PHY Measurements (RSRP/RSRQ/RSSI/SINR) in LTE 4G
The measurement quantities the UE reports, and how RSSI, RSRP, RSRQ and SINR relate in LTE.
Your phone is always sizing up the cells around it, and it does so with a small family of numbers: RSSI, RSRP, RSRQ and RS-SINR. The trick to understanding them is that they are not unrelated readings — they are one measurement chain, from raw total power down to a clean per-reference-signal value, and then up again to a quality ratio. This page is grounded in TS 36.214 (physical-layer measurement definitions) and TS 36.133 (measurement requirements, reporting ranges and mapping).
Introduction
LTE radio measurements are the physical-layer numbers the UE derives from the downlink to describe how strong and how clean each cell looks. The core family — RSSI, RSRP, RSRQ and RS-SINR — is defined in TS 36.214, with reporting ranges, mapping tables and accuracy requirements in TS 36.133, and the RRC configuration that governs them (MeasConfig, filtering, events) in TS 36.331.
These measurements run throughout the UE lifecycle. In RRC_IDLE they drive cell selection and reselection — deciding which cell to camp on. In RRC_CONNECTED they feed measurement reports that trigger handover. The same underlying quantities are also reused for open-loop power control (path loss = referenceSignalPower − CRS RSRP) and, via CQI, for link adaptation.
They matter because every mobility and coverage decision in the network ultimately rests on them. Choose the wrong metric — judge a congested cell by signal strength alone, or react to a momentary fade — and the UE ends up on the wrong cell or ping-pongs between two. Understanding what each number actually measures, and how it is filtered before the network acts, is what makes those decisions correct.
On this page
Why the UE Measures at All
In plain words: think of the UE as a driver constantly glancing at road signs. It is not enough to know a town is nearby (signal strength) — you also need to know whether the road there is clear or jammed (quality). One glance can't tell you both, so LTE hands the UE several different "signs" and lets the network decide which one to steer by.
An LTE UE is mobile and the radio environment changes constantly — you walk, you turn a corner, a truck rolls past, a neighbouring cell suddenly floods the band with traffic. The network can only keep you on the best serving cell if the UE keeps reporting how strong and how clean each cell looks. Those reports drive two big decisions: which cell to camp on (cell selection and reselection in RRC_IDLE) and when to hand you over (handover in RRC_CONNECTED). Both decisions rest on the same physical-layer quantities, just consumed differently.
A set of physical-layer power and quality measurements the UE derives from the downlink, chiefly from the Cell-specific Reference Signals (CRS). The reported quantities are RSRP and RSRQ; RSSI is the total-power term inside RSRQ; RS-SINR was added later as a reportable quality metric.
Coverage decisions need a clean signal-strength number; quality decisions need to know how much interference and load sit on top of that signal. One number cannot express both, so LTE defines several and lets the network choose which to act on.
The UE isolates the CRS resource elements spread across the band, averages their power, and combines that with total received power over the same symbols to produce the reported quantities per TS 36.214, filtered per TS 36.331 before being reported.
One-line map: RSSI is everything received, RSRP is the clean signal part, RSRQ ties the two together as a quality ratio, and RS-SINR asks how the reference-signal power compares to interference plus noise.
Two ideas make the rest of this page fall into place. First, every one of these quantities is anchored to a defined measurement bandwidth — a contiguous block of resource blocks the UE is told (or allowed) to measure over. Second, the raw physical-layer value is not what the network usually acts on: it is a Layer-1 sample that gets averaged again at Layer-3 before a report is sent. Keep those two ideas in mind and the definitions below stop being a list of formulas and become a single, coherent pipeline.
