5G-NR/O-RAN Power-Saving Features For User Data

Introduction

Power-saving features in 5G (including O-RAN deployments) are needed mainly to improve energy efficiency and network sustainability.

1.      Reduce operational cost (OPEX): Base stations—especially massive MIMO O-RUs—consume significant electricity, so reducing unnecessary transmissions lowers energy bills for operators.

2.      Handle variable traffic load: Network traffic varies over time (night vs peak hours). Power-saving allows parts of the radio (RF chains, antennas, symbols, or slots) to sleep when traffic is low.

3.      Improve network sustainability: Lower power consumption reduces carbon footprint, which is a major goal for telecom operators and regulators.

4.      Increase equipment lifetime: Running hardware continuously at full power causes heat and wear; power-saving modes reduce thermal stress and extend hardware life.

5.      Enable dense 5G deployments: With many small cells and massive MIMO radios, energy-efficient operation becomes essential for scalable deployments.

5G RAN power-saving features: 

In 5G O-RAN, power-saving can be achieved by dynamically reducing transmission activity when slots carry only user plane (U-plane) data and no control signals. The O-DU scheduler can identify such slots and instruct the O-RU to disable unnecessary RF chains, antenna elements, or symbol transmissions. Techniques like symbol/slot blanking, partial antenna muting, and beamforming reduction help lower power consumption without affecting control signaling reliability. These optimizations are coordinated through O-RAN fronthaul control (C-plane) and management policies from the SMO/Non-RT RIC, enabling energy- efficient RAN operation.

3GPP Rel-18 introduced a cell-level DTX/DRX framework that lets a gNB silence downlink transmissions during configured “off” periods. In these silent intervals the gNB may pause PDSCH and even PDCCH (so the UE need not monitor control channels) without affecting broadcast or paging. The O-RAN Alliance also defines advanced sleep modes (ASMs) and partial array shutdown to reduce O-RU power. For example, the O-RAN white paper notes that low traffic can be handled by deactivating part of the TX/RX antenna array, cutting O-RU power. All these features aim to maximize silent time without breaking 5G timing or missing mandatory signals.

User-plane-only transmission patterns: Certain slots or symbols may carry only user-plane PDSCH (plus necessary DMRS) with no new control information or broadcasts. Examples include semi-persistent scheduling (SPS) or large data bursts where only PDSCH is active. In such cases the RU is transmitting user data continuously and the DU could potentially silence those intervals to save energy. For instance, 3GPP TR 38.864 discusses scenarios where the gNB turns off just data transmission during DTX “OFF” periods, continuing only minimal reference signals. Identifying these data-only slots (no PDCCH/SSB/SIB) enables mechanisms to stop or blank the RU’s transmitter selectively.

Candidate Power-Saving Mechanisms

1.Cell-Level DTX/DRX (Periodic Silence): 

A periodic DTX/DRX cycle is configured at cell level so that entire subframes or slots are suppressed. During the non-active span, no PDSCH/PUSCH is scheduled and the O-RU enters a reduced-power state.

Configure a periodic DTX/DRX pattern (Rel-18 NES) so that entire subframes or slots are off for PDSCH/PUSCH. During the “off” period the gNB/O-RU ceases data transmission (and possibly reference signals), letting the RU enter a low-power state. This is a standards-supported feature: TS 38.300 (Rel-18) specifies that a cell can have active and non-active durations, during which the UE need not monitor PDCCH or SPS. In practice, the DU would refrain from scheduling any user plane in the non-active span.

DU Scheduler Role

The O-DU refrains from scheduling any user-plane traffic in the non-active span. The O-CU/DU signals the pattern via RRC and the Near-RT RIC may trigger or adjust the cycle via the E2 interface.

O-RU Power State

The O-RU ceases data and reference signal transmission during the silence window. RF chains and DSP pipelines are gated, enabling meaningful PA / active component power reduction.

