DAPS Handover
How Dual Active Protocol Stacks Eliminate the Handover Gap
Mobility in cellular systems has historically relied on a break-before-make model: the user equipment releases the source cell, then synchronizes and performs random access on the target. That sequence leaves a user-plane gap that, in 5G NR, typically measures 30 to 60 ms depending on subcarrier spacing, random-access timing, and RRC reconfiguration latency. For voice over NR, cloud gaming, and time-sensitive industrial traffic, even a 40 ms gap can produce audible glitches or missed control deadlines. DAPS handover removes that gap by duplicating the lower layers of the protocol stack so the device can operate on both cells during the transition.
During a DAPS event the UE keeps the source MAC, RLC, and physical layer running while it establishes the target link. Downlink data continues from the source until the target cell is fully synchronized and acknowledged. The key enabler is a single common PDCP entity that sits above the two separate lower stacks: it performs robust header decompression, in-order delivery, and duplicate detection across packets arriving from both the source and target paths, so the application sees one continuous stream. Once the network confirms the target, the source bearer is released and the device collapses back to a single active stack.
The cost is hardware and processing. Running two active links means roughly double the baseband load during the overlap and an RF front end capable of receiving, and at the switching boundary transmitting, on both carriers without desensitization. Release 16 therefore restricts DAPS to a single target cell, allows uplink on only one cell at a time, and is most readily deployed for intra-frequency or closely spaced band pairs. Devices without the required dual receive and transmit capability simply fall back to conventional or conditional handover.
Interruption Time and Uplink Switching
Tint ≈ TRRC + TRACH + Tsync ≈ 30 to 60 ms
DAPS user-plane interruption:
Tint,DAPS ≈ 0 ms (downlink overlap on source & target)
Uplink switching delay (single Tx):
TUL ≈ TRACH,target + TTA (~ a few ms, one cell at a time)
PDCP reordering window:
t-Reordering ≥ max(RTTsource, RTTtarget) so out-of-order packets from both links are recovered
Where TRRC = reconfiguration latency, TRACH = random-access time, Tsync = target downlink sync, TTA = timing-advance acquisition. The common PDCP entity hides the residual lower-layer transition from the application.
Handover Scheme Comparison
| Scheme | Source link during HO | Interruption time | Target cells | Primary benefit | 3GPP release |
|---|---|---|---|---|---|
| DAPS handover | Kept active | ~0 ms | Single | Zero data gap | Rel-16 |
| Conditional handover (CHO) | Released at execution | 30 to 60 ms | One or more candidates | Robust triggering | Rel-16 |
| Make-before-break | Kept until target ready | Reduced (low tens of ms) | Single | Lower gap, no full DAPS cost | LTE / NR mobility enhancement |
| Legacy break-before-make | Released first | 30 to 60 ms | Single | Lowest UE complexity | Baseline |
Frequently Asked Questions
How does DAPS handover achieve 0 ms interruption time?
Legacy break-before-make handover detaches from the source before synchronizing to the target, giving a 30 to 60 ms gap while random access and RRC reconfiguration complete. DAPS keeps the source MAC, RLC, and PHY active while the UE does random access on the target. Downlink data keeps flowing from the source, and one common PDCP entity reorders packets arriving from both links, so the application-layer interruption is effectively 0 ms.
What is the difference between DAPS handover and conditional handover?
They solve different problems and can be combined. Conditional handover (CHO) pre-provisions target configurations and lets the UE execute autonomously when an RSRP or hysteresis-margin condition is met, cutting handover failures near the cell edge. DAPS keeps the source link alive during the switch to remove the data gap. 3GPP permits a candidate cell to be both a CHO and a DAPS target, giving robust triggering and near-zero interruption together.
Why is DAPS handover limited to a single target cell and certain band combinations?
Two active protocol stacks force the UE to receive (and at the switching point transmit) on both carriers at once, doubling baseband load and stressing the RF front end. To bound this, Release 16 specifies a single DAPS target, uplink to one cell at a time with a defined switching point, and favors intra-frequency or closely spaced bands so inter-carrier desensitization stays manageable. Devices lacking dual simultaneous receive and transmit fall back to conventional handover.