Documente Academic
Documente Profesional
Documente Cultură
1 INTRODUCTION
Non-GBR
100
10-1
Non-GBR
300
10-1
Non-GBR
300
10-1
Non-GBR
300
10-1
QCI
1
2
Voice, video
(live streaming),
interactive
gaming
Video (buffered
streaming)
TCP-based
(www, e-mail),
chat, FTP, p2p
file sharing,
progressive
video
The same like
QCI 8
3 SIMULATION MODEL
4.1 Scenario 1
According to Table 2, two LTE implementation scenarios
of the proposed adaptable turbo decoder are examined
using two example bit rates (1 and 30 Mbps). For 1 Mbps
data rate an LTE uplink physical channel is considered
(PUSCH Physical Uplink Shared Channel). Additionally, for 30 Mbps data rate an LTE downlink physical channel (PDSCH Physical Downlink Shared Channel) is considered in our analysis. The maximum values of the data
rates for PUSCH and PDSCH channels are 50 and 100
Mbps, respectively, according to [7], [8], [9].
Fig. 2. FER (or BLER) vs. Eb/N0 for AWGN channel, various small frames, SOVA and log-MAP (s=0.7).
2016 JOT
www.journaloftelecommunications.co.uk
channel
PUSCH
PDSCH
Frame size
Max data
rate Mbps
50
Chosen data
rate - Mbps
100
30
sova
td msec
0.48
LogMAP
td
msec
1.34
Max
delay
msec
100
sova
FER
LogMAP
FER
Max
FER
Pr. alg.
40
5.8*10-
5.5*10-
10-1
None
1120
13.4
37.6
100
2.8*10-
9.8*10-
10-1
101.06
100
1.7*10-
7*10-2
10-1
LogMAP
Either*
3008
36.09
4800
57.6
161.2
100
1.1*10-
7.5*10-
10-1
SOVA*
6114
73.72
206.4
100
8*10-2
8*10-2
10-1
SOVA
At the following Tables and for each scenario, five example frame (or block) lengths (40, 1120, 3008, 4800 and
6144 bits) are used similarly to [12], [13]. For the calculation of latency for SOVA and log-MAP we use the formulas which are given in [4], [5], [6]. For each data rate, all
possible 9 QCIs are applied, with each QCI corresponding
to a possible application, according to Tables 1 and 2.
As can be seen in Table 3 for QCIs 1, 5 and 6 SOVA is
the best turbo decoding algorithm choice since it can satisfy both the FER and latency criteria for large frames.
The asterisk indicates that the FER achieved with SOVA
is very close to the limit (for frame lengths of 3008 and
4800 bits). For a frame of 3008 bits either algorithm can be
used, while for a small frame of 1120 bits, log-MAP satisfies the latency and FER criterion. On the other hand, for
the smallest frame length of 40 bits neither algorithm can
be applied. According to Table 1, a possible application
for QCI 1 is conversational voice, for QCI 5 IMS signaling,
while for QCI 6 voice, live streaming video and interactive gaming.
The analysis is similar for QCI 2 (Table 4). The only
change compared with Table 3 is the looser latency limit
frame
- bits
sova
td msec
0.48
LogMAP
td
msec
1.34
Max
delay
msec
150
40
1120
sova
FER
LogMAP
FER
Max
FER
Pr. alg.
5.8*10-
5.5*10-
10-1
None
13.4
37.63
150
2.8*10-
9.8*10-
10-1
LogMAP
Either*
36
101.06
150
1.7*10
4800
57.6
161.2
150
1.1*10-
7.5*10-
10-1
SOVA*
6114
73.7
206.4
150
8*10-2
8*10-2
10-1
SOVA
10
-1
3008
7*10
-2
sova
td msec
Max
delaymsec
0.48
LogMAP
td
msec
1.34
40
1120
sova
FER
LogMAP
FER
Max
FER
300
5.8*10-
5.5*10-
10-1
None
13.4
37.63
300
2.8*10-
9.8*10-
10-1
3008
36.09
101.06
300
1.7*10
4800
57.6
161.28
300
1.1*10-
6114
73.72
206.4
300
8*10-2
7*10
-2
7.5*102
8*10-2
10
-1
10-1
10-1
Pr.
