CafeTeleRadio Optimization SRPO-600 Enroll now ↗
‹ Day 2 · Accessibility Course / Day 2 / 2.1 / 5G NSA Accessibility
NSA
3x
2.1 · 5G NSA · EN-DC

5G NSA Accessibility — SgNB Addition, in Full

In NSA (EN-DC, Option 3x) the UE keeps its LTE anchor for control plane and core (EPC) and gains a 5G leg by adding a secondary gNB. Samsung measures this as one KPI — SgNB Addition Success Rate (EndcAddSucc / EndcAddAtt, Family 5103) — split into a preparation phase (over X2 / E1 / F1) and an execution phase (NR PSCell access). This page walks it per stage: KPI → counters → failure counters → optimization, every name verbatim from the Samsung 5G NR Counter Description.

5103SGNB_ADD_PER_GNB family
2Phases: Prep · Execution
11EndcAdd failure counters
5085SCG-failure causes
EPCNo 5GC in NSA
The Samsung NSA accessibility KPI — SgNB Addition Success Rate

NSA accessibility is the LTE anchor chain (RRC/S1/E-RAB — see the LTE page) plus the NR-leg addition. The NR leg is one procedure with a preparation half (network side, over X2/E1/F1) and an execution half (over the air, NR PSCell). Samsung counts all three points in Family 5103 SGNB_ADD_PER_GNB:

Family 5103 — the three SgNB-Addition success rates
KPIFormula (Samsung counters)PhaseTypical target
SgNB Addition SR (end-to-end)EndcAddSucc ÷ EndcAddAtt × 100whole procedure≥ 99%
SgNB Addition Preparation SREndcAddPrepSucc ÷ EndcAddAtt × 100preparation (X2/E1/F1)≥ 99.5%
SgNB Addition Execution SREndcAddSucc ÷ EndcAddPrepSucc × 100execution (NR PSCell)≥ 99%
From a live Samsung network — the operator’s daily report tracks SgNb addition Success Rate % with its raw numerator/denominator, e.g. a cell at 100% (247 ÷ 247) and a loaded cell at 99.99% (19399 ÷ 19400). The split into Prep vs Execution tells you instantly whether a loss is network-side (X2/E1/F1/CU) or radio-side (NR PSCell).
The SgNB-Addition flow — where each counter pegs

On a B1 measurement report (NR neighbour better than threshold) the master eNB fires an X2AP SgNB Addition Request. The gNB admits and replies Addition Request Acknowledge (preparation done); the eNB reconfigures the UE, which performs random access on the NR PSCell; the UE returns Reconfiguration Complete and the eNB signals SgNB Reconfiguration Complete (execution done).

B1 → EndcAddAtt → PrepSucc → NR PSCell RACH → EndcAddSuccFamily 5103 peg points
SgNB Add ReqReconfig
UE + MeNB
B1 report → decide add
① EndcAddAtt
X2 SgNB Add Req
attempt
② EndcAddPrepSucc
Add Req Acknowledge
③ EndcAddSucc
Reconfig Complete
NR PSCell up
Execution (air)
NR PSCell RACH · Tdcoverall
Prep = ①→② (X2/E1/F1) · Execution = ②→③ (NR air)Control plane stays on LTE; the NR leg carries user plane only. No 5GC.
Read the counter name — the EN-DC failure-domain decoder

Every SgNB-Addition failure counter is named EndcAdd[Prep]Fail_<domain>. The suffix pins the failure to one interface/element — so the counter tells you which team owns the fix before you open a log.

Suffix → domain → first place to look
SuffixDomainMeaningFirst place to look
_MenbFailMeNB / X2SgNB Release Request from the MeNB, or the X2 link failedX2 (EN-X2) link, MeNB config, B1 trigger
_CpFailgNB-CU-CPFailure of gNB-CU-CP (control) blocks / gNBCU-CP health, NR admission, gNB load
_UpFail / _UpTimeoutgNB-CU-UP · E1Failure of the gNB-CU-UP or the E1 link (or an E1 message timeout)E1 link & CU-UP (user-plane) transport
_DuFail / _DuTimeoutgNB-DU · F1Failure of the gNB-DU or the F1 link (or an F1 message timeout)F1 / fronthaul & DU health
_TdcoverallAir · NR PSCellTdcoverall timeout — no SgNB Reconfiguration Complete during executionNR coverage / PSCell RACH, B1 threshold, Tdcoverall
!
Prep vs Execution is the first cut — rank the phase first: if Preparation SR is the loser, the fault is network-side (X2/E1/F1/CU) and no RF change will help; if Execution SR is the loser (dominated by EndcAddFail_Tdcoverall), it is NR coverage / PSCell access. Only then read the domain suffix.
Preparation phase — counters & failure causes (Family 5103, verbatim)

Preparation is the network handshake: the MeNB’s SgNB Addition Request (attempt) to the gNB’s Addition Request Acknowledge (prep success), spanning the E1 (CU-UP), F1 (DU) and CU-CP internals.

