Power Control in 5G NR

1. Introduction

Power control is one of the least visible but most influential mechanisms in cellular radio systems. In 5G NR, it directly impacts:

• Uplink throughput and latency

• Cell-edge user experience

• Inter-cell interference

• UE battery life

• Massive MIMO beam efficiency

• Network energy consumption (Green RAN goals)

Unlike LTE, 5G NR operates with:

• Very wide bandwidths

• Beam-based transmission

• Dynamic TDD

• Cloud-native and O-RAN architectures

As a result, power control in 5G NR is no longer a simple formula — it is a multi-loop adaptive system involving both UE intelligence and gNB scheduling algorithms.

Power Control is a smart mechanism in 5G networks that automatically adjusts how much transmission power your phone uses when sending data to the cell tower. Think of it as an intelligent volume control system for wireless communication.

Just like you wouldn't shout at someone standing right next to you, your phone shouldn't use maximum power when it's close to a cell tower. Power control ensures your device uses just the right amount of power - not too much, not too little.

How Does It Work?

Power control is a continuous, automatic process involving constant communication between your phone and the cell tower:

• Measurement: Your phone constantly measures the signal strength from the tower and calculates how far away it is (called "path loss").

• Instructions: The cell tower analyzes the signal quality and sends power control commands to your phone, telling it to increase or decrease transmission power.

• Adjustment: Your phone immediately adjusts its power level - this can happen hundreds of times per second as conditions change.

• Feedback Loop: The tower monitors the received signal quality and continues to fine-tune the power commands in real-time.

 2. Power Control Philosophy:

Power control is split between two control domains:

Control Loop	Controlled By	Purpose	Time Scale
Open Loop Power Control	UE	Baseline power estimation	Slow
Closed Loop Power Control	gNB	Fine correction	Fast

Key idea

The UE estimates how much power it should use,while the gNB continuously corrects how much power it actually uses.

3. Open Loop Power Control (OLPC)

3.1. What is Open Loop Power Control?

Open loop power control allows the UE to autonomously calculate it’s transmit power using downlink measurements.This ensures the UE can transmit even before tight feedback loops are established.

General Uplink Power Formula

For uplink channels such as PUSCH, PUCCH, and PRACH:

P_TX = min(P_MAX, P0 + α × Pathloss + ΔTF + f(i))

Explanation of Each Term

• P_MAX

  Maximum UE transmit power (typically 23 dBm for handheld UEs).

• P0

  A configurable power offset defined by the network.

  It represents the target received power at the gNB.

• α (Alpha) – Pathloss Compensation Factor

  Controls how much of the pathloss is compensated by the UE.

  • α = 1 → full compensation

  • α < 1 → partial compensation

• Pathloss (PL)

  Estimated by the UE using downlink reference signals (SSB / CSI-RS).

• ΔTF (Transport Format Offset)

  Extra power required for:

  • Higher modulation orders

  • More layers

  • Higher coding rates

• f(i) - PUSCH Power Control Adjustment State

  This term represents the power adjustment by TPC (Transmit Power Control) command.

It is derived from:

• f(i - i₀ ): The adjustment state from the last transmission.

• ∑_m=0^|D_L^|-1 δ_PUSCH,b,f,c(m,l): The cumulative power control commands (δ_PUSCH) received via TPC (Transmit Power Control) for the current scheduling interval.

• δ_PUSCH,b,f,c(m,l) indicates the transmit power control commands provided by the gNB to refine the UE's uplink transmission power. 

3.2. Pathloss Estimation in 5G NR (Beam-Based)

In 5G, pathloss estimation is beam-dependent, unlike LTE.

Pathloss = Reference Signal Power – Measured RSRP

Key points:

• Reference signal may change per beam

• Beam switching causes sudden UL power changes

• FR2 pathloss varies rapidly due to blockage

This is why uplink power instability is more common in 5G than LTE.

What Is the “Reference Signal Power”?

The reference signal power is not something the UE guesses. It is explicitly signaled by the network.

In 5G NR, the reference signal used for pathloss estimation is typically:

• SSB (Synchronization Signal Block), or

• CSI-RS (Channel State Information Reference Signal)

The gNB transmits these signals with a configured power per beam, and the UE is informed of that power via broadcast or RRC signaling.

