Hyper Open Edge Cloud

Case Study: Stabilizing VoLTE Performance in Production

This case study walks through an end-to-end approach to stabilizing VoLTE in a live network. We explain how to set the reference signal correctly; align Pa/Pb with multi-port RS behavior, and tune uplink power control (mainly P0) so CQI, BLER, throughput, and UE Tx power all make sense.
  • Last Update:2026-03-03
  • Version:001
  • Language:en

Background

Live field tests show degraded VoLTE (gaps/robot voice), elevated UE Tx power, and less-than-ideal BLER/throughput. We deployed according to the operator power plan; some parameters (output power / p_a / p_b / P0 / RS / bandwidth / antenna ports) were inherited or mapped from a legacy site. Rather than assume a single culprit, we follow an end-to-end flow of power plan to performance to isolate each factor cleanly.

Small misalignments anywhere can hurt VoLTE:

  • RS declaration (SIB2 referenceSignalPower)
  • DL data vs RS (p_a), and multi-port RS scaling (p_b)
  • UL power control (P0, alpha, TPC) and UL grant size M
  • Scheduler/precoding using CQI (which is RS-based)
  • DPC
     

Pilot configuration

    • DPC: off
    • Output power: up to 40 W × (4T or 2T)
    • SIB2 referenceSignalPower (RS(SIB2)): +15 dBm
    • p_a = 0, p_b = 0 (legacy mapping)
    • P_max: 23 dBm
    • P0-NominalPUSCH: −90 dBm
    • alpha: 1.0 (alpha al1)
    • Actual RS EPRE at RU output (RS(actual)): unknown

Power Plan to Performance

RS(SIB2) and RS(actual)

Phones estimate path loss from RS and SIB2:
PL(est) = RS(SIB2) − RSRP = PL(true) + [RS(SIB2) − RS(actual)]

SIB2 does not change DL transmit power. It informs UL power control through path-loss:

  • If RS(SIB2) > RS(actual):
    UE over-estimates PL => targets higher UL power by ~ alpha*[RS(SIB2) − RS(actual)]
    Expected: louder PUSCH/PUCCH/PRACH, UL interference/noise-rise risk; DL RSRP/SINR unchanged.
  • If RS(SIB2) < RS(actual):
    UE under-estimates PL => targets lower UL power by ~ alpha*|RS(SIB2) − RS(actual)| dB
    Expected: low UL SNR, high PUCCH BLER, high PRACH ramps; DL may suffer indirectly via missed ACK/NACK or CQI.

Key point: SIB2 mismatch doesn’t move DL RSRP/SINR directly; it affects DL indirectly via UL control/feedback and cell-wide interference.

  • Working baseline: RS(SIB2) = +15 dBm (legacy value).
  • Ensure method: conducted check (RSRP + known attenuation) to set SIB2 = RS(actual).
  • Keep p_b fixed when comparing profiles in lab; p_b changes can shift RS level.

p_a / p_b and what the UE “sees”

  • PDSCH EPRE = RS EPRE(fixed) + p_a (per port, per RE, at the transmitter).
  • The UE “composite” sees RS on one port per RE, but PDSCH across N_tx ports. Roughly:
    PDSCH = RS (true) + p_a + 10·log10(N_tx)
  • For 4T:
    p_a = 0 => PDSCH = RS (true) + 6 dB at the UE;
    p_a =−6 => PDSCH = RS (true) .

Why align them? CQI is RS-based.

  • If PDSCH ≫ RS:
    CQI is pessimistic (under-modulation unless scheduler climbs by ACKs).
  • If PDSCH ≪ RS (very negative p_a):
    CQI is optimistic (BLER/HARQ rise until MCS backs off).

The goal is PDSCH SNR = RS SNR at the UE, then let p_b keep average symbol power flat.