RSSI — Total Wideband Power
The Received Signal Strength Indicator (RSSI, formally E-UTRA Carrier RSSI) is the most inclusive number: the total linear-average received power the UE observes across the measurement bandwidth. Crucially it lumps everything together — the wanted serving-cell signal, co-channel interference from other cells, adjacent-channel leakage, and thermal noise from the receiver front end. In LTE, TS 36.214 defines it as measured over the OFDM symbols that contain CRS of antenna port 0, and it spans N resource blocks of bandwidth.
That "on CRS-bearing symbols" detail matters. RSSI is not a free-running total measured over the whole subframe; it is the average received power in the specific OFDM symbols the reference symbols live in, because those are exactly the symbols whose power the RSRQ ratio needs to compare against. Measuring RSSI and RSRP over the same symbols is what makes the ratio meaningful.
Because it includes interference and noise, a high RSSI is not automatically good news — a cell edge drowning in interference can show a large RSSI. That is exactly why RSSI is rarely reported on its own in LTE; its main job is to serve as the denominator that turns RSRP into a quality figure. The value grows with measurement bandwidth (more RBs, more collected power), so it is meaningless without knowing the N it was measured over.
Mental model: RSSI is the volume of the whole room — your friend talking plus every other conversation plus the air-conditioner hum. It tells you how loud it is, not how well you can hear the one voice you care about.
RSRP — The Clean Per-RE Signal
The Reference Signal Received Power (RSRP) strips out everything except the wanted signal. TS 36.214 defines it as the linear average over the power contributions of the resource elements that carry cell-specific reference signals (CRS) within the considered measurement bandwidth. Because the CRS pattern is known exactly — its resource-element positions follow deterministically from the Physical Cell ID the UE found during cell search — the UE can isolate those REs and measure their power cleanly. So RSRP is a per-resource-element quantity, reported in dBm, not a wideband sum.
The linear-average power of the CRS resource elements, expressed in dBm as a per-RE value. A pure signal-strength metric with interference and noise excluded by construction.
Coverage and mobility decisions need a stable measure of how strong the cell is that does not swing with instantaneous load or with channel bandwidth. Averaging over per-RE signal power gives exactly that.
The UE locates the port-0 CRS REs, averages their linear power over the measurement bandwidth and over time, and reports it against the mapping in TS 36.133.
Because it is a per-RE value, RSRP does not depend on how wide the channel is: a 5 MHz and a 20 MHz cell of identical coverage report similar RSRP. That bandwidth-independence is what makes it the primary metric for the S-criterion in cell selection and for populating handover measurement reports. Typical values run from roughly -80 dBm (excellent, close to the cell) down to about -120 dBm (poor, near the edge), with usable service generally holding down to around -110 to -115 dBm.
A subtlety worth internalising: RSRP and RSSI are both derived from the same received waveform but answer different questions. RSRP asks "how much power is in the known reference tones?"; RSSI asks "how much power is in the whole symbol?". The SVG below shows this directly — the tall spikes are the CRS REs that feed RSRP, while the entire shaded band (spikes plus everything between them plus the noise floor) feeds RSSI.
Measurement bandwidth: both RSRP and RSSI are defined over a minimum of 6 resource blocks (the width of the LTE synchronisation signals / narrowest channel). The network can widen the reference-signal averaging via allowedMeasBandwidth broadcast in SIB; a wider bandwidth means more CRS REs averaged, so a steadier RSRP. Below 6 RB the estimate would be too noisy to trust.
RSRQ — Folding In Load and Interference
RSRP tells you the signal is strong; it says nothing about how much junk sits on top. The Reference Signal Received Quality (RSRQ) fills that gap by relating the clean per-RE signal to the total received power, scaled by the number of resource blocks so the two quantities are comparable:
Here N is the number of resource blocks of the E-UTRA Carrier RSSI measurement bandwidth — i.e. the same N the RSSI was measured over. The factor is essential: RSRP is a per-RE power while RSSI is a wideband power summed over many RBs, so multiplying RSRP by N puts numerator and denominator on the same footing before the division. Both RSRP and RSSI must be measured over the same set of resource blocks for the ratio to be valid.
