An Algorithm of DL Cell throughput based on MCS

level in LTE network

Li Panxing, Wang Jing, Feng Jie
The First Branch of China International Telecommunication Construction Group Design Institute Co. Ltd
Baoding, China

AbstractIn order to estimate the DL cell average average throughput which based on the probability distribution
throughput of LTE network, the paper proposes an algorithm of of different SINR (Signal to Interference plus Noise Ratio),
LTE throughput based on MCS level distribution which can be whose core algorithm is shown as Eq. 1:
used to estimate the DL cell capacity under different load
conditions. The algorithm is convenient for operation and useful
in guiding network planning and optimization, the simulation
results show that the algorithm is effective and the effectiveness
cell average throughput = P Throughput_ i
i

doesnt depend on scene. (1)

Where Pi is the probability associated with SINR.
Index TermsLTE, MCS, Throughput. Throughput_i is the throughput which calculated based on
SINR, its calculated by different modulation and coding
I. INTRODUCTION scheme which according to SINR as [5] describes. But [5]
According to the latest reports of GSA (Global Mobile didnt describe the calculation method in detail, and the cell
Suppliers Association) released on April 9, 2015, there are 393 average throughput still obtained by network simulation in [5].
operators have commercially launched LTE in 138 countries, Besides, The MCS selection cant keep up with real channel
and in LTE network construction, capacity planning is a key change based on the SINR only, it must be depending on the
link. The reasonable capacity planning can control the cost and BLER (Block Error Ratio).
evaluate the network construction effectively, while the Using idea of [5], the paper presents an estimating method
throughput is one of the two key parameters for capacity of DL(downlink) cell average throughput which based on MCS
planning in LTE network [1]. The throughput can be divided level distribution, it mainly applies to the distribution of full
into peak throughput, cell average throughput and edge user downlink resources and can be used to calculate the cell
throughput, in which the cell average throughput reflects the capacity under different load in actual network. Its useful in
bearing efficiency of base stations under a certain network load guiding network planning and optimization. The innovation of
and user profile. It is influenced by resources, network load, the study as for [5]is described as follows:1. Pi will be change
user distribution and scheduling algorithms, etc. and the to the distribution ratio based on MCS level; 2.The query
calculation of cell average throughput is necessary for network methods of TBS under different MCS level will be described in
planning and optimization. detail.3.The effectiveness of the algorithm will be verified
It has two methods to calculate the throughput in LTE through network simulation.
system[2]: The first one is calculating based on the physical II. ALGORITHM OF CELL THROUGHPUT BASED ON MCS LEVEL
layer bandwidth and overhead, as in [3] and [4];the other is
based on the MAC (Modulation and Coding Scheme )layer The throughput can be divided into the application layer
data block size , as in [2].In which [3] gives the throughput throughput, IP layer throughput, PDCP (Packet Data
capacity calculation methods and results from different points Convergence Protocol) layer throughput, RLC layer throughput,
of RB bandwidth configuration, normal subframe ratio, the MAC layer throughput and physical layer throughput from
special subframe ratio ,etc. [4] introduces the system frame highest-level to lowest according to different layers of the
structure to TD-LTE throughput, and it also takes the control protocol stack. The reason for the differences between the
information into account. The simulation results of peak above throughputs is the overhead and retransmission, the
throughput show that the algorithm is correct. [2] gives the higher level throughput is less than the lower. The paper study
peak throughput Calculation method based on MAC layer data the MAC layer throughput mainly, the other layer throughput
block size. However, the above documents are focused on the can be calculated through the MAC layer throughput.
study of peak throughput, but the cell average throughput is far In actual network, the chief considerations of MAC layer
less than the theoretical peak throughput due to the influence of throughput include available PRB(Physical Resource Block),
load and interference in reality. Most of the estimates of the TBS(Transport Block Size), UE Category and code rate,
cell average throughput are based on network simulation Where the available PRB is decided by bandwidth, e.g. there
currently, only [5] refers to a calculation method of the cell are 100PRBs in 20Mhz; the transport block size is determined

