2. BACKGROUND
2.2 LTE
2.2.4 HARQ
HARQ (hybrid-ARQ) is a technique combined with ARQ (Automatic
Repeat-reQuest) and FEC (Forward Error-Correction). Through ARQ, packets’ error can be detected, and those damaged packets may be recovered via FEC. HARQ provides more reliable data transmission.
In LTE, HARQ can be divided into two parts, HARQ entity and HARQ process, and HARQ entity maintains many HARQ processes. We will detail HARQ entity first and HARQ process second.
2.2.4.1 HARQ Entity
The unit which HARQ entity handles is TB, and HARQ entity will
dispatch/receive TBs to/from HARQ processes. HARQ entity also directs HARQ information which contains TB size, HARQ process ID, NDI, RV, etc. to the corresponding HARQ process, then the designated HARQ process follows HARQ information to transmit/receive data. In other words, HARQ entity controls HARQ processes’ activities, and HARQ processes just execute on the commands. Figure 2-13 depicts the relationship between HARQ entity and HARQ processes.
Figure 2-13 Relationship between HARQ entity and HARQ processes
2.2.4.2 HARQ Process
HARQ process can be divided into two groups, and each is corresponding to receiving side and transmitting side. There are many processes in each group and those processes are independent.
Receiving side processes reserve incoming codeword until the codeword can be decoded correctly by CRC and FEC. If the codeword cannot be decoded correctly, receiver saves the codeword in the rx process as figure 2-13, and replies HARQ NACK to sender. When receiving retransmission of the same codeword with different RV, the previous incorrectly decoded codeword will be combined with the
retransmission codeword, generating a more robust codeword to be decoded again. Rx process will repeat the above procedures which include combining codeword,
decoding, replying HARQ NACK until decoding codeword correctly, then delivers the resulting TB to the upper layer and replies HARQ ACK to the sender.
On transmitting side, HARQ processes add CRC, FEC on TB to generate codeword, and then deliver the codeword to the next layer. After sending codeword, HARQ process waits for ACK/NACK reply, if it receives ACK, it discards the saved TB and get a new TB to transmit, otherwise, it retransmit the original TB with
18
different RV as figure 2-13, until receiving ACK or retransmission times exceeds retransmission threshold, and reports this situation to upper layer to handle this problem.
2.2.4.3 HARQ Process Number
HARQ Process Number is related to HARQ RTT, which depends on processing time and DL/UL configuration. DL/UL configuration is shown is Table 2-1.
Processing time is approximately 3 ms, and HARQ RTT is calculated by the longest time interval between two transmissions on the same HARQ process.
Take TDD DL/UL configuration 0 and FDD for example separately. Figure 2-14 demonstrates different situations on HARQ RTT. For FDD, HARQ RTT is constantly 8 ms. At 1st ms eNB sends downlink data and meanwhile UE receives the data within 1 ms, after 3 ms processing, UE sends HARQ ACK at 5th ms, then eNB receives HARQ ACK and then waits 3 ms processing time to decode. Until 9th ms eNB can know the decode results then decide to do retransmission or send the next TB by the same HARQ process. From 1st ms to 9th ms, there are 8 blanks in this period without any downlink transmission, so we need 8 HARQ processes to fill the blanks for both downlink and uplink on UE and eNB, and it will be the same situation on uplink. For TDD DL/UL configuration 0, although HARQ RTT varies, the longest HARQ RTT is 11 ms. In different situations UE sends uplink data at 1st and 2nd ms, and eNB should sends back HARQ ACK at 5th and 6th ms respectively, however 5th and 6th time slot both belong to UL, so in both situation, eNB can only send back HARQ ACK at 8th ms, resulting in that the next time for issuing an uplink transmission on the same HARQ process is at 12th ms. Here we see the reason why the RTT varies. Then we take the longest HARQ RTT as our base to assign TX/RX HARQ process number.
From 1st to 12thms, there are 7 uplink slots and 4 downlink slots. In UE, the number of
TX HARQ PROCESS is 7, number of RX HARQ PROCESS is 4, and for eNB, it is just the opposite on UE.
Uplink-downlink
Table 2-1 TDD DL/UL configuration
Figure 2-14 HARQ RTT
TX HARQ PROCESS use stop-and-wait mechanism, each TX HARQ PROCESS is independent. The TX HARQ PROCESS which completes its
transmission first will retrieve the next TB in DLSCH. We make an example in Table 2-2, supposing using FDD, so there are 8 TX HARQ PROCESS and HARQ RTT is 8 ms. In table 2-2, the left column is the situation of tx process, showing that a TB Y is sent by a tx process #P N times. On the contrary, right column is the situation of rx process, showing that a TB Y is received by rx process #P, and D is decode results.
20
Notice that tx process #0 and #3 both failed at the first transmission in TTI #0 and TTI
#3, respectively, so the both tx process should issue a retransmission after a HARQ RTT, until TTI #9 and TTI #11, respectively. At this time, tx process #0 and #3 both have a successful retransmission, so they can get a new TB for next transmission after a HARQ RTT to TTI#17 and TTI#19.
Table 2-2 HARQ Process Stop-and-Wait Scenario