NES Objective

Network Energy Saving (NES) targets BS energy reduction proportional to silence duty cycle. At 50% off-time, theoretical PA savings reach ~40–50% depending on hardware and load conditions.

Coexistence

Active UEs are pre-informed via RRC signalling. SPS / configured grant users skip their occasions in the non-active window. Emergency/paging traffic uses predefined wake-up signals (WUS).

Feasibility/Latency: 

Fully feasible with Rel-18 base stations;

wake-up latency depends on configuration (e.g. slot-boundary). Using periodic DTX will increase user-plane delay (packets wait for the next on period), but this is acceptable for low-to-moderate delay workloads.

Standards compliance: 

Cell DTX is already standardized for RRC_CONNECTED UEs[1], so it can be enabled via RRC signaling without new specifications.

2.Symbol-level Advanced Sleep Modes (O-RU-ASMs): 

Exploit O-RAN RU sleep modes on a per-symbol basis. The DU (via the O-RAN fronthaul C-plane) can send a “Go-to-sleep” command for specific OFDM symbols, using the Section-Type4 message with a symbol mask. For example, when only PDSCH is scheduled in a symbol, the DU can instruct the RU to enter ASM1 or ASM2 for that symbol and immediately wake up at the next control symbol. In ASM1 the PA is shut off (≈30% power saving) with only ~37 µs wake-up overhead; deeper ASMs (ASM2/3) shut additional components at the cost of longer wake-up (~0.5–5 ms). The DU can use these ASMs dynamically whenever a symbol carries only user data.

O-RAN ASM Concept

The DU exploits O-RAN fronthaul C-plane sleep commands at per-symbol granularity. When only PDSCH is scheduled in a symbol, the DU can instruct the O-RU to enter ASM1/2/3 for that symbol and wake immediately at the next control symbol.

Section Type 4 Message

The DU uses the O-RAN WG4 Section-Type 4 C-plane message with a symbol mask to target specific OFDM symbols. This provides sub-slot granularity control of O-RU sleep states without altering the UE data path.

Dynamic Scheduling

The DU can apply ASMs dynamically every TTI — whenever a symbol carries only user-plane data and no control. The decision is per-slot, per-symbol, and can be adapted based on load, latency, and QoS requirements.

Feasibility Criteria

Requires an O-RU with fine-grained ASM hardware support and fast wake-up circuitry. ASM1 is broadly feasible (<50 µs overhead). Deeper modes (ASM2/3) suit longer data-only bursts where latency budget allows.

Latency Overhead

ASM1: ~37 µs wake-up — fits within one OFDM symbol at 30 kHz SCS. ASM2: ~0.5 ms. ASM3: ~5 ms. Delay must not exceed the guard window before the next control symbol.

Power Savings

ASM1 shuts the PA (~30% cell power saving). ASM2 additionally gates driver and pre-driver stages. ASM3 powers down ADC/DAC and clock trees, achieving up to ~60–70% symbol-level saving at cost of longer wake latency.

3.Data Burst Consolidation: 

Modify the DU’s scheduler to cluster user-plane traffic into bursts, creating longer idle intervals. Instead of sending small PDSCH blocks every slot, the DU holds non-urgent packets and then transmits them in one or few slots at higher throughput. This “burst scheduling” or packet aggregation makes more symbols completely free of traffic.

For example, one could align PDSCH transmissions to fill entire slots and leave intervening slots empty. This concept is explicitly suggested in Rel-18 studies: cell DTX is achieved by aggregating user-data into active periods and keeping the rest silent.

Instead of sending small PDSCH blocks every slot, the DU holds non-urgent packets and transmits them in one or few consolidated burst slots at higher throughput, leaving intervening slots completely idle.

3GPP Rel-18 studies explicitly suggest: Cell DTX is achieved by aggregating user-data into active periods and keeping the rest silent. Burst consolidation is the DU-side mechanism that directly enables cell-level DTX.

Scheduler Modification

The O-DU scheduler is modified with a burst aggregation engine and a hold timer. Non-real-time packets are buffered until a threshold is met, then transmitted in a dense, high-MCS burst across filled slots.

