Sie sind auf Seite 1von 17

Graduate Paper

Kate Gleason College of Engineering


Rochester Institute of Technology
Dubai Silicon Oasis
Dubai, U.A.E.

Hybrid Automatic Repeat Request in LTE


Chowdhury Mizan Mahmood Taher Dr. Muhieddin Amer
Student of Electrical Engineering Professor of Electrical Engineering

MCS – Modulation and Coding Scheme


ABSTRACT NACK – Negative Acknowledgement
OFDMA – Orthogonal Frequency Division
Wireless communication systems are Multiple Access
continuing to grow and evolve amidst high PER – Packet Error Rate
performance, expectations and requirements. RTT – Round Trip Time
Therefore the 3GPP has considered LTE to ensure SAW – Stop and Wait
future competitions. Obtaining accurate channel SINR – Signal to Interface Noise Ratio
quality is not feasible due to various errors. These SNR – Signal to Noise Ratio
errors tend to degrade the system throughput. The TTI – Transmission Time Interval
Hybrid ARQ (HARQ) comes into use in such UE – User Equipment
situations to provide fast re-transmission and
lowering the error rate in the physical link. A HARQ I. INTRODUCTION
mechanism allows one to mend the errors using
techniques such as Incremental Redundancy and Data transmissions in wireless channels are
Chase Combining along with certain protocols subject to errors because of variations in the signal
depending on the timing and modulation selection. In quality received. Link Adaptation can handle such
this paper, the various HARQ schemes will be errors to some degrees. However, counteractions
discussed. To save the radio resources, the number of cannot be done to receiver noise and interference that
HARQ retransmission in the downlink may be are unpredictable. Therefore a Forward Error
limited. Therefore adequate result using an algorithm Correction is used in all wireless systems. The main
proposed in an earlier research work has been principle beyond forward error-correction coding is
provided in this paper to make more efficient to introduce redundancy in the transmitted signal. In
utilization of the radio link. This paper also focuses this, the parity bits are added to the information bits
on the performance of HARQ schemes in the prior to transmission. These parity checks are
OFDMA downlink operation. The throughput and computed from the information bits using a method
PER of these schemes are compared based on the given by the coding structure used.
buffering memory available.
The other approach to handle transmission errors
NOMENCLATURE is to use Automatic Repeat Request. In this approach,
the receiver employs an error detection code to detect
ARQ – Automatic Repeat Request the received packet contains error or not. A positive
ACK – Acknowledgement Acknowledgement (ACK) is sent by the receiver to
BLER – Block Error Rate the transmitter when no error is detected in the
CC – Chase Combining packet. In case of an error occurring, the receiver
CRC – Cyclic Redundancy Check discards the received error packet and transmits a
DL - Downlink negative Acknowledgement (NACK) to the
eNB – e Node Base Station transmitter. Thus the transmitter re-transmits the
HARQ – Hybrid Automatic Repeat Request information after receiving the NACK.
IR – Incremental Redundancy

Copyright © 2013 Rochester Institute of Technology


2

Most of the modern wireless systems, including


LTE deploy Hybrid ARQ (HARQ), which uses a Thus, the transmitter retransmits the missing
combination of Forward Error Coding (FEC) and packet. It can be seen that the stop-and-wait protocol
ARQ scheme in which unsuccessful attempts are requires the receiver to buffer at most one packet that
used in FEC decoding instead of being discarded. is currently being decoded. High transmission delays
The received packets are discarded and the receiver are a major disadvantage for SAW protocol. This is
requests retransmissions of corrupted packets. because the transmitter has to wait for ACK/NACK
feedback before proceeding to the next transmissions.
The first proposal of Hybrid ARQ was in [14] Therefore, the waiting times can be quite long due to
and since then, numerous publications have appeared a combination of transmission delays as well as
(see literature [13] and references therein). In receiver processing times. In the LTE system, the
principle, any error-detection and error correction packet transmission time is only one subframe (1
code can be used. But most of the practical hybrid ms). But it requires seven subframe (7 ms) waiting
ARQ schemes rely on cyclic redundancy check code time before the packet in error can be retransmitted
for error detection and convolution or turbo codes for using Hybrid ARQ.
error correction.

A. ARQ Protocols: Go-Back-N

The ARQ can be divided into three categories: Waiting time problem associated with the
 Stop-And-Wait (SAW) SAW protocol is solved using the Go-Back-N. In this
 Go-Back-N and case, the transmitter does not wait for the ACK to be
 Selective repeat protocols. received from the receiver and keeps sending a
number of packets specified by a window size as
SAW shown in Figure 2.
Stop-and-wait ARQ is the simplest type of
Scheme. A transmitter sends one packet at a time.
After sending a packet, the transmitter waits for
acknowledgment (ACK) or negative
acknowledgment (NACK) and does not send any new
packets until it receives either an ACK or a NACK as
shown in Figure 1. In case of successful transmission,
the receiver sends an ACK and the transmitter
transmits the next information. On the other hand if
decoding of a packet is failed, a NACK signal is sent
by the receiver.

