Hybrid automatic repeat request (HARQ) is an important component of LTE and its evolution. In this blog, we shall present the salient features of the HARQ with respect to TD-LTE. Significant differences exist between the HARQ implementation in FDD systems as compared to the TD-implementation. We will highlight the differences and also discuss impacts of some of the HARQ features unique to TD-LTE. We will first explore the various timing related differences between FDD and TD-LTE. We will then outline the difference in the maximum number of HARQ processes in both systems. Finally, we shall address ACK bundling and multiplexing in TD-LTE.
In TD-LTE, the HARQ process is different from that in the FDD implementation of LTE. In the case of FDD, for a transmission on subframe #n, an ACK/NACK message is sent on subframe # n+4. The reason for the 4 subframe delay in the transmission of a ACK/NACK message is due to the processing delay of about 3 ms at the receiver. If it is a NACK, the retransmission is scheduled on subframe # n+8 for UL transmissions while the DL retransmission can be asynchronous.
In TD-LTE, the time association between the data transmission and the ACK/NACK cannot be maintained due to the variable numbers of DL and UL subframes being present in a frame. The UL and DL delay between data and ACK is dependent on the TDD configuration chosen. Hence, a fixed delay between a transmission and the HARQ ACK/NACK is not possible in TD-LTE.
In TDD, the delay between the transmission and the HARQ ACK/NACK depends on both the TDD configuration and the subframe in which the data was transmitted. A fixed delay cannot be assured as subframes are allocated to DL and UL depending on the configuration. For example, in configuration 1 which is shown below in Figure 1, there are some DL subframes for which the nearest UL subframe (greater than a separation of 4 or more subframes) is 7 subframes away. In the Figure 1 shown below, this can be seen clearly for the DL data transmission case.
The tables 1 and 2 below show the ACK/NACK delays for the DL and UL case respectively when the original data transmission occurred in a certain subframe no. n. It can be noticed that the delays are dependent on the configuration and the subframe in which the original data transmission has occurred. For example, in configuration1, the delay from the eNB side can range from 4-6 subframes while the delays from the UE side can be from 4-7 subframes. These tables have been fixed by the standard and all implementations are expected to follow these numbers for the respective cases.
Hence, the delay for transmission of ACK/NACK can range from 4 subframes (which is the same as in FDD) to 7 subframes for the DL HARQ and from 4 subframes to 13 subframes for the UL HARQ. The higher number in the UL case is due to the larger DL asymmetry in the configuration.
Number of HARQ Processes:
In FDD systems, multiple HARQ processes can be activated in both DL and UL directions. Due to the fixed timing associations between the original and the retransmission on the UL, the maximum number of HARQ processes has been fixed at 8. However, in TDD due to the variable delays experienced with respect to ACK/NACK arrivals as well as the retransmissions, the numbers of HARQ processes are dependent on the configuration. The maximum number of HARQ processes allowed per configuration for DL and UL is clearly specified in the table 3 below. Typically, the retransmissions can be scheduled 4 subframes (or more) after the reception of a NACK message.
A sample scenario is shown below for configuration 1 (both the DL and UL HARQ) to highlight the number of HARQ processes possible in both cases. It can be seen that the number of HARQ processes is dependent on various factors like DL/UL and the configuration used.
ACK Bundling and Multiplexing:
In TD-LTE, the number of DL and UL subframes in a frame is not equal. For instance, configurations 0-5 have more DL subframes as compared to UL subframes. Consequently, data transmission from multiple DL subframes needs to be acknowledged in a certain UL subframe and vice-versa for certain cases for the UL data transmission. The transmission of multiple ACK/NACK messages in UL or DL subframes is a unique feature of TD-LTE as compared with FDD due to the above mentioned scenario
For example, in configuration 1, DL data in SF#0 and SF#1 are ACKed in SF#7 to avoid further delays. In each configuration, the ACK/NACK response window is clearly specified keeping in mind both the delay as well as balance in the number of ACK/NACKS to be sent in a subframe. Two methods have been specified in the standard to deal with this issue and they are ACK bundling and ACK multiplexing, respectively. For the eNB, transmission of multiple separate ACKs/NACKs is not a major problem as the eNB is not power constrained and can send multiple ACK/NACKs in a subframe. Such a method wherein separate ACKs/NACKs are sent (in the same subframe) for transmissions in different subframes is called as ACK multiplexing. Thus, in DL ACK/NACK transmission, it is by default ACK/NACK multiplexing.
However, in the UL, the UEs might be power constrained depending on their location within the cell. For UEs which are not constrained by power and coverage multiple PUCCH messages relating to each HARQ process can be generated, i.e., separate ACK/NACK messages can be sent. For UEs in coverage constrained situations, ACK bundling is used as fewer bits need to be sent in this case. Typically, an AND operation on all the ACK/NACK messages is performed and sent. For example, in configuration no.1, data in SF# 0 and SF# 1 are acknowledged in SF# 7. A logical AND of these messages is taken and sent. As shown figure 5, if there is a ACK and a NACK, a NACK is sent. This implies that both transport blocks will be retransmitted by the eNB as it does not have individual ACK status of the DL transmission. This is the disadvantage of ACK bundling as more retransmissions than what is necessary might be sent. The main advantage of the ACK bundling method is due to the fact that the limited power resources of a UE will now be used only in a fewer resource blocks as compared to a larger number of resource blocks (RBs). This can improve the UL coverage of the UE as there will be more power per RB that will be available to the UE in the case of ACK bundling.
In summary, HARQ in TD-LTE has significant differences as compared to FDD. In general, the HARQ round trip times can be larger in TD-LTE especially for some strongly asymmetric configurations as compared to FDD. Some results have shown that heavy DL asymmetric configurations need not translate into proportionately higher DL downlink performance. Moreover, the delays can also vary between DL and UL for the same configuration. Typically, the delay performance of TD-LTE is expected to be worse than that of a FDD system. The maximum number of HARQ processes varies between DL and UL and with configuration and can be larger than the number available in FDD.