Idle Slot Creation

Aligned PDSCH bursts create entire slots free of traffic. These idle slots directly enable O-RU sleep modes (ASM1–3), Cell-Level DTX/DRX cycles, and RF chain power gating at the O-RU level.

Throughput Equivalence

Aggregate user throughput is preserved. Higher MCS in burst slots compensates for fewer scheduled slots. Peak rate per UE increases; average rate and FTP/HTTP goodput remain equivalent or improve due to lower overhead ratio.

QoS Constraints

Only non-urgent, delay-tolerant traffic (eMBB, FTP, video) is consolidated. URLLC, VoNR, and SPS bearers bypass the burst buffer and are transmitted on-demand. QoS class gating is mandatory.

Feasibility/Latency: Purely a scheduling strategy, so it’s implementable in the DU without hardware changes. It increases latency (since data is delayed to form bursts) and would need careful buffer management.

Standards compliance: This is an internal DU algorithm (non-standard), but it leverages the existing cell DTX framework. It does not require new signaling; it simply changes how the DU schedules UE data under the configured DTX pattern.

4.Carrier/Cell Switch-Off (SSB-less SCell): 

If carrier aggregation or multiple cells are used, completely shut off an entire carrier or small cell that carries only user-plane data bursts. In practice, this means not transmitting PDSCH (or even SSB) on that cell, letting it sleep.

3GPP Rel-18 introduced the “SSB-less SCell” feature where an SCell can omit SSB and most transmissions. Similarly, one could apply a cell shutdown for data-only traffic.

When carrier aggregation or multiple cells are deployed, an entire carrier or small cell carrying only user-plane data bursts can be completely shut off — no PDSCH, no SSB — letting the O-RU enter full sleep.

Rel-18 SSB-less SCell

3GPP Rel-18 introduced SSB-less SCell where an SCell can omit SSB and most transmissions. The SCell remains configured in RRC but ceases beacon-level signalling, enabling deeper O-RU power gating than dormancy alone.

SCell Dormancy

Prior to Rel-18, SCell dormancy (Rel-16) allowed the SCell to suspend PDCCH monitoring. SSB-less extends this by also removing the SSB overhead, making the silence complete and RF-clean for the O-RU hardware.

Traffic Offloading

Before shutdown, the O-CU/DU migrates all UE bearers back to the PCell or remaining SCells. The shutdown SCell's frequency band is vacated entirely. UEs perform seamless bearer re-mapping without session interruption.

O-RU Power State

With SSB-less SCell or full cell shutdown, the associated O-RU can power off PA, ADC/DAC, PLL, and RF front-end entirely — equivalent to ASM3 but at the cell level. Maximum power saving per RU hardware unit achieved.

Re-activation Latency

Re-activating an SSB-less SCell requires RRC reconfiguration + SCell activation (~20–100 ms). Full cell restart with SSB is slower (~100 ms–few seconds). The Near-RT RIC must predict traffic demand ahead of time.

Signalling Sequence — SCell Switch-Off & Re-activation

1.Near-RT RIC(E2/O1): Low Load Detected on SCell

xApp monitors SCell PRB utilisation via E2 SM-KPM. Load falls below threshold for T_idle duration. xApp issues CONTROL message to O-CU: initiate SCell switch-off policy for target cell. 

2.O-CU(F1-C / NG-C): Bearer Migration to PCC / Remaining SCells

O-CU triggers S1/NG bearer modification. All active DRBs on the target SCell are remapped to PCC or other SCells. PDCP reordering ensures lossless transfer. UE is notified via RRC reconfiguration. 

3.O-DU(RRC / F1-C): RRC SCell Deactivation / SSB-less Config

For SSB-less mode: O-DU sends RRCReconfiguration with dormancyGroupForSCell and SCell-SSBless-r18 IE. UE stops monitoring SCell PDCCH and SSB. For full shutdown: SCell is released from UE RRC context.

