CafeTeleRadio Optimization SRPO-600 Enroll now ↗
‹ Day 3 · Connected-Mode Mobility Course / Day 3 / 3.1 / 5G NSA Mobility
NSA
5G
3.1 · 5G NSA · EN-DC

5G NSA Mobility — PSCell Change · Anchor Handover · the Add-Back Choreography

An EN-DC UE moves on two radios at once — so NSA mobility is three coupled planes: the NR leg moves (PSCell change — Family 5109 intra-SN, 5110 inter-SN), the LTE anchor moves (the EN-DC twins of every LTE handover leg — Families 748/749/750, counted separately for EN-DC-capable UEs), and the choreography between them: every anchor move must decide whether the SgNB is kept, changed or re-added (5090/5095/5096, LTE 706/708/709/710, modification 5112). Even refusals are counted (906: PSCell change blocked by VoLTE conflict or radio-link outage). Per plane: KPI → counters → failure causes → optimization — every name verbatim.

5109intra-SN PSCell chg
5110inter-SN PSCell chg
748·749·750ENDC anchor HO
5090·95·96add-back at HO
906refusals VolteConf/RLO
793EN-DC idle LB
The NSA mobility KPI set — three planes, three success rates
The KPIs (family-derived, counters verbatim)
KPIFormula (Samsung counters)Read it as
Intra-SN PSCell Change SREndcIntraChgSucc ÷ EndcIntraChgAtt × 100 (Fam 5109)NR-leg moves inside one SgNB — beam/cell change, no X2 SgNB change
Inter-SN PSCell Change SREndcInterChgSrcSucc ÷ EndcInterChgSrcAtt × 100 (Fam 5110, source view)The NR leg moves to a different SgNB — the heavy procedure
Anchor EN-DC HO SR(IntraEnbSucc + InterX2OutSucc + InterS1OutSucc) ÷ (…Att) over Fams 748/749/750The LTE composite, recounted for EN-DC-capable UEs only
SgNB survival at anchor HOEndcRelByMenbMcgMobUeCtxInd (5092) ÷ EndcRelByMenbMcgMob (5113)How often the NR context survives the anchor move (context-kept path)
Add-back SR after MCG HOEndcAddLteHoNoSgnbChgSucc ÷ …Att (5090) · EndcAddMnHoIntra/InterEnbSucc ÷ …Att (5095/5096)The leg returning after the anchor moved
Refusal pressureFRChgRefuse_VolteConf + FRChgRefuse_RLO (Fam 906) vs change attemptsPSCell changes the policy refused — deliberate, but count them
Why three planes — a user crossing a city block can trigger: a PSCell change (NR moved, LTE stayed), an anchor HO with add-back (LTE moved, NR re-added), or both. Each books in different families on different nodes. Reading only one plane makes the other two invisible — the classic NSA mobility mistake.
The choreography — what happens when an EN-DC UE moves
the three paths of an EN-DC move — and where each booksdual ledger
EN-DC UE moves
MeNB decides the path
B1/A3 evidence · 742
① NR leg only
PSCell change 5109 / 5110
② Anchor only
ENDC HO 748/749/750 + keep SgNB (5090)
③ Both move
MCG HO + SgNB change (708 IntraMenbToMenbSgnbChg)
Books on SgNB
EndcIntra/InterChg* · refusals 906
Books on MeNB
748-753 twins · 5090 add-back
Books on both
5112 mod · 5113/5092 rel · 706/708
Guards
Tdcoverall fails
RLO / VolteConf refusals
the MeNB is the choreographerevery path starts with an LTE-side decision — NR-side failures often have anchor-side causes (_MenbFail).
Read the counter name — the PSCell-change decoder
Grammar → meaning → first place to look
TokenMeaningFirst place to look
EndcIntraChg* / EndcInterChgSrc*PSCell change inside one SgNB / across SgNBs (source view)Intra healthy + inter failing = the X2/SgNB-to-SgNB path, not radio
…PrepFail_ / …Fail_The same prep/exec split as every Samsung HO familyPrep = target resources/anchor · Exec = UE/air/guard timer
_MenbFailThe anchor failed the procedure — unique to NSAThe LTE ledger (748-753, X2 health) — an NR symptom with an LTE cause
_UpTimeout/_UpFail · _DuTimeout/_DuFailCU-UP / DU didn't deliver during the changeSame E1/F1 treatment as the SA pages
_TdcoverallThe EN-DC overall guard timer expired mid-procedureX2 latency; raise Tdcoverall only on counted evidence
_RrcToThe UE never confirmed on the NR sideNew PSCell coverage/beam — trigger placement
FRChgRefuse_* / FRChgRelease_* (906)PSCell change refused (kept old) / released (leg dropped) by policy: VolteConf (voice active) or RLO (UL radio-link outage)Policy intent + the retainability page's UL-outage levers
EndcAdd<path>* (5090/5095/5096)SgNB addition because of an LTE HO — no-SgNB-change / intra-eNB / inter-eNB pathsAdd-back failing = EN-DC users lose 5G at every anchor move
!
The NSA mobility superpower — _MenbFail appearing in the NR families is the cross-ledger arrow: the SgNB says "the anchor broke it". Correlate with the LTE ENDC-HO twins (748-753) and X2 health (943/944 link-time families) — a PSCell-change "problem" is frequently an anchor-HO or X2 story.
Plane ① — PSCell change (Families 5109 · 5110 · 906)

