Hybrid Automatic Repeat Request in 3G Long Term Evolution

Zheng Kan,Wang Ling,Wang Wenbo

(School of Telecommunications Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, China)

Abstract:The standardization of 3G Long Term Evolution(LTE)has widely progressed and has received extensive attention.Various advanced wireless transmission technologies,including Hybrid Automatic Repeat Request(HARQ),will be adopted in the system.HARQ can be classified as being synchronous or asynchronous by different instants of retransmission.Synchronous HARQ implies that retransmissions for a certain HARQ process are restricted to occur at known time instants,and the process number can be derived from the sub-frame.For asynchronous HARQ,transmissions for a certain HARQ process may occur at any time.Explicit signaling of the HARQ process number is therefore required.HARQ has become one of the critical techniques of 3G LTE because it can effectively compensate the effects of time variation and multi-path fading of wireless mobile channel on signal transmission.It will be better-designed with the development of 3G LTE.

I n the pursuit of meeting the demand for higher-speed services,3GPP kicked off the 3G Long Term Evolution(LTE)standardization work in December 2004［1］.The purpose of promoting LTEis to retain the competitiveness of 3G system in the next 10 years,and achieve all-around improvement in its performance and functionality.Major factors that 3G LTE pores over include lower time delay,higher subscriber data rate,larger system capacity and coverage,and reduced operational cost.Therefore,the radio interface and wireless network architecture should be improved to meet such requirements.

In order to combat time variation and multi-path fading of the wireless mobile channel on signal transmission,3G LTE adopts the error control methods based on Forward Error Correction(FEC)and Automatic Repeat Request(ARQ)to reduce the bit error rate of the system and guarantee the quality of service.Although the FEC scheme allows a low time delay,it suffers a decrease in throughput due to code redundancy generated;ARQ achieves an ideal throughput in the case of low bit rate,but it will generate higher time delay,and is unsuitable for delivering real-time services.Hybrid Automatic Repeat Request(HARQ)is a combination of these two methods,developed to overcome their drawbacks.In the HARQ scheme,the ARQ system contains an FEC sub-system.ARQ will not be applied if the error correction capability of FEC is able to correct the errors;the ARQ feedback channel will be used to send a retransmission request of the bit error group only when FEC fails to correct the errors normally.

The effective combination of ARQ and FEC can not only achieve higher reliability than using only the FEC system,but it also has higher system throughput than using only the ARQsystem.As demands for high data rate and highly reliable services are soaring,HARQ has become one of the criticaltechnologies for wireless communication system,and therefore,it is being researched thoroughly.It willdefinitely be applied in the 3G LTEsystem.

1 ARQ Protocols

Typical ARQ protocols include Stop-And-Wait(SAW),Go-Back-N(GBN)and Selective-Repeat(SR)［2］.

1.1 SAW

The transmitter sends a data packet,and stops to wait for acknowledgement(ACK)from the receiver.The receiver checks the data packet,and sends back ACK if it received the data packet correctly;otherwise,it sends back negative acknowledgement(NACK).Upon receiving the ACKsignal,the transmitter sends new data,or it resends the previous data packet.The channelis idle while awaiting acknowledgements,without transmitting any data.

This scheme is simple to implement because the receiver and the transmitter operate the same data packet at the same time,and it requires little signaling overhead and low receiver buffer.

However,no data sent while waiting for acknowledgements leads to huge waste of resources,especially when the time delay of channel transmission is very long.Thus,the SAWmode results in low channel utilization and low system throughput.

A simple demonstration of SAWARQ is shown in Figure 1.

1.2 GBN

In the transmission system adopting GBN ARQ protocol,when a data packet is sent,the transmitter will go on sending other data packets without waiting for an ACK.The receiver feedbacks the ACK or NACK related to each data packet,accompanying the grouping numbers of the data packets,to the transmitter.Upon receiving a NACK,the transmitter will resend the N data packets in the group containing erroneous data packet,as shown in Figure 2.The receiver then receives the data packets in sequence,and when it receives an erroneous data packet,even if followed by a correct data packet,it has to discard the correct data packet,and resend ACK.

▲Figure 1. Operating demonstration of SAW ARQ protocol.

▲Figure 2. Operating demonstration of GBN ARQ protocol.

Compared with SAW,GBNimproves system throughput because the transmitter sends data packets continuously while also increasing system signaling overhead for the extra data packet number.Besides,as the receiver only receives data in sequence,data previously transmitted correctly have to be resent(due to an error data packet occurring before these data packets are grouped).Therefore,as a result,channel utilization is decreased.

1.3 SR

In order to improve channel utilization,the SRprotocol only resends the erroneous data packets,and the receiver won't receive data packets in sequence.