ICSPCC2015
978-1-4799-8920-1/15/$31.00 2015 IEEE
by TBS tables mentioned by 3GPP TS36.213;The UE into TBS_L2 using the mapping rule shown in Table 7.1.7.2.2-
Category determines Maximum number of bits of a transport 1[7].
block received within a TTI; the code rate is defined as the 3) calculate the bits for each sub-frame
number of downlink information bits (including CRC bits) In particular, if the transport blocks are transported in
divided by the number of physical channel bits on PDSCH, DwPTS of the special subframeTBS should be zoomed 0.75
The UE may skip decoding a transport block in an initial compared with the normal subframe, because the duration
transmission if the code rate is higher than 0.930.The period of DwPTS is shorter than normal subframe[8].
Computation of throughput is substantially as follows: firstly, 4) Calculate bits transported in one radio frame, then
SINR is converted into CQI (Channel quality indicator) by UE calculate the throughput.
which ensure the BLER under 10%;secondly, UE sends CQI to
eNodeB and eNodeB selects the appropriate MCS level, at last, III. CALCULATION EXAMPLE
data transports on appropriate transport block which selected In this example, conditions of computation are given as
according to the MCS level. The code rate should be follows: system: TD-LTE; bandwidth: 20MHz; DL: UL: 2: 2;
considered during MCS level selecting, and the UE Category the special subframe ratio: 10: 2: 2; transmission mode: dual-
will be considered during TBS selecting. The algorithm stream.
neednt consider code rate because of it bases on MCS whose Calculation steps:
step is after MCS level selecting. From the UE Category table
1) Table I shows that at there are 100PRB in 20Mhz
we can know that Level 1 and Level 2 are limited under single
bandwidth.
transmit antenna port; and level3 is limited under two transmit
2) Assume that the MCS level distributes as table II (Due
antenna ports in downlink. But level3 is limited in the case of
20MHz bandwidth and 100PRB, in actual network, the to limitations on space, parts of data are shown in tabel II)in
resources are shared by multiple users, Cases of full resources this example,then we can obtain TBS distribution which shown
given to single user are extremely rare.so we leave UE limited in table II too.
considerations aside as an estimation method.
Steps of the estimation method of cell average throughput TABLE II. MCS LEVEL DISTRIBUTION
based on the MCS level are as follows: MCS index Mi TBS index
TBS_L1
Obtain the transmission bandwidth configuration (NRB) NPRB=100)
according to channel bandwidth from Table I [6]. 0 3.45% 0 2792

1 3.45% 1 3624
TABLE I. TRANSMISSION BANDWIDTH CONFIGURATION NRB IN E-UTRA 4584
CHANNEL BANDWIDTHS 2 3.45% 2
3 3.45% 3 5736
Channel
bandwidth 1.4 3 5 10 15 20 4 3.45% 4 7224
BWChannel [MHz] 8760
5 3.45% 5
Transmission
6 15 25 50 75 100 6 3.45% 6 10296
bandwidth
configuration NRB 7 3.45% 7 12216
1) Count the MCS level distribution Mi(i=0,228)from 8 3.45% 8 14112
actrual network or network simulation results. 15840
9 3.45% 9
2) Calculate the equivalent TBS according to Eq. 2.
10 3.45% 9 15840