The NR leg's own mobility. Intra-SN (5109) is the light procedure — same SgNB, new PSCell. Inter-SN (5110) rebuilds the leg at another SgNB — every interface can fail it. Family 906 counts the changes policy refused: verbatim FRChgRefuse_VolteConf (voice call active on the anchor — don't disturb), FRChgRefuse_RLO (UL outage — don't move a leg that's already suspended), and their FRChgRelease_* twins when the refusal ends in release.

①-a · Optimization playbook
SymptomDiagnosisLever
…Fail_RrcTo dominantNew PSCell not reachable — trigger too late or weak target beamNR A3/A5 trigger for PSCell change; handover-margin-nr 3→1–2 dB dense (deck); per-beam RACH at the target (5243)
…PrepFail_MenbFail / Fail_MenbFailThe anchor failed the X2 leg of the changeLTE ledger: 748-753 health, X2 SCTP (943/944); this is LTE work
…Fail_TdcoverallGuard expiry — slow X2 or peerX2 latency first; Tdcoverall raise only on counted evidence (same rule as retainability page)
Inter-SN ≪ intra-SN SRThe SgNB-to-SgNB path (X2 + target SgNB resources)Target SgNB CAC/capacity; X2 between SgNBs; check 5110 target-side twin if licensed
FRChgRefuse_RLO climbingChanges blocked because the leg is in UL outageFix the outage first (retainability page's en-dc-ul-radio-link-outage-* set) — the refusal is a symptom
FRChgRefuse_VolteConf highVoice-first policy blocking NR movesVerify intent; if VoLTE sessions are long, the NR leg may stay pinned to a weakening PSCell — consider release-vs-keep policy per market
①-b · The 5109 failure taxonomy — every prep/exec sub-counter decoded

Family 5109 names the culprit inside the counter. Read it in two gates: Prep SR = EndcIntraChgPrepSucc / EndcIntraChgAtt (nothing sent to the UE yet — failures are node/transport), then Exec SR = EndcIntraChgSucc / EndcIntraChgPrepSucc (UE is reconfiguring — failures are air-interface or MeNB). The interface is encoded in the peg: Up = CU-UP / E1 · Du = gNB-DU / F1 · Menb = X2 / MeNB · Cp = gNB-CU-CP. And the suffix style tells the mechanism: *Timeout = the answer never came (latency / overload) · *Fail = an explicit failure or reject (fault / alarm / resource) — timeouts want capacity+transport work, fails want the alarm cleared first.