alg.
sova
td msec
40
LogMAP
Either*
1120
3008
1.2
Either*
4800
1.92
Either
6114
sova
td msec
0.48
LogMAP
td
msec
1.34
Max
delay
msec
50
40
1120
13.44
37.63
3008
36.09
4800
6114
40
10-1
LogMAP
SOVA*
5.5*10-
10-1
None
100
2.8*10-
9.8*10-
10-1
3.36
100
1.7*10
LogMAP
Either*
5.37
100
1.1*10-
6.88
100
0.44
1.25
3008
1.2
7.5*10-
10-1
SOVA*
4800
8*10-2
10-1
None
6114
10-1
None
50
2.8*10-
9.8*10-
10-1
101.06
50
1.7*10-
7*10-2
57.6
161.28
50
1.1*10-
73.72
206.4
50
8*10-2
5.8*10-
1120
5.5*102
Max
FER
0.01
5.8*10-
LogMAP
FER
Max
delay
msec
150
Pr. alg.
2.45
sova
FER
LogMAP
td msec
0.04
Max
FER
1.25
sova
td msec
LogMAP
FER
0.44
frame
- bits
sova
FER
Max
delay
msec
100
4.2 Scenario 2
For scenario 2 the data rate is 30 Mbps and Eb/No is
0.6 dB. In our analysis we also consider the simulation
results presented in Fig. 2 and 3. As can be seen in Table 7
for QCIs 1, 5 and 6 for all large frame lengths both algorithms are equally suitable. For the smallest frame of 40
bits neither algorithm can be applied. For the other small
frame of 1120 bits, log-MAP is proposed because SOVA
cannot satisfy the FER limit of 10 . The asterisk again indicates that SOVA FER is very close to the FER budget
(10 ), thus we assume that the FER criterion is satisfied.
8*10
-2
7*10
-2
10
7.5*10-
10-1
8*10
-1
-2
10
-1
Either*
Either
sova
FER
LogMAP
FER
Max
FER
Pr. alg.
5.8*10-
5.5*10-
10-1
None
150
2.8*10-
9.8*10-
10-1
3.36
150
1.7*10-
7*10-2
10-1
LogMAP
Either*
1.92
5.37
150
1.1*10-
7.5*10-
10-1
Either*
2.45
6.88
150
8*10-2
8*10-2
10-1
Either
sova
td msec
40
0.01
LogMAP
td msec
0.04
Max
delay
msec
300
sova
FER
LogMAP
FER
Max
FER
5.8*10-
5.5*10-
10-1
-1
1.25
300
2.8*10
3008
1.2
3.36
300
1.7*10-
7*10-2
10-1
LogMAP
Either*
4800
1.92
5.37
300
1.1*10-
7.5*10-
10-1
Either*
6114
2.45
6.88
300
8*10-2
8*10-2
10-1
Either
10
None
0.44
9.8*10
Pr.
alg.
1120
-1
-1
Pr.
alg.
0.01
LogMAP
td msec
0.04
sova
td msec
0.01
LogMAP
td msec
0.04
Max
delay
msec
50
sova
FER
LogMAP
FER
Max
FER
40
5.8*10-
5.5*10-
10-1
None
1120
0.44
1.25
50
2.8*10-
9.8*10-
10-1
3.36
50
1.7*10-
7*10-2
10-1
LogMAP
Either*
3008
1.2
4800
1.92
5.37
50
1.1*10-
7.5*10-
10-1
Either*
6114
2.45
6.88
50
8*10-2
8*10-2
10-1
Either
Pr.
alg.
Scenario 1
TABLE 11: QOS, PROPOSED ALGORITHM FOR QCIS 1, 5, 6, SCENARIO 1.
Fig. 4. FER (or BLER) vs. Eb/N0 for uncorrelated Rayleigh fading channel, various small frames, SOVA and logMAP (s=0.7).
frame
- bits
sova
td msec
0.48
LogMAP
td
msec
1.34
Max
delay
msec
100
40
sova
FER
LogMAP
FER
Max
FER
Pr. alg.