4. O-DU → O-RU(eCPRI / M-Plane): Sleep Command via eCPRI C-Plane

O-DU ceases all U-plane IQ sample delivery to the SCell O-RU. For SSB-less: C-plane signals ASM2/3 sleep mode. For full off: carrier shutdown command issued. O-RU powers down PA, RF chain, and clock domains.

5.O-RU(O-RU Internal): Full Power Gate — RF Silence Achieved

SCell O-RU enters full sleep: PA off, DAC/ADC off, PLL gated, fronthaul kept alive for wake-up. Estimated power saving: 60–90% of SCell O-RU consumption. PCC remains active; UE throughput maintained via PCC.

6. Re-activation: RIC Predicts Traffic Demand Increase

xApp detects PCC load rising / scheduled burst approaching. Issues wake-up policy T_ahead ms before needed. O-RU ramps up (20–100 ms for SSB-less, longer for full cell). O-DU re-adds SCell via RRCReconfiguration. CA throughput restored.

Feasibility/Latency: Feasible when traffic on one carrier is low. Switching a carrier off/on can incur millisecond-level reconfiguration or require triggering RRC (unless using O-RAN deep sleep).

Standards compliance: SSB-less is in Rel-18, so shutting an entire cell (or CC) is aligned with 3GPP’s energy features. It would need coordination with higher layers to resume transmissions.

5.UE-DRX Alignment with DTX: 

Coordinate cell silence with UE-side DRX. Ensure that whenever the cell goes data-silent, the scheduled UEs are in DRX “off” state so they’re not missing packets. For example, align the cell’s DTX cycle so that UEs would be asleep or idle anyway. This means the DU only schedules UEs in their active-on windows.

Scheduler Window Gating

The O-DU scheduler is modified to only schedule UEs in their DRX active-on windows. If a UE's DRX on-duration overlaps with the cell's DTX active period, it is eligible for scheduling. Otherwise, it is skipped and the cell remains silent.

DTX Cycle Alignment

The gNB configures the cell DTX cycle period T_DTX as a multiple of the UE DRX cycle T_DRX. This guarantees that every cell-active window coincides with at least one DRX active window for each UE group, ensuring no packet is stranded.

Multi-UE Grouping

UEs with different DRX offsets are grouped such that their combined active windows span the entire cell DTX active period. The DU assigns DRX configurations to maximise the overlap with planned DTX active slots.

O-RAN Integration

The Near-RT RIC xApp monitors per-UE DRX parameters and cell DTX patterns via E2. It can dynamically adjust DTX cycle offsets or DRX configurations (via E2 SM-RRC) to maintain or improve alignment as traffic changes.

Compound Power Gain

When aligned, both the O-RU and the UE RF chain are simultaneously in low-power state during the same silence window. This doubles the energy-saving benefit: gNB-side DTX saving + UE-side DRX saving occur concurrently.

Feasibility/Latency: This reduces wasted silent time (UE not expecting data) and avoids buffer overflow. UE DRX patterns are configurable via RRC, so aligning patterns is feasible with cooperative scheduling.

Standards compliance: 3GPP already allows indicating DRX cycles and aligning them with cell DTX. It simply requires that UEs supporting NES DTX be managed accordingly (no new spec needed).

6.ML-Optimized DTX Scheduling: Use machine learning (as in recent studies) to predict traffic and dynamically configure DTX periods for maximum savings. A DU software (e.g. in the near-RT RIC) could learn the optimal DTX on/off pattern given load and QoS, switching the RU on/off adaptively. For example, a DRL(Deep Reinforcement Learning) agent might turn on data only when a threshold of buffered bits is reached, maximizing silent time otherwise.

Instead of fixed DTX patterns, a machine learning model in the Near-RT RIC continuously predicts traffic load and configures DTX periods dynamically — adapting on/off cycles to maximise silence while satisfying per-UE QoS constraints.

