Dual Connectivity: EN-DC, Bearer Types, and Talking to Two Nodes at Once

Carrier aggregation lets a UE use several carriers from one base station. Dual connectivity lets it use carriers from two different base stations simultaneously — potentially on different sites, using different radio technologies, connected by ordinary backhaul.

That last clause is the whole point. CA needs a single scheduler coordinating carriers within slot timing, which means one node. DC splits the problem at PDCP, where millisecond-scale delays between the two nodes are tolerable.

It's also how every NSA 5G network in the world works — so if you've used 5G, you've used dual connectivity.

---

Master and secondary

The two nodes are not peers, and the asymmetry is the key to understanding everything else.

The Master Node (MN) owns the RRC connection. It's the only node with an RRC relationship to the UE, it handles the control-plane connection to the core, and it decides when to add, modify, or release the secondary node. Its cell group is the MCG (Master Cell Group), and its primary cell is the PCell.

The Secondary Node (SN) provides additional radio resources. Its cell group is the SCG (Secondary Cell Group), and its primary cell is the PSCell (Primary Secondary Cell).

The collective term SpCell (Special Cell) means "PCell or PSCell depending on which group you're talking about" — it appears constantly in the specifications and means nothing more exotic than that.

The SN can have its own RRC entity in some variants, allowing SN-initiated reconfiguration of its own resources. But the MN retains overall control, and losing the PCell drops the connection regardless of how healthy the SCG is. Losing the PSCell triggers SCG failure, which is recoverable — the MN just reconfigures or releases the SN.

---

The four MR-DC variants

MR-DC (Multi-Radio Dual Connectivity) is the umbrella term. Which variant you have depends on two things: which technology is master, and which core it connects to.

EN-DC is by far the most deployed — it is NSA 5G. LTE anchors, NR adds capacity, EPC underneath. The NR node is formally an en-gNB, connected to the eNB over X2, not Xn.

NGEN-DC and NE-DC are transitional configurations on a 5G Core. NE-DC in particular is useful for operators with strong NR coverage wanting to keep LTE as a capacity or coverage supplement.

NR-DC is two NR nodes, and it has a specific practical use: pairing an FR1 node with an FR2 node. Mid-band provides reliable coverage, mmWave adds enormous capacity where it reaches. Given how easily mmWave gets blocked, having a robust FR1 anchor isn't optional.

---

The three bearer types

This is where dual connectivity gets genuinely interesting, and it's the part most explanations skip.

Each radio bearer is configured independently as one of three types, determined by where its PDCP entity lives and which cell groups carry it.

MCG bearer — PDCP at the master node, transmitted only over the master's cell group. Ordinary single-node operation.

SCG bearer — PDCP at the secondary node, transmitted only over the secondary's cell group. Traffic goes directly from the core to the SN, bypassing the MN entirely.

Split bearer — one PDCP entity feeding RLC entities in both cell groups. PDCP decides packet by packet which leg to use.

The split bearer is the mechanism that delivers aggregate throughput across two nodes. PDCP holds the reordering buffer, so packets arriving out of order across two legs with different delays get reassembled correctly before reaching higher layers.

Termination point

Any bearer type can terminate at either node — giving combinations like "MN-terminated split bearer" or "SN-terminated MCG bearer." That terminology sounds baroque until you see what it's for.

In EN-DC, an SN-terminated split bearer is the standard configuration for high-throughput traffic. PDCP sits in the NR node, which receives data from the EPC directly and decides how much to send over NR versus routing to the LTE node.

Why terminate at the NR node? Because the alternative makes the LTE eNB handle the full 5G data rate, and it usually can't. This is precisely the difference between deployment Option 3 (MN-terminated, LTE bottleneck) and Option 3x (SN-terminated, no bottleneck) — and it's why essentially every commercial NSA network uses 3x.

---

Flow control

A split bearer creates a problem that doesn't exist elsewhere: the PDCP entity must decide how to divide traffic between two legs whose capacity it can't directly observe.