Figure 2 – Go-Back-N

When a negative acknowledgment (NACK)


for a missing packet is received, the transmitter starts
retransmitting packets starting from the missing
packet. Here also buffering is done at most one
packet at the receiver. Due to the small memory
Figure 1 – SAW
3

available for buffering, it can be noted that after


sending a NACK, the receiver ignores all subsequent
packets until it receives retransmission for the
missing packet. The main drawback of the Go-Back-
N approach is duplicate transmissions as the
transmitter retransmits some packets that are already
successfully decoded at the receiver.

Selective Repeat ARQ:

Duplicate transmission and waiting time is Figure 4 – HARQ classification


eventually solved by Selective Repeat. In selective
repeat ARQ protocol, the sending process continues Figure 4 lists the various types of HARQ
to send a number of packets specified by a window transmission based on the timing and adaptation. The
size even after loss of packet as shown in Figure 3. 4 types namely, Synchronous Non-Adaptive,
Synchronous Adaptive, Asynchronous Non-Adaptive
and Asynchronous Adaptive, have been described in
the following lines.

Synchronous Non-Adaptive

As the name suggests, in synchronous


HARQ protocol, the retransmissions happen at fixed
time intervals. With N = 8, if the first subblock (SB)
is transmitted in time subframe number 0, the first
retransmission attempt can only take place in
subframe number 8 and similarly the second
retransmission in subframe number 16 as shown in
figure 5.

Figure 3 – Selective Repeat Request

In this system, the receiving process will


continue to accept and acknowledge packets sent
after an error occuring. Therefore buffering takes
place in both transmitter and receiver. We note that
the selective repeat ARQ protocol retransmits only
the missing packets and therefore avoids the
duplicate transmissions problem of Go-Back-N. In
addition, as the packets can be sent continuously, Figure 5 – Synchronous Non-Adaptive [4]
unlike stop-and-wait, there is no waiting time
problem. The main benefit of this type of HARQ is
that the control information needs to be transmitted
B. HARQ transmission based on timing and along with the first subblock only. However, a major
Adaptation drawback is that the retransmitted subblocks cannot
be scheduled at time-frequency resources
experiencing good channel conditions at the time of
retransmissions. Moreover, the MCS and resource
4

format cannot be adapted at the time of


retransmission according to the prevailing channel
conditions.

Synchronous Adaptive

Figure 6 shows an example of Synchronous


Adaptive HARQ. It allows one to change the
resource allocation and MCS information for
retransmissions while maintaining the time
synchronicity.

Figure 7 – Asynchronous Non-Adaptive [4]

The main drawback of this scheme is limited


flexibility. The retransmissions resource allocation,
MCS and MIMO formats cannot be adapted.

Asynchronous Adaptive

An asynchronous adaptive hybrid ARQ


scheme [6] provides full flexibility for
retransmissions. Retransmissions are done the same
way as the original transmissions as shown in Figure
Figure 6 – Synchronous Adaptive [4] 8. Both the Adaptation and Timing are adjusted
according to the channel quality. Therefore,
In this case, the control information is sent retransmission timing, resource allocation, MCS and
along with the retransmissions as the resource MIMO formats are all adapted according to the
allocation, MCS and MIMO precoding can change prevailing channel and resource conditions at the
for retransmissions. The synchronous adaptive time of retransmission.
hybridARQ scheme thus allows scheduling
retransmissions at frequency resources experiencing
good channel conditions at the time of the
retransmissions and hence recuperates some
frequency-selective scheduling gains.

Asynchronous Non-Adaptive

An asynchronous non-adaptive hybrid ARQ


protocol allows one to schedule retransmissions in
time as shown in Figure 7. In such HARQ, resource
allocation, MCS and MIMO formats kept the same as
the initial transmission. Only the control information
carrying UE ID, hybrid ARQ process and redundancy
version is carried with every retransmission. When
the prevailing channel quality is good, the time-
domain channel sensitive scheduling can be
Figure 8 –Asynchronous Adaptive
performed for retransmissions.
One major drawback of this HARQ is that the
full control information needs to be sent with
retransmissions. Even if the timing, resource
allocation, MCS and MIMO formats of the
5

retransmissions are unchanged relative to the first is a new packet transmission or to retransmission for
transmission, the control information in a previous packet. The resource assignment indicates
asynchronous adaptive HARQ needs to be which time frequency resource units are allocated to
transmitted. The receiver decodes a packet when it the UE. The modulation indicates one of the
receives the control information, which indicates the supported modulations. The payload size or transport
presence of a transmission. block size gives the data information block size. The
hybrid ARQ information consists of hybrid ARQ
process number, redundancy version and new data
C. Flexibility v/s Overhead indicator. The MIMO control information includes
information on transmission rank and precoding, etc.

