SA vs NSA & Deployment Options in 5G NR
Standalone vs Non-Standalone, EN-DC, and the deployment options that shaped 5G rollout.
There are two fundamental ways to run a 5G network: let the gNB stand on its own talking to a 5G core (Standalone, SA), or bolt a 5G radio onto an existing 4G network as a capacity booster (Non-Standalone, NSA). Almost every early 5G launch was NSA; SA is where the network becomes truly "5G" end to end, from the radio all the way to the core.
Introduction
SA and NSA are not two different radios — they are two different deployment architectures for the same NR air interface. The question each answers is structural: which core does the radio connect to (the 4G EPC or the 5G 5GC), and which node owns the control plane (the LTE eNB or the NR gNB). 3GPP captured the choices as a numbered set of deployment options — the architecture split is described in TS 38.300 and the multi-connectivity rules in TS 37.340.
You meet this distinction the moment a UE powers on. In NSA the device attaches to LTE first, is authenticated by the EPC, and only then has an NR carrier added underneath it as extra bandwidth. In SA the device attaches directly to the 5GC over NR, registers with the AMF, and never needs LTE to make or hold the connection. That single difference ripples into everything downstream: voice (VoLTE vs VoNR), latency, network slicing, battery life, and even how the phone behaves in idle mode.
Because the differences are architectural rather than message-by-message, the way to master this topic is to hold two axes in your head — anchor and core — and read every option, every dual-connectivity flavour, and every bearer type as a point on those two axes.
On this page
Why SA and NSA both exist
In plain words: NSA is like bolting a turbocharger onto a car you already own — the original engine still does the steering, braking and idling (LTE holds the control plane), and the turbo just adds raw power (the NR carrier adds data throughput). SA is a brand-new car built from the ground up: every system, including the "engine control unit" (the 5G core), is native 5G.
The concrete reason both exist is timing of investment. When 5G arrived, operators already ran mature LTE networks anchored on the 4G core (EPC), but almost none had deployed a 5G core (5GC) yet. Ripping out a working core to launch 5G would have been slow and expensive. NSA solved that by letting NR ride on top of the existing LTE control plane and EPC, delivering the one benefit customers immediately notice — peak throughput — without a new core. SA exists because NSA's dependence on LTE structurally caps what 5G can do: the features that define 5G (network slicing, URLLC-grade latency, a service-based core, and eventually switching LTE off) all require the 5GC and an all-NR control plane.
A pair of architecture choices — SA (Option 2: NR to the 5GC) and NSA (Option 3: NR bolted onto an LTE anchor on the EPC) — defined by which core the radio uses and which RAT owns signalling.
NSA lets an operator sell "5G" fast by reusing LTE + EPC; SA unlocks the deeper 5G feature set once the 5GC is in place.
NSA uses EN-DC dual connectivity — LTE Master Node for control, NR Secondary Node for a boosted user plane. SA uses a single gNB that carries both planes to the 5GC over the NG interface.
The big picture: two roads to 5G
When operators first deployed 5G, most of them already ran a mature LTE network and did not yet have a 5G core (5GC). So 3GPP defined a set of deployment options — the architecture split is in TS 38.300 and the multi-connectivity rules in TS 37.340 — that let you introduce NR gradually rather than replacing everything at once. The two you must know first are Option 2 (SA) and Option 3 (NSA), but the full menu runs from Option 1 to Option 7.
Option 2. The gNB connects only to the 5G core (5GC) over the NG interface (NG-C/N2 to the AMF, NG-U/N3 to the UPF). The UE uses NR for both the control plane (signalling) and the user plane (data). Pure 5G — no LTE required for the UE to attach, register, or move data.
Option 3 family. An LTE eNB is the anchor and stays connected to the 4G core (EPC) over S1. The gNB (called an en-gNB here) is added as a data booster, tied to the eNB over X2. The UE keeps its control-plane lifeline on LTE and gets extra throughput from NR.
Speed to market. NSA reuses existing LTE coverage and the EPC, so an operator can light up NR carriers for capacity without deploying a whole new core. SA follows once the 5GC is in place and features such as network slicing, URLLC and true low latency are wanted.