Attempt / preparation-success + the 6 preparation-fail counters
CounterDomainVerbatim description
EndcAddAttattemptNumber of SgNB Addition Request messages received from the MeNB.
EndcAddPrepSuccprep OKNumber of SgNB Addition Request Acknowledge messages transmitted to the MeNB after an SgNB Addition.
EndcAddPrepFail_UpTimeoutCU-UP · E1Call releases because of an E1 message timeout during the SgNB Addition Preparation procedure.
EndcAddPrepFail_DuTimeoutDU · F1Call releases because of an F1 message timeout during the SgNB Addition Preparation procedure.
EndcAddPrepFail_MenbFailMeNB · X2Call releases because of an SgNB Release Request from the MeNB or the X2 link failure during preparation.
EndcAddPrepFail_CpFailCU-CPCall releases because of a failure of gNB-CU-CP or gNB blocks during preparation.
EndcAddPrepFail_UpFailCU-UP · E1Call releases because of a failure from the gNB-CU-UP or the E1 link during preparation.
EndcAddPrepFail_DuFailDU · F1Call releases because of a failure from the gNB-DU or the F1 link during preparation.
Inside EndcAddPrepFail_CpFail — the CU-CP sub-cause taxonomy

CpFail is a bucket, not a diagnosis. SVR25B breaks it into named sub-counters — each one is a different team's problem. Rule of thumb: CallAlloc* = capacity/license · CpOverload = CU-CP load · SpCellSel* = "no usable NR cell for THIS UE" (capability, coverage, DU state) · UpSelectionFail = no CU-UP · ErabConfigurationFail = bearer config. The execution-phase twin EndcAddFail_CpFail means the E1/F1 modification failed after the UE was already told — same buckets, later timestamps.

The sub-causes (verbatim) → what to check → the fix
Sub-counterMeaning (verbatim basis)CheckFix
…CpFail_CallAllocFailVnmb / _CallAllocFailVccbCU-CP call-allocation resource / license exhaustedRPC license headroom · CU-CP call count vs dimensioningGrow license/capacity or shed load — this is a hard budget, not a tuning knob
…CpFail_CpOverloadCU-CP overload rejected the additionCU-CP CSL (Family 5100) · overload thresholdsOffload anchors / add capacity; the reject is the node protecting itself
…CpFail_SpCellSelFailure_UeNotSupCellFreqBand / _ChannelBw / _90MhzThe UE cannot use the target NR cell — band / bandwidth capability mismatchUE capability mix vs NR band/BW planProvide an alternative NR FA or adjust the BW profile — a planning fix, not a threshold
…CpFail_SpCellSelFailure_NrcgiNoCellSelectable / _NoNsaCellSelectableNo NSA-capable NR cell selectable where the add firedIs the NR cell up? NSA allowed? Is b1-threshold-nr-rsrp so low it adds where there is no real NR?Fix NR coverage/config, or raise the B1 gate so additions only fire on real coverage
…CpFail_SpCellSelFailure_DuOverload / _EnergySavingTarget DU overloaded or the cell is sleepingDU load · energy-saving schedule vs traffic hoursRelieve the DU; exclude ES cells from the add window (policy)
…CpFail_UpSelectionFailNo gNB-CU-UP selectable for the callAUPF pool state · E1 up · CU-UP capacityRestore/scale the CU-UP side
…CpFail_ErabConfigurationFailThe E-RAB configuration was invalid/unsupportedWhich QCI fails · PDCP/RLC profile for it · MeNB↔SgNB config alignmentConfig audit across both nodes for that bearer profile
i
How to run it — ① rank the sub-counters (Pareto) · ② one dominant sub-cause = one owner (license / CU-CP / planning / RF / CU-UP / config) — hand it with the counter as evidence · ③ SpCellSel* family climbing together = the B1 gate is adding UEs where no serviceable NR cell exists — check b1-threshold-nr-rsrp (61 ⇒ ≈−95 dBm) before touching anything else · ④ prove on EndcAddPrepSucc/EndcAddAtt recovery with fam 5208 (denied NSA) flat.
Execution phase — counters & failure causes (Family 5103, verbatim)

Execution is the over-the-air half: after preparation, the eNB reconfigures the UE, which does random access on the NR PSCell. End-to-end success pegs EndcAddSucc; anything that stops the UE reaching the NR cell pegs an EndcAddFail_*.