For example:

• The gNB may transmit SSB Beam #3 with a power of −60 dBm per resource element.

• This value is known to the UE as the reference transmit power.

This is the “expected” power before any propagation loss.

How the UE Measures RSRP

The UE measures the Reference Signal Received Power (RSRP) of the currently selected beam. This measurement is beam-specific.

For instance:

• UE measures SSB Beam #3

• Measured RSRP = −92 dBm

This measurement already includes:

• Free-space pathloss

• Shadowing

• Beamforming gain

• Body loss (hand, head)

• Blockage effects (especially in FR2)

Pathloss Calculation – Numerical Example

Let’s put this together with real numbers.

Assume:

• Reference signal transmit power (SSB beam) = −60 dBm

• Measured RSRP at UE = −92 dBm

Then:

Pathloss = (−60) − (−92)Pathloss = 32 dB

This 32 dB becomes the input to the UE’s open loop power control formula.

If the UE later switches to a different beam with a different measured RSRP, the calculated pathloss changes immediately.

n LTE, pathloss estimation was relatively stable because:

• Reference signals were transmitted quasi-omnidirectionally

• The UE did not frequently change its serving “direction”

In 5G NR:

• Each beam has its own spatial gain

• Different beams arrive at the UE with different power levels

• Beam switching is frequent, even for stationary UEs

Consider this practical example:

• Beam A RSRP = −90 dBm

• Beam B RSRP = −96 dBm

If the reference transmit power is the same, then:

• Pathloss via Beam A = 30 dB

• Pathloss via Beam B = 36 dB

From the UE’s perspective, the environment suddenly looks 6 dB worse, even though the UE has not moved at all.

3.3. Alpha (α): Partial vs Full Pathloss Compensation

α Value	Network Behavior	Typical Deployment
0.4–0.6	Low interference, lower cell-edge power	Dense small cells
0.7–0.9	Balanced	Urban macro
1.0	Full compensation	Coverage-limited rural

Operator reality:Most live networks avoid α = 1 due to uplink interference rise.

The gNB informs the UE about alpha (α) through RRC signaling, not through MAC or PHY. Once configured, alpha remains static until reconfigured and is not signaled dynamically per transmission.

RRCReconfiguration

 └─ radioBearerConfig

    └─ physicalConfigDedicated

       └─ uplinkConfig

          └─ pusch-Config

             └─ pusch-PowerControl

                └─ alpha

Note : Alpha is not broadcast in SIBs. It is UE-dedicated configuration.

Example:

PUSCH_PowerControl_Config:

  p0-NominalWithGrant = -76 dBm

  alpha = alpha08 -> UE decodes this to alpha=0.8

  deltaTF = enabled

  maxRank = 2

UE Log Example – Power Calculation Using Alpha

UL_PWR_CALC:

Pathloss = 94 dB

Alpha = 0.8

Alpha × Pathloss = 75.2 dB

P0 = -76 dBm

DeltaTF = 2 dB

TPC = +1 dB

Calculated TX Power = 18.2 dBm

If alpha were 1.0, the same pathloss would result in:

TX Power ≈ 23 dBm (near Pmax)

4. Closed Loop Power Control (CLPC)

Why Closed Loop Is Required

Open loop power:

• Cannot react to fast fading

• Cannot observe real BLER

• Cannot manage MU-MIMO fairness

 The gNB therefore applies TPC (Transmit Power Control) commands.

How TPC Is Sent

• Via DCI on PDCCH

• Via MAC Control Elements for long-term tuning

Typical adjustment granularity:

Accumulated Power Offset (Relative Power Control)

In accumulated mode, each TPC command adds to or subtracts from the UE’s existing power offset. The UE remembers the previous offset and updates it incrementally.

Simply, “Increase power by +1 dB from wherever you are now.”

The UE maintains an internal variable, often called a power control accumulator. Each new TPC command modifies this value.