One-liner: SA = NR radio + 5G core, control and data both on NR. NSA = NR radio bolted onto an LTE anchor + 4G core, control on LTE, data mostly on NR.
The key mental split is which core the radio talks to and which radio owns the control plane. Every option below is just a different answer to those two questions. Standalone options (a single RAT to a single core) are Options 1, 2 and 5. The rest are dual-connectivity options where two radios serve the UE simultaneously and one of them anchors signalling.
The full 3GPP deployment options menu
3GPP originally sketched Options 1 through 8; Options 6 and 8 were dropped early (an eNB or NR on the EPC in combinations that made no commercial sense). The survivors are the ones you will meet in real networks and in interviews. The two properties that define each option are the anchor (which node the UE reads for control-plane signalling) and the core it connects to.
| Option | Anchor / Master | Core | Dual connectivity | In plain words |
|---|---|---|---|---|
| 1 | eNB | EPC | None | Plain LTE. The baseline everyone started from — no NR at all. |
| 2 (SA) | gNB | 5GC | None | Pure standalone NR. Control and user plane both on NR. |
| 3 / 3a / 3x | eNB | EPC | EN-DC | The common NSA. LTE master, NR (en-gNB) secondary, on the 4G core. |
| 4 / 4a | gNB | 5GC | NE-DC | NR master, LTE (ng-eNB) secondary, on the 5G core. |
| 5 | ng-eNB | 5GC | None | LTE eNB upgraded to attach to the 5G core, no NR yet. |
| 7 / 7a / 7x | ng-eNB | 5GC | NGEN-DC | LTE master + NR secondary, both on the 5G core. |
Read the table as a migration story. Option 1 is where you begin: legacy LTE on the EPC. Option 3 keeps that EPC but adds an NR booster — the cheapest way to advertise "5G" because the core is untouched. Everything with a 5GC (Options 2, 4, 5, 7) is a later step that requires the new core to exist. Option 5 connects an upgraded LTE node (ng-eNB) to the 5GC as a stepping stone. Options 4 and 7 flip or share the anchor once NR coverage is dense. Option 2 is the destination: NR standing on its own.
Spec anchor: the deployment options and overall architecture are in TS 38.300; the multi-connectivity procedures (MN/SN roles, cell groups, bearer types) live in TS 37.340; the EN-DC RRC handling is split across LTE RRC (TS 36.331) and NR RRC (TS 38.331). Options 6 and 8 were studied and abandoned.
Multi-RAT Dual Connectivity (MR-DC)
All the dual-connectivity options are instances of one umbrella concept: MR-DC — Multi-Radio Access Technology Dual Connectivity. It means the UE is connected to two nodes of possibly different radio technologies (LTE and NR) at the same time, with one node acting as the Master Node (MN) and the other as the Secondary Node (SN). The MN owns the control-plane relationship with the core; the SN mainly adds user-plane capacity. Which RAT plays which role, and which core they attach to, gives the four named flavours.
| MR-DC type | Master (MN) | Secondary (SN) | Core | Inter-node interface | Option |
|---|---|---|---|---|---|
EN-DC | LTE eNB | NR en-gNB | EPC | X2 | 3 / 3a / 3x |
NE-DC | NR gNB | LTE ng-eNB | 5GC | Xn | 4 / 4a |
NGEN-DC | LTE ng-eNB | NR gNB | 5GC | Xn | 7 / 7a / 7x |
NR-DC | NR gNB | NR gNB | 5GC | Xn | (within Option 2) |
A useful way to remember the names: read them as "master-secondary." EN-DC is E-UTRA (LTE) master, NR secondary. NE-DC is NR master, E-UTRA secondary. NGEN-DC is "next-gen E-UTRA" (an ng-eNB, i.e. LTE on the 5G core) master with NR secondary. NR-DC is NR master and NR secondary — two NR nodes, typically an FR1 anchor plus an FR2 mmWave leg for extra capacity, used entirely inside an SA (Option 2) network on the 5GC.
