Methods for Compression of Feedback in
Adaptive Multi-carrier 4G Schemes
Víctor P. Gil Jiménez
Ana García Armada
M. Julia Fernández-Getino García
University Calos III de Madrid
Spain
Thomas Eriksson
Tony Ottosson
Arne Svensson
Chalmers University of Technology
Sweden
Outline
•
•
•
•
•
OFDMA for Downlink in 4G.
Feedback information.
Opportunistic Feedback.
System Description.
Algorithms for compression:
– Time correlation
– Frequency correlation
– Time-Frequency correlation
• Results.
• Conclusions.
COST 289 Final Workshop. Göteborg 2007
OFDMA for downlink in 4G
COST 289 Final Workshop. Göteborg 2007
Frequency
• Several techniques for
Downlink.
• OFDMA (Orthogonal
Frequency Division Multiple
Access) is one of the
candidates.
• The BS (Base Station)
allocates the different
user’s transmission across
the orthogonal frequencies.
Time
Feedback Information
• The Adaptive Modulation selects the adequate
Modulation and Coding Scheme (MCS) according to
instantaneous channel conditions for each terminal.
• Channel is usually known (estimated/predicted) at
the receiver.
• The receiver should feedback this information.
• It may be highly rate-demanding.
Reduction & Compression
COST 289 Final Workshop. Göteborg 2007
Opportunistic Feedback
• The scheduler usually selects the best
user(s) for transmitting. Maximum
Throughput criterion.
• It is a waste of resources if a terminal with
a bad channel feeds back its data (it will
never be served except for fairness
policies).
• The BS broadcasts a minimum quality for
feeding back.
COST 289 Final Workshop. Göteborg 2007
Opportunistic Feedback
Some Results
Prob of finding a sub-carrier able to tx >= 4 bits
1
0.9
0.8
Probability
0.7
0.6
Load Factor 0 %
Load Factor 10 %
Load Factor 20 %
Load Factor 30 %
Load Factor 40 %
Load Factor 50 %
Load Factor 60 %
Load Factor 70 %
0.5
0.4
0.3
0.2
0.1
0
10
15
20
25
SNR (dB)
30
35
UMTS Vehicular A Channel
COST 289 Final Workshop. Göteborg 2007
40
(Load factor = %
subcarr occupied
when user arrives)
System Description
Data
S/P
Adap.
Mod.
IFFT
P/S
CP
Terminal
CP
S/P
FFT
Ch. Est.
Feedback Decode/Expand
COST 289 Final Workshop. Göteborg 2007
Adaptive
Demod.
Base Station
P/S
Data
Adapt.
Feedback Encode/Compress
System Description
Parameters
• Downlink OFDMA.
• 512 sub-carriers (416 useful for data).
• Sub-carriers divided into groups of 8 sub-carriers
during 12 OFDM symbols = a chunk.
• 52 chunks in total.
• The same MCS in the same chunk but different
(possibly) among them. Adaptive Modulation at
chunk level.
• Available modulations: BPSK – 256 QAM.
• UMTS channel models: vehicular A, pedestrian A.
COST 289 Final Workshop. Göteborg 2007
System Description
A Chunk
Frame
Frequency
Chunk
12
OFDM
Symbols
8 subcarriers
Time
COST 289 Final Workshop. Göteborg 2007
Algorithms for compression
• Huffman coding for
compression. Needs
well conditioned data.
• Using Time correlation.
• Using Frequency
correlation.
• Using Time-Frequency
correlation.
(BPSK, QPSK, … 64QAM)
COST 289 Final Workshop. Göteborg 2007
Algorithms for Compression
Time and Frequency Correlation
Feed-back difference between actual and former chunk
Time
COST 289 Final Workshop. Göteborg 2007
Frequency
Frequency
Algorithms for Compression
Iterative Time - Frequency
b2t 1
b1t 1
b2t
t
1
b
t 1
0
b0t
b
1
1
1
2
.
.
8
8
.
1
?
Time
1
P1
Huffman code 0
Huffman code 1
…
Huffman code 63
COST 289 Final Workshop. Göteborg 2007
P2
…
P7
P8
Design a Huffman
code for each
combination bit-1,
bi-1t:
64 codes
Frequency
Algorithms for Compression
Block Time - Frequency
1
b2t 1
b1t 1
b
1
b0t 1
b0t
1
b2t
t
1
?
P1
P2
…
P7
P8
Time
1
1
1
1
1
2
.
.
8
8
8
.
Huffman code 0
Huffman code 1
…
Huffman code 511
COST 289 Final Workshop. Göteborg 2007
Design a
Huffman code
for each
combination
bi-1t-1, bit-1 ,
bi+1t-1 :
512 codes
Results
Comparison to other algorithms
30 users. UMTS Vehicular A channel. 50 km/h and SNR = 20 dB
4.5
4
3.5
Througput [bits/s/Hz]
•
3
2.5
2
1.5
MCS Feedback (R = 3)
1-bit quantization [Johansson, Floren]
SNR-limited Feedback [Gesbert]
Our proposal
1
0.5
0
0.5
1
1.5
2
Rate Feedback [bits/chunk]
COST 289 Final Workshop. Göteborg 2007
2.5
3
Results
UMTS Vehicular A Channel
30 users
Uncoded
Frequency (50 km/h)
Frequency (100 km/h)
Time (50 km/h)
Time (100 km/h)
Iterative (50 km/h)
Iterative (100 km/h)
Block (50 km/h)
Block (100 km/h)
3
2.8
Mean Rate Feedback [bits/chunk bits]
•
2.6
2.4
2.2
2
1.8
1.6
1.4
1.2
1
10
12
14
16
18
20
22
SNR [dB]
COST 289 Final Workshop. Göteborg 2007
24
26
28
30
Results
UMTS Pedestrian A Channel
30 users
Uncoded
Frequency
Time
Iterative
Block
3
2.8
Mean Rate Feedback [bits/chunk bits]
•
2.6
2.4
2.2
2
1.8
1.6
1.4
1.2
1
10
12
14
16
18
20
22
SNR [dB]
COST 289 Final Workshop. Göteborg 2007
24
26
28
30
Results
Robustness
100 km/h. AWGN Feedback Channel. Different Refresh Rate
1
0.8
Normalized Throughput
•
0.6
0.4
SNR =
SNR =
SNR =
SNR =
SNR =
SNR =
0.2
0
5
0 dB (ITF)
10 dB (ITF)
20 dB (ITF)
0 dB (BTF)
10 dB (BTF)
20 dB (BTF)
10
15
20
25
Refreshing Frequency (frames)
COST 289 Final Workshop. Göteborg 2007
30
35
Conclusions
• Opportunistic feedback allows 50 % of reduction in feedback
data.
• The compression of feedback data by using time, frequency
or both correlation is feasible and offers compression in the
range of 50 % more.
• Time-Frequency techniques exhibit better performance than
the others (in general).
• Compression algorithms jointly with Opportunistic feedback
allow reductions of more than six times.
• Once the codes are designed and stored, complexity is
almost negligible.
• Time – Frequency algorithms need refresh for robustness.
• Future OFDMA systems are closer to be implemented.
COST 289 Final Workshop. Göteborg 2007
Thank you very much
COST 289 Final Workshop. Göteborg 2007