Current_TPC_Offset = Previous_TPC_Offset + TPC_Command

Practical Example

Assume:

• Initial TPC offset = 0 dB

TPC commands received over time:

• Slot 1: +1 dB → offset = +1 dB

• Slot 2: +1 dB → offset = +2 dB

• Slot 3: −1 dB → offset = +1 dB

UE log view:

TPC_CMD = +1 dB

Accumulated_TPC = +1 dB

TPC_CMD = +1 dB

Accumulated_TPC = +2 dB

TPC_CMD = -1 dB

Accumulated_TPC = +1 dB

The UE gradually increases or decreases transmit power based on link conditions.

Absolute Power Offset (Set-and-Hold Power Control)

In absolute mode, the TPC command sets the power offset to a specific value, rather than adding to the previous one.

In simple terms: “Set your power offset to +3 dB — ignore what it was before.”

The UE discards any previously accumulated offset and replaces it with the new value.

Conceptually:

Current_TPC_Offset = TPC_Command_Value

This offset is then applied directly to uplink transmissions.

Practical Example

Assume:

• Current TPC offset = +4 dB

gNB sends an absolute TPC command:

• TPC = +1 dB (absolute)

UE behavior:

TPC_CMD = ABSOLUTE +1 dB

Previous Offset = +4 dB

New Offset = +1 dB

The UE immediately drops its uplink power by 3 dB.

5. Power Control per Uplink Channel

5.1 PUSCH Power Control

Most complex and dynamic.

Factors considered by gNB:

• UL SINR

• HARQ retransmissions

• BLER target (typically ~10%)

• UE Power Headroom

• Rank and MCS

• Beam quality

Real network behavior

• Retransmissions receive higher power

• MU-MIMO users get conservative power

• Cell-edge users hit P_MAX frequently

Example 1: PUSCH Power Calculation

P0 = -78 dBm

Alpha = 0.8

Pathloss = 90 dB

Delta_TF = 2 dB

TPC = +1 dB

=> TX Power = 17 dBm

5.2 PUCCH Power Control

PUCCH carries control information, not data:

• HARQ ACK/NACK

• CSI

• Scheduling Requests

Hence:

• Reliability > throughput

• Conservative power offsets

• Often fixed configuration

PUCCH failures cause false DL retransmissions, heavily impacting throughput.

5.3 PRACH Power Control

PRACH uses power ramping:

PRACH Power = Initial Power + (Ramp Step × Attempt Number)

Typical values:

• Ramp step: 2–6 dB

• Max attempts: 8–64

PRACH Power Ramp calculation:

Attempt 4:

Initial Power = -110 dBm

Ramp = 4 dB

TX Power = -94 dBm 

Misconfigured PRACH power leads to:

• RACH failures

• Congestion storms

• NSA dual-connectivity attach issues

6. Power Headroom Report (PHR)

Power Headroom Report (PHR) is one of the most critical uplink feedback mechanisms in 5G NR, yet it is often treated as a secondary detail. In reality, PHR is the scheduler’s reality check. It tells the gNB whether the UE can physically support the uplink grants being assigned.

Many uplink “mystery issues” — low throughput, repeated retransmissions, unstable MU-MIMO — trace back to ignored or misunderstood power headroom information.

What Is Power Headroom?

Power headroom represents how much additional transmit power the UE still has available beyond its current uplink transmission.

Conceptually:

PH = P_MAX − Current TX Power

Where:

• P_MAX is the UE’s maximum allowed transmit power (typically 23 dBm for handheld UEs)

• Current TX Power is the actual power used for the current uplink transmission

A high PH means the UE has margin to transmit more power if needed.A low PH means the UE is already close to saturation.

 Why Power Headroom Exists at All

Without PHR, the gNB could easily schedule uplink resources that look good on paper but are impossible for the UE to power in reality.

PHR exists to ensure:

• Uplink grants are power-feasible

• UE battery and thermal limits are respected

• Scheduler decisions are grounded in physical reality

In 5G NR, where bandwidths are wide and MCS can change rapidly, this feedback becomes even more important than in LTE. 

There are three types of PHRs in 5G NR.

Type-1 PHR: Single Serving Cell Power Headroom:

• Type-1 PHR reports the power headroom for one serving cell, considering the UE’s current uplink transmission on that cell.

• This is the basic and most common PHR type.