Anchor rule: the letter before the dash in the MR-DC name is the master (control-plane anchor). E-N-DC → E-UTRA anchors. N-E-DC → NR anchors. The core is EPC only for EN-DC; every other MR-DC type sits on the 5GC. Correspondingly, the inter-node interface is X2 only for EN-DC and Xn for all the 5GC-based types.
MN/SN, cell groups and bearer types
A handful of paired terms show up constantly in every dual-connectivity discussion, and they are worth nailing down once because they apply to all four MR-DC types.
| Term | Meaning |
|---|---|
| MN / SN | Master Node / Secondary Node. The MN holds the control-plane connection to the core; the SN adds capacity. In EN-DC the MN is the LTE eNB, the SN is the NR en-gNB. |
| MCG | Master Cell Group — the serving cells provided by the MN: the PCell plus its SCells. |
| SCG | Secondary Cell Group — the serving cells provided by the SN: the PSCell plus its SCells. |
| MCG bearer | A radio bearer whose radio resources come only from the MCG. |
| SCG bearer | A radio bearer whose radio resources come only from the SCG. |
| Split bearer | One bearer carried over both cell groups: a single PDCP entity feeds RLC legs in both the MCG and the SCG. Aggregates throughput and adds robustness. |
| MN/SN-terminated | Which node hosts the PDCP that anchors the bearer toward the core. This decides the core-side user-plane path (Option 3 vs 3a/3x). |
The two attributes — termination (which node's PDCP anchors the bearer) and type (MCG / SCG / split, i.e. which cell groups carry it) — are independent, so in principle they multiply out to six bearer combinations (MN-terminated MCG/SCG/split and SN-terminated MCG/SCG/split). In EN-DC a key detail is the PDCP flavour: an SN-terminated bearer uses NR PDCP even though the MN is LTE, which is exactly why EN-DC needs the eNB software upgraded to host or interwork with NR PDCP.
The control plane is what makes MR-DC coherent: the UE has exactly one RRC connection, terminated at the MN. In EN-DC that RRC runs over LTE (E-UTRA RRC, TS 36.331); the SN's NR configuration is delivered embedded inside an LTE RRCConnectionReconfiguration as an NR RRCReconfiguration (the nr-SecondaryCellGroupConfig). In the 5GC-based options the master RRC is NR RRC (TS 38.331). The SN can generate its own RRC messages, delivered to the UE either through the MN (the split SRB1 path) or directly over SRB3, an optional signalling radio bearer straight to the SN — but the master always stays in charge of the connection. The PSCell (Primary Secondary Cell) is the special cell in the SCG that the UE runs random access against when the SN is added or changed (via reconfigurationWithSync).
Why "non-standalone": in EN-DC the control-plane anchor is always the LTE eNB. The gNB only adds a Secondary Cell Group for user-plane capacity — it cannot bring a UE from idle to connected on its own. The UE reads LTE for sync and signalling and treats NR as a bolt-on carrier. That dependency is precisely what the word "non-standalone" describes.
NSA in detail: the Option 3 variants (3, 3a, 3x)
NSA Option 3 is built on EN-DC. The three variants — 3, 3a and 3x — differ only in where the user-plane bearer is terminated and how it reaches the core. "Terminated" means which node holds the PDCP layer that anchors the bearer toward the EPC. All three share the same control plane: the LTE eNB as master over X2-C to the en-gNB, with user-plane data riding S1-U (to the core) and X2-U (between the nodes).
| Option | Bearer type | User-plane path to core | In plain words |
|---|---|---|---|
| 3 | MN-terminated | All S1-U data lands at the eNB; the eNB forwards some over X2-U to the en-gNB | 4G node is the funnel; it hands part of the traffic to 5G. The eNB backhaul can bottleneck. |
| 3a | SN-terminated | Bearers split at the EPC: some go straight to the eNB, others straight to the en-gNB over their own S1-U | The core decides per bearer which node carries it. No X2 user-plane split. |
| 3x | SN-terminated, split at SN | The bearer's S1-U terminates at the en-gNB; the en-gNB's NR PDCP can route some data back to the eNB over X2-U | Most deployed. The 5G node anchors the boosted bearer and uses LTE for extra legs/coverage. |
Why 3x won. The en-gNB normally has much fatter backhaul than a legacy eNB, and in 3x the high-rate split bearer lives on the NR side, anchored by NR PDCP. That avoids the eNB becoming a throughput bottleneck the way plain Option 3 can, while still letting the LTE anchor add a slower leg for coverage and reliability. Option 3a was rarely used because splitting purely at the core gives you neither the flow-control flexibility of 3/3x nor a single fast anchor. So the industry converged on 3x for NSA and treats it as "the" NSA option. A practical consequence is that in 3x the en-gNB runs the PDCP for the split bearer and applies flow control down the X2-U leg to the eNB, so the operator can tune how much traffic rides the slower LTE path versus the fast NR path per radio conditions.
