Measurement Gaps in LTE 4G

Introduction

A measurement gap is a periodic, network-scheduled window — 6 ms long in LTE — during which a UE performs no serving-cell transmission or reception, so that its single receiver is free to retune and measure a cell on a different frequency or radio technology. It exists purely because of a hardware reality: a one-receiver device physically cannot be tuned to two carriers at once, so the network must carve out protected time rather than let the UE vanish at random.

Gaps sit at the heart of connected-mode mobility. In RRC_CONNECTED the eNB, not the UE, controls when the UE looks elsewhere. The gap is what makes inter-frequency and inter-RAT measurement possible at all; the neighbour levels gathered inside gaps are what feed the reporting events (A3/A4/A5, B1/B2) that ultimately trigger a handover. Without gaps, the UE could measure only the carrier it is already camped on.

Because gaps cost airtime — every 6 ms measuring is 6 ms not exchanging data — the network configures them deliberately, arms them when the serving signal weakens, and releases them when the hunt is over. Understanding the gap is therefore both a configuration skill (which pattern, which offset, delivered in which IE) and a debugging skill (why is a UE not reporting a neighbour it should be able to see?).

Why a Single-RF UE Needs Gaps

Radio measurements come in two flavours from the UE's point of view. Intra-frequency measurements are of neighbour cells on the same carrier the UE is already camped on — it can gather those in the normal course of receiving, because its receiver is already tuned there. Inter-frequency (a different LTE carrier) and inter-RAT (UMTS/WCDMA, GSM, or, in later releases, NR) measurements are the hard case: the neighbour lives on a different centre frequency, and a single-receiver UE has to physically retune its RF front end to look at it.

While the receiver is retuned elsewhere, it is deaf to the serving cell. It cannot decode PDCCH, cannot receive PDSCH, and must not transmit either — anything the eNB schedules to that UE during the window is simply lost. The network cannot allow that to happen at random, unknown moments, so it schedules the absences itself. In effect it tells the UE, "for these 6 ms, on this recurring cadence, you are excused — go and measure." That agreed, protected window is a measurement gap.

There is a subtlety worth stating up front. Whether a gap is needed is partly a UE-capability question, not purely a geometry question. Some UEs signal, via InterFreqBandInfo/needForGaps capability fields, that for particular band combinations they can measure a given inter-frequency carrier without interrupting the serving cell — for example when a second receive chain or a wide enough RF can cover both. Where the capability says no gap is required, the eNB simply does not configure one. The rest of this page assumes the common case: a single-RF UE that genuinely must go dark to look elsewhere.

Gap Configuration — MeasGapConfig

Gaps are configured through the measGapConfig field of the measConfig IE, delivered inside an RRCConnectionReconfiguration on SRB1 (DL-DCCH). It is a simple setup / release CHOICE: setup installs a gap pattern, release tears it down. The only real parameter carried under setup is gapOffset, and its type encodes two pieces of information at once — which pattern to use, and where in the radio frame the gap falls.

gapOffset is itself a CHOICE between gp0 and gp1: choosing gp0 selects Gap Pattern 0, choosing gp1 selects Gap Pattern 1. The integer value you assign is the gap offset itself, which fixes the exact SFN and subframe where each gap begins, so gaps land in a known, collision-free place relative to the frame boundary. Because the offset range differs with the pattern's period, gp0 takes a value 0–39 (one per subframe within the 40 ms period) and gp1 takes 0–79 (within the 80 ms period). The two patterns are not free-form; they are fixed rows in TS 36.133, so any UE and any eNB agree on exactly what "Pattern 0" and "Pattern 1" mean.

Concretely, the gap's start is derived from the offset by the relation the standard fixes between the pattern and the frame timing: a gap subframe occurs when the current system frame and subframe satisfy the pattern's periodicity counted from the configured gapOffset. In plain terms — the offset says "the first gap of each period starts here," and the period (MGRP) then repeats it. This lets the eNB place a UE's gaps where its own scheduler has the least to lose, and (with multiple UEs) stagger different UEs' gaps so they do not all vanish in the same subframes.

Gap Pattern 0 and Gap Pattern 1

Two quantities define each pattern. The Measurement Gap Length (MGL) is how long each individual gap lasts — 6 ms in both LTE patterns. The Measurement Gap Repetition Period (MGRP) is how often a gap recurs — 40 ms for Pattern 0 and 80 ms for Pattern 1. So Gap Pattern 0 is "6 ms every 40 ms" and Gap Pattern 1 is "6 ms every 80 ms." The 6 ms length is deliberately generous: it must cover retuning out, the measurement of the target carrier, and retuning back to the serving cell — all inside the one window.

Note the duty cycle, which is the whole trade-off in one number. Pattern 0 spends 6 ms out of every 40 ms measuring — roughly 15% of the time the UE is unavailable to the serving cell. Pattern 1 spends 6 ms out of every 80 ms, roughly 7.5%. Pattern 0 finds and refreshes neighbour measurements about twice as fast; Pattern 1 costs the serving cell about half as much airtime. The measurement-period and reporting-delay requirements in TS 36.133 scale accordingly: with a longer MGRP the UE simply gets fewer samples per unit time, so the standard allows it correspondingly longer to deliver a valid cell measurement.

What Happens on the Air — a Timeline

Picture the serving carrier as a stream of 1 ms subframes. Most of the time the UE is receiving and transmitting normally. Then a gap arrives: for 6 ms the UE goes quiet on the serving cell, retunes, measures the neighbour, and retunes back. With Pattern 0 this repeats every 40 ms; with Pattern 1, every 80 ms.