UE measures RSRP on RS => Path-loss PL (box 4)

  • With RS(SIB2) fixed, RSRP & RS-SINR should not change when only p_a changes.
  • Phones estimate path loss from RS and SIB2:
    PL(est) = RS(SIB2) − RSRP = PL(true) + [RS(SIB2) − RS(actual)]

Downlink data branch (left side)

  • ​​​​​If p_a is 6 dB too high on 4T (pₐ=0 vs target −6):
    • UE-side PDSCH SNR = RS-SNR + 6 dB
    • CQI (RS-based) pessimistic => initially conservative MCS, very low BLER; throughput may lag unless scheduler ramps by ACKs.
    • DL control (PDCCH) not affected by p_a.
  • If pₐ is 6 dB too low: expect BLER↑/reTx↑ at same CQI until adaptation backs MCS off.

Recommended for 4T: p_a = −6 dB, and choose p_b to keep average symbol power flat across RS/non-RS symbols.

Uplink control branch (right side)

UL target (simplified):

P_UL = 10*log10(M) + P0 + alpha*PL + ...  (capped at UE max)
alpha = 1.0 maps any SIB2 error 1:1 into UL target.

Quick Summary

  • SIB2 too high (RS(SIB2) > RS(actual)): UEs too loud (high UL power by α·Δ), PRACH ramps fewer steps, PUCCH robust, UL interference rises; DL RSRP/SINR unchanged.
  • SIB2 too low (RS(SIB2) < RS(actual)): UEs too quiet (low UL power by α·|Δ|), high PUCCH BLER, high PRACH ramps, low UL SNR; DL hurt indirectly via missed ACK/NACK/CQI.
  • p_a too high on 4T (e.g., 0 instead of −6): PDSCH = RS(true) + 6 dB at UE; CQI pessimistic, BLER very low, possible under-utilization unless scheduler ramps.
  • p_a too low: data weaker than pilots => high BLER until MCS backs off; VoLTE becomes fragile in fades.

Scheduler & precoding

  • CQI is RS-based. Aligning PDSCH SNR = RS SNR (via p_a) makes CQI-driven MCS selection predictable.
  • p_b impacts RS per-port scaling in 4T; keep it fixed in A/B so RS-based measures (RSRP/SINR) stay stable.
  • Improvement on the eNB software side: precoding algorithm questions 

Conclusion

We need to start by locking RS (constant, truthful), then align PDSCH with RS for 4T (p_a=−6, proper p_b=-1), and finally tune UL (P0, alpha) once the UL path loss is unbiased. This order removes ambiguity, keeps risk low, and directly targets VoLTE’s sensitivity to bursty delay (HARQ) and control-channel robustness (PUCCH/PRACH). DPC is recommended to be off for deployment.

 

Abbreviations:

  • RU: Radio Unit
  • eNB: LTE base station
  • UE: User Equipment
  • RS / CRS / CSI-RS: Downlink reference signals
  • EPRE: Power per Resource Element
  • RSRP: Reference Signal Received Power
  • SINR / SNR: Signal to Interference plus Noise Ratio
  • SIB2: SystemInformationBlockType2
  • referenceSignalPower: Absolute CRS EPRE at RU output
  • PDSCH: Physical Downlink Shared Channel
  • PUSCH: Physical Uplink Shared Channel
  • PUCCH: Physical Uplink Control Channel
  • PRACH: Physical Random Access Channel
  • Pa/p_a: Data power offset relative to RS
  • Pb/p_b: RS scaling across multiple transmit ports
  • P0 Nominal PUSCH: UL baseline target for PUSCH
  • alpha: UL path loss compensation factor
  • Pmax/P_max: UE maximum transmit power
  • PRB: Physical Resource Block
  • M: Number of UL PRBs in a grant
  • CQI: Channel Quality Indicator
  • RI: Precoding Matrix Indicator(PMI), Rank Indicator
  • MCS: Modulation and Coding Scheme
  • HARQ: Hybrid Automatic Repeat reQuest
  • BLER / iBLER: Block Error Rate, instantaneous BLER
  • PHR: Power Headroom Report
  • QoE: Quality of Experience
  • VoLTE: Voice over LTE
  • BW: Bandwidth
  • MIMO 2T / 4T: 2 or 4 downlink transmit ports
  • PL: Path loss, approximately referenceSignalPower - RSRP
  • TPC: Transmit Power Control
  • DPC: Dynamic Power Control