Execution success + the 5 execution-fail counters
CounterDomainVerbatim description
EndcAddSuccsuccessNumber of SgNB Reconfiguration Complete messages received from the MeNB after SgNB Addition.
EndcAddFail_TdcoverallAir · NR PSCellCall releases because of a Tdcoverall timeout for not receiving an SgNB Reconfiguration Complete (or an SgNB Release Request) during the completion procedure after preparation.
EndcAddFail_MenbFailMeNB · X2Call releases because of an SgNB Release Request from the MeNB or X2 link failure during the completion procedure.
EndcAddFail_CpFailCU-CPCall releases because of a failure of gNB-CU-CP blocks during the SgNB Reconfiguration procedure.
EndcAddFail_UpFailCU-UP · E1Call releases because of a failure from the gNB-CU-UP or the E1 link during the completion procedure.
EndcAddFail_DuFailDU · F1Call releases because of a failure from the gNB-DU or the F1 link during the completion procedure.
i
The signature that matters — a high EndcAddFail_Tdcoverall with a healthy Preparation SR means the network prepared the leg fine but the UE never reached the NR PSCell — pure NR coverage / PSCell RACH, or B1 firing in coverage the NR cell can’t hold. That is the single most common EN-DC accessibility fault.
SCG failure operation — why an established NR leg drops (Family 5085)

Once EN-DC is up, the NR leg can still fail. Family 5085 SCG_FAIL_OPER classifies every SCG failure by exact cause and by the operation it triggered (Count / Suspend / Resume) — the retainability companion to the accessibility KPI.

SCG failure causes (verbatim) — DL & UL
Counter (cause)Meaning
SCGFailOper_DL_t310Expiry_CountSCG failures caused by DL T310 expiry (radio-link failure on the NR leg)
SCGFailOper_DL_syncReconfFailure_CountDL Synchronization Reconfiguration Failure
SCGFailOper_DL_randomAccessProblem_CountDL Random Access Problem on the NR PSCell
SCGFailOper_DL_rlcMaxNumRetx_CountDL exceeding RLC maximum retransmission
SCGFailOper_DL_srb3IntegrityFailure_CountDL SRB3 Integrity Failure
SCGFailOper_DL_scgReconfFailure_CountDL SCG Reconfiguration Failure
SCGFailOper_UL_outOfSync_CountUL Out-of-Sync
SCGFailOper_UL_rlcMaxNumRetx_CountUL exceeding RLC maximum retransmission
SCGFailOper_UL_srb3IntegrityFailure_CountUL SRB3 Integrity Failure
Suspend / Resume — each cause also has _Suspend and _Resume variants, plus SCGFailOper_MrBased_Suspend/Resume (measurement-report-based). Governed by t310-timer / t312-timer ([ACPF] fast-rlf-recovery) and handled by FGR-BC1202 SCG Failure Handling for EN-DC. A DL-T310 / random-access-problem spike here echoes the same NR-coverage root cause as EndcAddFail_Tdcoverall.
SgNB-Addition Optimization Studio — counters → grade → weakest phase → fix

Enter the Family-5103 counters for a cell; the studio computes the three success rates, grades them, finds whether the loss is in preparation or execution, and prints the domain-specific fix from the largest failure counter. All levers are real Samsung MOs / features.

SgNB-Addition Optimization Studiolive
Family 5103 · SGNB_ADD_PER_GNB
Family 5085 · SCG_FAIL_OPER (retainability check)
SgNB Addition Success Rate
%
Prep × Execution
Recommended action
Diagnostic engine — failure Pareto & ranked action plan
Failure Pareto — who is eating the SRcomputed live from the studio counters
Ranked action plan — fix in this order
The EN-DC failure-routing fabric — domain × phase

Every SgNB-Addition failure counter is one wire from its domain (which element/interface owns the fault) to the phase where it pegs. Click a domain to light its counters across both phases, a phase to see everything that fails there, or click any counter chip for its full optimization card.