E. Hybrid ARQ with Soft Combining

The main objective of soft combining is to store


received packet in a buffer memory and combining
with the data retransmission to obtain a single packet
that is much more reliable than its original
constituent. The simplest form of such scheme was
proposed by Chase [5] in 1985. Therefore it was
named Chase Combining. This scheme involves the
retransmission by the transmitter of the same coded
data packet. The receiver decodes these packets and
combines these multiple copies of the transmitted
Figure 9 – Flexibility v/s Overhead [4]
packet.
Depicted in Figure 9 is flexibility versus
overhead trade off for various hybrid ARQ schemes.
The synchronous non-adaptive scheme occupies the
lowest overhead as well as provides the lowest
flexibility. On the other hand, the asynchronous
adaptive scheme provides the best flexibility at the
expense of the largest overhead. The synchronous
adaptive scheme and the asynchronous non-adaptive
scheme provide some flexibility such as avoiding
resource conflicts with persistent allocations with
intermediate overhead. The overhead for the
synchronous adaptive scheme is expected to be larger
than the asynchronous adaptive scheme as resource
allocation generally contributes the most to the total
overhead. Figure 11 – Chase Combining [16]

D. Scheduling grant control message Chase combining can be seen as additional


repetition coding as each retransmission is an
identical copy of the original transmission. Therefore
the accumulation of received Eb/N0 for every
retransmission occurs, due to the fact that no new
Figure 10 – Scheduling grant control message redundancy is being transmitted.
contents [4]
Incremental redundancy is another type of
As mentioned earlier the Hybrid ARQ sends soft combined Hybrid ARQ. In this scheme,
control information along with the data. The Figure progressive parity packets are sent in each
10 shows all the fields that the control message transmission of the packet, instead of sending simple
contains. UE ID indicates the User Equipment (UE) repeats of the coded data packet. The decoder later
for which the data transmission is intended. The new combines all the transmissions and decodes the
data indicator (NDI) is used to indicate if a subblock packet at a lower code rate.
6

is used in the downlink while synchronous adaptive


hybrid ARQ is employed in the uplink. The new data
indicator (NDI) field in the uplink scheduling grant is
used to indicate if the grant is for a retransmission of
a previous transmission or grant for a new transport
block transmission.

If the control message is received with the NDI


bit toggled, this means that eNB is scheduling a new
uplink transmission. On the other hand, if NDI is not
toggled, this means a retransmission of the previous
transmission attempt. Moreover, if no uplink
scheduling assignment is received while an ACK is
received on the Physical Hybrid Automatic Repeat
Request Indicator Channel (PHICH), this indicates
Figure 12 – Incremental Redundancy [16] successful transmission of the uplink transport block.

It is proved that incremental redundancy can


give superior performance due to coding gain at
retransmissions. However, this gain comes at the B. Number of Hybrid ARQ processes
expense of additional UE complexity because the
buffering required in the case of incremental The N-channel SAW deployed in LTE
redundancy is higher than in the case of Chase consists of number of channels or number of HARQ
combining. processes depending on the buffering and delays.
This is defined as the number of HARQ processes
The LTE system supports the operation of Chase that can be initiated at a given time. This Number is
combining as well as incremental redundancy. given by:

( )
II. Analysis of HARQ in 3GPP LTE [ ] (1)

Previously we have seen the working of 2


In this equation:
main Hybrid ARQ techniques, Chase Combing and
Incremental Redundancy. The effectiveness of these
= Propagation time between the eNB and UE
2 techniques has been studied and analyzed
extensively. This section will look into the general
performance characteristics of each of these schemes. = Subblock transmission time
The overall performance of both the schemes, in
terms of BLER for respective SINR will be studied = Processing time of UE
and analyzed thoroughly. The delay in system and the
effect of HARQ on the system delay are also looked = Transmission time of Ack/Nack
into.
= Processing time of eNB
In the following subsections of this section,
a suitable system model will be implemented. Before Figure 13 shows the relationship between all
that we shall see how the Hybrid ARQ works in LTE the above parameters represented in a Round Trip
system. After that the HARQ model for this Time (RTT)
simulation will be considered. Finally the results of
the simulations are presented.

A. Hybrid ARQ in LTE system

In LTE system incremental redundancy (IR)


based hybrid ARQ with Chase combining as a special
case of IR is employed. In terms of timing and
adaptivity, asynchronous adaptive (AA) hybrid ARQ
Figure 13 – RTT [4]
7

It can be seen that the propagation time for


the cell sizes between eNB and UE is much smaller
than the subblock transmission and the time required
for processing. Therefore this propagation time can
be neglected. Compromising between latency and
signal overhead, the is chosen to be one subframe
(1ms). This was because a smaller subblock
transmission time will only accommodate smaller
information on the transport block. Thus it requires a Figure 14 – HARQ in LTE [4]
larger overhead.
C. Channel Quality Indicator
, the UE processing time, is selected
based on the compromise of latency and complexity. Channel Quality Indicator (CQI) is a quantity of 4-
This is due to the UE complexity increases for a bits which indicate the maximum rate of data that can
smaller value of . The eNB processing time be handled by the mobile. This quantity depends on
accounts for the decoding of the ACK/NACK by the the signal to interface plus noise ration on the
eNB and also the scheduling for the new transport receiving end, as a high data rate is achieved at high
block. and in LTE are selected equal to 1 SINR. But advanced receivers can be exceptional by
subframe (3ms). providing high data rate at low SINR as this factor
also depends on the implementation of the receiver.
is the time required for the transmission
of ACK/NACK. The value of this is considered 1
subframe (1ms) in the LTE system.

Parameter Symbol Value


Propagation Negligible
Time
Subblock 1ms
Transmission
Time
UE Processing 3ms
Time
Ack 1m
Transmission
Time
eNB Processing 3ms Table 2 – CQI table [19]
Time
Table 1 – RTT parameters for LTE [4] Table 2 shows the CQI table in terms of downlink
modulation scheme and the coding rate. The base
Table 1 summarizes all the values used in station uses this received CQI in order to calculate
the Round Trip Time (RTT). Using all the values the modulation and coding scheme. And it uses only
from the table in the equation for , the 1 modulation and coding scheme per transport block
following numerical value can be obtained. in the downlink transmission.