PREPARATION phase — verbatim cause → what to check → the fix
Counter (5109)Verbatim root causeCheckFix
EndcIntraChgPrepFail_UpTimeoutE1 message timeout in prepCU-UP CPU/load · E1 SCTP latency/dropsRelieve CU-UP, fix E1 transport — capacity/transport, not a radio knob
EndcIntraChgPrepFail_DuTimeoutF1 message timeout in prepDU CPU/PRB/admission · F1 SCTP · cell not in energy-savingRelieve DU, fix F1
EndcIntraChgPrepFail_MenbFailX2 link failure in prepend-point-x2 … sctp-state · x2-sctp-stop=false · x2-failover-timerHeal X2 — read the anchor's term-point-to-enb/sctp-state suffix for the reason (comm-lost vs setup-timeout vs sctp-oos)
EndcIntraChgPrepFail_CpFailgNB-CU-CP block failure in prepCU-CP CSL (5100) · admission (5148) · license/RPC headroomRelieve CU-CP; SW fault → escalate with evidence
EndcIntraChgPrepFail_UpFailCU-UP or E1 explicit failure in prepAUPF alarms · E1 association · resource pool (7004)Clear the CU-UP/E1 alarm first
EndcIntraChgPrepFail_DuFailgNB-DU or F1 explicit failure in prepDU cell fault · F1 down · target PRB/RACH/CCE admissionRestore DU/F1, relieve target admission
EXECUTION phase — the big three + the twins
Counter (5109)Verbatim root causeCheckFix (verbatim knobs)
EndcIntraChgFail_RrcToNo RRCReconfigurationComplete — UE failed random-access to the new PSCell (air interface)Target RSRP/SINR at the change point · target PRACH/beams · did A3 fire too late?Trigger earlier: a3-pscell-change-offset-delta (0, −30..30) + hysteresis/TTT; give RA time: ue-timer-const-info/t304 ms1000→ms2000 only if RA is slow-but-succeeds — t304-first on a weak target just delays the drop
EndcIntraChgFail_TdcoverallTdcoverall expiry — no X2 SGNB MODIFICATION CONFIRM from the MeNBX2 RTT · MeNB load · LTE-anchor RLFtimer-info/x2-dc-overall (3000, 100..65535) → 4000–5000 only if X2 latency is genuinely high; raising the timer masks, it doesn't cure
EndcIntraChgFail_MenbFailX2 SGNB MODIFICATION REFUSE / SGNB RELEASE REQUEST from the MeNB, or X2 downLTE anchor RLF/coverage · MeNB CAC/load · X2 healthThe root lives on the LTE side — fix anchor coverage / MeNB admission, not NR knobs
EndcIntraChgFail_UpTimeout / _DuTimeout / _CpFail / _UpFail / _DuFailE1/F1 timeouts and explicit CU-CP/CU-UP/DU failures during executionSame buckets as prep, at exec timestampsIdentical runbooks to the prep twins — group by interface and sum both phases per interface
i
Analysis order — ① split prep-heavy vs exec-heavy · ② within prep, group by interface (one interface dominating = one node/link; broad spread = generic CU/DU overload → capacity) · ③ within exec: RrcTo → air · Tdcoverall → X2/MeNB latency · MenbFail → LTE anchor · ④ prove the fix on Succ/Att plus no new drops in 5085/5105/5106, and watch the change-time ledger — Family 5228 splits it into IntraSNSPcellChgSRB3Tot/Cnt (NR air via SRB3), IntraSNSPcellChgMCGSRBTot/Cnt (X2 + LTE SRB), InterSNSPcellChgTot/Cnt (SGNB CHANGE REQUIRED→CONFIRM) — whichever average creeps names the slow leg. Inter-SN twin: family 5110 repeats the whole taxonomy as EndcInterChgSrc* and adds EndcInterChgSrcPrepFail_X2LinkFail.
①-c · Band-change recipe — LTE 1800 anchor stays, NR leg moves 3500 → 800

The classic coverage move: the UE keeps its LTE 1800 anchor and swaps the NR leg from n78 to the low band. In Samsung terms that is an inter-frequency EN-DC PSCell change — counted in 5109 if the same gNB serves both bands, 5110 if a different gNB serves 800 ("intra/inter-SN" is about the node, never the frequency). The trap: every enable in this chain ships OFF — until you flip them, EndcIntraChgAtt stays at zero and the UE rides n78 into the ground.