EN-DC Failure-Routing Fabric
Scenarios
The EN-DC feature pipeline — every accessibility feature (SVR25B)

The NR leg is added through four gates — trigger → prep transport → SgNB Addition → keep the leg. Each node is a real Samsung FGR feature; click any node to open its card: what it is, benefit, how it works, the recommended optimization parameters (MO · default · recommended), the optimization technique and the counters it moves.

EN-DC Accessibility Feature Pipelineclick a feature node to open its detail card
!
Order matters — a UE gets its NR leg via B1 trigger → X2/E1/F1 preparation → SgNB Addition → SCG-failure recovery. A Preparation loss is a transport/feature-config issue; an Execution loss is NR coverage. When SgNB-Addition SR drops, walk the pipeline left-to-right.
Optimization technique & parameters (verbatim)
EN-DC accessibility levers
LeverWhat it controlsTechnique
B1 event threshold (NR RSRP)When the MeNB decides to add the NR legSet around the NR cell-edge RSRP. Too high → EN-DC never sets up despite good NR; too low → added in coverage the NR cell can’t hold → EndcAddFail_Tdcoverall & SCG drops. Pair with hysteresis / time-to-trigger.
Tdcoverall (completion timer)Execution supervision — time to receive SgNB Reconfiguration CompleteGoverns when EndcAddFail_Tdcoverall fires; widen only if the air path is marginal but recoverable — the real fix is NR coverage / PSCell RACH.
t310-timer / t312-timer ([ACPF] fast-rlf-recovery)NR-leg radio-link-failure detection & fast recoveryBalances holding a weak SCG vs fast re-establishment; drives the SCGFailOper_DL_t310Expiry_* counters.
preamble-trans-max (NR PSCell RACH)Max preamble transmissions for NR PSCell accessRaise where PSCell access is coverage-limited; combine with CFRA for the added PSCell to cut execution failures.
FGR-BC1202 · SCG Failure Handling / FGR-CC5300 · MCG Failure HandlingHow SCG/MCG failures are recoveredEnsure enabled so an SCG failure is recovered (suspend/resume) rather than dropping the whole EN-DC leg.
X2 (EN-X2) linkPreparation transport MeNB↔gNBFix flaps / SCTP to cut EndcAddPrepFail_MenbFail; check E1 (CU-UP) and F1 (DU) transport for the _UpFail/_DuFail counters.
!
Work the phase, then the domain — Preparation loss = network/transport (X2/E1/F1/CU) — no RF change helps. Execution loss (Tdcoverall) = NR coverage / PSCell RACH / B1 too aggressive. The anchor LTE accessibility (RRC/S1/E-RAB) must also be healthy — an NSA cell can have perfect SgNB-Addition SR and still fail if the LTE anchor is failing.
B1-threshold trade-off explorer — the master EN-DC lever, interactive

The B1 threshold decides where the NR leg is added. Both failure directions are counted by the same families — drag the slider and watch which real counters a mis-set threshold feeds. Directional physics only: the counters and consequences are Samsung-exact; your network's exact RSRP number comes from the NR cell-edge survey.

B1 Trade-off Explorerdrag
◂ aggressive (low threshold — add NR early)conservative (high threshold — add NR late) ▸
The 7-step SgNB-Addition optimization MOP
1
Baseline — pull Family 5103 per cell

Collect EndcAddAtt / EndcAddPrepSucc / EndcAddSucc and all 11 failure counters for the busy hour. Compute the three SRs: end-to-end, Preparation, Execution.

Fam 5103 SGNB_ADD_PER_GNB · daily report "SgNb addition Success Rate %" Skip this and you tune blind — a healthy aggregate can hide one bad PSCell.
2
Split the loss — Preparation vs Execution

Prep SR = EndcAddPrepSucc ÷ EndcAddAtt; Exec SR = EndcAddSucc ÷ EndcAddPrepSucc. The lower one owns the loss: Prep = network-side, Exec = radio-side.

phase split — the first cut of every EN-DC investigation If wrong: RF teams chase transport faults and transport teams chase RF — weeks lost.
3
Rank the failure counters — the Pareto

Within the losing phase, rank the failure counters (the diagnostic engine above does this). The top counter's suffix names the owning domain: _MenbFail=X2 · _CpFail=CU-CP · _UpFail=E1 · _DuFail=F1 · _Tdcoverall=air.