The eNB knows exactly when these holes occur — it configured them via the gapOffset — so its scheduler simply avoids allocating PDSCH/PUSCH to that UE during the gap subframes. TS 36.133 defines precisely how the gap interacts with the rest of the physical layer: HARQ processes that would fall in the gap are suspended and resumed around it, periodic reporting such as CQI/SRS that collides with the gap is dropped for those subframes, and DRX timers are handled so that nothing is silently lost. The point is that everything about the gap is deterministic on both sides.

The Throughput Cost, and When No Gap Is Needed

Gaps are not free. Every 6 ms the UE spends measuring is 6 ms it is not exchanging data with the serving cell, so per-UE throughput dips whenever gaps are active — noticeably more with Pattern 0 (≈15% duty) than with Pattern 1 (≈7.5%). Both downlink and uplink take the hit, because during the gap the UE neither receives PDSCH nor is allowed to transmit PUSCH/PUCCH. The eNB scheduler also has to work harder: it must route real traffic around the recurring holes for that specific UE while still serving everyone else normally.

Because of this cost, gaps are configured only when needed. If the UE has only intra-frequency neighbours to track, no gap is set up at all — it measures those without retuning, in the ordinary flow of reception. The eNB typically arms gaps when the serving signal has weakened enough that it is worth looking for somewhere better to go — frequently triggered by an A2 "serving becomes worse than threshold" event, which is the network's cue to start (or intensify) inter-frequency and inter-RAT measurements. When the neighbour hunt is finished, or the serving signal recovers (an A1 "serving better than threshold" event), the eNB can release the gap and hand the airtime straight back.

How Gap Measurements Feed the Reporting Events

Gaps produce raw measurements; those measurements only matter because they drive reporting events. The neighbour RSRP/RSRQ (or the RAT-specific equivalent) that the UE gathers inside its gaps is evaluated against the entry/exit criteria configured in reportConfig. When a criterion is satisfied for the duration of timeToTrigger, the UE fires a MeasurementReport on SRB1, which is what actually lets the eNB decide on a handover.

So the chain is: gap → measure neighbour → event (A3/A4/A5 or B1/B2) → MeasurementReport → handover decision. Without gaps, the inter-frequency and inter-RAT events could never gather the data they test. It is worth being precise about the division of labour: A1 and A2 concern the serving cell only and never need a gap (the UE is always listening to serving); they are the events that commonly arm and disarm gap-based measurement. A3/A4/A5 evaluate inter-frequency LTE neighbours; B1/B2 evaluate inter-RAT neighbours. The A5 and B2 events are the "both conditions" events — serving has degraded and a good target exists — and are the classic triggers for a coverage-driven handover or an inter-RAT redirection.

Carrier Aggregation Considerations

Carrier aggregation complicates the "when is a gap needed?" question, because a CA UE already spans several carriers at once. The key idea from TS 36.133 is that a carrier that is part of the UE's own aggregated set does not require a gap to be measured — the UE is already receiving it. So a frequency that is an inter-frequency neighbour for one UE may be an activated SCell for a CA UE, and for that CA UE it needs no gap at all.

Gaps become necessary again only when the CA UE must measure a carrier that is outside its configured/activated set — a genuine inter-frequency neighbour it is not currently aggregating, or an inter-RAT target. There is an important asymmetry: a measurement gap in LTE applies to all serving cells of the UE simultaneously. When the gap opens, every configured component carrier goes quiet together, because the single RF has retuned away. That is why gaps are comparatively expensive for a CA UE — the interruption is not to one carrier but to the whole aggregated pipe. Later releases introduced per-CC and per-UE gap flexibility to soften this, but the baseline behaviour is one gap, all cells dark.

Reading Gap Configuration in the Logs

On the UE side, a measurement gap shows up as an ASN.1 decode of the RRCConnectionReconfiguration that installed it — specifically the measConfig → measGapConfig subtree. This is the single most useful thing to check first: is the gap actually there, which pattern is it, and where does it start?

Representative UE RRC log (ASN.1 decode) — illustrative, values vary by vendor/build:

A Note on NR

5G NR keeps the same core idea — a single RF still cannot be in two places at once — but generalises it. Instead of two fixed LTE patterns, NR defines per-FR gaps (a gap for FR1, a gap for FR2, or a single per-UE gap) so a UE need not go dark on both frequency ranges at once, and it pairs gaps with the SMTC (SS/PBCH Measurement Timing Configuration) window that tells the UE exactly when the target cell's SSBs appear — so the gap can be aimed precisely at the beacon rather than left open blindly, and NR also offers a richer set of gap lengths and periods than LTE's two rows.

Summary

A measurement gap is the network's deliberate, scheduled loan of the UE's single receiver: a 6 ms window in which the UE goes silent on the serving cell to retune and measure a carrier or RAT it otherwise cannot reach. LTE offers exactly two shapes — Gap Pattern 0 (6 ms / 40 ms, ≈15% duty, fast) and Gap Pattern 1 (6 ms / 80 ms, ≈7.5% duty, light) — selected and timed by the gapOffset CHOICE inside measGapConfig.

Intra-frequency needs no gap; inter-frequency and inter-RAT do (unless UE capability says otherwise). The measurements gaps produce feed the A3/A4/A5 (inter-freq) and B1/B2 (inter-RAT) events, whose MeasurementReport is what actually triggers a handover. When debugging "the UE won't hand over to a neighbour it should see," walk the chain in order: is there a measObject for that frequency, a reportConfig and measId binding it, and — crucially — a measGapConfig letting the UE go look? The missing link is usually the gap, and every other IE will look fine while it is absent.