near tothe CRLBs


Numerical results: SER vs. K

Parameter setting:• IQ mismatch α=0.1 (10%), θ=π/18 (10°)• K is the number of received samples• 64-QAM• SNR 1/σn

2=25dB

Existing method:

22 2 *1, 1 1, 2,

22 2 *2, 1 2, 1,

1 1

1 1

k k k k k

k k k k k

w y w y w

w y w y w


Numerical results: IIR vs. K

Parameter setting:• K is the number of received samples• SNR 1/σn

2=25dB• 64-QAM

Image Rejection Ratio (IIR):2

2IIR

( ) 0.1, 10

( ) 0.04, 4

a

b


DVB-T system

Matlab Simulink to create a DVB-T environment.


IQ imbalance in DVB-T systems


IQ imbalance in DVB-T systems


IQ level estimation in DVB-T system


BER after decoding


MIMO SystemsMIMO Systems


MIMO system model

Input-output equation of the LTI MIMO system:

No

ky

1kn

1kx

2kx

iNkx

1ky

2ky

1,1kh

1, oNkh

1,2kh 2

kn

oNkn

,i oN Nkh

2,2kh

,1iNkh

,

,0

max i jk m k m k i j

m

y H x n

1,1,1 1,2

2,1 2,2 2,1

,1 ,2 ,

i

o o o i

Nk k k

k k kk

N N N Nk k k

h h h

h h h

h h h

H

k k k y Hx n non-dispersive channel


The linear DF MMSE equalizer

The MIMO DF equalizer output:

*

* *

1,2

1, 1,

0 0 0

0

0

0i i i

H

N N N

b

b b

B

H H H H H z W y B x W Hx W n B x

Feedforward-based equalizer:

0 i iN NB 0Successive-cancellations:

1,1,1 1,2

2,1 2,2 2,1

,1 ,2 ,

i

i i i i

N

H

N N N N

g g g

g g g

g g g

G W H

(1)z (1)x̂

(2)z (2)x̂

(3)x̂(3)z

( , ) ( , )*k i i kg b


The linear DF MMSE equalizer

The MIMO DF equalizer output:

H H H H H z W y B x W Hx W n B x


0 0i iN NB

Multistage equalizer:

1,1,1 1,2

2,1 2,2 2,1

,1 ,2 ,

i

i i i i

N

H

N N N N

g g g

g g g

g g g

G W H

(1)z (1)x̂

(2)z (2)x̂

(3)x̂(3)z

( , ) 0k kb


Fast synthesis of DF equalizers

The DF equalizer optimization reduces to a least-square problem under the assumption of uncorrelated input and noise:

2min

xAx c

Multistage equalizer:

Successive-cancellations:


o

T

n N

HA c e

I

1 2 1 |o

T

n N

HA e e e c e

I

is the unit vectore

1 1 1 |i

o

T

Nn N

HA e e e e c e

I


Fr d̂

estimandum

observation vector

d

r

11 12

21 22

ˆ

ˆ

T T

T T

d F F r

rF Fd

widely linear processing

ˆ

ˆ

T T

T T

d F F r

rF Fd

linear processing

ˆ Hd F r

T T T Tj F r F r F r F r

ˆ d ˆ d

Widely linear processing (WL)


*

*

[ ]

[ ]

T

dd

T

rr

E

E

R dd 0

R rr 0

Widely linear processing (WL)

Question: When does the WL processing allow one to achieve a performance gain over the conventional linear processing?

WLF

LF optimum


The WL DF MMSE equalizer

The WL DF equalizer employs both a WL feedforward filter and a WL feedback filter to exploit the non-circular properties of the signals to be processed.

(1: ,1: ) (1: , 1: ) (1: , 1: )

1:(1: ,1: ) (1: , 1: ) (1: , 1: )o r o r i o r i

r io r o r i o r i

N n N n N N n N

n NN n N n N N n N

y nxH

xy H H H n

xy H H H n

By assuming that nr (0≤nr≤Ni) components of x are real-valued, the MIMO channel input-output equation can be rewritten as follows:

The WL DF equalization is defined as the DF equalization operating on the new channel model.


Multistage equalizers

The following multistage equalizer structures have been considered:

feedforward-basedequalizer

(FFB)

successive-cancellationsequalizer

(SC)

successive-cancellationsBLAST equalizer

(BLAST)

multistage equalizer (MS)MS-SC

multistage equalizer (MS)

MS-BLAST

multistage equalizer (MS)

MS-PARALLEL

x̂

x̂

x̂

ˆ ( .)iterx

ˆ ( .)iterx

ˆ ( .)iterx


Numerical results: SER vs. iter.

2

60, =60 (4-PAM)

60

116dB

i r

o

n

N n

N

SNR


Dynamic addressing-based routing scalable Dynamic addressing-based routing scalable protocols for MANETprotocols for MANET


Presentation of the activity

Scope:

To single out scalable routing protocols able to face with the dynamic topology of MANET.

Approach:

Dynamic Addressing.

Goal:

To sepatate the unique terminal identifier, with no location information in a mobile network, from the routing address, which embeds a topological information.


Implementation Address Allocation: distributed mechanism able to assign to

nodes an address which reflects the node relative position, without resorting to flooding.

Routing: proactive hierarchical multi-path protocol, able to explore multiple paths towards a destination with a limited overhead.

Address lookup: distribuited hash-based service (DHT) for providing the mapping between node identifier and routing address.


Results

Normalized packet delivery ratio versus the number of nodes.


Conclusions

IQ imbalance in RF systems:• The blind estimation of the receiver IQ parameters has been

addressed.• The effects of IQ imapairments in DVB-T systems has been

presented.

MIMO systems:• The DF equalizer design has been re-examined in terms of the

conventional least-square problem.• The WL processing has been combined with the DF strategy to

exploit the properties of rotationally variant signals.• Three multistage DF equalizer structures have been compared.

Routing protocols for MANET:• An augmented tree-based routing protocol has been proposed to

handle scalable ad-hoc networks.


Thank you

Documents

PRIN 2005 WOMEN Project Research unit: Università degli Studi di Napoli Federico II 5 th Meeting of WOMEN Project Trento, February 12, 2008 Fabio Sterle