The enable chain (all verbatim SVR25B, all default disable)
StepKnob (verbatim MO leaf)Default → setFeature
Master enable (SPID)mobility-config-gnb/spid-based-inter-frequency-pscell-change-enableddisable → enableFGR-CC5001
Master enable (freq-priority)mobility-config-gnb/fp-based-inter-frequency-pscell-change-enableddisable → enableFGR-CC5001
Target-band report A4long-periodic-measurement/en-dc-spid-inter-frequency-nr-pscell-change-purpose-a4-enableddisable → enableFGR-MC1014
…or safer A5 (serving-below + target-above)…-a5-enableddisable → enableFGR-MC1014
"n78 weak → measure 800"a2-report-config/a2-purpose = en-dc-activate-inter-frequency-purpose(0)configure on the high bandFGR-CC5010
"n78 recovered → stop"a1-report-config/a1-purpose = en-dc-deactivate-inter-frequency-purpose(0)configureFGR-CC5010
Bias toward 800…/a3-spid-inter-frequency-pscell-change-offset-delta0 (−30..30)FGR-CC5001
i
Flow — UE on 3500 → A2 (serving weak, activate-inter-frequency purpose) → gNB configures NR inter-freq measurement on 800 → A4/A5 (800 good) → PSCell change → booked in 5109/5110; success-rate KPI Succ/Att, speed KPI from 5228. Note this is the connected-mode MR-based path; the SCG-band steering at addition time (scg-band-selection-mode, FGR-LC0704) solves the same problem earlier — right band from the start beats changing later.
Plane ② — Anchor handover, EN-DC edition (Families 748/749/750/752/753 · 802 · 742)

Every LTE leg from the LTE mobility page exists again for EN-DC-capable UEs — same grammar (IntraEnbAtt…), separate families. Why: an EN-DC anchor HO carries extra freight (SgNB context, X2 forwarding), so its SR deserves its own ledger. Family 802 counts inter-frequency HOs made for EN-DC — the anchor moved to a better EN-DC band on purpose. Family 742 IRAT_NR_MR_CLASSIFICATION classifies the B1 measurement evidence steering it all.

i
Playbook — compare the EN-DC composite (748/749/750) with the general composite (17/18/20): a systematic gap = the EN-DC freight is the problem (X2 load, SgNB choreography), not LTE mobility itself. 802 attempts without SgNB additions following = the steering moved anchors but the B1 layer never fired — check the B1 thresholds (accessibility page) and scg-band-selection-mode. Idle twin: Family 793 EN_DC_IDLE_LB (EnDc_IdleLbCnt/UECount/TargetEarfcnDL) steers idle EN-DC-capable UEs to anchor-friendly carriers before they even connect — with the anchor-selection-enable threshold set (LTE LB deck, verbatim: high-load-threshold, delta-anchor-threshold-cell-load, anchor-area-option-for-Idle-endc, nr-priority-option-for-idle-endc).
Plane ③ — the add-back choreography (Families 5090 · 5095 · 5096 · 5112 · 706/708)

When the anchor moves, the SgNB must be kept, modified or re-added — and Samsung counts each path from both sides:

③-a · The choreography families (verbatim)
FamilyPathCounters (grammar verbatim)
5090 (NR)Intra-LTE HO, SgNB unchanged — the cheapest pathEndcAddLteHoNoSgnbChgAtt/PrepSucc/Succ + PrepFail/Fail_<Up|Du|Menb|Cp|Tdcoverall>
5095/5096 (NR)SgNB addition because of intra/inter-eNB anchor HOEndcAddMnHoIntraEnb* / EndcAddMnHoInterEnb* — the add-back after a move
5112 (NR)SgNB modification with MCG mobility — leg adjusted, not droppedEndcModByMenbMcgMobAtt/PrepSucc/Succ + fails
706 (LTE)SgNB change (SCG mobility) — the LTE viewSgnbChange_AddAtt/PrepSucc/Succ · Sgnb_AddPrepFail_DcPrepTO/_Reject · Sgnb_AddFail_RrcReconfigTO/_InternalFail/_Rre/_UeCtxtRel
708/709/710 (LTE)SgNB add/mod/release due to MCG mobility — per anchor pathIntraEnbToMenb_Add* · InterEnbToMenb_Add* · IntraMenbToMenbSgnbChg_Add*
The retention chain — for a moving EN-DC user the 5G experience = anchor HO SR × (keep-share + add-back SR × (1−keep-share)). Maximize the keep path first (5090 + context-kept 5092 — no interruption at all), then the add-back SR. An excellent add-back SR with a low keep-share still means a 5G blip at every anchor move — the choreography targets, in order: keep > modify > re-add > release.
Interactive — the MCG-move decision explorer

Pick the move and the conditions; see which of the three paths fires, what books on each ledger, and the 5G-continuity verdict.