Same ideas, different core: Options 4a and 7x are the 5GC analogues of this split idea — 4a is SN-terminated in NE-DC (LTE secondary), and 7/7a/7x mirror 3/3a/3x for NGEN-DC, only over Xn instead of X2. Learn the 3/3a/3x pattern once and it transfers.
Option 3x vs Option 2, side by side
The clearest way to see the difference between NSA and SA is to draw them next to each other. On the left is EN-DC Option 3x: an LTE eNB anchoring control on the EPC, an en-gNB carrying the boosted user plane, and X2 tying them together. On the right is Option 2 SA: a single gNB doing everything against the 5GC.
eNB is the Master Node on the EPC and holds the control plane; the NR en-gNB is the Secondary Node anchoring the boosted user plane, with a dashed S1-U to the core and an X2 link to the eNB. Right: SA Option 2. A single gNB carries both control and user plane to the 5GC over NG.LTE ↔ NR: the interface names shift with the core. NSA/EN-DC keeps the LTE-era names — S1-MME/S1-U to the EPC, X2-C/X2-U between nodes, and RRC in E-UTRA RRC (TS 36.331). SA/Option 2 uses the 5G names — NG-C (N2, to the AMF) and NG-U (N3, to the UPF), Xn between gNBs, and RRC in NR RRC (TS 38.331). Same job, renamed for the 5G system architecture.
Why NSA first, and what changes on the way to SA
Operators started with NSA Option 3x for a simple reason: it was the fastest, cheapest way to sell "5G." The EPC and the LTE control plane were already deployed and paid for, so adding NR carriers meant new radios and an X2 link — not a whole new core. NSA immediately delivered the headline benefit customers noticed, peak throughput, by aggregating an LTE leg and an NR leg into a split bearer. But NSA cannot deliver the deeper 5G promises: network slicing, URLLC-grade latency, service-based core architecture, and energy savings from switching LTE off all require the 5GC and an all-NR control plane. That is why the roadmap is Option 3x → Option 2, sometimes stopping at intermediate 5GC options along the way.
| Dimension | NSA (Option 3x) | SA (Option 2) |
|---|---|---|
| Core | EPC (4G) | 5GC (5G, service-based) |
| Control plane | LTE RRC at the eNB (TS 36.331) | NR RRC at the gNB (TS 38.331) |
| NAS registration | 4G EPS NAS to the MME | 5G NAS to the AMF |
| Time to deploy | Fast — reuses LTE + core | Slower — needs new core |
| Peak throughput | High (LTE + NR aggregated) | High (NR); no LTE leg |
| Latency | Limited by LTE anchor + EPC | Low — native NR + 5GC |
| Network slicing | No | Yes (S-NSSAI end to end) |
| Voice | VoLTE on the LTE anchor | VoNR, with EPS fallback to LTE where VoNR is not yet deployed |
| UE battery | Two radios active — higher drain | Single NR radio — lower drain |
What changes for the UE. In NSA the device must be an EN-DC-capable UE: it maintains two active radio legs, reads LTE for all signalling, and only ever gets NR as a secondary. Voice runs on LTE (VoLTE). In SA the UE registers directly with the AMF in the 5GC using 5G NAS, does all signalling over NR RRC, uses VoNR for voice with EPS fallback to LTE where VoNR is not yet available, and can use a single active radio — which usually improves battery life. Idle-mode behaviour changes too: an SA UE camps on NR and reselects between NR and LTE cells, rather than always camping on LTE as in NSA.