Send too much to one leg and its RLC buffer builds up, adding latency and eventually overflowing. Send too little and that leg's capacity goes unused.

The solution is downlink data delivery status reporting over X2-U or Xn-U. The receiving node tells the PDCP-hosting node how much data it has buffered and how much it has successfully delivered. PDCP paces accordingly.

Getting this tuning right is one of the more finicky parts of DC deployment. Badly tuned flow control produces a specific, recognisable symptom: throughput noticeably lower than the better leg alone, because reordering delay across mismatched legs stalls the PDCP buffer. If you see split-bearer throughput underperforming single-leg, flow control is the first place to look.

---

Adding a secondary node

The procedure follows a consistent shape:

The MN configures the UE to measure candidate SN frequencies. Based on reports, it decides to add an SN and sends an SN Addition Request over X2 or Xn, including the bearer configuration it wants.

The SN admits or rejects — it may refuse on load grounds — and returns its own RRC configuration for the SCG.

The MN passes that configuration to the UE inside an RRC reconfiguration message. The UE performs random access on the PSCell to synchronise, and confirms completion.

Two details worth knowing. The SN's configuration is transported inside the MN's RRC message rather than sent directly — the MN owns the RRC connection, so all reconfiguration flows through it. And random access on the PSCell is required because the UE has no prior timing relationship with the secondary node.

---

DC vs carrier aggregation

They're frequently confused and the distinguishing question is simple: how many schedulers?

CA achieves tighter coordination and better aggregate efficiency, because one scheduler sees everything and can allocate across carriers within a single slot decision.

DC tolerates real-world transport. The two nodes can be kilometres apart with tens of milliseconds of latency between them, because PDCP-level splitting doesn't require slot-level coordination.

They compose. A UE can run CA within the MCG and within the SCG, aggregating four LTE carriers on the master and three NR carriers on the secondary. Device capability signalling for this gets genuinely complicated.

---

What it's actually for

Throughput. The headline use: aggregate capacity across two nodes and two technologies. This is what NSA 5G delivers.

Mobility robustness. Because the MN maintains the RRC connection, losing the SN degrades throughput rather than dropping the call. In early 5G deployments with patchy NR coverage, this was arguably more valuable than the throughput gain — it let operators deploy NR opportunistically without coverage risk.

Coverage. A low-band master gives reliable coverage; a high-band secondary gives capacity where available. Best of both, without depending on the high band.

Reliability via PDCP duplication. For URLLC, the same packet can be sent over both legs. That's diversity across two independent radio paths and two sites — one of the strongest mechanisms available for the final nines of reliability, and considerably more robust than retransmission on a single link.

Fast NR introduction. Historically the most important. EN-DC let operators launch 5G without deploying a 5G Core.

---

The mental model

MN owns RRC. SN adds resources. Lose the PCell, lose the connection; lose the PSCell, recover.

MCG/SCG are the cell groups; PCell/PSCell their primaries; SpCell means whichever applies.

Four variants: EN-DC (the NSA one), NGEN-DC, NE-DC, NR-DC.

Three bearer types: MCG, SCG, split — and any can terminate at either node.

Option 3x = SN-terminated split bearer, which is why the LTE node isn't a bottleneck.

CA = one scheduler, tight sync. DC = two schedulers, tolerates real backhaul.

PDCP duplication over both legs is a serious reliability tool for URLLC.

Dual connectivity is the mechanism that let 5G launch on LTE's back — and it remains useful well beyond that, wherever two nodes are better than one and the backhaul between them isn't ideal.

---

Further reading

• NSA vs SA architecture — where EN-DC fits in deployment
• Carrier aggregation explained — the single-node alternative
• 3GPP TS 37.340 — Multi-connectivity, overall description
• 3GPP TS 38.323 — PDCP specification, split bearer and duplication
• 3GPP TS 36.423 / 38.423 — X2AP and XnAP, SN addition procedures