5.5*10-
5.1*10-
10-1
None
1120
13.4
37.6
100
1.3*10-
7*10-2
10-1
Either*
3008
36.09
101.06
100
7.4*10-
6.8*10-
10-1
Either*
-1
SOVA
4800
57.6
161.2
100
6.3*10
6114
73.72
206.4
100
1.3*10-
5.8*10
2
7.4*102
10
10-1
SOVA*
Fig. 5. FER (or BLER) vs. Eb/N0 for uncorrelated Rayleigh fading channel, various large frames, SOVA and logMAP (s=0.7).
frame
- bits
sova
td msec
0.48
LogMAP
td
msec
1.34
Max
delay
msec
150
40
1120
sova
FER
LogMAP
FER
Max
FER
Pr. alg.
5.5*10-
5.1*10-
10-1
None
13.4
37.63
150
1.3*10-
7*10-2
10-1
Either*
-1
Either*
SOVA
3008
36
101.06
150
7.4*10
4800
57.6
161.2
150
6.3*10-
5.8*10-
10-1
150
-1
6114
73.7
206.4
1.3*10
1
6.8*10
7.4*10
2
10
10
SOVA*
1120 bits, either algorithm is proposed (the asterisk indicates that SOVA FER is very close to the limit). Finally, for
the smallest frame length of 40 bits neither algorithm is
proposed due to the high FER achieved which is far away
from the limit.
TABLE 13: QOS, PROPOSED ALGORITHM FOR QCIS 3, 7, 8, 9 SCENARIO 1.
frame
- bits
sova
td msec
40
0.48
LogMAP
td
msec
1.34
1120
13.4
37.63
Max
delaymsec
sova
FER
LogMAP
FER
300
5.5*10-
5.1*10-
10-1
None
300
1.3*10-
7*10-2
10-1
Either*
Max
FER
Either
36.09
101.06
300
7.4*10
4800
57.6
161.28
300
6.3*10-
5.8*10-
10-1
Either
6114
73.72
206.4
300
1.3*10-
7.4*10-
10-1
Either*
10
-1
3008
6.8*10
Pr.
alg.
sova
td msec
40
0.48
LogMAP
td
msec
1.34
1120
13.44
37.63
Max
delay
msec
50
sova
FER
LogMAP
FER
Max
FER
Pr.
alg.
5.5*10-
5.1*10-
10-1
None
50
1.3*10-
7*10-2
10-1
Either*
-3
SOVA
3008
36.09
101.06
50
7.4*10
4800
57.6
161.28
50
6.3*10-
5.8*10-
10-3
None
6114
73.72
206.4
50
1.3*10-
7.4*10-
10-3
None
6.8*10
10
5.2
Scenario 2
For scenario 2 the data rate is 30 Mbps and Eb/No is 2
dB. Our analysis is based on the simulation results presented in Fig. 4 and 5. As can be seen in Tables 15, 16, 17
and 18 for all QCIs, for all frame lengths both algorithms
are equally suitable apart from the very small frame
length of 40 bits where neither algorithm can be applied.
It is obvious that both algorithms satisfy the latency and
FER limits for this scenario. Here, the asterisk indicates
that the FER achieved with SOVA is very close to the FER
budget.
sova
td msec
0.01
LogMAP
td msec
0.04
Max
delay
msec
100
40
1120
sova
FER
LogMAP
FER
Max
FER
5.5*10-
5.1*10-
10-1
None
0.44
1.25
100
1.3*10-
7*10-2
10-1
Either*
Either
Either
1.2
3.36
100
7.4*10
4800
1.92
5.37
100
6.3*10-
5.8*10-
10-1
100
-1
6114
2.45
6.88
1.3*10
1
7.4*10
2
10
-1
3008
6.8*10
10
Pr.
alg.
Either*
sova
td msec
0.01
LogMAP
td msec
0.04
Max
delay
msec
150
40
1120
sova
FER
LogMAP
FER
Max
FER
5.5*10-
5.1*10-
10-1
None
0.44
1.25
150
1.3*10-
7*10-2
10-1
Either*
Either
Either
1.2
3.36
150
7.4*10
4800
1.92
5.37
150
6.3*10-
5.8*10-
10-1
150
-1
6114
2.45
6.88
1.3*10
1
7.4*10
2
10
-1
3008
6.8*10
10
Pr.
alg.
