Improved scheme of Open loop Transmit Diversity (STTD) in the DS- - - PowerPoint PPT Presentation

1

Improved scheme of Open loop Transmit Diversity (STTD) in the DS- CDMA case

• S. MAYRARGUE, CEA-LETI
• T. CLESSIENNE, France Telecom R&D

IWCT’05 6-10-th June 2005

IWCT’O5 6-10-th June 2005 2

Outline

• STTD (Spatio-temporal transmit diversity or « Alamouti scheme ») in

the flat fading case

• STTD: extension to frequency selective channels in UMTS
• Problems with UMTS approach in terms of intra-cell interference
• Conditions that spreading codes should fulfill in order to obtain full

benefit of STTD for frequency selective channels in a CDMA system

• Simulation results
• Conclusion

IWCT’O5 6-10-th June 2005 3

STTD- flat fading case

• Spatio-temporal Transmit diversity original scheme
• yn = h1 sn - h2 sn+1* + bn
• yn+1 = h1 sn+1 + h2 sn*+ bn+1

After matched filtering :

• +
• −

=

• +

+ + * 1 n n * 1 n n * 1 * 2 2 1 * 1 n n

b b s s h h h h y y

h1 h2

• sn+1*, sn*

sn, sn+1 yn, yn+1

Time : n, n+1

h1 h2

• sn+1*, sn*

sn, sn+1 yn, yn+1

• −

+

• +

+ =

• −

=

• +

+ + + * 1 n n 1 * 2 2 * 1 * 1 n n 2 2 2 1 2 2 2 1 * 1 n n 1 * 2 2 * 1 1 n n

b b h h h h s s h h h h y y h h h h R R

IWCT’O5 6-10-th June 2005 4

STTD: extension to frequency selective case, CDMA

• STTD introduced in UMTS standard « open loop- diversity » downlink

together with the Rake receiver

• each branch of the Rake receiver corresponds to a flat fading case
• STTD flat fading is applied on each branch

hj

*

data Spreading

h1

despreader + despreader +

h1

*

detection

h2 hj

despreader

delays

hL

IWCT’O5 6-10-th June 2005 5

Problem with the previous approach

• CDMA is a MULTI CODE transmission system
• Simultaneously transmitted codes are orthogonal
• Multipath propagation induces orthogonality loss

Rake receiver is impaired by MAI (Multiple Access Interference) STTD implemented as above creates additional MAI caused between codes transmitted during different time instants of the STTD symbol pair.

IWCT’O5 6-10-th June 2005 6

Example with 2 codes

dn, dn+1 sn s n+1

• dn+1

*, dn *

• sn+1

*

s n

*

Spreading codes are time dependent after scrambling Cn

k = diag ( spreading seq. (n))* Walsh (k)

Cn

l = diag ( spreading seq. (n))* Walsh (l)

Walsh (l) Walsh (k) scrambling Channel 1 Walsh (l) Walsh (k) scrambling Channel 2

yn, yn+1

spreading

IWCT’O5 6-10-th June 2005 7

Example with 2 codes- Time domain

dn sn

• dn+1

*

• sn+1

*

Yn(τ) = (sn h 1 (t) -sn+1

* h2 (t) ) * Cn k (t) + (dn h 1 (t) - dn+1 * h 2 (t) ) * Cn l (t)

Yn+1(τ) = (sn+1 h1(t) + sn

* h2 (t)) * Cn+1 k (t) + (dn+1 h1 (t) - dn * h2 (t)) * Cn+1 l (t)

Cn

l

Cn

k

h 1(t) h2(t)

yn

Cn

l

Cn

k

IWCT’O5 6-10-th June 2005 8

Examples with 2 codes Frequency domain

• For notational simplicity, equations are considered in the frequency

domain, which is equivalent to assuming a cyclic prefix, and to assimilate the Rake receiver to a matched filtering.