28
ETBS = TBSj i
i =0
( ) Mj (2) 3) Calculate the equivalent TBS
The transmission mode is dual-stream in this example, and
there are 100RB in 20MHz, so we translate the TBS which
Where ETBS: equivalent TBS; corresponding to 100RB in table 7.1.7.2.1-1(TBS_L1) into
i: MCS index number TBS_L2 as shown in table III (due to space limitations,
j: TBS index number. The mapping relation of i and j is intercept part of the data):
shown in 3GPP TS36.213 (Table 7.1.7.1-1: Modulation and According to Eq.2, ETBS is
TBS index table for PDSCH) 58082.8bit/ms(5544*0.0345+7224*0.0345+31740*0.034
TBSj: TBS which is shown in 3GPP TS36.213. in the case 5)
of single-stream transmission, the value is shown as table
7.1.7.2.1-1: Transport block size table (dimension 27110), in TABLE III. TBS_L1 TO TBS_L2
the case of dual-stream transmission, for 1 <= NPRB <= 55, the
NPRB=100
TBS is given by the (ITBS, 2NPRB) entry of Table 7.1.7.2.1-1; for ITBS
TBS_L1 TBS_L2
56 <= NPRB <= 110 ,a baseline TBS_L1 is taken from the
0 2792 5544
(ITBS,NPRB) entry of Table 7.1.7.2.1-1, which is then translated
1 3624 7224
 NPRB=100 environment dense urban
ITBS
TBS_L1 TBS_L2 Station number 12
2 4584 9144 station spacing 502m
3 5736 11448 system TD-LTE
4 7224 14688
frequency 1.8GHz
5 8760 17568
bandwidth 20MHz
6 10296 20616
7 12216 24496 DL:UL 3:1
8 14112 28336 special subframe ratio 10:2:2
9 15840 31704 CFI 1
10 17568 35160 transmission mode AMS

4) Calculate the DL cell average throughput 2) Simulation results
The DL cell average throughput is 31.94Mbps The paper use Atoll to make predictions of DL cell
((58082.8*2+58082.8*0.75)*2*100/1000000). throughput under different traffic load which submitted by
Where the first 2means there are 2 DL normal sub- Monte Carlo simulation in scene I, then compare the prediction
frames in a half-frame in this example; the second 2means results between Atoll and the algorithm as shown in Fig. 2.
there are 2 half-frames in a radio frame; 35.0
(58082.8*2+58082.8*0.75)*2 means the bits transport in a s
p 30.0
b
radio frame(10ms); then*100/1000000 in order to converse M
t( 25.0
u
the unit to Mbps. p
h
g 20.0
u
o
r 15.0
h
IV. SIMULATION t
ge
ar
10.0
e
In order to verify the validity of the algorithm, the paper va 5.0
ll
e
c 0.0
establishes two simulation scenes in Atoll(an wireless network L
D 15% 21% 33% 41% 50% 61% 75% 85% 94% 100%
engineering software)first, then compares the cell average DL traffic load

throughputs under different traffic load between Atoll and the DL throughput by the algorithm DL throughput by Atoll
algorithm as shown below.
The traffic load used in the paper is submitted by Monte Fig. 2. Comparison of cell average throughputs between Atoll and the
Carlo simulation which sets four services: High Speed Internet, algorithm
Mobile Internet Access, Video Conferencing and VoIP.
The details are shown in table V:
A. Simulation under sceneI
1) Scene description TABLE V. DETAILS DATA
In scene I, the paper sets 12 base stations in dense urban, DL throughput
DL traffic Equivalent DL throughput
the distributions of base stations and the network load TBS(bit)
by the
by Atoll(Mbps)
configurations are shown as Fig. 1 and table . algorithm(Mbps)
15% 36477 27.4 32.7
21% 32602 24.5 28.9
33% 27937 21.0 24.3
41% 26271 19.7 22.8
50% 24339 18.3 21.1
61% 23109 17.3 20.0
75% 21712 16.3 18.7
85% 21155 15.9 17.9
94% 20349 15.3 17.5
100% 20200 15.2 17.4

B. Simulation under scene

1) Scene description
In scene II, the paper sets 11 base stations in another dense
Fig. 1. Base station distribution of scene I
urban, and change DL: UL to 2:2. The distributions of base
stations and the network configurations are shown as Fig.3 and
tableVI.
TABLE IV. NETWORK CONFIGURATIONS OF SCENE I
 DL throughput
DL traffic Equivalent DL throughput
by the
load TBS(bit) by Atoll(Mbps)
algorithm(Mbps)
22% 31778 17.5 20.8
32% 28319 15.6 18.3
44% 24983 13.7 15.9
51% 23572 13.0 15.0
60% 22237 12.2 14.1
69% 21064 11.6 13.3
82% 19923 11.0 12.6
92% 19158 10.5 12.1
100% 18818 10.4 11.8