DRL Agent

A Deep Reinforcement Learning (DRL) agent learns the optimal DTX policy by observing buffer depth, load, and QoS metrics as the state, issuing ON/OFF commands as actions, and receiving energy-QoS trade-off rewards.

Buffer-Threshold Policy

A key learned strategy: the agent turns TX on only when buffered bits exceed a threshold, then transmits at high MCS to drain the buffer quickly, then returns to silence. This maximises the idle ratio while bounding packet delay.

Near-RT RIC Deployment

The DRL agent runs as an xApp in the Near-RT RIC. It receives per-cell KPMs via the E2 interface (buffer occupancy, PRB usage, UE RSRP, QoS violations) and issues DTX pattern reconfiguration commands to the O-DU every 10–100 ms.

Traffic Prediction

A companion LSTM or Transformer prediction module forecasts short-term traffic (next 10–100 ms). The DRL agent uses this forecast as part of its state, enabling proactive DTX extension before load drops and early TX activation before bursts.

Compound Gain

ML-optimized DTX outperforms fixed patterns by 15–30% additional energy saving (per recent studies) at the same QoS level. The agent learns traffic periodicity, load correlation, and QoS class behaviour to find policies that static rules cannot discover.

Feasibility/Latency: This is a software-driven scheme on top of the existing DTX mechanism. It has high implementation complexity and introduces decision latency (training and inference delay), but can approach optimal saving in simulations.

Standards compliance: No new radio protocol is needed – it uses the same DTX controls – but it depends on a learning-enabled RAN controller. It is not standardized, but it aligns with 3GPP’s study of energy-aware scheduling.

Summary

Mechanisms that align with 3GPP/O-RAN specs are most straightforward:

Cell DTX/DRX and ASM1-based symbol sleep are natively supported and should be deployed first. Traffic aggregation scheduling and AI-based DTX can yield large savings (studies show ~30–45% cell power reduction) but at the cost of higher latency and custom algorithms.

In summary, recommendation:

(1)   Enable Rel-18 Cell DTX/DRX patterns for cells with light data, as per TS 38.300;

(2)   Implement DU-RU coordinated sleep commands (O-RAN Section-Type4) to silence RU on pure-data symbols;

(3)   Use advanced sleep modes (PA off) aggressively for idle symbols;

(4)   Aggregate and schedule user-plane traffic to enlarge silent periods.

Next steps include RU firmware support for fast wake/sleep, DU scheduler enhancements, and possibly RIC/xApp development to optimize patterns.

By combining these methods, an O-RAN deployment can significantly cut transmitter-on time for data-only slots while respecting 3GPP timing and user-QoS requirements.

References

1.      TS 138 300 - V18.1.0 - 5G; NR; NR and NG-RAN Overall description; Stage-2 (3GPP TS 38.300 version 18.1.0 Release 18), https://www.etsi.org/deliver/etsi_ts/138300_138399/138300/18.01.00_60/ts_138300v180100p.pdf

2.      arxiv.org https://arxiv.org/pdf/2501.15853

3.      Potential Energy Savings Features in O-RAN - white paper 2025-01, https://mediastorage.o-ran.org/white-papers/O-RAN.SuFG.Potential%20Energy%20Savings%20Features%20in%20O-RAN%20white%20paper%202025-01.pdf

4.      X. Mao et al., IEEE ICC 2026 (Cell DTX/DRX)[9]; NEC/Kairos et al., IEEE TWC (2025) (RU sleep modes)[8][5]; ns-3/O-RAN RU model (sleep current)[10].

5.      Network Energy Savings Study for NR | PDF | Energy Conservation | Radio, https://www.scribd.com/document/833079169/TR-38-864-on-NW-DTX-DRX-v0-Rapp-good

6.      [2507.21385] Deep Reinforcement Learning-based Cell DTX/DRX Configuration for Network Energy Saving, https://arxiv.org/abs/2507.21385

7.      RU Energy Modeling for O-RAN in ns3-oran, https://arxiv.org/html/2509.10978v1