• Yn (f) = (H1(f) sn – sn+1* H2(f)) Cn,

k (f)+ (H1 (f) dn – dn+1*H2 (f)) C n l (f)

• Yn+1 (f) = (H1(f) sn+1 + sn* H2(f)) C n+1

k (f) + (H1 (f) dn+1 - dn*H2(f)) Cn+1 l (f)

• +
• =
• −

+

• −

=

• +

+ + + + + + + + * 1 n n * 1 n n * 1 n n * l 1 n * 1 * l 1 n * 2 l n 2 l n 1 * 1 n n * k 1 n * 1 * k 1 n * 2 k n 2 k n 1 * 1 n n

d d s s d d ) f ( C ) f ( H ) f ( C ) f ( H ) f ( C ) f ( H ) f ( C ) f ( H s s ) f ( C ) f ( H ) f ( C ) f ( H ) f ( C ) f ( H ) f ( C ) f ( H ) f ( Y ) f ( Y

) f ( ) f (

int

H H

IWCT’O5 6-10-th June 2005 9

Example with 2 codes Matched filtering

• Matched filtering the previous equation gives
• In order to avoid sn+1 (resp. sn) interfere with estimate of sn (resp.

sn+1), we have to have

• « usual » MAI is created by dn (resp dn+1) interfering on sn (resp. sn+1).

« extra » MAI would be created by dn+1 (resp. dn) interfering on sn (resp. sn+1). Avoiding this « extra » MAI, we should have

df d d df s s df ) f ( Y ) f ( Y R R

B * 1 n n B * 1 n n B * 1 n n 1 n n

) f ( ) f ( ) f ( ) f ( ) f (

• +
• =
• =
• +

+ + +

int

* * *

• matrix

diagonal df

B

) f ( ) f (

=

• int

*

• matrix

identity // df

B

) f ( ) f (

• *

IWCT’O5 6-10-th June 2005 10

Dependencies on codes

Clearly, this matrix is proportional to Identity iff This matrix is diagonal if Cn+1

k (f) = Cn k * (f) for all codes

NB This second condition implies the first one

2 k 1 n 2 k n

C C

+

=

df C H C H ) C C ( H H ) C C ( H H C H C H df

2 k n 2 2 2 k 1 n 2 1 2 k n 2 k 1 n 2 * 1 2 k n 2 k 1 n 2 * 1 2 k 1 n 2 2 2 k n 2 1 B

) f ( ) f (

• +

− − + =

+ + + +

• *

df ) f ( C C H ) f ( C C H ) C C C C ( H H ) C C C C ( H H ) f ( C C H ) f ( C C H df

* l n * k n 2 2 l 1 n * k 1 n 2 1 l n * k n * l 1 n k 1 n * 2 1 l n * k n * l 1 n k 1 n 2 * 1 * l 1 n * k 1 n 2 2 l n * k n 2 1 B

) f ( ) f (

• +

− − + =

+ + + + + + + +

• int

*

IWCT’O5 6-10-th June 2005 11

What does it mean?

• Cn+1

k (f) = Cn k * (f) in the frequency domain means

Cn+1

k (t) = Cn k * (T-t) in the time domain

Codes on each symbol of the STTD pair should be time reversed and conjugate from each other. Both orthogonal codes and scrambling sequence should have this property. Otherwise, additional MAI arises. This is the case in UMTS!!

IWCT’O5 6-10-th June 2005 12

Simulation results for UMTS

• Comparison of «usual » STTD with STTD with modified codes
• SF =32
• Mobile speed = 3km/h
• Channel 3 of 3GPP
• Ior/Ioc = 10 dB, 20 dB (ratio between intra cell interference caused by
• ther codes, and extra cell interference, modelled by gaussian noise)
• Channel estimation using CPICH
• 2 sets of simulations: with and without coding. Coding is a turbo

coding with rate = 1/3

IWCT’O5 6-10-th June 2005 13

Simulation results 1/3

IWCT’O5 6-10-th June 2005 14

Simulation results 2/3

IWCT’O5 6-10-th June 2005 15

Simulation results 3/3

IWCT’O5 6-10-th June 2005 16

Conclusion

• A simple modification in the spreading code allows a 1dB

improvement on STTD performances, compared to the usual UMTS implementation