mmWave & 5G

Configured Grant

Q: What is the difference between Type 1 and Type 2 configured grant?

Type 1 configured grant is fully provisioned by RRC signaling. The complete set of parameters, including time and frequency allocation, periodicity, modulation and coding scheme, and the start symbol, is delivered in the rrc-ConfiguredUplinkGrant information element, and the grant becomes active as soon as RRC configures it. There is no Layer 1 activation step. Type 2 configured grant separates configuration from activation: RRC provisions the periodicity and most parameters, but the resource is dormant until a DCI scrambled with the CS-RNTI activates it (and a separate DCI deactivates it). Type 2 therefore behaves like semi-persistent scheduling and lets the gNB start, stop, and reconfigure the allocation with low-latency Layer 1 signaling instead of slower RRC reconfiguration. Both types eliminate the scheduling request and dynamic grant round trip.

Q: How much uplink latency does configured grant save compared with dynamic scheduling?

Dynamic scheduling requires the UE to send a scheduling request on PUCCH, wait for the gNB to respond with an uplink grant in DCI, and only then transmit on PUSCH. That SR-to-grant exchange typically adds several milliseconds depending on PUCCH periodicity and processing timelines. With configured grant the resource is already allocated, so the UE can transmit on the next configured occasion as soon as data arrives. With a short periodicity (for example 2 symbols at 30 kHz subcarrier spacing, roughly 71 microseconds) and PUSCH repetitions for reliability, configured grant supports the URLLC user-plane budget of 1 ms one-way at 99.999 percent reliability, which dynamic scheduling cannot reliably meet.

Q: What periodicities and parameters can be set for a configured grant?

Periodicity is defined in symbols and ranges from very short values (2, 7, or 14 symbols) up to many slots, the exact set depending on subcarrier spacing. At 30 kHz SCS a 14-symbol period equals one 0.5 ms slot, while a 2-symbol period is about 71 microseconds. Beyond periodicity, the configuration specifies the time-domain allocation (start symbol and length via SLIV), frequency-domain resource blocks, MCS, number of HARQ processes (nrofHARQ-Processes), the repetition factor (repK of 1, 2, 4, or 8), the redundancy version pattern (repK-RV), and the power control parameters. For Type 2, the timeDomainOffset and the activating DCI together fix the first transmission occasion.

Q: How does HARQ work without dynamic feedback on a configured grant?

On a configured grant the UE derives the HARQ process ID from the timing of the transmission occasion rather than from a field in a DCI grant. The process is computed from the symbol index of the start of the transmission, the periodicity, and the number of configured HARQ processes. Reliability is improved by blind PUSCH repetitions: the UE sends repK copies (up to 8) across consecutive occasions following the configured redundancy version sequence, without waiting for any acknowledgment. The gNB combines the repetitions. Because there is no per-transmission grant, the configuredGrantTimer prevents the UE from reusing a HARQ process before the gNB has had the chance to request a retransmission.

/kuhn-FIG-yerd grant/

In 5G NR, an uplink resource allocation that the gNB pre-configures so a UE can transmit on periodic PUSCH occasions without first sending a scheduling request and waiting for a dynamic grant. Type 1 is provisioned entirely by RRC and active immediately; Type 2 is configured by RRC but activated and released by a DCI scrambled with the CS-RNTI, much like semi-persistent scheduling. By removing the request-and-grant round trip, configured grant shortens uplink access from several milliseconds to a single configured period (as short as ~71 μs at 30 kHz SCS), making it a cornerstone of URLLC traffic that demands 1 ms latency at 99.999% reliability. Blind PUSCH repetitions (repK up to 8) add diversity without per-packet feedback.

Category: mmWave & 5G

Types: Type 1 (RRC), Type 2 (DCI)

Min Period: 2 symbols (~71 μs)

How Grant-Free Uplink Scheduling Works in 5G NR

Standard NR uplink scheduling is dynamic: when a UE has data to send, it requests a resource by transmitting a scheduling request on PUCCH, the gNB responds with an uplink grant carried in a DCI message, and the UE finally transmits on PUSCH. That handshake is fine for elastic mobile broadband, but it imposes a fixed cost of several milliseconds before the first bit can leave the device. For deterministic industrial control, motion synchronization, and other time-critical services, that startup delay alone can blow the latency budget. Configured grant, standardized in 3GPP TS 38.321 (MAC) and TS 38.214 (physical layer), removes the handshake by giving the UE a standing reservation of uplink resources that recur on a fixed period.

Two flavors exist. Type 1 is fully described by the rrc-ConfiguredUplinkGrant information element: time-domain offset, start symbol and length, frequency resource blocks, modulation and coding scheme, number of HARQ processes, and repetition factor are all signaled by RRC, and the grant is live the instant RRC applies it. Type 2 keeps the heavy configuration in RRC but leaves the resource dormant until an activation DCI scrambled with the CS-RNTI fires it up, and a deactivation DCI tears it down. Because activation and release ride on Layer 1 signaling rather than RRC reconfiguration, Type 2 lets the network adapt the allocation quickly while still avoiding the per-packet grant.