Trade-off in one line: NSA buys time-to-market and peak speed by leaning on LTE; SA unlocks the real 5G feature set (slicing, low latency, VoNR, LTE shutdown) at the cost of deploying a new core and migrating devices.
⚠ Common pitfalls / gotchas
- "EN-DC needs no eNB upgrade" is false. An SN-terminated (3x) bearer runs NR PDCP, so the LTE
eNBsoftware must be upgraded to interwork with it — legacy eNBs cannot do EN-DC unchanged. - Confusing anchor with core. The MR-DC name's first letter is the anchor, not the core.
NGEN-DChas an LTE (ng-eNB) anchor but sits on the5GC— it is not NSA on the EPC. - Assuming SA always means lower latency for voice. Where
VoNRis not yet deployed, an SA voice call triggersEPSfallback to LTE, which can add setup delay — sometimes worse than a mature VoLTE-on-NSA network. - Band-combination limits. EN-DC needs a UE-supported
DC_x_nyband combination; a UE may support NR standalone on a band yet not the specific EN-DC pairing, so NR simply never gets added as SCG.
Summary
Reduce the whole topic to two axes: which core (4G EPC vs 5G 5GC) and which anchor owns the control plane (LTE eNB vs NR gNB). SA (Option 2) is NR to the 5GC with both planes on NR; NSA (Option 3) is NR bolted onto an LTE anchor on the EPC via EN-DC, control on LTE and data mostly on NR.
The dual-connectivity flavours are just points on those axes: read the MR-DC name's first letter as the anchor, remember EN-DC is the only one on the EPC (over X2) while the rest sit on the 5GC (over Xn). Within NSA, 3/3a/3x differ only in where the bearer terminates and reaches the core, and 3x — SN-terminated with NR PDCP splitting back to the eNB — is what the industry actually deployed.
The migration story is 3x → 2: NSA for fast, cheap capacity on the existing core; SA once the 5GC is in place, to unlock slicing, low latency, VoNR and eventually LTE shutdown.
Quick Q&A
Q. In one sentence, SA vs NSA?
A. SA (Option 2) is a gNB on the 5GC carrying both control and user plane on NR; NSA (Option 3) uses an LTE eNB anchor on the EPC for control while the en-gNB adds NR user-plane capacity via EN-DC.
Q. Name the four MR-DC types and their masters.
A. EN-DC (LTE master, NR secondary, on EPC), NE-DC (NR master, LTE secondary, on 5GC), NGEN-DC (ng-eNB master, NR secondary, on 5GC), and NR-DC (NR master and NR secondary, on 5GC). The letter before the dash is the master.
Q. What is the difference between Options 3, 3a and 3x?
A. It is where the bearer terminates and reaches the core. 3 = MN-terminated, all data via the eNB. 3a = SN-terminated, bearers split at the EPC. 3x = SN-terminated with the split done at the en-gNB, which routes some data to the eNB over X2 — this is the most deployed.
Q. In EN-DC, which node runs the control plane and why does it matter?
A. The LTE eNB (the Master Node), over LTE RRC. That dependency is what makes it "non-standalone" — the NR en-gNB cannot bring a UE from idle to connected on its own and relies on the LTE anchor for signalling to the EPC. The SN's NR config is even delivered embedded inside an LTE RRCConnectionReconfiguration.
Q. What is the difference between an MCG, SCG and split bearer?
A. An MCG bearer uses radio resources only from the Master Cell Group; an SCG bearer only from the Secondary Cell Group; a split bearer uses both, with one PDCP entity feeding RLC legs in each group to aggregate throughput.
Q. Why did operators start with NSA and move to SA?
A. NSA (Option 3x) reused the existing LTE anchor and EPC for fast, cheap capacity, but only SA (Option 2) with the 5GC unlocks slicing, URLLC, VoNR and LTE shutdown — so the migration path is 3x to 2.
Where to go next
Now that you know how NR sits alongside LTE and how the control plane is anchored, follow the interfaces and procedures that make dual connectivity and standalone access work: