CafeTeleRadio Optimization SRPO-600 Enroll now ↗
‹ Day 2 · Retainability Course / Day 2 / 2.3 / 5G NSA Retainability
NSA
5G
2.3 · 5G NSA · EN-DC (Option 3x)

5G NSA Retainability — SCG Failure · SgNB Release, in Full

In EN-DC the NR leg can die without the call dying — the LTE anchor holds the connection while the SCG (Secondary Cell Group) fails, suspends, resumes or releases. So NSA retainability is its own discipline: the UE reports every NR-leg failure via SCGFailureInformationNR (over LTE RRC!), Samsung books it twice — on the SgNB (CU-CP Families 5085, 5106, 5100) and on the MeNB (LTE Families 717, 719, 788) — and the drop KPI must subtract the releases that are healthy mobility. Per ledger: KPI → counters → failure causes → optimization — every name verbatim.

5085SCG_FAIL_OPER ×23
5106SgNB-init release
5105MeNB-init release
5100CSL cause ledger
717·788LTE-anchor twins
5120E-RAB REL per gNB
The NSA drop KPI — count the failures, subtract the healthy mobility

The reporting headline is Session PS Drop Rate % (the operator's OMC recommended-KPI set). Underneath sits the deck-verbatim 5G Context Drop Rate — note what it does: adds the radio failures, subtracts MCG-mobility releases (the anchor moving is not a drop), and puts user inactivity in the denominator (a leg that fell asleep is not a drop either).

The NSA retainability KPIs (verbatim)
KPIFormula (Samsung counters)Read it as
5G Context Drop Rate [%] (deck, verbatim)((CauseInternalT310Expiry + CauseInternalRlcMaxNumRetx + EndcRelBySgnbMr + EndcRelByMenb − EndcRelByMenbMcgMob) ÷ (CauseInternalT310Expiry + CauseInternalRlcMaxNumRetx + EndcRelBySgnbMr + EndcRelByMenb − EndcRelByMenbMcgMob + CauseInternalUserInactivity)) × 100abnormal ends ÷ (abnormal + inactivity ends)
Session PS Drop Rate [%] (operator OMC)Nom ÷ Denom per cell from the recommended-KPI report — the NSA twin of "SgNb addition Success Rate"network-reported session loss
SCG failure intensity (derived)Σ SCGFailOper_*_Count (5085) ÷ SgNB Addition Succ (5103)failures per added leg — typical target <2%
Rescue share (derived)Σ *_Resume ÷ Σ *_Suspend (5085)how many suspended legs came back — the NSA rescue rate
Why the subtraction matters — EndcRelByMenbMcgMob (Fam 5113) counts SgNB Release Requests with cause MCG mobility: the LTE anchor is handing over, so the NR leg is released by design and re-added after — Family 5092 even counts the variant with "UE context kept" indicator. Counting those as drops would punish your busiest, best-moving cells. Same logic as LTE's normal-vs-abnormal split, EN-DC edition.
How the NR leg dies — and who writes it down

The SCG fails on the DL (the UE detects: T310 expiry, RACH problem, RLC max-retx, reconfig/integrity failure) or on the UL (the gNB detects: out-of-sync, UL RLC max-retx). The UE reports via SCGFailureInformationNR — carried over the LTE RRC connection — and the MeNB/SgNB then suspend, resume or release the leg. Every step pegs on both sides of X2.

SCG failure → SCGFailureInformationNR (via LTE!) → suspend / resume / releasedual ledger
SCG fails
DL: UE detects
UL: gNB detects
SCGFailureInformationNR
UE → MeNB over LTE RRC
cause carried verbatim
Decision
5085: Count → Suspend
fast-rlf-recovery window
RESUMED
SCGFailOper_*_Resume
LEG RELEASED
5106 EndcRelBySgnb_*
LTE call usually survives
The LTE-anchor ledger (same events)
717 EnDc_RelRequest/Require · 788 EnDc_ConnRelease_*
the report travels over LTE — a dead NR leg can still be reportedSuspend→Resume is the NSA rescue: FGR-BC1202 SCG Failure Handling keeps the leg configured and revives it without a full re-addition.
Read the counter name — the EN-DC release decoder

Three grammars cover all of NSA retainability. Learn them once and every counter self-explains:

Grammar → meaning → first place to look
PatternWho / whatMeaningFirst place to look
EndcRelBySgnb_DL_*UE detected (via SCGFailureInformationNR)DL-side SCG failure: t310Expiry · randomAccessProblem · rlcMaxNumRetx · syncReconfFailure · scgReconfFailure · srb3IntegrityFailureNR coverage/beam at the PSCell edge; NR t310/RACH
EndcRelBySgnb_UL_*gNB detectedUL-side failure: outOfSync · rlcMaxNumRetx · srb3IntegrityFailureUL link budget / interference — the en-dc UL-outage levers
SCGFailOper_<cause>_Count / _Suspend / _Resume (5085)the rescue machineSame causes, three checkpoints: reported → leg suspended → leg revivedResume÷Suspend ratio per cause — FGR-BC1202 health
EnDc_RelRequest_* (717, LTE)MeNB asks (X2 SgNB Release Request)Anchor-driven: ErabRelease · UeContxtRel · UserInactive · VolteConf · DL_RLF_t310Expiry…Anchor policy & mobility — mostly deliberate
EnDc_RelRequire_* (717, LTE)SgNB demands (X2 SgNB Release Required)NR-side driven: DcOverallTO · BearerNotSupp · UserInactive · UL_RLF_RadioConnection_UeLostThe NR leg's own failures, seen from LTE
CauseInternal* (5100 CSL)CU-CP call summaryThe per-release cause ledger — T310Expiry, RlcMaxNumRetx, RandomAccessProblemverbatim: "the UE transmits an SCGFailureInformationNR message following T310 expiry"KPI numerator — the drop formula reads this family
!
Request vs Require — on X2, RelRequest = the MeNB initiates (anchor decision), RelRequire = the SgNB initiates (NR-side necessity). The same release lands in Family 5105 (MeNB-init) or 5106 (SgNB-init) on the NR side, and in Family 717 on the LTE side. If the two ledgers disagree in trend, you're looking at an X2 signalling problem, not a radio one.
Ledger ① — SCG Failure Operation (Family 5085) & the CSL causes (Family 5100)

The family (verbatim): "statistics for handling SCG failure per SCG failure cause" — 23 counters: for each cause a _Count, and for the recoverable causes a _Suspend and _Resume pair. This triple is the rescue instrumentation: Count says it happened, Suspend says the leg was parked instead of killed, Resume says it came back.

①-a · The cause set — with its CSL twin (verbatim descriptions)
Cause (5085 triple)SideCSL twin (5100) — verbatim meaning
SCGFailOper_DL_t310Expiry_*DL · UECauseInternalT310Expiry: release because "the UE transmits an SCGFailureInformationNR message following T310 expiry"KPI numerator
SCGFailOper_DL_randomAccessProblem_*DL · UECauseInternalRandomAccessProblem: random access problem indication by SCG MAC
SCGFailOper_DL_rlcMaxNumRetx_*DL · UECauseInternalRlcMaxNumRetx: SCG RLC reached max retransmissions — KPI numerator
SCGFailOper_DL_syncReconfFailure_* / _scgReconfFailure_*DL · UECauseInternalScgReconfigFailure: reconfiguration failure of the NR RRC Reconfiguration message
SCGFailOper_DL_srb3IntegrityFailure_*DL · UECauseInternalSrb3IntegrityFailure: SRB3 IP check failure — security, not radio
SCGFailOper_UL_outOfSync_* / _UL_rlcMaxNumRetx_*UL · gNBgNB-detected uplink loss — the half the UE cannot report; pairs with the en-dc-ul-radio-link-outage levers
SCGFailOper_MrBased_Suspend / _ResumeMR · policyCauseInternalMeasurementReportBasedRelease: release on an event A2 measurement report — coverage-based suspension, deliberate
①-b · Optimization playbook — dominant cause → diagnosis → lever
Dominant causeDiagnosisTechnique / lever
DL_t310ExpiryPSCell coverage/beam collapse — the NSA analogue of the LTE RLF dropNR coverage at the PSCell edge (SSB beams, tilt); NR t310/n310 widening for fade-rich cells; B1 threshold hygiene so weak PSCells aren't added at all
DL_randomAccessProblemSCG RACH breaks after addition — UL coverage or RACH config on NRNR RACH levers (preamble-trans-max, power-ramping-step, prach-configuration-index); check per-beam RACH (5243) on the PSCell
DL/UL_rlcMaxNumRetxSustained bad link the ARQ can't bridgeNR RLC depth (gnb/ue-max-retransmission-threshold — deck rec 8), BLER targets (dl-target-bler 8% / ul 10%), MCS caps under congestion
UL_outOfSyncUL dies first — classic NSA edge (UE TX power split between LTE+NR)The EN-DC UL-outage machinery: en-dc-ul-radio-link-outage-support enable, …-count-threshold +50%, …-resume-max-tbs-threshold +20% (deck-verbatim moves)
MrBased_*A2-based suspends — policy workingIf excessive: the A2/B1 thresholds are fighting (add-drop ping-pong) — widen the hysteresis between addition B1 and release A2
srb3IntegrityFailureSecurity/config, not radioKey-refresh & SRB3 configuration audit; escalate as SW issue if systematic
Ledger ② — SgNB-initiated release (Families 5106 · 5111 · 5114)

The family (verbatim): "statistics for the SgNB initiated SgNB release for each cell and MeNB" — the NR node giving the leg up, cause-split with the same DL/UL grammar, plus two deliberate causes. EndcRelBySgnbMr (Fam 5111, KPI numerator) counts "SgNB Release Required messages transmitted to the MeNB" on measurement-report grounds.

②-a · The 12 causes of 5106 (verbatim names)
CounterClassMeaning
EndcRelBySgnb_DL_t310Expiry / _DL_randomAccessProblem / _DL_rlcMaxNumRetxradio · DLThe UE-reported SCG failures that ended in release (not resume)
EndcRelBySgnb_DL_syncReconfFailure / _DL_scgReconfFailure / _DL_srb3IntegrityFailureconfig · securityReconfiguration / integrity failures — procedure-level causes
EndcRelBySgnb_UL_outOfSync / _UL_rlcMaxNumRetx / _UL_srb3IntegrityFailureradio · ULgNB-detected uplink loss ending in release
EndcRelBySgnb_MrBasedSuspendResumepolicyMR-based suspend/resume flow ending in release
EndcRelBySgnb_UserInactivitynormalLeg released because it fell idle — healthy, keep out of the drop numerator
i
Playbook — compute the DL:UL ratio of 5106. DL-heavy = PSCell coverage/config (UE sees it first) → beams/tilt/B1. UL-heavy = uplink budget (gNB sees it first) → UL-outage levers, check UE TX-power sharing (LTE+NR simultaneous). _UserInactivity dominant = traffic profile, not a fault — but if the LTE ledger shows EnDc_RelRequire_UserInactive diverging from it, the two inactivity timers (LTE vs NR) are fighting: align them.
Ledger ③ — MeNB-initiated release (Families 5105 · 5113 · 5092) & the LTE-anchor twins (717 · 788)

The anchor's view. Family 5105 (verbatim): "MeNB initiated SgNB release for each cell and MeNB" — including EndcRelByMeNB_radioConnectionWithUeLost: the LTE anchor lost the UE entirely, taking the NR leg with it. On the LTE side, Family 717 splits the same X2 procedure into RelRequest_*/RelRequire_* causes, and Family 788 books the release of the whole EN-DC-capable RRC connection with the full LTE 25-cause grammar.

③-a · MeNB-initiated causes — both sides of X2 (verbatim)
CounterLedgerMeaning
EndcRelByMenb (5105)NR sideSgNB Release Request messages received from the MeNB — the anchor asked
EndcRelByMeNB_radioConnectionWithUeLost (5105)NR sideThe anchor lost the UE — an LTE drop cascading into the NR leg; fix the LTE page's problem, not NR
EndcRelByMenbMcgMob (5113)NR sideCause MCG mobility — the anchor is handing over; subtracted from the drop KPI
EndcRelByMenbMcgMobUeCtxInd (5092)NR sideMCG mobility with UE context kept — the leg survives the anchor HO (SN-terminated bearer kept)
EnDc_RelRequest_UserInactive / _VolteConf / _ErabRelease / _UeContxtRel / _InternalResouFail (717)LTE sideWhy the anchor asked: inactivity, VoLTE conflict (voice arrives → NR leg dropped by policy), bearer/context teardown, internal resource fail
EnDc_RelRequire_DcOverallTO / _BearerNotSupp / _UL_RLF_RadioConnection_UeLost (717)LTE sideWhy the SgNB demanded: Tdcoverall expiry (the EN-DC guard timer), unsupported bearer, UL RLF with UE lost
EnDc_ConnRelease_* (788, 25 causes)LTE sideThe whole EN-DC-capable connection released — LTE grammar (RrcHcTo/UpMacUEInact/S1ap*/CpPreemption/MmeOverload…)
SgNB Release due to SCG / MCG Mobility / Re-establishment (707 · 710 · 711)LTE sideMobility-caused releases, split by which side moved — plus re-establishment-triggered
The VoLTE conflict counter — EnDc_RelRequest_VolteConf is unique to NSA: a voice call arriving on the LTE anchor can evict the NR leg (UE capability / power sharing). If it dominates, that's a policy choice — verify the intended behaviour (keep EN-DC during VoLTE vs drop) matches the configuration, and check the UE capability mix before "fixing" it.
Ledger ④ — the bearer view (Families 5120 · LTE 719 · 948 · 5220)

Sessions are bearers. Family 5120 (NR): "E-RAB Release per gNB" — 12 counters. LTE Family 719 books EN-DC E-RAB releases by initiator (byMme / byEnb / bySgnb) per bearer type — MCG, SCG, or split — and Family 948 covers the split-bearer specifics. Family 5220 (DRB Release and Erase) is the NR DRB-level twin.

i
Playbook — read 719 per endc_bearerType: SCG-bearer releases tracking 5106 = radio story already told; split-bearer releases without SCG failures = the PDCP-split path (X2-U capacity, flow control); bySgnb spikes = NR-side, byMme = core-driven (same treatment as the LTE page's by-MME family). Bearer-type divergence is the fastest NSA triage cut after DL:UL.
Interactive — the SCG-failure flow explorer

Pick a failure cause and the recovery policy; watch the flow route through report → suspend → resume/release and read exactly which counters peg on the NR and LTE ledgers. This is the 5085 triple, animated.

SCG Failure Flow Explorerlive
The failure
User impact
Counters pegging — NR ledger
Counters pegging — LTE anchor ledger
NSA Retainability Studio — the drop formula, live

Enter the five formula counters plus the rescue pair; the studio computes the deck-verbatim 5G Context Drop Rate, the rescue share, and the healthy-mobility share — then routes you to the dominant cause.

NSA Retainability Studiolive
The formula counters (5100 · 5111 · 5105 · 5113)
The rescue machine (5085) & context
5G Context Drop Rate (deck formula, verbatim)
%
Recommended action
The NSA release routing fabric — root cause × ledger

Every EN-DC release counter routed from its root-cause domain to the ledger where it books — NR CU-CP families on top of the LTE-anchor twins. Click a domain, a ledger, a scenario, or any counter chip for its optimization card.

NSA Release Routing Fabric
Scenarios
The NSA retainability parameter set — the EN-DC-specific levers (deck-verbatim)

NSA shares the NR radio levers with SA (BLER/HARQ/RLC — full table on the SA page); what's unique to EN-DC is the UL-outage machinery, the guard timer, and the mobility couplings:

EN-DC-specific levers (deck-verbatim recommendations)
ParameterDefaultDeck recommendationWhy / target scenario
en-dc-ul-radio-link-outage-supportdefaultEnabled"Improve mobility robustness for NSA" — arms the UL-outage detection that feeds UL_outOfSync handling
en-dc-ul-radio-link-outage-count-thresholdnetwork defaultIncrease by 50%"More tolerance for temporary path issues" — fewer premature UL-outage suspends
en-dc-ul-radio-link-resume-count-thresholdnetwork defaultDecrease by 30%"Faster path recovery" — resume the leg sooner
en-dc-ul-radio-link-resume-max-tbs-thresholdnetwork defaultIncrease by 20%"Faster recovery from path issues"
Tdcoverall (EN-DC guard)release defaultraise only on EnDc_RelRequire_DcOverallTO evidenceThe overall EN-DC procedure guard — its expiries are counted verbatim in Fam 717
handover-margin-nr3 dB1–2 dB in dense areas"Less aggressive handovers in dense areas" — steadier PSCell choice
time-to-trigger (NR events)480 ms160 ms high-mobility"Faster handovers in high mobility scenarios" — before the SCG dies
B1 (add) vs A2 (release) thresholdsplankeep a clear hysteresis gapAdd-release ping-pong manufactures MrBased_* suspends and 5111 releases
gnb/ue-max-retransmission-threshold (RLC)release default8Deck's RLC retainability set — bridges bad patches before rlcMaxNumRetx fires
dl-target-bler / ul-target-bler~10%8% / 10%Deck RRC & BLER optimization table — link adaptation headroom
!
The NSA-specific trap — the UE's transmit power is shared between LTE and NR. A "5G drop problem" that is UL-heavy (UL_outOfSync, EnDc_RelRequire_UL_RLF…) at cell edge is often a power-split problem: the LTE anchor is consuming the budget. Check LTE UL health and the power-sharing configuration before touching NR timers.
The 6-step NSA retainability MOP
1
Baseline both ledgers

NR: 5085 + 5106 + 5105/5113 + 5100 causes + 5120. LTE: 717 + 719 + 788. Compute the deck drop formula and the rescue share (Resume÷Suspend).

dual ledger — always both sides of X2 Skip this and an X2 signalling problem masquerades as a radio one.
2
Subtract the healthy releases

MCG-mobility (5113/5092), user inactivity (both sides), VoLTE-conflict policy, MR-based suspends that resumed. What remains is the true drop mass.

the formula's own subtraction, extended If wrong: your busiest cells look like your worst cells.
3
Cut DL vs UL

DL-heavy (t310Expiry, randomAccessProblem, reconf) = PSCell coverage/config — the UE saw it first. UL-heavy (outOfSync, UL rlcMaxNumRetx) = uplink budget/power-split — the gNB saw it first.

the DL:UL ratio of 5106/5085 If wrong: you tilt antennas for an uplink power problem.
4
Check the rescue machine

Per cause: Suspend high + Resume low = the leg parks but never returns — work the resume thresholds (en-dc-ul-…-resume-*) and FGR-BC1202 config before chasing the failure itself.

Resume ÷ Suspend per cause (5085) If wrong: you prevent 10 failures while 100 rescues stay broken.
5
Apply ONE lever

DL radio → PSCell RF + NR t310/RACH. UL → the four en-dc-ul-outage levers (deck moves: +50% / −30% / +20%). Mobility → handover-margin-nr, TTT, B1/A2 gap. Policy → align inactivity timers, verify VolteConf intent.

one change per soak If wrong: unattributable results.
6
Verify on both ledgers + the anchor

The cause must fall on 5085/5106 and its 717 twin; SgNB Addition SR (5103) must hold (don't fix drops by not adding legs); LTE-anchor drop rate must hold (Fam 24 — the other page).

guards: 5103 addition SR · LTE Fam 24 If wrong: the NSA "gain" was bought by breaking LTE.

NSA retainability is two ledgers, one leg, and a rescue machine.

Count the SCG failures, subtract the healthy mobility, read DL vs UL, and keep Resume÷Suspend high. Then compare with LTE's three ledgers and SA's clean single-node chain.

Sources: Samsung 5G NR Counter Description (CU-CP ACPF — Families 5085 SCG_FAIL_OPER, 5106/5111/5114 SgNB-initiated release, 5105/5113/5092 MeNB-initiated release, 5100 CUCPCSL_PER_GNB incl. the verbatim CauseInternalT310Expiry/RlcMaxNumRetx/RandomAccessProblem descriptions, 5120 ERAB_REL_PER_GNB, 5220 — via the operator NR KPI workbook) · Samsung LTE eNB Counter Description (EN-DC Families 717/719/788/948, 707/710/711 — via the operator LTE KPI workbook) · 5G NR Retainability Optimization SA deck (5G Context Drop Rate formula verbatim, en-dc-ul-radio-link-outage recommendations, mobility table) · operator OMC recommended-KPI report (Session PS Drop Rate). EN-DC procedures per 3GPP TS 37.340. Targets marked "typical" are operator practice; counters, families and the formula are vendor-exact.