frequency diversity through hopping results in performance gain (fig. 1). expectably, performance is...
TRANSCRIPT
Frequency diversity through hopping results in performance gain (fig. 1). Expectably, performance is better over larger delay spread channel VehB.
Naturally, performance is improved as the interleaving depth is increased (fig. 2). However, the gain from employing frequency hopping as opposed to plain block interleaving is reduced. This is because the attainable frequency diversity between two time slots decreases with their time difference Δt, due to the multiplicative factor Jo(2πfDΔt) in the channel correlation function.
The effect of the Doppler spread depends on the channel (fig. 3).
Substantial gain can be attained if only a subset of the hopping patterns is used, such that any hop is at least 20 tones (fig. 4).
FH-OFDM for Mobile Broadband Wireless AccessFH-OFDM for Mobile Broadband Wireless AccessKostas Stamatiou, John G. Proakis
AbstractAbstract We are studying the deployment of Frequency Hopped OFDM
(FH-OFDM) at the downlink of a cellular mobile radio system. Each user in a cell is assigned a number of tones by the base-station, which change over time according to a predetermined hopping pattern. The construction of the hopping patterns is based on orthogonal latin squares, which have desirable properties regarding intra-cell interference avoidance and inter-cell interference averaging.
In this work, the performance in terms of the bit-error-probability is evaluated analytically for two coded modulation schemes, i.e. TCM and BICM, combined with a block interleaver, without taking into account the adjacent cell interference. Our objective is to quantify the effect of the channel parameters, the hopping pattern selection and the interleaver depth on the performance of a coded FH-OFDM link.
MotivationMotivation
OFDM
Orthogonal multiple access: No intra-cell interference (potential for higher capacity than CDMA)
High data rates without the need for equalization (simple receiver design)
Frequency Hopping
Frequency diversity (gain through coding)
Inter-cell interference averaging (higher frequency reuse)
Candidate technology for 802.20
FH-OFDM Downlink FH-OFDM Downlink
Each mobile is assigned a number of tones, according to its
demand in bandwidth
The tones change over time according to a hopping pattern
MS 2BS
MS 1
1 2 1 1
2 1 2 1
1 2 1 1 2
2 1 1 1
1 2 1
1 1 2 1
1 2 1 2
2 1 1
1 1 1
1 2f
t
Ban
d wid
th
OFDM symbol
Interference Interference
Different hopping patterns are assigned to adjacent cells in order
to average inter-cell interference
1
1
1
1
1
1
1
1
1
1
1 3 2
2 1 3
1 3 2
3 1 2
3 1 2
2 1 3
2 1 3
3 1 2
2 3 1
2 1 3
1
2
3
Cell 1 Cell 2
Parameters Parameters
Carrier frequency 2 GHz
Bandwidth 1.25 MHz
OFDM symbol duration, Ts 100 μsec
Cyclic prefix duration, Tcp ~ 11,1 μsec
Tone spacing, 1/T 11.25 KHz
FFT size, N’ 128
Number of tones used, N 113
Period of hopping sequences
113
Tcp T
Ts
t
OFDM symbol
Channel Model Channel Model Channel impulse / frequency response:
where gk(t) are WSS, independent, CN (0,2σk2)
Jakes’ model for each tap (fD is the Doppler frequency)
Correlation function:
Noise is AWGN
M
kkk tgtg
1
)()();(
M
k
fjk
ketgtfG1
2)();(
tfJttgtgE Dkkk 22)]()([ 02*
M
k
fjkDoGG
ketfJttffGtfGEtf1
22* )2(2);();();(
Vehicular Test Environments Vehicular Test Environments
UMTS macro-cellular channel model also considered for 802.20
Vehicular Test Environments (TE) VehA and VehB
Latin Squares Latin Squares For α = 1,…,N-1 define an NxN matrix Aα by setting
Aα(i,j) = αi + j (modN)
where i,j = 0,…,N-1 and N is a prime number. The set {Aα} is a family of N-1 mutually orthogonal latin squares.