C. The Comprehensive Analysis

From simulation under scene I and scene II we can see: a).
Fig. 3. Base station distribution of scene with the increasing of load, the cell average throughput would
reduce as shown in Fig.1 and Fig.2, the reason is that the same
frequency is deployed in LTE network, in the same frequency
TABLE VI. NETWORK CONFIGURATIONS OF THE FIRST ENVIRONMENT network, with the increasing of neighbors load, the
environment dense urban interference will be greater while the cell throughput will be
reduced accordingly. Therefore, the throughput phase of
Station number 11
network planning should match with the expected load. b).form
station spacing 590m
table V and table VI we can see that the results between the
system TD-LTE algorithm and Atoll are similar, the slightly differences are due
frequency 1.8GHz to the differences between algorithms. The simulation results
bandwidth 20MHz show that the algorithm is effective and the effectiveness
DL:UL 2:2 doesnt depend on scene.
special subframe ratio 10:2:2 V. CONCLUSION
CFI 1
The paper proposed an estimation method of DL cell
transmission mode AMS
average throughput based on MCS in LTE network. The cell
capacity under the different loading quantity can be predicted
2) Simulation results by the algorithm. And the simulation results show that the
The comparison of cell average throughputs between Atoll algorithm is effective and the effectiveness doesnt depend on
and the algorithm is shown in Fig. 4. scene. The next work is to optimize the algorithm using the
30.0 measured data of actual network.
s
p 25.0
b
M
(t ACKNOWLEDGEMENTS
u 20.0
p
h
g
u 15.0
o
I would like to express my gratitude to my colleagues who
r
th
e 10.0 have helped me during the writing and modification. And also I
agr
e
va 5.0
would like to thank reviewers for their careful reading and
ll
e
c 0.0
constructive suggestions.
L 13% 22% 32% 44% 51% 60% 69% 82% 92% 100%
D
DL traffic load REFERENCES
DL throughput by the algorithm DL throughput by Atoll [1] Li zhengmao,Wang Xiaoyun.TD-LTE Application and
Practice.Beijing,China:Post&telecom press,2014.
Fig. 4. Comparison of cell average throughputs between Atoll and the [2] Zhu Xuetian,An Xiaodong,Gao Yu and Ma yunfei. TD-LTE
algorithm Analysis and optimization of wireless performance,
Beijing,China:publishing house of electronics industry,2014.
The details are shown as table VII: [3] Xiao Qinghua. The Quantitative analysis of TD-LTE system
throughput, Designing Techniques of Posts and
TABLE VII. DETAILS DATA Telecommunications.2013.3,pp16-21.
DL throughput [4] Shi Kui,Li Xiaolin and Jin Yu.Throughput and influence factor
DL traffic Equivalent DL throughput of throughput in LTE network, Mobile communication.
by the
load TBS(bit) by Atoll(Mbps) 2013,pp110-114.
algorithm(Mbps)
13% 37793 20.8 25.1
[5] Huawei technologies Co. Ltd. The special training of DLTE [7] 3GPP TS 36.213 V10.8.0: Evolved Universal Terrestrial Radio
wireless network planning analysis and solution,pp94-95. Access (E-UTRA); Physical layer procedures.
[6] 3GPP TS 36.101 V9.14.0 Evolved Universal Terrestrial [8] Erik Dahlman,Stefan Parkvall,Johan Skld.4G LTE/LTE-
Radio Access (E-UTRA);User Equipment (UE) radio Advanced for mobile broadband. Beijing,China:Post&telecom
transmission and reception. press,2012.
`