Either*
sova
td msec
40
0.01
LogMAP
td msec
0.04
Max
delay
msec
300
sova
FER
LogMAP
FER
Max
FER
5.5*10-
5.1*10-
10-1
None
0.44
1.25
300
1.3*10
3008
1.2
3.36
300
7.4*10-
6.8*10-
10-1
Either
4800
1.92
5.37
300
6.3*10-
5.8*10-
10-1
Either
6114
2.45
6.88
300
1.3*10-
7.4*10-
10-1
Either*
10
-1
1120
7*10
-2
Pr.
alg.
Either*
sova
td msec
0.01
LogMAP
td msec
0.04
Max
delay
msec
50
40
1120
0.44
1.25
3008
1.2
4800
1.92
6114
2.45
sova
FER
LogMAP
FER
Max
FER
5.5*10-
5.1*10-
10-1
None
50
1.3*10-
7*10-2
10-1
Either*
3.36
50
7.4*10-
6.8*10-
10-1
Either
5.37
50
6.3*10-
5.8*10-
10-1
Either
50
-1
6.88
1.3*10
1
7.4*10
2
10
Pr.
alg.
Either*
CONCLUSION
In this work we show that an adaptable SOVA/logMAP turbo decoder can be implemented in the next
generation standard LTE. Considering BLER or FER
performance, latency, pipeline turbo decoder architecture, single input-output antennas, five different frame
(block) lengths and two example service scenarios (data rates), we propose a turbo decoding algorithm for
AWGN and uncorrelated Rayleigh fading channels.
The proposed turbo decoding algorithm for each frame
length also depends on Eb/N0, the data rate and the
QCI.
As our analysis has shown for AWGN channels, for
large data rates (scenario 2 with 30 Mbps), either
SOVA or log-MAP are proposed for all QCIs and
frame lengths since they both satisfy the FER and latency constraints. There are two exceptions: the first is
the very small frame length of 40 bits, where neither
algorithm can be applied for both scenarios. The second exception is the small frame length of 1120 bits
where log-MAP is proposed.
For small data rates and for scenario 1 (1 Mbps), either algorithm is the proposed choice for QCIs with
loose latency limits (300 msec) apart from the small
frame of 1120 bits where log-MAP is proposed. On the
other hand, for QCIs with tight latency limits (50
msec), SOVA represents the proposed algorithm for
medium and large frames (3008 and 4800 bits), whereas for small frames (1120 bits) log-MAP is preferred.
For even larger frame lengths (6114 bits) no algorithm
is proposed. For QCIs with medium latency limits (100
and 150 msec) SOVA is proposed for large frames,
whereas for medium frames (1120 bits) either algorithm is proposed. Furthermore, for the small frame of
1120 bit, log-MAP is proposed.
As our analysis has shown for uncorrelated Rayleigh
fading channels, for large data rates (scenario 2 with 30
Mbps), either SOVA or log-MAP are proposed for all
QCIs and frame lengths since they both satisfy the FER
and latency constraints. The exception is the very small
frame length of 40 bits, where neither algorithm can be
applied (for both scenarios).
For small data rates and for scenario 1 (1 Mbps), either algorithm is the proposed choice for QCIs with
loose latency limits (300 msec). On the other hand, for
QCIs with tight latency limits (50 msec), SOVA represents the proposed algorithm for medium frames (3008
bits). For larger frame lengths no algorithm is proposed. For QCIs with medium latency limits (100 and
150 msec) SOVA is proposed for large frames, whereas
for medium and small frames (1120 and 3008 bits) either algorithm is proposed.
The research findings described above are similar with
the observations in [12] for uncorrelated Rayleigh fading channel, where BER performance is considered.
For [13] again BER is considered and the comparison
with the work presented here shows that in [13] logMAP is proposed for high data rates and all frames,
whereas in this work we show that log-MAP is pro-
posed for all data rates and small frames. Future work
should focus on correlated Rayleigh fading channel
with different mobile terminal speeds.
ACKNOWLEDGMENT
This research has been co-financed by the European Union (European Social Fund - ESF) and Greek national
funds through the Operational Program "Education and
Lifelong Learning" of the National Strategic Reference
Framework (NSRF) Research Funding Program: ARCHIMEDES III: Investing in knowledge society through
the European Social Fund.
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]