( ) D. System Model
[ ]
Given the SINR and MCS, the BLER is
Therefore 8 HARQ processes have been derived based on early models put forward by
obtained from this analysis. Furthermore, it can be Chawla and Goldsmith [7, 8]. The following equation
seen that an ACK/NACK response for a given is used in deriving the BLER
subframe is transmitted in n+4 subframe. This is
illustrated in figure 14. (2)
8

= SINR, valid for Frame K received


from transmitter
M = Modulation coding scheme used

Where ,
Random assigned
The efficiency of modulation code rate is number checked
given by [9]

(3) No K>effective Yes


BLER
Can be written as
Send NACK Send ACK

(4)

= Signal received power


= Noise density
= Bandwidth Return to main

The implementation was carried out in a


single cell with users in varying number. It has been
assumed that 25RBs are used in a cell. One eNodeB Figure 15 – Flowchart of ACK/NACK process
is needed to control the cell. Users report the SINR to
the eNodeB. A flowchart of the ACK/NACK process is
depicted in the figure 15. It can be seen that if the
The HARQ process used is asynchronous random assigned number is below a threshold BLER,
HARQ process in the downlink and a synchronous the packet seemed to be corrupted or contain error.
HARQ process in the uplink. Both these processes Therefore a NACK is being sent in this case.
have been discussed earlier. Asynchronous will allow
eNodeB to transmit whenever it has packet available HARQ uses both Chase Combining and
to be scheduled. While a synchronous HARQ can Incremental Redundancy. Therefore models of these
only transmit in a fixed time slot. 2 methodologies are looked into on the LTE type
environment.
A Channel Quality Indicator (CQI) concept
is used here which maps the SINR to the transmitted
data bits and the modulation coding scheme. Based
on some previous work, a simplified version of the
CQI table is used here in the implementation. This
CQI is used by eNodeB to determine the modulation
and coding scheme.

E. Algorithm and HARQ model

HARQ helps in delivering service levels.


Priority to packets needing transmission can be
provided by tying a scheduler closely with HARQ.
9

NACK received These 2 flowcharts explain the operations that take


place in these methodologies.

The CQI is mapped using some simplified


Retransmit Yes values for this model. SINR is included in here. The
<=4 eNodeB uses the CQI to calculate the MCS. Table 1
depicts the downlink SINR to data rate mapping.
Drop code
rate to next
No level down

Send packet
with parity bits
Discard the to scheduler
packet and
notify
scheduler
Finish
Table 3 – Downlink SINR to Data Rate
mapping [17]
Figure 16 – Flowchart of Incremental
Redundancy
F. Simulation and System Parameters

C++ programming language has been used


NACK received here for the coding and compiling of the simulation.
Therefore a faster execution of the program was
possible.

LTE standards were emulated without


introducing complexity to the system. But some extra
requirements had to be added. Firstly, the eNodeB
Yes needs to wait for 8ms before it can send a
Retransmit retransmission of the packet [3]. The time taken to
Re-queue
<=4 send an Ack/Nack is 1ms by the UE. The eNodeB
packet for
and UE takes 3ms each in processing the frame.
transmission
Therefore a total of 8ms is required for the data
transmission to the receiving of the Ack/Nack by
No eNodeB.
Send packet to
scheduler for Secondly, parallel running of the HARQ
retransmission process was made ensured. It has been proved in part
B of section II that the number of HARQ processes
used in LTE is 8. Therefore 8 parallel HARQ
Discard the
processes are kept running. Different time slots are
packet and allocated for the transmitter to run several HARQ
notify scheduler Finish processes. So the transmitter is able to send frames to
the same corresponding user but with different
HARQ processes.
Figure 17 – Model of Chase Combining
In the system parameter, we consider a
The working of chase combining as well as single cell environment consisting of 1-40 users. At a
incremental redundancy was previously discussed. particular instance, not every user has data to send.
10

Data frames can be lost en-route or corrupted due to


random fluctuations created in the channel. The
probability of ACK being interpreted with NACK is
[10]. Therefore it has been excluded from the
simulation.

As seen earlier in the flowchart, in case of a


NACK, a packet will go through the process of being
re-scheduled and finally getting discarded if the
threshold time is exceeded.

G. Results

Results were obtained after running several Figure 19 – System Delay performance [1]
batches of files, with each file being repeated in the
same simulation 20 times. It can be seen from figure 19 that
introducing HARQ to the system has a huge impact
A contrast between Chase Combining and on the system delay. A difference of roughly 50ms is
Incremental Redundancy is obtained first. This spotted from the graph. Thus the usage of HARQ
simulation was done for 64 QAM. increases the delay in the system.

Now the question arises whether using


HARQ will be beneficial or not? The answer can be
found by studying the simulated figure 20 of
throughput versus the BLER between HARQ and no
HARQ.

