HARQ Processes

HARQ uses a stop and wait protocol. When a transmission has been made, the transmitting entity stops and waits until it receives an acknowledgment (ACK) or negative acknowledgement (NACK) back from the destination before transmitting the next block of data or retransmitting the same data block. In either case (ACK or NACK), the transmitting entity is required to schedule and process the next transmission within a specific time period.

For LTE frequency-division duplex (FDD) on the uplink, this time has been set to eight 1-ms subframes. Since it only takes one subframe to transmit the data, this results in seven subframes of unutilized bandwidth. To fully utilize this bandwidth, LTE uses multiple HARQ parallel processes offset in time from each other. Each process transmits a block of data. By the time its next transmission allocation arrives, it will have already received the ACK or NACK from the receiving entity and created the next packet for (re)transmission (Fig. 1).

It should be noted that LTE time-division duplex (TDD) supports a configurable number of HARQ processes and the timing requirements aren’t fully defined as of 36.213 v8.3.0.

For FDD, there are exactly eight uplink HARQ processes, while the downlink can have up to eight. Downlink HARQ processes can be transmitted in any order without fixed timing (asynchronous HARQ), whereas each uplink HARQ process is assigned to a specific subframe (synchronous HARQ). The user equipment (UE) transmits within the same HARQ process every eighth subframe.

Synchronous Versus Asynchronous HARQ

LTE uses asynchronous Type-II HARQ transmission on the downlink. This means the receiver doesn’t know ahead of time what’s being transmitted (or when), so the HARQ process identifier and the RV must be sent along with the data. The RV specifies which combination of data, ED, and FEC bits is being sent to the UE.

This is done through the physical downlink shared channel (PDSCH) resource allocation messages sent on a physical downlink control channel (PDCCH) simultaneous to the corresponding PDSCH transmission. The advantage of this scheme is that the scheduling algorithm has considerable freedom in deciding which UEs are sent data during any subframe (Fig. 2).

In contrast, LTE uses synchronous HARQ transmission on the uplink. This means that the eNodeB knows exactly which HARQ process and RV the UE will transmit ahead of time. So, this information doesn’t have to be included in the PDCCH message providing the uplink scheduling information to the UE.

Synchronous HARQ can be used because the UE transmits the same HARQ process every eighth subframe. Because retransmissions of a HARQ process are associated with previous transmissions based on the eight-subframe delay, the scheduling in the uplink is not quite as flexible as that in the downlink.

Adaptive Modulation and Coding

Adaptive modulation and coding (AMC) attempts to match the transmissions from a HARQ process to the channel conditions. Under strong signal conditions, less redundancy and/or a higher-order modulation format is employed in the initial transmission, enabling a higher user data rate for a given bandwidth. Under weak signal conditions, more redundancy bits are used and/or a lower-order modulation format is used to improve the probability of reception.

However, this lowers the user data rate. If the error rate is zero, then it is likely that too much protection is being applied. Alternatively, if insufficient protection is applied, the same data will be retransmitted, effectively wasting valuable network resources. The ideal situation is where data throughput is maximized at an error rate that is relatively low but greater than zero.

As with W-CDMA, the LTE eNodeB decides on the modulation coding scheme (MCS), depending on information the UE sends in the channel quality indicator (CQI). LTE, though, is more complicated than W-CDMA/HSPA, which requires only a single CQI value to be transmitted.

Unlike WCDMS/HSPA, LTE allows a single shared channel transmission to occur on a subset of the possible subcarriers. Given the wide bandwidths provided by LTE, some of the subcarriers could be in fading nulls at the same time others can be received clearly. CQI information from the UE provides channel information measured per subband or wideband.

The network’s scheduling algorithm can use subband CQI to assign subchannel resources for optimum transmission. If the channel characteristics change considerably after the initial transmission, the Medium Access Control (MAC) layer is free to assign a different set of subchannels on subsequent transmissions, or to abort that transmission and start a new one using a more appropriate modulation and coding scheme.

LTE uses a clever algorithm to implement incremental redundancy and adaptive coding (Fig. 3). The systematic bits from the turbo encoder are interleaved and placed into a circular buffer called the soft buffer. The redundancy bits are then interleaved and placed after the systematic bits. All the redundancy bits are included in the soft buffer used on uplink transmissions, but upper layers define the number of redundancy bits included for downlink transmissions.

Bits are copied from the buffer starting at a position that depends on the RV. The starting position for RV_n is approximately n/4 of the way around the circular buffer, plus a fixed offset of two interleave rows. The number of bits pulled from the circular buffer for each RV depends on the target code rate.

For poor channel conditions, the code rate approaches 0.1, in which case the entire soft buffer is transmitted multiple times each RV. In excellent channel conditions, the code rate approaches 0.92, which means the number of bits transmitted in each RV is slightly more than the number of bits in the transport block.