EndcAddPrepFail_* · EndcAddFail_* — suffix = domain = team If wrong: fixing the #3 counter first wastes the maintenance window.
4
Rule out deliberate rejects

Check Tx_X2apSgNBAdditionRequestRejectByOvld (gNB admission shedding) and Energy-Saving cell status. A reject-by-design is capacity/policy, not a fault — enable endc-addition-with-switching-off-cells and retrial avoidance (cac-reject-cause 1/2) first.

Fam 5123 · cac-reject-cause · endc-addition-with-switching-off-cells If wrong: you "fix" RF on a cell that is simply admission-limited.
5
Apply ONE lever

Execution loss → raise B1 threshold toward the NR cell-edge (+ hysteresis/TTT), verify PSCell CFRA / preamble-trans-max. Prep loss → fix the owning link (X2/E1/F1) or CU health. One change per soak.

B1 · Tdcoverall · preamble-trans-max · X2/E1/F1 transport If wrong: two levers at once = you can't attribute the gain (or the regression).
6
Soak & verify — same counters, same hour

Re-measure Family 5103 over the same busy hour. The target counter must fall and the phase SR must rise; check SCGFailOper_* (Family 5085) did not rise — a B1 set too low converts addition failures into SCG drops.

before/after on Fam 5103 + Fam 5085 guard If wrong: an SR gain that ships SCG failures is a net loss for the user.
7
Lock in & watch retainability

Document the change; keep watching SgNB-Release (Fam 5105/5106 — FGR-CC5005 view) and the Blacklisted-Cell events. A cell that adds legs it cannot keep needs coverage work, not more parameter tuning.

FGR-CC5005 · FGR-BC1202 Blacklist · MENB/SGNB_INIT_SGNB_REL If wrong: ping-pong add/release burns signalling and battery — the KPI looks fine, the user suffers.
The X2AP message counters behind it (Family 5123, verbatim)

Preparation rides X2AP between MeNB and gNB. Family 5123 X2AP_MSG_UE_PER_GNB counts each message — the raw evidence behind the SgNB-Addition KPI.

Key X2AP SgNB-Addition messages
CounterVerbatim description
Rx_X2apSgNBAdditionRequestNumber of X2 SgNB ADDITION REQUEST messages received from the MeNB.
Tx_X2apSgNBAdditionRequestAcknowledgeNumber of X2 SgNB ADDITION REQUEST ACKNOWLEDGE messages transmitted by the SgNB.
Tx_X2apSgNBAdditionRequestRejectNumber of X2 SgNB ADDITION REQUEST REJECT messages transmitted by the SgNB.
Tx_X2apSgNBAdditionRequestRejectByOvldNumber of X2 SgNB ADDITION REQUEST REJECT messages transmitted because of overload.
Rx_X2apSgNBReconfigurationCompleteNumber of X2 SgNB RECONFIGURATION COMPLETE messages received from the MeNB.
i
Overload tie-in — Tx_X2apSgNBAdditionRequestRejectByOvld rising means the gNB is shedding EN-DC adds under load — an NR-capacity signal, not an RF one. Correlate with gNB UE-count (Family 5115 UE_NUM_PER_GNB).

NSA = anchor + a leg — prepare it, then reach it.

Read the LTE anchor first, then the SgNB-Addition split: preparation (X2/E1/F1/CU) vs execution (NR PSCell / Tdcoverall). To see native NR accessibility — RRC, Registration and PDU-Session on the 5GC — go to 5G SA.

Sources: Samsung 5G NR Counter Description — Family 5103 SGNB_ADD_PER_GNB, 5085 SCG_FAIL_OPER, 5123 X2AP_MSG_UE_PER_GNB (counter names & verbatim descriptions, via the operator KPI workbook & SVR25A CU/DU Counter Descriptions) · 5G RAN Feature Description SVR25B (SgNB Addition procedure, Tdcoverall, t310/t312 fast-rlf-recovery, FGR-BC1202 SCG Failure Handling) · live operator KPI report (SgNb addition Success Rate). EN-DC procedure per 3GPP TS 37.340. Targets are typical operator values; counters, families and mechanics are vendor-exact.