Figure 18 – Chase Combining v/s Incremental


Redundancy [1]

The figure 18 clearly shows Incremental Figure 20 – System Throughput versus BLER [1]
Redundancy outperforming Chase Combining.
Previous works on this issue was studied earlier in Figure 20 clearly shows that using HARQ
literature [11]. has better throughput when BLER is around or
higher. Thus HARQ provides higher throughput at
System delay is an important aspect of data the expense of system delay.
transmission. We create a situation where no HARQ
is used, i.e. the transmitter doesn’t wait for the
ACK/NACK and no re-transmission of packet takes
place. This situation has been simulated against using
Incremental Redundancy and Chase Combining.
11

III. Limiting HARQ retransmission in value. It is for eNodeB to decide of how


the downlink for poor radio condition much limitation of retransmission should be
applied with what value of CQI.
HARQ is implemented by the MAC module
called the HARQ entity in LTE. The techniques by  As seen earlier, the HARQ retransmission
which HARQ helps to recover data have been uses the asynchronous adaptive in the
discussed in the previous sections. But it does so with downlink. So the full control information
the expense of consuming radio resources. In this has to be sent with each transmission
section, a proposal based on a previous work is leading to significant signaling overhead on
analyzed to limit the maximum HARQ PDCCH. Thus, additional resources can be
retransmission in order to save radio resources for saved by avoiding overhead signaling when
UEs in the state of poor radio link conditions. the numbers of retransmissions are reduced.

The Physical Downlink Control Channel  Even after all the 3 retransmission, more
(PDCCH) is used by eNodeB to signal about the likely chances exist for the receiver to fail in
allocation of resources to be shared by UEs. The decoding the data packet in a poor radio link
resources allocated to the PDCCH can be varied. But condition. Thus, the radio resources for the
in case of amount of resources being too small, the attempt are wasted in the retransmissions.
UL and DL data schedulers will not be able to But this scenario can be minimized if lesser
schedule all the UEs needed to be served. Whereas if retransmissions are used leading to lesser
the amount is too large, then the resources that could wastage during the poor channel link
have been used for data transmission will be wasted.
These allocations of resources to the PDCCH are B. Results
addressed in the literature [12].
The simulation was done using a LTE Link
A. Proposal and Analysis Level Simulator with poor radio link conditions.
HARQ retransmissions have been performed from 0
In order to recover data correctly, the HARQ to 3 in the poorest radio link condition of CQI values
retransmissions consume radio resources. However, 1 and 2. The following values in tables are taken
the maximum number of retransmissions in downlink from the simulated outputs of SINR versus BLER
is proposed to be limited and made less than 3 to save plots [2].
radio resources with poor radio conditions for the Analysis for CQI 1
UEs.
BLER
The following points will justify this proposal. SNR 0 tx 1 re tx 2 re tx 3re tx
-9.5 0.83 0.72 0.61 0.6
 At a poor radio link condition, an UE does
-9 0.8 0.67 0.57 0.58
not have the ability to make good use of the
resources. The uses of resources are done -8 0.7 0.53 0.5 0.5
less efficiently and thus leading to lower cell -7.2 0.6 0.43 0.42 0.42
throughput. Therefore the resources that can
be saved from reduced HARQ
-6 0.42 0.4 0.33 0.33
retransmissions can be utilized properly by -5 0.35 0.3 0.27 0.27
the UE during the good radio link and thus -4 0.25 0.2 0.2 0.2
improving cell throughput.
-3 0.15 0.15 0.15 0.15
 The eNodeB has the ability to easily -2 0.1 0.1 0.1 0.1
implement reduction of the number of -1 0.06 0.06 0.06 0.06
HARQ retransmissions in downlink. The UE
sends the CQI report to the eNodeB
0 0.04 0.04 0.04 0.04
indicating the quality of the downlink radio 1 0.02 0.02 0.02 0.02
link. Whenever the CQI value is too low, the 2 0 0 0 0
eNodeB can simply reduce the maximum
number of HARQ retransmissions. The 3 0 0 0 0
eNodeB may use a mapping of maximum 4 0 0 0 0
number of HARQ retransmissions with CQI Table 4 – SINR versus BLER in CQI 1
12

Figure 21 – Plot of the table 4 Fig 22 – Plot for table 5

CQI of 1 and 2 occupy the lowest values in


terms of radio conditions. Therefore CQI of 1 is the
Analysis for CQI 2
poorest radio link. The table 4 and table 5 shows that
BLER the Block Error Rate improves with higher
retransmissions. Same scenario can be seen in the
SNR 0 tx 1 re tx 2 re tx 3re tx graph plots of figure 21 and figure 22. But it’s not
-9.5 0.95 0.9 0.82 0.74 very substantial. Therefore, a sacrifice of this
improvement can be justified allowing saving of
-9 0.93 0.85 0.7 0.7 resources as mentioned previously.
-8 0.9 0.8 0.65 0.6
In short, the simulation result reiterates the
-7.2 0.85 0.68 0.6 0.58 proposal that has suggested earlier in [2] by this
-6 0.7 0.57 0.5 0.52 simulated model that HARQ retransmission can be
minimized to 1 or at best 2 with poor radio link
-5 0.6 0.43 0.4 0.4 especially when the network is overloaded. It can
-4 0.45 0.33 0.33 0.33 also lead to a better scheduling methodology for DL
resource allocation in LTE.
-3 0.38 0.25 0.25 0.25
-2 0.23 0.2 0.2 0.2 IV. Performance Evaluation of various
HARQ schemes
-1 0.19 0.16 0.16 0.16
0 0.1 0.1 0.1 0.1 To evaluate the performance of HARQ, we
have classified them into 4 types of schemes. The
1 0.05 0.05 0.05 0.05 Type–I is the simplest where a retransmission request
2 0.02 0.02 0.02 0.02 is sent to the transmitter through the use of CRC
(Cyclic Redundancy Check) and the corrupted packet
3 0.01 0.01 0.01 0.01 is discarded. After this, the transmitter will engage in
4 0 0 0 0 retransmitting the same packet until it is successfully
Table 5 – SINR versus BLER in CQI 2 decoded or until the maximum limit of retransmission
is reached.