Example: N = 50 1 2 3 4
1 2 3 4 0
2 3 4 0 1
3 4 0 1 2
4 0 1 2 3
Cell A1
0 1 2 3 4
2 3 4 0 1
4 0 1 2 3
1 2 3 4 0
3 4 0 1 2
Cell A2
Properties Properties
The rows and columns of Aα represent the OFDM tones and time slots respectively
If an element of Aα is assigned to a user, a N-periodic hopping pattern is constructed for that user
Hopping patterns within cell are orthogonal
There can be only one collision between any pair of hopping patterns in two different cells (interference randomization)
Δf between two consecutive tones in the hopping pattern of any user in cell Aα is either -(α-1)modN or N-(α-1)modN
TCM TCM Base-station with Trellis Coded Modulation (TCM)
TCM
EncoderTCM
Encoder
Block
Symbol
Interleaver
(depth β)
Block
Symbol
Interleaver
(depth β)
Hopping pattern
generatorHopping pattern
generator
Coded streams of
other users
Symbol
mapperSymbol
mapper
One tone/user/OFDM symbol
Latin squares, N = 113
Rate 2/3, 8-state Ungerboeck code with 8-PSK (Lmin = 2)
OFDM Mod.
OFDM Mod.
Binary
SourceBinary
Source
Performance - TCMPerformance - TCMThe bit-error-probability is approximated by
where
E is the set of dominant error events (L 3, S = 4)
is the number of bit errors for the symbol error
event e (obtained from code error-state-diagram)
H is the hopping pattern event of span S
Perfect channel estimation
Block interleaver span α >> S
Eb PwEP
eH Hee )|()(
2
1
)(ew
BICM BICM
Rate 1/2, 8-state Convolutional Code with 16-QAM
Higher diversity order (dfree = 4), but half the trellis complexity
Same information rate (20kbps)
Conv.
EncoderConv.
Encoder
Binary
SourceBinary
Source
Block Bit
Interleaver
(depth βb)
Block Bit
Interleaver
(depth βb)
OFDM Mod.
OFDM Mod.
Hopping pattern
generatorHopping pattern
generator
Coded streams of
other users
Gray
mapperGray
mapper
Bit-interleaving and Gray mappingBit-interleaving and Gray mapping
c1
c2
c3
cα
cα+1
cα+2
cα+3
c2α
c(β-1)α+1
c(β-1)α+2
cβα
c(β-1)α+2
c2α+1
c2α+2
c2α+3
c3α
c3α+1
c3α+2
c3α+3
c4α
16-QAM symbol x1
α
βb
16-QAM symbol x2
0010 0110
01110011
01010001
01000000
1010
1011
1001
1000
1110
1111
1101
1100
dmin
αβb = codeword length
α >> dfree
Performance - BICMPerformance - BICM
The bit-error-probability is approximated by the expression
where P(dfree|H) is the probability of the only minimum distance binary error event (1000111), for given hopping pattern event H
Generally a very complicated problem to compute P(dfree|H) (very complex metrics)
Simplification: For high SNR, probability of erroneous decision on a bit transmitted in symbol xk is dominated by the closest neighbor of xk with the complementary bit
This neighbor is unique for Gray mapping
)]|([ HH freeb dPEP
Results - TCM Results - TCM
Base-station with Bit-interleaved Coded Modulation (BICM)
Results - BICM Results - BICM
Fig. 1 – TCM performance Fig. 2 – Effect of interleaving
Fig. 3 – Effect of Doppler spread Fig. 4 – Effect of hopping pattern selection
Conclusions - TCM Conclusions - TCM
Conclusions - BICM Conclusions - BICM The BICM scheme outperforms the TCM one (fig. 5), since the diversity order of the convolutional code (dfree = 4) is higher than the diversity order of the Ungerboeck code (Lmin = 2, which is the minimum symbol error event length). Moreover, while the information rate achieved by both schemes is 2 bits/symbol, the trellis complexity of BICM is half that of TCM.
Increasing the interleaver depth leads to greater performance gain than in the TCM case (fig. 5). Inversely, frequency hopping as opposed to plain block interleaving also provides more significant gain for BICM compared to TCM (compare fig. 6 and fig. 1,2).
The above observations are related to the fact that the higher the diversity order of a code, the more ‘difficult’ it is for it to be acquired for a given degree of channel correlation. Additional time or frequency diversity thus leads to greater performance gain, as the diversity order of the code is increased.
Future work Future work
Consider a multi-cellular system with frequency re-use factor 1
Effect of interference randomization on the outage capacity
Deployment of MIMO techniques to increase data rate, increase diversity and suppress the interference
Imperfect channel estimation
Iterative decoding (e.g. LDPC)
Fig. 5 – Comparison of BICM and TCM Fig. 6 – Effect of frequency hopping on BICM performance