Reliability without feedback is achieved through blind repetition. The UE transmits the same transport block repK times (1, 2, 4, or 8) across consecutive occasions, cycling through the configured redundancy version sequence, so the gNB can soft-combine copies even if some are lost. The HARQ process number is not signaled per transmission; instead the UE computes it from the occasion timing, and the configuredGrantTimer guards against reusing a process before any retransmission request can arrive.

Periodicity and Latency Budget

The single most important configured-grant parameter is periodicity, expressed in symbols. Shorter periods place more transmission opportunities per slot, cutting the worst-case wait between a packet arriving in the buffer and the next usable occasion. At 30 kHz subcarrier spacing one slot is 0.5 ms (14 symbols), so a 2-symbol periodicity yields a fresh opportunity roughly every 71 μs, while a full-slot period gives one every 500 μs. The trade is overhead: very short periods reserve large amounts of uplink capacity that may go unused if traffic is sporadic.

HARQ Process Derivation

For a configured grant the HARQ process ID is a function of the transmission timing rather than a DCI field. The MAC layer maps the start of the transmission to a process index using the configured periodicity and the number of HARQ processes, ensuring each periodic occasion deterministically selects a process so the gNB and UE stay aligned without explicit signaling.

Configured-Grant HARQ Process ID:
HARQ ID = ⌊ CURRENT_symbol / periodicity ⌋ mod nrofHARQ‑Processes
where CURRENT_symbol = (SFN × numberOfSlotsPerFrame + slot) × symbolsPerSlot + symbol

Occasion Spacing (time):
T_period = periodicity_(sym) × (1 / (14 × 2^μ)) ms
μ = numerology index (μ=1 → 30 kHz SCS, 14 sym = 0.5 ms slot)

Worst-case access latency (no SR/grant):
L_access ≈ T_period + T_proc,UE (vs. dynamic: T_SR + T_grant + T_PUSCH-prep)

Example: μ=1, periodicity = 2 symbols → T_period ≈ 71 μs; repK = 4 adds blind diversity for 99.999% reliability.

Configured Grant vs. Dynamic and Semi-Persistent Scheduling

Attribute	Configured Grant Type 1	Configured Grant Type 2	Dynamic Grant	Downlink SPS
Direction	Uplink	Uplink	Uplink & Downlink	Downlink
Activation	RRC only (immediate)	RRC config + DCI (CS-RNTI)	Per-packet DCI grant	RRC config + DCI
SR needed	No	No	Yes	N/A
Min periodicity	2 symbols	2 symbols	On demand	10 slots typical
Repetitions (repK)	1, 2, 4, 8	1, 2, 4, 8	DCI-controlled	Configurable
Typical use	Periodic URLLC, fixed sensors	Adaptive URLLC, VoNR-like	eMBB, bursty traffic	Periodic downlink (voice)

Common Questions

Frequently Asked Questions

What is the difference between Type 1 and Type 2 configured grant?

Type 1 is fully provisioned by RRC in the rrc-ConfiguredUplinkGrant IE (time and frequency allocation, periodicity, MCS, start symbol) and is active as soon as RRC applies it, with no Layer 1 activation step. Type 2 keeps configuration in RRC but stays dormant until a DCI scrambled with the CS-RNTI activates it, and a separate DCI deactivates it. Type 2 behaves like semi-persistent scheduling, letting the gNB start, stop, and adjust the allocation with fast Layer 1 signaling. Both eliminate the scheduling-request and dynamic-grant round trip.

How much uplink latency does configured grant save compared with dynamic scheduling?

Dynamic scheduling needs an SR on PUCCH, a grant in DCI, then PUSCH, adding several milliseconds depending on PUCCH periodicity and processing timelines. With a configured grant the resource is pre-allocated, so the UE transmits on the next occasion as soon as data arrives. A 2-symbol period at 30 kHz SCS (~71 μs) plus blind repetitions lets configured grant meet the URLLC budget of 1 ms one-way at 99.999% reliability, which dynamic scheduling cannot reliably hit.

What periodicities and parameters can be set for a configured grant?

Periodicity is defined in symbols, from 2, 7, or 14 symbols up to many slots depending on subcarrier spacing; at 30 kHz SCS, 14 symbols equals one 0.5 ms slot and 2 symbols is about 71 μs. The configuration also sets the SLIV time allocation, frequency RBs, MCS, nrofHARQ-Processes, repetition factor repK (1, 2, 4, or 8), the repK-RV redundancy pattern, and power control. For Type 2, the timeDomainOffset and activating DCI fix the first occasion.

How does HARQ work without dynamic feedback on a configured grant?

The UE derives the HARQ process ID from the timing of the occasion, computed from the start symbol, periodicity, and number of configured HARQ processes, rather than from a DCI field. Reliability comes from blind repetitions: the UE sends repK copies (up to 8) over consecutive occasions following the configured redundancy version sequence, and the gNB soft-combines them. The configuredGrantTimer prevents reusing a HARQ process before a retransmission can be requested.

Related Terms

← Configuration Management Configured NSSAI →