Example, A UE connected to one NR cell reports:

• P_MAX = 23 dBm

• Current PUSCH power = 20 dBm

• Type-1 PHR = 3 dB

The scheduler knows it can safely increase PRBs or MCS if needed. 

Type-2 PHR: Per Serving Cell Power Headroom (Carrier Aggregation)

• Type-2 PHR reports separate power headroom values for multiple serving cells when carrier aggregation (CA) is used.

• Instead of a single value, the UE reports one PHR per uplink carrier.

 Type-2 PHR is used when:

• Multiple uplink carriers are active

• Each carrier may have independent scheduling

• gNB needs per-carrier power awareness 

Example, UE with two uplink carriers reports:

• Cell-1 PHR = 4 dB

• Cell-2 PHR = 1 dB

The gNB schedules more aggressive uplink grants on Cell-1 and keeps Cell-2 conservative.

Type-3 PHR: Multiple PHR per BWP / Power Domain

Type-3 PHR allows the UE to report multiple power headroom values within the same serving cell, typically per:

• Bandwidth Part (BWP), or

• Power domain / carrier group

Example, UE reports:

• BWP-1 (wide, high SCS): PHR = 0 dB

• BWP-2 (narrow, low SCS): PHR = 5 dB

The scheduler prefers BWP-2 for uplink data to avoid power saturation. 

PHR Type	Scope	Why Used	Typical Scenario
Type-1	Single cell	Basic power awareness	Single-cell NR
Type-2	Per serving cell	CA power balancing	UL carrier aggregation
Type-3	Per BWP / domain	Fine-grained control	Dynamic BWP switching

How the UE Computes Power Headroom

The UE internally knows:

• Its maximum transmit capability (P_MAX)

• The power required for the current PUSCH/PUCCH transmission

It simply subtracts the two and reports the remaining margin.

Important nuance:The UE does not assume ideal conditions. It accounts for:

• Current MCS and number of layers

• Allocated PRBs

• Pathloss and alpha

• Closed loop TPC corrections

• UE thermal state (vendor-specific)

This means PHR reflects real, usable margin, not theoretical capability.

How Does the UE Inform the gNB About PHR?

Power Headroom is reported using a MAC Control Element (MAC CE).

MAC CE Type

In 5G NR, the UE sends:

• Single PHR or

• Multiple PHR (per serving cell / per BWP)

These are standardized MAC control elements.

When Does the UE Send PHR?

PHR transmission is triggered by:

• Periodic timer expiry

• Significant change in power headroom

• Uplink grant transmission

• BWP switch

• UE reaching power saturation

The gNB controls how often PHR is sent by configuring PHR timers via RRC.

PHR configuration parameters

• phr-PeriodicTimerDefines how often the UE sends a periodic Power Headroom Report to the gNB, ensuring the scheduler has up-to-date UE power capability information. Values: {10, 20, 50, 100, 200, 500, 1000} ms

• phr-ProhibitTimerSpecifies the minimum time the UE must wait after sending a PHR before it is allowed to send another PHR, preventing excessive reporting. Values: {0, 10, 20, 50, 100, 200, 500, 1000} ms

• phr-Tx-PowerFactorChangeSets the transmit power change threshold that triggers an immediate PHR when the UE’s required uplink power changes significantly. Values: {dB1, dB3, dB6, infinity}

• phr-ModeOtherCGIndicates whether power headroom for the other carrier group (in dual connectivity) is reported based on real measured power or a virtual estimated value. Values: {real, virtual}

• multiplePHRControls whether the UE reports separate power headroom values per serving cell or BWP instead of a single aggregated report. Values: enabled/disabled

UE Log Example: PHR Generation

A realistic UE log snapshot might look like this:

MAC_PHR_TRIGGER:

P_MAX = 23 dBm

Current_PUSCH_Power = 21 dBm

Calculated_PHR = 2 dB

Trigger = Periodic

Internally, the UE encodes this value into a compact MAC CE format and transmits it along with uplink data.

How PHR Is Encoded

PHR is not sent as raw dB values.It is quantized and encoded as an index representing a power range.

Example:

• PHR index 0 → PH < 0 dB

• PHR index 1 → 0–2 dB

• PHR index 2 → 2–4 dB

• …

• PHR index N → High margin

The exact mapping is standardized, and the gNB decodes it back into a usable power margin estimate.

gNB Side: How PHR Is Decoded

When the gNB receives an uplink MAC PDU, it parses the MAC CEs before processing the data.

A gNB-side trace might show:

MAC_CE_DECODE: 

PHR received

PHR_Index = 1

Decoded_PHR ≈ 2 dB

Serving Cell = NR_Cell_1

The decoded PHR value is then stored in the UE context and immediately made available to the uplink scheduler.

How the gNB Uses PHR in Scheduling

PHR acts as a hard constraint for the scheduler.

If the scheduler ignores PHR, uplink instability is almost guaranteed.

When PHR is low, the gNB typically reacts in the following ways:

• It reduces the MCS, favoring robustness over throughput

• It reduces the number of PRBs assigned

• It avoids scheduling the UE in MU-MIMO groups

• It may favor PUCCH-only scheduling for control stability

These actions prevent the UE from being forced into under-powered transmissions.

Practical Scheduler Decision Example

Consider this scenario:

• UE reports PHR ≈ 1 dB

• Scheduler wants to assign 40 PRBs with 64QAM

If this grant is applied:

• Required TX power exceeds P_MAX

• UE transmits at capped power

• BLER increases

• Retransmissions follow

Instead, a PHR-aware scheduler:

• Reduces PRBs to 20

• Drops MCS to 16QAM

• Achieves stable decoding with lower peak throughput

This trade-off is intentional and correct.

UE Log Example: PHR Impact on Scheduling

UE-side observation:

PHR = 1 dB

UL Grant Reduced

MCS changed from 21 → 16

PRBs changed from 40 → 20

gNB-side observation:

Scheduler Decision:

PHR constraint active

UL grant scaled down

This correlation is one of the most useful debug patterns for uplink issues.

7. SRS-Based Power Control

Closed-loop power control using PUSCH decoding works well after data is already scheduled. But the gNB also needs a way to predict uplink channel quality before scheduling and without relying on past HARQ failures. That is where Sounding Reference Signals (SRS) come in.

SRS gives the gNB a clean, interference-aware measurement of the uplink channel, independent of data modulation. Using SRS, the gNB can estimate pathloss, SINR, beam quality, and spatial characteristics before deciding how aggressively to schedule the UE.

In practice, SRS-based power control is proactive, while HARQ/TPC-based control is reactive.

SRS:

• Is transmitted with a known structure

• Uses configured power offsets

• Is beam-specific

• Is ideal for estimating uplink channel quality

This makes SRS especially valuable for:

• MU-MIMO pairing

• Beam selection

• Initial power target estimation

• Mobility and beam switching scenarios

UE Log Example:

SRS Transmission

A realistic UE log snapshot might look like this:

SRS_TX_CONFIG: 

SRS_Resource_ID = 3

SRS_Beam_ID = 5

P0_SRS = -80 dBm

Alpha = 0.8

Pathloss = 90 dB

SRS_Offset = 3 dB

Calculated_SRS_TX_Power = 15 dBm

gNB Interpretation

On the gNB side:

SRS_MEASUREMENT: 

UE_ID = 12

Beam_ID = 5

Measured_SINR = 9 dB

Estimated_UL_Pathloss = 90 dB

Predicted_PUSCH_SINR (64QAM) = Marginal

SRS Influencing TPC

A few slots later, UE log shows:

TPC_CMD_RECEIVED = +1 dB

TPC_Source = PUSCH_DCI

Accumulated_TPC = +2 dB

New_PUSCH_TX_Power = 18 dBm

8. Practical UE-Side Power Control Algorithms

UEs do much more than the spec mandates:

Common UE optimizations:

• Power smoothing (avoid sharp spikes) - UEs avoid applying abrupt transmit power changes even if the calculated value jumps suddenly. Instead, power is ramped gradually to prevent RF instability, spectral splatter, and hardware stress.

Practical example:After a beam switch causes pathloss to increase by 8 dB, the UE increases uplink power over several slots rather than instantly jumping by 8 dB.

• PUCCH prioritization - UEs prioritize PUCCH power over PUSCH when total transmit power is limited. Control signaling is considered more critical than data throughput.

Practical example:When the UE approaches P_MAX, PUCCH power remains stable for reliable HARQ ACKs, while PUSCH power is reduced, leading to lower uplink data rates but stable DL performance.

• Thermal power back-off- When the UE temperature rises due to sustained uplink transmission, the UE reduces its maximum allowed transmit power to protect hardware components.

Practical example:During prolonged uplink-heavy traffic (video upload), the UE caps TX power at 20 dBm instead of 23 dBm, even though radio conditions remain unchanged.

• Battery-aware behavior - UEs adjust transmit power behavior based on battery state and power-saving modes. This behavior is entirely implementation-specific.

Practical example:With low battery or battery saver mode enabled, the UE limits uplink power and accepts lower uplink throughput to conserve energy.

• EN-DC power split protection - In EN-DC (LTE–NR dual connectivity), the UE must share its total transmit power between LTE and NR uplinks. The UE actively limits NR power to maintain LTE anchor stability.

Practical example:During simultaneous LTE UL control signaling and NR data upload, the UE caps NR PUSCH power, resulting in lower NR uplink throughput despite good NR signal quality.

This explains why:

Two UEs in the same cell can show very different UL behavior.

9. gNB-Side Power Control Algorithms

Typical gNB control loop:

1. Measure UL SINR and BLER -The gNB measures per-UE SINR and decoding success for each uplink transmission.

2. Compare with target BLER - Observed BLER is compared against the configured target (commonly around 10%).

3. Issue TPC corrections - small power up/down commands are sent to fine-tune UE transmit power.

4. Adjust MCS and RB allocation - The scheduler adapts modulation, coding rate, and resource block allocation.

5. Monitor PHR feedback - Power headroom reports are used to ensure grants remain power-feasible.

Advanced features:

• Beam-specific power offsets -The gNB applies different power targets for different beams to compensate for varying beam gains and coverage characteristics.

Practical example:A weaker edge beam is configured with a higher effective power target than a strong central beam.

• Cell-edge biasing - Cell-edge UEs are intentionally biased to receive more conservative scheduling and power treatment to maintain link stability.

Practical example:A cell-edge UE is assigned fewer PRBs with slightly higher power rather than many PRBs at insufficient power.

• Interference-aware power caps - The gNB limits uplink power for certain UEs to control overall interference, even if individual link quality would allow higher power.

Practical example:During UL-heavy TDD slots, gNB caps UL power to protect neighboring cells from interference rise.

• AI-assisted optimization (modern RAN) -AI-based algorithms analyze long-term trends and dynamically tune power control parameters such as alpha, P0, and scheduling bias.

Practical example:During peak hours, AI reduces alpha to limit interference and improves overall cell stability without changing UE behavior.

10. LTE vs 5G NR Power Control – Key Differences

Aspect	LTE	5G NR
Transmission	Cell-based	Beam-based
Bandwidth	≤20 MHz	≤400 MHz
Alpha usage	Mostly static	Dynamic per BWP
Power loops	Simpler	Multi-loop
PHR usage	Limited	Scheduler-critical
MU-MIMO	Limited	Core feature
TDD impact	Fixed	Dynamic TDD

 11. Common Power Control Issues in Live Networks

Issue	Root Cause
High UL BLER	α too low
UE overheating	Sustained P_MAX
RACH failures	Poor PRACH ramp
Poor cell-edge UL	P0 mis-tuned
MU-MIMO instability	Power imbalance

 12. Key Takeaways

• Power control is not a single equation

• UE and gNB both apply intelligent heuristics

• PHR is central to uplink scheduling

• LTE assumptions break in 5G

• AI-driven power control is the future

13. References

1.      3GPP TS 38.213 “NR; Physical layer procedures for control”

2.      3GPP TS 38.321 “NR; Medium Access Control (MAC) protocol specification”

3.      3GPP TS 38.331 “NR; Radio Resource Control (RRC) protocol specification”

4.      5G NR | 5g New Radio Standard | Qualcomm

5.      https://www.nokia.com/bell-labs/publications-and-media/publications/5g-new-radio-and-lte-uplink-coexistence

6.      https://www.ericsson.com/en/blog/2022/11/uplink-power-control-reinforcement-learning