Next is the Type-I with Chase Combining


where the packets with errors are stored in a buffer
and the values are combined according to the weights
of signal to noise ratio. Type-II is the Full
Incremental Redundancy which operation has been
discussed earlier in section I. Type –III is Partial IR.
13

In Partial IR the retransmitted packets are chase


combined with previous packets.

A. Channel Modeling

The Orthogonal Frequency Division


Multiple Access (OFDMA) is implemented in the
downlink of LTE. This is a multiple access scheme
which is based on OFDM where data transmission
takes place on different subcarriers to different users.
More detailed parameters for the downlink
transmission can be found in literature [15]. Figure
23 depicts the proposed model of LTE OFDMA
system employing Hybrid ARQ. As can be seen, a
Cyclic Redundancy Check (CRC) is used in the Table 6 – LTE OFDMA downlink parameters [18]
receiving end to enable error detection. And the
parameters used for this simulation is given in the The simulation involved here uses the
table 6. Spatial Channel Model Extension for 3GPP. Such
type of channel model is used by the European
WINNER project. This Model Extension utilizes 3
environments namely Suburban Macro, Urban Macro
Input data
data
Decoded

and Urban Micro. For this simulation Urban Macro


has been used.

The maximum number of retransmission is 3


CRC as discussed earlier. In, LTE 3 data modulations are
CRC
Encoder supported namely QPSK, 16QAM and 64QAM. For
Check
this simulation, 3 MCS are considered as shown in
the table 7.
Ack/Nack Turbo
Turbo
Encoder
Decoder
MCS Modulation Coding Rate
1 QPSK 0.5
Puncturing
De puncture
2 16QAM 0.75
3 64QAM 0.75
Bit- interleaving Table 7 – Modulation and Coding schemes (Edited
Bit- De
interleaving from [18])

Demodulation
Modulation As can be seen from the table 7, the lowest MCS
carries the lowest modulation as well as the lowest
coding rate. As the MCS increases, the Modulation
IFFT and Coding rate increases. For this particular
FFT
simulation only 3 MCS values have been considered

Add CP B. Enhancement of HARQ schemes:


Remove Channel
CP
The M-QAM constellation comprises of 2
Figure 23 – LTE OFDMA system model components. These are the in-phase (I) and the
[18] quadrature (Q). Using the Gray encoding, these
components are mapped to a complex symbol.

For 16-QAM, every four bits (I1Q1I2Q2)


represent a symbol of constellation which is made to
transmit over the communication channel. The
14

unequal error protection exists among the four bits of


constellation which make up the 16-QAM symbol. It can be seen how the data are shifted in
Among these constellation bits, there is high error each re-transmission. Main merit being that there is
probability in least significant bit (LSB) compared no hassle of using a buffer and only shifting
with the most significant bit (MSB). operation is done.

Taking an example, it takes 3 times more C. Simulated Metrics:


perturbation in the real or imaginary part than the 3 rd
bit of the 16 QAM constellation symbol for the first For the evaluation of performance of the
bit of the same symbol to be demodulated different schemes, 2 methods of metrics are
erroneously. Cutting the story short, the MSBs are considered, Packet Error Rate (PER) and the
three times more reliable than LSBs in inducing bit Throughput.
error. Therefore, bits are rearranged so that lower
protected bits are given high protection during the PER is defined as the residual packet error
retransmission to compensate for the above issue. rate after a maximum number of retransmission.
Thus in the table 8 the rearrangement of constellation
for 16-QAM is given Throughput is defined as the number of
correct bits per channel that are received. The
measuring unit is bits per second. It can be written in
the form of equation as:

( )
( )

Here R is the transmitted bit rate and N is


the average number of retransmissions.

Table 8 – Constellation Rearrangement for


16QAM [18] D. Simulated results:

Such a similar rearrangement can be applied Referring to the table of modulation and
to 64-QAM for achieving the maximum benefit. coding schemes, various results have been simulated.
These simulated results were summarized in the form
OFDM subcarriers in a frequency selective of a table for comparing the PER and Throughput
channel are prone to distortion. This leads to different between the various techniques. The values were
received signal quality. Fading occurs in pairs of obtained from simulated outputs [18]. The first table
subcarriers which are uncorrelated given that they are shows the gradual change in PER at a specific SNR
separated wider than coherence bandwidth in the for all the MCS values.
frequency domain. Therefore code bits are assigned
to subcarriers in retransmission to achieve frequency PER Analysis
diversity. The method is implemented by shifting the
code bits with an appropriate step which is larger
ARQ MCS 1 MCS 4 MCS 6
than the channel coherent bandwidth. Types PER SNR PER SNR PER SNR
(dB) (dB) (dB)
A sub-carrier rearrangement scheme is Type 1 -
deployed in this experiment as shown in the figure
24.
Simple ARQ 0.4 5 0.8 10 1 10
Type 1 - with
CC 0.05 5 0.2 10 0.6 10
Type 2 - Full
IR 0.02 5 0.009 10 0.006 10
Type 3 -
Partial IR 0.02 5 0.02 10 0.01 10
Figure 24 - Subcarrier Rearrangement [18] Table 9 – PER Summary
15