MCG-Move Decision Explorerlive
5G continuity verdict
NR / SgNB ledger
LTE / MeNB ledger
NSA Mobility Studio — three planes, graded
NSA Mobility Studiolive
Moving-user 5G continuity (anchor SR × retention)
%
Recommended action
The EN-DC mobility parameter set (deck-verbatim)
The levers unique to NSA mobility (LTE radio levers → LTE page · NR radio levers → SA page)
ParameterDefault / typicalDeck guidance (verbatim where quoted)
handover-margin-nr3 dB"1–2 dB (lower value) — less aggressive handovers in dense areas" → steadier PSCell choice
time-to-trigger (NR events)480 ms"160 ms — faster handovers in high mobility scenarios"
B1 (add) vs A2 (release/change) thresholdsplanKeep a clear gap — add/change ping-pong manufactures PSCell churn (Fam 742 classifies the MR evidence)
anchor-selection-enable + high-load-threshold / delta-anchor-threshold-cell-loadenable in EN-DCIdle anchor steering (Fam 793) — "steer EN-DC capable idle UEs to less loaded, suitable anchor carriers" (LB deck verbatim); pair with anchor-area-option-for-Idle-endc, nr-priority-option-for-idle-endc, load-balancing-idle-mode-anchor-prefer-earfcn-list
scg-band-selection-mode + scg-band-threshold-capacity / -num-uescg-band-select-capacity · 70–80%"Avoid adding an SCG leg that will perform poorly due to NR congestion" (LB deck verbatim)
Tdcoverallrelease defaultThe guard behind every _Tdcoverall fail — X2 latency first, raise only on evidence
en-dc-ul-radio-link-outage-*Owns the FRChgRefuse_RLO story — full set on the NSA retainability page (cross-ref, not duplicated)
The EN-DC mobility feature stack — full detail
NSA Mobility Feature Pipelineclick a feature node to open its detail card
The 5-step NSA mobility MOP
1
Baseline all three planes

5109/5110 PSCell changes, 748-750 anchor twins, 5090/5095/5096 add-back, 5112 mods, 5113/5092 releases, 906 refusals. Compute the continuity product.

both ledgers, three planes Skip this and one plane's problem masquerades as another's.
2
Find the losing plane

PSCell SR low = NR-leg mobility. Anchor twins below the general LTE composite = EN-DC freight. Keep-share low = choreography. Refusals high = policy/outage.

the continuity product's weakest factor If wrong: NR trigger tuning for an X2 problem.
3
Decode the cross-ledger arrows

_MenbFail on NR families → the anchor/X2. _Tdcoverall → guard/latency. _RrcTo → new-PSCell radio. FRChgRefuse_RLO → the UL-outage machinery (retainability page).

suffix = which node owns the fix If wrong: two teams tune two ends of one failure.
4
Apply ONE lever on the owning plane

NR trigger: handover-margin-nr / TTT / B1-A2 gap. Choreography: prefer keep>modify>re-add (5090 path health, context-kept share). Steering: anchor-selection + scg-band thresholds.

one lever per soak If wrong: unattributable results.
5
Verify the product + guards

Continuity product up; SgNB addition SR (5103, accessibility) held; NSA drop formula (retainability page) held; LTE composite (17/18/20) untouched.

guards across all three Day-2/Day-3 pages If wrong: mobility "fixed" by adding fewer legs.

NSA mobility is a choreography — keep > modify > re-add > release.

Three planes, two ledgers, one continuity product. The anchor choreographs; the counters name which side dropped the beat.

Sources: Samsung 5G NR Counter Description (CU-CP ACPF — Families 5109 INTRA_SN_PSCELL_CHANGE, 5110 INTER_SN_PSCELL_CHANGE_SRC, 5090/5095/5096 SgNB-addition-by-HO, 5112 MENB_INIT_SGNB_MOD_MCGMOB, 5113/5092 — counters verbatim) · Samsung LTE eNB Counter Description (Families 748/749/750/752/753 ENDC_HO, 802 INTER_FREQ_HO_FOR_EN_DC, 793 EN_DC_IDLE_LB, 906 NOT_ALLOWED_PSCELL_CHANGE, 742 IRAT_NR_MR_CLASSIFICATION, 706/708/709/710 SCG/MCG mobility SgNB families, 943/944 X2 link time — verbatim) · 5G NR Mobility/Retainability Optimization SA decks (handover-margin-nr, TTT high-mobility values) · LTE Throughput Optimization deck LB chapter (anchor-selection & scg-band-selection parameters, verbatim). EN-DC per 3GPP TS 37.340. The B1 addition trigger is covered on the accessibility NSA page; the UL-outage and drop machinery on the retainability NSA page — cross-referenced, not duplicated.