Uplink Power Control in LTE 4G

Introduction

Uplink power control is the set of rules that decides, subframe by subframe, how much power an LTE UE puts into each thing it transmits. It is a physical-layer function parameterised by RRC and driven live by the eNB through TPC commands, and it applies independently to the data channel (PUSCH), the control channel (PUCCH), the sounding reference signal (SRS) and the random-access preamble (PRACH).

It runs everywhere in the UE lifecycle. The very first thing a UE transmits — the PRACH preamble from RRC_IDLE — is already power-controlled by an open-loop estimate; once connected in RRC_CONNECTED, every PUSCH and PUCCH transmission is set by the combined open-loop plus closed-loop formula. There is no point in the UE's transmit path where power is simply "full blast."

It matters because the LTE uplink is interference-limited and battery-limited at the same time. Too little power and the eNB cannot decode you; too much and you waste battery and raise the noise floor for every neighbouring cell reusing the same frequencies. Power control is the mechanism that holds both in balance, and understanding it explains a large share of uplink coverage, capacity and battery behaviour.

Why Uplink Power Must Be Controlled

The uplink is where power control really matters. Many UEs in many cells all transmit on the same frequencies at the same time, and every one of them is a source of interference to the others. If a UE transmits harder than it needs to, three bad things happen at once: it wastes its own battery, it raises the interference floor for every neighbouring cell, and it can even swamp the eNB's receiver relative to weaker UEs on adjacent resources.

Open-Loop: Compensating for Path Loss

The open-loop component is the UE's own estimate of how much power it needs, made without any per-command feedback from the eNB. Its job is to undo the path loss (PL) between UE and eNB — the further you are, the more signal you lose on the way, so the more you must transmit.

The UE derives PL entirely from the downlink. The eNB broadcasts the transmit power of its Cell-specific Reference Signal in the IE referenceSignalPower (inside SIB2 / RadioResourceConfigCommon), and the UE measures how strong those reference symbols actually land as CRS RSRP. The difference is the path loss: PL = referenceSignalPower − higher-layer-filtered RSRP. Because it is measured on the downlink CRS, this estimate is available even before the UE has sent anything, which is exactly why open-loop can set the very first PRACH preamble.

Two RRC-signalled parameters then shape the open-loop level. P0 is the target received power per resource block — effectively "how strong I want your signal to land at my antenna." It is itself the sum of a cell-wide part and a UE-specific part: P0 = P0-NOMINAL-PUSCH + P0-UE-PUSCH, the first broadcast to everyone, the second a per-UE offset in RRCConnectionReconfiguration. Both are signed integers in dBm-per-RB units: p0-NominalPUSCH spans roughly −126…+24 dBm in 1 dB steps, and p0-UePUSCH is a smaller trim of about −8…+7 dB. The fractional path-loss compensation factor α (alpha), an enumerated value in the set {0, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0}, decides how much of the measured path loss the UE actually compensates for. With α = 1 the UE fully compensates: every UE lands at the same received power regardless of distance. With α < 1 the UE only partly compensates, so distant UEs land a little weaker than near ones — this is fractional power control, a deliberate trade of cell-edge throughput for lower inter-cell interference and higher total capacity. Typical deployments sit around α = 0.7–0.8.

Finally there is a bandwidth term. Power is quoted per resource block, so a UE granted M_PUSCH resource blocks must scale its total power up by 10·log10(M_PUSCH) — transmitting over ten RBs needs about ten times the total power of transmitting over one, at the same per-RB density. This is the reason a UE at the edge often gets a narrow grant: concentrating its limited power into fewer RBs raises the per-RB energy and keeps decoding possible.

Closed-Loop: TPC Corrections and f(i)

Open-loop gets you close, but it can't see everything — fast fading, receiver load, or a slightly wrong path-loss estimate all leave a residual error. The closed-loop component lets the eNB correct that error directly. It watches your actual received quality and sends Transmit Power Control commands (TPC) telling you to nudge up or down.

Each TPC command is a small field carried in the DCI on PDCCH. For PUSCH it rides either inside your own uplink grant (DCI format 0, a 2-bit TPC field) or in a group-common power-control DCI (DCI format 3 or 3A) that packs TPC bits for many UEs at once, each UE reading its own bits at a configured tpc-Index. DCI format 3 carries 2-bit commands; DCI format 3A carries 1-bit commands. The group-common formats are scrambled with a dedicated TPC-PUSCH-RNTI (or TPC-PUCCH-RNTI for the PUCCH loop) so a UE knows which DCI to look inside.

These corrections feed an accumulated term written f(i). The behaviour depends on the RRC flag accumulationEnabled:

• Accumulative mode (accumulationEnabled = true): each command adds to the running total, so f(i) = f(i−1) + δ_PUSCH. The typical accumulative steps are −1, 0, +1 and +3 dB (the four values a 2-bit TPC field encodes). The loop has memory and drifts the power gradually to the right place, correcting slow drift with tiny nudges.
• Absolute mode (accumulationEnabled = false): each command sets the correction directly, f(i) = δ_PUSCH, chosen from a wider set −4, −1, +1, +4 dB. There is no memory; every command is a fresh absolute offset.

Both halves run together every subframe. The open-loop branch turns a downlink measurement into a base level; the closed-loop branch folds in the eNB's running corrections; the two are summed, the result is clipped to P_CMAX, and that becomes the actual transmit power. The eNB then measures what actually arrived and issues the next TPC, so the loop keeps chasing the target as the radio channel drifts. The figure below traces that full cycle — downlink CRS to path-loss estimate, through both branches, to the clipped transmission and back.

The PUSCH Power Formula

Put both halves together and you get the PUSCH power expression from TS 36.213. Everything is in dBm, computed per subframe i, and the whole thing is capped at the UE's configured maximum P_CMAX:

Read it left to right, then look up each term in the table below. P_CMAX is the ceiling — the UE never transmits above it, and when the inner value exceeds P_CMAX the UE is power-limited and clips. The highlighted group 10·log10(M_PUSCH) + P0_PUSCH + α·PL is the open-loop estimate. ΔTF is a transport-format (MCS-dependent) offset, and f(i) is the accumulated closed-loop correction from all the TPC commands so far.

PUCCH and SRS Power Control

Control information cannot afford the fractional-power gamble — if the eNB misses your HARQ-ACK or SR, the whole scheduling loop stalls. So PUCCH uses its own parameters and always fully compensates path loss (effectively α = 1). Its power formula, again capped at P_CMAX, is:

The pieces mirror PUSCH but with control-specific twists. P0_PUCCH = P0-NOMINAL-PUCCH + P0-UE-PUCCH. There is no M bandwidth term because PUCCH always occupies one RB per slot. Δ_{F_PUCCH}(F) is a format-dependent offset, signalled as deltaF-PUCCH-FormatX, that accounts for how much information the chosen PUCCH format carries — Format 1/1a/1b (ACK/SR) versus Format 2/2a/2b (CQI) versus Format 3 all have different reference levels. h(n) is a term that scales with the number of information bits carried (for CQI and multi-bit ACK), so a report with more bits gets more power. ΔTxD is added only with transmit diversity, and g(i) is the PUCCH closed-loop accumulator, fed by TPC in DCI format 1/1A/2/2A/2B/2C (downlink assignments) or in DCI format 3/3A.

SRS (Sounding Reference Signal) reuses the PUSCH machinery. Its power is the PUSCH open-loop plus closed-loop value adjusted by a dedicated offset P_SRS_OFFSET, so P_SRS = min{ P_CMAX, P_SRS_OFFSET + 10·log10(M_SRS) + P0_PUSCH + α·PL + f(i) }. It shares the same f(i) as PUSCH — sounding is only useful if it lands at a power the eNB can relate back to the data channel it is meant to characterise.

PRACH Preamble Power and Ramping

The PRACH case is special because there is no connection yet — the eNB can't send TPC to a UE it hasn't heard from. So the preamble uses open-loop plus a self-driven ramp: the "closed loop before there is a loop."

The UE computes the first preamble's transmit power from a broadcast target and its own path-loss estimate:

preambleInitialReceivedTargetPower is the wanted received level of the preamble at the eNB, broadcast in SIB2 (RACH-ConfigCommon / PowerRampingParameters); it is an enum stepping from about −120 dBm up to −90 dBm in 2 dB steps. Adding PL fully compensates the link. If no Random Access Response arrives within the RAR window, the UE assumes it was too quiet, steps the power up by powerRampingStep (an enum of 0, 2, 4 or 6 dB) and retransmits. This repeats up to preambleTransMax attempts (an enum such as n3, n4, n6 … n200); if it still fails, the UE declares a random-access problem to higher layers. The count of transmissions so far, PREAMBLE_TRANSMISSION_COUNTER, is what multiplies the ramp step.

Power Headroom and Scaling Under P_CMAX

The eNB schedules the uplink blind to the UE's real power budget unless the UE tells it. That is the job of the Power Headroom Report (PHR), a MAC control element that reports the difference between P_CMAX and the power the UE would have used for its current grant. The reported value is quantised to a range of roughly −23 dB to +40 dB in 1 dB steps.

A Type 1 PHR reports headroom for PUSCH alone: PH = P_CMAX − { 10·log10(M_PUSCH) + P0_PUSCH + α·PL + Δ_TF + f(i) }. A positive value means the UE has spare power (the eNB can grant more RBs or a higher MCS); a negative value means the formula demanded more than P_CMAX and the UE is clipping — the eNB should pull back. Reporting is triggered by periodicPHR-Timer, or by prohibitPHR-Timer plus a path-loss change exceeding dl-PathlossChange. (LTE-A adds a Type 2 PHR that also covers PUCCH when PUSCH and PUCCH are sent together.)

When the UE really is out of headroom, something has to give — and the rules are strict about what. With carrier aggregation the UE has one P_CMAX across all serving cells, and the total requested power over simultaneous PUSCH/PUCCH transmissions can exceed it. LTE then applies power scaling with a fixed priority order:

• PUCCH is protected first — if PUCCH and PUSCH would together exceed P_CMAX, the PUSCH power is scaled down to make room, because control must survive.
• PUSCH carrying UCI (uplink control multiplexed onto data) is prioritised over PUSCH carrying only data.
• Remaining PUSCH powers across component carriers are scaled by a common factor w (0 ≤ w ≤ 1) so their sum fits under the total budget.

Summary

LTE uplink power control sets, every subframe, how loud each uplink signal transmits. Open-loop builds a base level from the UE's own downlink path-loss estimate (referenceSignalPower − CRS RSRP), a target P0, the fractional factor α, and the bandwidth term 10·log10(M_PUSCH). Closed-loop then folds in the eNB's live TPC corrections through f(i), accumulative or absolute per accumulationEnabled. The sum is always clipped to P_CMAX.

PUCCH uses the same skeleton but fully compensates path loss and adds format- and bit-count-dependent offsets, because losing control breaks scheduling. SRS rides the PUSCH loop with an offset. The PRACH preamble, having no connection to lean on, uses open-loop plus self-driven ramping. When the UE runs out of budget, the PHR warns the eNB and a strict scaling order (PUCCH › PUSCH-with-UCI › PUSCH data) decides what survives under a single P_CMAX. The through-line: reach the eNB at its target quality, and no louder.

Quick Q&A