Therefore,sufficient buffer size at the receiver is required to store the decoded packet groups that are unable to be sent out in sequence.Because the receiver must know the sequence numbers prior to combining data packets,the encoding of the sequence numbers has to be separated from that of the data and be more reliable to overcome whatever error occurs in the data at any time.This leads to higher requirement for signaling.

By comparison,SRhas the highest channel utilization while it demands the largest buffer size and signaling overhead.Asimple demonstration of SR ARQ is shown in Figure 3.

The 3G LTEsystem will adopt the SAWprotocol.It is simple and reliable,has small signaling overhead,and requires less buffer size at the receiver.However,this protocol has low channel utilization efficiency.To avoid such disadvantage,3G LTEemploys N-channel SAWprotocol,in which,the transmitter operates in parallel N sets of different SAWprotocols in the channels and utilizes the gaps between different channels to send data and signaling alternately.N-channel SAWimproves channelutilization.

2 Basic HARQ Types

According to the content of retransmission,HARQ is divided into three main schemes in 3GPPstandards and recommendations［3］:HARQ-I,HARQ-IIand HARQ-III.

2.1 HARQ-I

HARQ-Iis the traditional HARQ scheme.It merely introduces the error correcting code into ARQ,that is,to-be-sent packets are added with cyclic redundancy check(CRC)bits and are encoded using FEC.The receiver decodes the packets using FEC and conducts CRC verification.The receiver discards erroneous packets,feedbacks NACK to the transmitter and asks for resending the data packet identical to the previous frame.

Generally,in order to prevent continuous retransmissions of data packets from a certain subscriber due to long-term slow fading of the channel,the restriction on maximal retransmission times is set at the physical layer to avoid huge waste of channel resources.If the receiver still fails to correctly decode and reaches the maximal retransmission times(3 in 3G LTEsystem),the data packet is determined as transmission error and is discarded.Then the receiver notifies the transmitter to send a new data packet.

This scheme simply discards erroneous data packets,without fully utilizing their useful information of previous transmission.Therefore,the performance of HARQ-Imainly relies on the error correction capability of FEC.

▲Figure 3. Operating demonstration of SR ARQ protocol.

2.2 HARQ-II

HARQ-IIis also called incremental redundancy scheme.In this scheme,the information bit is encoded,and then the encoded check-bit is punctured in a certain cycle and is transmitted in sequence to the receiver according to the bit rate compatibility rule.Instead of being discarded,the erroneous packets received are combined with additional resent packets for subsequent decoding;the resent data,with additional redundancy information,is not a simple duplicate of the previously sent data.The receiver implements combined decoding at each retransmission,and all bits previously received are combined with lower bit rate codes,thus increasing the coding gain and achieving incremental redundancy.Redundancies at each retransmission are different,and the resent data can only be decoded after combining with the received data.

2.3 HARQ-III

HARQ-IIIis an improved incremental redundancy retransmission scheme.It employs complementary deletion of data packets sent.The data packets can be decoded separately,or can be combined with a coding packet with larger redundancy information for combined decoding.

Based on different redundancy versions of retransmission,HARQ-IIIis further divided into two types.One is HARQ-IIIwith a single redundancy version.All retransmission redundancy versions are identical to the first transmission,implying that the formats and contents of the retransmitted packets are the same as those of the first transmission.The decoder at the receiver combines these multiple copies of the transmitted packet weighed by the received Signal-to-Noise Ratio(SNR).

Time diversity gain is thus obtained.The other is HARQ-IIIwith multiple redundancy versions.All retransmission redundancy versions are different.The deletion method for encoded redundant bits is well designed to make the deleted codes complementary and equivalent.As a result,the combined codes can cover the bits in FEC code,thus enabling the decoded information to be more comprehensive and to facilitate correct decoding.

Figure 4 shows the transmission processes of various HARQ schemes when the encoding rate is Rc=2/3.

In 3GPP,all member companies have reached some preliminary consensus:the uplink or downlink in the 3G LTE system will adopt the HARQ retransmission policy based on Incremental Redundancy(IR)retransmission scheme;the system also supports Chase Combining(CC),seen as a specialcase of IR.

▲Figure 4. Specific transmission processes of various HARQ schemes, with Rc =2/3.

3 Synchronous and Asynchronous HARQ

HARQ can be classified as being synchronous or asynchronous by different instants of retransmission.Synchronous HARQ implies that(re)transmissions of a HARQ process are restricted to occur at known time instants.As the receiver can envision the transmission instant,the sequence number of HARQ process can be obtained from the sub-frame number,eliminating the need for extra signaling overhead for indication.Asynchronous HARQ implies that(re)transmissions for a certain HARQ process may occur at any time.As the receiver can not envision the transmission instant,the HARQ processing sequence number must be transmitted accompanying the data.