From Table 9, Observation can be made that


Type III and type II outperforms the rest of the type
of schemes by having the lowest PER in terms of
Throughput Analysis
SNR for all the MCS. In MCS 1, both the IR
techniques are well above CC as they occupy the ARQ MCS 1 MCS 4 MCS 6
highest decoding gain obtained over retransmission.
Types (MBPS) SNR (MBPS) SNR (MB PS) SNR
This occurs at a lower code rate as this MCS has a
code rate of ½. For coding rate ½, Full IR no longer (dB) (dB) (dB)
gains over Partial IR as both methods are successful Type 1 -
in reducing the coding rate of mother code to 1/3 Simple
with just one retransmission. Type I with CC is still ARQ 3.6 5 2.5 10 7 15
better than the type I with simple ARQ. Type 1 -
The values under MCS 4 show a clear view
with CC 5.5 5 8 10 16 15
about the behavior of the scheme as the MCS is Type 2 -
increased. Type I with CC is more prominently better Full IR 5.8 5 17 10 25 15
than the Type I with Simple ARQ. But Again the Type 3 -
Type II and Type III out performs both of these Partial
schemes. It can also be observed that Type II, being
IR 5.8 5 15 10 23 15
full IR, has a slight advantage over the Type III in
this case.
Table 10 - Throughput Summary
As the MCS is increased to the maximum
according to our table, we can see that the result For a lower value of MCS, it can be seen
becomes more prominent. Full IR has an edge and that again the Type II and Type III overcome the rest
outperforms the other HARQ schemes in terms of of the schemes in terms of throughput this time. This
reliability. This is clearly seen under the MCS 6 is shown as a summary in table 10. Type I with CC is
which is the highest MCS used. also far ahead compared to simple ARQ.

At a high coding rate (3/4), it takes 6 From the values under MCS 4 and MCS 6,
retransmissions by the Partial IR to reduce the coding Full IR occupies the highest throughput performance.
rate down to the mother code by transmitting all Partial IR comes next in throughput performance
required redundancy bits. But Full IR reduces the with less memory requirements. Chase Combining on
coding rapidly because more redundancy bits are sent the other hand has relatively poor throughput
in its transmission. performance, still better than the simple ARQ. But in
the next section of the simulated results, it can be
The PER values in all 3 simulated MCS seen that the CC scheme benefits more from the
prove that Full IR has to offer the maximum coding frequency and constellation diversity.
gain at the expense of highest buffer requirement.
Chase Combining is easier to implement compared to E. Enhanced HARQ performance:
Incremental redundancy with low memory
requirement. The type I with simple ARQ offers the Previously in the section of Enhanced
worst performance but with least complexity and no HARQ Schemes, the rearrangement of constellation
memory requirement. and the sub carrier rearrangement schemes have been
defined. Based on these 2, the performance of Chase
While considering the throughput Combining is investigated.
performance of all these schemes, we observe a
similar pattern in the conclusion compared to the
PER
16

interestingly seen that by the addition of enhanced


parameters, the performance gap between the Full IR,
PER and Throughput for Chase Combining Partial IR and CC has been reduced to approximately
2dB.
CC Types SNR(dB) PER Throughput
(MBPS)
CC 10 0.18 8 V. CONCLUSION AND FUTURE
WORK
CC - Subcarrier
Rearrangement 10 0.04 12 The variation in received signal quality depends
CC – Constellation on reflection, refraction, scattering, and diffraction.
Rearrangement 10 0.07 11 Thus the RF signals are prone to errors. The various
suitable techniques that can be used as an error
CC - Combined 10 0.01 13 detecting scheme are demonstrated in this paper with
the desired results obtained.
Table 11 – PER and Throughput for Chase
Combining LTE systems employ incremental redundancy
(IR) based hybrid ARQ with chase combining as a
The table 11 shows the comparison of special case of IR. Asynchronous Adaptive (AA)
various enhanced schemes in terms of Chase hybrid ARQ in the downlink and Synchronous
Combining for MCS 4. Clear observations can be Adaptive hybrid ARQ in the uplink is used, in regard
seen that the inclusion of constellation rearrangement to timing and adaptivity. The Round Trip Time
and subcarrier rearrangement improves the equals 8ms with 3ms processing time for each UE
performance of the system. For the same SNR, the and eNB.
PER for the combination of CC techniques
outperforms the conventional CC by a good margin. The introduction of Hybrid ARQ into the system
Even the individual performances of these techniques puts a huge impact on the system delay. It increases
are better than the CC alone. Similar variation can be the delay in the system by roughly 50ms. But it offers
spotted in the throughput of the system. gain in the throughput performance leading to
Comparatively the subcarrier rearrangement has a improvisation in the quality of service for cell-edge
slightly upper hand to the constellation users. Thus, it provides better throughput in the
rearrangement. expense of system delay.

