Description
Hybrid Automatic Repeat Request (HARQ) is a fundamental error control mechanism employed in the physical layer of 3GPP radio access technologies, including UMTS (HSPA), LTE, and 5G NR. It operates by integrating two classical error control methods: Forward Error Correction (FEC) and Automatic Repeat Request (ARQ). The 'hybrid' nature stems from this combination. In operation, the transmitter sends a data packet encoded with FEC. The receiver attempts to decode it. If decoding fails, instead of discarding the corrupted packet, the receiver stores it and sends a Negative Acknowledgement (NACK) back to the transmitter. Upon receiving a NACK, the transmitter sends a retransmission. The receiver then combines the soft information (e.g., log-likelihood ratios) from the initial transmission and the retransmission before attempting decoding again. This process, known as soft combining, significantly improves the probability of successful decoding compared to treating each transmission independently.
HARQ is implemented using multiple parallel processes, known as HARQ processes, to maintain continuous data flow. Each process handles the transmission and potential retransmission of one transport block. While one process is waiting for an acknowledgement (ACK/NACK), another process can be transmitting new data, thus hiding the round-trip time latency. The protocol is managed by the Medium Access Control (MAC) layer, which handles the generation of ACK/NACK feedback, scheduling of retransmissions, and management of the HARQ buffers. The physical layer is responsible for the actual encoding, modulation, and the soft combining operation.
Key variants include Chase Combining, where identical copies of the packet are retransmitted, and Incremental Redundancy (IR), where each retransmission contains different parity bits, effectively increasing the code rate with each attempt. HARQ is tightly coupled with adaptive modulation and coding (AMC). The initial transmission uses a modulation and coding scheme (MCS) selected based on channel quality indicators (CQI). HARQ provides a second line of defense if the channel degrades unexpectedly after the MCS is selected. Its role is absolutely critical for achieving the high reliability and spectral efficiency targets of modern cellular systems, as it allows the system to operate closer to the capacity limit of the channel by efficiently recovering from errors.
Purpose & Motivation
HARQ was created to address the fundamental challenge of reliable data transmission over inherently unreliable and time-varying wireless channels. Traditional ARQ schemes, which simply discard erroneous packets and request retransmissions, are inefficient for wireless links due to high latency and wasted bandwidth. Pure FEC schemes, which add heavy redundancy to correct errors, become inefficient under good channel conditions. The purpose of HARQ is to synergistically combine the best of both: the proactive error correction capability of FEC to handle common channel variations, and the reactive error recovery of ARQ to handle deep fades or unexpected interference, but in a much more efficient manner than standalone ARQ.
Its introduction in 3GPP Release 5 with High-Speed Downlink Packet Access (HSDPA) was a pivotal moment for enabling high-speed mobile broadband. Prior 3G systems relied on RLC-layer ARQ, which had higher latency and was less efficient for real-time services. HARQ, operating at the physical/MAC layer with much shorter round-trip times, drastically reduced retransmission delay and improved throughput. This was essential for supporting latency-sensitive applications like voice over IP and interactive video. The evolution through LTE and 5G NR has further refined HARQ to support more complex scenarios like carrier aggregation, massive MIMO, and ultra-reliable low-latency communication (URLLC), where its fast and reliable error correction is a cornerstone technology.
Key Features
- Combines FEC and ARQ for efficient error control
- Employs soft combining (Chase Combining or Incremental Redundancy) at the receiver
- Utilizes multiple parallel HARQ processes to hide round-trip latency
- Operates at the MAC/Physical layer for minimal retransmission delay
- Works in tandem with Adaptive Modulation and Coding (AMC)
- Supports both synchronous and asynchronous operation modes
Evolution Across Releases
Introduced with HSDPA in UMTS. Established the foundational HARQ architecture with multiple stop-and-wait processes, Chase Combining, and tight integration with the Node B scheduler for fast retransmissions in the downlink, marking a shift from RLC-layer ARQ to physical-layer retransmissions.
Enhanced for LTE, introducing support for both Chase Combining and Incremental Redundancy (IR). Defined synchronous, adaptive HARQ for the uplink and asynchronous, adaptive HARQ for the downlink, providing greater scheduling flexibility. Integrated with OFDMA/SC-FDMA.
Fundamentally redesigned for 5G NR to support diverse numerologies and service requirements (e.g., eMBB, URLLC). Introduced a fully asynchronous and adaptive HARQ design for both uplink and downlink. Enhanced feedback mechanisms with support for multi-bit HARQ-ACK and increased number of HARQ processes to accommodate wider bandwidths and shorter slot durations.
Defining Specifications
| Specification | Title |
|---|---|
| TS 21.905 | 3GPP TS 21.905 |
| TS 25.101 | 3GPP TS 25.101 |
| TS 25.212 | 3GPP TS 25.212 |
| TS 25.221 | 3GPP TS 25.221 |
| TS 25.222 | 3GPP TS 25.222 |
| TS 25.301 | 3GPP TS 25.301 |
| TS 25.302 | 3GPP TS 25.302 |
| TS 25.308 | 3GPP TS 25.308 |
| TS 25.309 | 3GPP TS 25.309 |
| TS 25.319 | 3GPP TS 25.319 |
| TS 25.321 | 3GPP TS 25.321 |
| TS 25.331 | 3GPP TS 25.331 |
| TS 25.420 | 3GPP TS 25.420 |
| TS 25.427 | 3GPP TS 25.427 |
| TS 25.430 | 3GPP TS 25.430 |
| TS 25.766 | 3GPP TS 25.766 |
| TS 25.823 | 3GPP TS 25.823 |
| TS 25.912 | 3GPP TS 25.912 |
| TS 26.267 | 3GPP TS 26.267 |
| TS 26.268 | 3GPP TS 26.268 |
| TS 26.926 | 3GPP TS 26.926 |
| TS 28.841 | 3GPP TS 28.841 |
| TS 36.104 | 3GPP TR 36.104 |
| TS 36.116 | 3GPP TR 36.116 |
| TS 36.117 | 3GPP TR 36.117 |
| TS 36.133 | 3GPP TR 36.133 |
| TS 36.141 | 3GPP TR 36.141 |
| TS 36.201 | 3GPP TR 36.201 |
| TS 36.216 | 3GPP TR 36.216 |
| TS 36.300 | 3GPP TR 36.300 |
| TS 36.302 | 3GPP TR 36.302 |
| TS 36.306 | 3GPP TR 36.306 |
| TS 36.314 | 3GPP TR 36.314 |
| TS 36.322 | 3GPP TR 36.322 |
| TS 36.331 | 3GPP TR 36.331 |
| TS 36.747 | 3GPP TR 36.747 |
| TS 36.791 | 3GPP TR 36.791 |
| TS 36.825 | 3GPP TR 36.825 |
| TS 36.863 | 3GPP TR 36.863 |
| TS 36.938 | 3GPP TR 36.938 |
| TS 37.105 | 3GPP TR 37.105 |
| TS 37.901 | 3GPP TR 37.901 |
| TS 38.133 | 3GPP TR 38.133 |
| TS 38.201 | 3GPP TR 38.201 |
| TS 38.212 | 3GPP TR 38.212 |
| TS 38.331 | 3GPP TR 38.331 |
| TS 38.521 | 3GPP TR 38.521 |
| TS 38.551 | 3GPP TR 38.551 |
| TS 38.762 | 3GPP TR 38.762 |
| TS 38.811 | 3GPP TR 38.811 |
| TS 38.824 | 3GPP TR 38.824 |
| TS 38.830 | 3GPP TR 38.830 |
| TS 38.838 | 3GPP TR 38.838 |
| TS 38.878 | 3GPP TR 38.878 |
| TS 38.889 | 3GPP TR 38.889 |