As retransmission in synchronous HARQ occurs at a fixed time,the synchronous HARQ without additional processing sequence numbers can only support one HARQ process at a time.In principle,HARQ operation may support the occurrence of multiple HARQ processes at the same time;thus,extra signaling overhead is needed for synchronous HARQ to indicate the sequence number of HARQ process.However,asynchronous HARQ itself may support transmissions of multiple processes simultaneously at a time instant.

Additionally,in a synchronous HARQ scheme,the transmitter may choose not to fully utilize all retransmission instants.For example,in order to support the HARQ process with higher priority,the process pre-allocated to the instant should be stopped.Extra signaling is also needed,as a result.

HARQ schemes can further be classified into non-adaptive and adaptive in terms of retransmission attributes.The transmission attributes include Resource Block(RB)allocation,modulation mode,transport block size,and transmission duration.Adaptive transmission implies that the transmitter may change some transmission parameters according to practical channel state information during each retransmission.Therefore,the control signaling message containing the transmission parameters needs to be transmitted together with retransmitted packets.The changeable parameters are modulation mode,resource block allocation,and transmission duration.Non-adaptive transmission implies that all these transmission parameters are known to both the receiver and transmitter in advance.Therefore,it is unnecessary to send the control signaling message containing the parameters for retransmission in a non-adaptive system.

During retransmission,the transmission scheme,which can adaptively change the retransmitted packet formats and retransmission time instants according to channel environment,is called IR-based asynchronous adaptive HARQ operation.In this scheme,resources can be effectively allocated based on the environmental characteristics of a time-varying channel.It features flexibility and results in more system complexity.In each retransmission,the control signaling message containing the transmission parameters must be transmitted along with the data packets.

This will lead to extra signaling overhead.As for synchronous HARQ,the retransmitted packet formats and retransmission time instants are known in each retransmission,eliminating the need for extra signaling information.

Compared to asynchronous HARQ,synchronous HARQ has the following advantages:

(1)Control signaling overhead is small because the parameters are known in advance in each transmission process,and it is accordingly unnecessary to indicate the HARQ process number.

(2)The operation at the receiver has lower complexity in the non-adaptive system.

(3)The reliability of the control channel is improved.In the non-adaptive system,the signaling information of the control channel in retransmission is identical to that in the initialtransmission,in some cases.Therefore,soft information combination may be implemented at the receiver to enhance the decoding performance of the control channel.

According to the actual L1/L2 requirements,asynchronous HARQ may best address the following issues:

(1)Scheduling flexibility if fully adaptive HARQ technique is adopted,and if both localized and distributed subcarrier allocations are used

(2)Support for multiple HARQ processes at the same sub-frame

(3)Flexibility in scheduling of retransmissions

In the 3G LTEsystem,it has been decided that the downlink system in Evolved Universal Terrestrial Radio Access(E-UTRA)network will adopt the asynchronous adaptive HARQ technology［4］.In comparison to the synchronous non-adaptive HARQ technology,asynchronous HARQ can fully utilize the channel state information,which helps improve system throughput.Furthermore,asynchronous HARQ can avoid performance degradation resulting from resource allocation conflicts during retransmission.

Based on the strengths of asynchronous HARQ mentioned,and taking the problem of heavy signaling overhead into account,many vendors have proposed their methods to reduce the overhead of asynchronous HARQ system.For example,Samsung［5］proposed that if retransmission time instants and transmission parameters have changed,relevant control signaling must be transmitted;otherwise,similar to synchronous HARQ,the signaling message related to the transmission parameters does not have to be transmitted.When system performance is guaranteed,signaling overhead will be lowered.

As for its uplink,the 3G LTEsystem will adopt synchronous non-adaptive HARQ technology［6］.Although the asynchronous adaptive HARQ technology is more flexible in scheduling than synchronous non-adaptive HARQ,the latter requires less signaling overhead.The uplink is complex,and interferences from other cells are uncertain.Therefore,the base station is unable to precisely estimate the actual Signal to Interference plus Noise Ratio(SINR)values of various users.

For the Adaptive Modulation and Coding(AMC)system,on one hand,the system selects an appropriate modulation and coding scheme according to channel quality,and roughly selects the data rate;on the other hand,HARQ accurately adjust the coding rate based on channel conditions.The uncertainty of SINR values leads to inaccurate selection for Modulation and Coding Schemes(MCS)by the uplink;therefore,the HARQ technique is relied more to ensure reliable system performance.

As a result,the average transmission times for uplink are higher than that for downlink.In consideration of the control signaling overhead,the uplink prefers the synchronous non-adaptive HARQ.

4 Conclusions

With increasing demands for high-speed wireless multimedia services and shortage of wireless frequency spectrum resources,exploration for future highly-efficient mobile communications system becomes more and more important.The HARQ technology can well complement the effects of time variation and multi-path fading in wireless mobile channel on signal transmission.