At poor channel conditions, UE is unable to


make the best use of the radio resources. Thus the
proposal of limiting the HARQ retransmission in
poor radio conditions was stated in this paper. Results
showed that there is improvement in the BLER with
higher retransmission but is not very much
substantial. Therefore considering the points stated in
section III, it is feasible to employ retransmissions as
low as 1 or 2 during poor conditions allowing saving
of resources and leading to better scheduling
methodologies for Downlink resource allocation.

Results have shown that Incremental


Redundancy is better in terms of throughput
compared to the Chase Combining. Full IR achieves
the best performance in terms of PER and
Throughput but with the cost of high buffer
Figure 25 - PER for MCS 6 for Enhanced HARQ requirement. If we consider the cost of memory, CC
[18] will be favorable as it has a comparable performance
at a lower cost for lower MCS.
Figure 25 is the last simulated result of this
model. This graph shows the PER for Enhanced
HARQ for all the schemes in MCS 6. It can be
17

The Partial IR gives a good tradeoff between the Wireless Personal Communications, pp. 119-143,
cost of memory and performance. Thus we can 2000.
conclude the justification of LTE using this scheme.
[9] W.Stallings, Wireless Communications and
The use of frequency diversity and constellation Networks. New Jersey: Prentice Hall, 2005.
diversity along with the Hybrid ARQ schemes, offer
better throughput performance for various MCS. The [10] E. Dahlman, Stefan Parkvall, Johan Skold and
subcarrier rearrangement technique when coupled Per Beming, 3G Evolution, 2nd ed. Oxford, Great
with the constellation rearrangement is able to give a Britain: Academic Press, 2009.
significant improvement for existing Hybrid ARQ
schemes. This factor can be looked into as a future [11] J.-F. T. Cheng, "On the Coding Gain of
work for enhancing the performance and stability of Incremental Redundancy Over Chase Combining,"
error handling in radio links. GLOBECOM, pp. 107-112, 2003.

[12] P. Hosein, “Resource Allocation for the LTE


REFERENCES Physical Downlink Control Channel,” IEEE
GLOBECOM Workshops, 2009.
[1] Kumbesan Sandrasegaran, Scott Reeves, Huda
Adibah Mohd Ramli, and Riyaj Basukala, “Analysis [13] S. Lin and D. Costello, Error Control Coding,
of Hybrid ARQ in 3GPP LTE Systems”, 16th Asia- Prentice-Hall, Upper Saddle River, NJ, USA.
Pacific Conference on Communications (APCC),
2010. [14] J.M. Wozencraft, M. Horstein, Digitalised
Communication Over Two-way Channels, Fourth
[2] Mohammad T. Kawser, Nafiz Imtiaz Bin Hamid, London Symposium on Information Theory, London,
Md. Nayeemul Hasan, M. Shah Alam, and M. UK, September 1960.
Musfiqur Rahman, “Limiting HARQ
Retransmissions in Downlink for Poor [15] “3GPP; Technical Specification Group Radio
Radio Link in LTE”, International Journal of Access Network; Physical layer aspects for evolved
Information and Electronics Engineering, Vol. 2, No. UTRA (R7)’’, TR 25.814 V7.0.0 (2006-06) [Online].
5, September 2012. Available: http://www.3gpp.org/ftp/Specs/html-
info/25814.htm.
[3] 3GPP, "Evolved Universal Terrestrial Radio
Access (E-UTRA). [16] E. Dahlman, Stefan Parkvall, Johan Skold and
Per Beming, 4G LTE/ LTE-Advanced for Mobile
[4] F. Khan, LTE for 4G Mobile Broadband. New Broadband, Oxford, Great Britain: Academic Press,
York: Cambridge University Press, 2009. 2011.

[5] Chase, D., “Code combining - A maximum- [17] Kumbesan Sandrasegaran, et al., "Delay-
likelihood decoding approach for combining an Prioritized Scheduling (DPS) for Real Time Traffic
arbitrary number of noisy packets,” IEEE in 3GPP LTE System," presented at the WCNC 2010,
Transactions on Communications, vol. 33, pp. 385– Sydney, Australia, 2010.
393, May 1985.
[18] Kian Chung Beh, Angela Doufexi, Simon
Armour, “PERFORMANCE EVALUATION OF
[6] Das, A., Khan F. and Nanda, S., “A2IR: an HYBRID ARQ SCHEMES OF 3GPP LTE OFDMA
asynchronous and adaptive hybrid ARQ scheme for 3G SYSTEM”, The 18th Annual IEEE International
evolution,” 53rd IEEE VTS Vehicular Technology Symposium on Personal, Indoor and Mobile Radio
Conference, VTC 2001 Spring, Vol. 1, pp. 628–632, 2001. Communications (PIMRC’07)

[7] X. Qiu and K. Chawla, "On the Performance of [19] Christopher Cox, An Introduction to LTE, Wiley
Adaptive Modulation in Cellular Systems," IEEE Publication.
transactions on communications, vol. vol 47, pp.
884-895, 1999.

[8] M. Alouini and A. J. Goldsmith, "Adaptive


Modulation over Nakagami Fading Channels,"

Das könnte Ihnen auch gefallen