The result is a ratio, expressed in dB, and it is essentially always negative. When the cell is lightly loaded and interference-free, RSSI is dominated by the wanted signal and RSRQ climbs toward its best values — around -3 dB is close to the theoretical ceiling for a lightly loaded cell with a single transmit antenna port active. As neighbours light up and interference grows, or as the serving cell itself loads up with traffic on the PDSCH, RSSI rises while RSRP stays put, so RSRQ falls. That is the whole point: RSRQ moves with load and interference even when raw coverage is unchanged.
Why two metrics, not one: two cells can share the same RSRP (equally strong) yet have very different RSRQ — the one on a congested, interference-heavy carrier reports worse RSRQ. That is why event-triggered handover in loaded networks often leans on RSRQ, while pure coverage decisions lean on RSRP.
Rel-13 RSRQ extension
The classic definition above measures RSSI over the CRS-bearing OFDM symbols. Release 13 added an extended RSRQ definition in TS 36.214 that lets the UE measure the E-UTRA Carrier RSSI over all OFDM symbols of the downlink subframe (not only the reference-symbol symbols), signalled to the UE via measurement configuration. This matters because in networks using time-domain features — almost-blank subframes (ABS) for eICIC, or heavy DL/UL asymmetry — the CRS-symbol-only RSSI can misrepresent the true interference the UE will see on data symbols. The wideband-symbol variant gives a more representative quality figure. It also broadened the reporting range and finer step (see the reporting-range table), so devices could distinguish very good cells from merely good ones. The formula is unchanged; only the symbols that enter RSSI differ.
RS-SINR — Signal vs Interference and Noise
RS-SINR (Reference Signal Signal-to-Interference-plus-Noise Ratio, written RS-SINR) is the metric that best predicts how fast you can actually go, because throughput ultimately tracks how far the wanted signal stands above the interference-plus-noise floor. TS 36.214 defines it as the linear average over the power contribution of the resource elements carrying CRS divided by the linear average of the noise-and-interference power contribution over the same resource elements, within the measurement bandwidth. In short: reference-signal power over the interference-plus-noise measured on those same reference REs.
Two clarifications keep this straight. First, plain "SINR" as exposed in chipset field tools is historically a vendor-computed number and was not a 3GPP-reported quantity in early LTE releases; RS-SINR is the standardised, reportable version (added for later releases, with its own reporting range in TS 36.133). Second, RS-SINR feeds link adaptation: it is closely tied to how the UE picks the CQI it feeds back, which in turn drives the MCS the scheduler assigns.
Roughly, higher is better: above 20 dB is excellent (top-order MCS), around 13–20 dB is good, 0–13 dB is fair, and below 0 dB the wanted signal is weaker than interference-plus-noise and service degrades fast. Because it explicitly separates interference from signal, RS-SINR can be poor even where RSRP is healthy: a strong cell sitting under an equally strong neighbour has good RSRP but poor RS-SINR. That is precisely the "cell-edge in a dense network" condition where RSRP alone would mislead you.
Don't confuse them: RSRQ and RS-SINR both describe "quality," but RSRQ is a ratio of clean signal to total power (so it saturates as load falls), while RS-SINR is a ratio of signal to interference-plus-noise only (so it keeps climbing as interference falls). They correlate but are not the same number, and RS-SINR is the better throughput predictor.
Layer-1 vs Layer-3: Filtering the Raw Samples
The value TS 36.214 defines is a physical-layer (Layer-1) measurement — a single estimate the PHY produces every measurement period. But radio measurements are noisy: fast fading can swing an instantaneous RSRP by many dB slot to slot. If the network reacted to raw L1 samples it would trigger handovers on momentary fades — the dreaded ping-pong. So between the PHY estimate and the reported value sits a defined smoothing stage.
L1 filtering is the implementation-side averaging the PHY does across the measurement bandwidth and a measurement period to produce one sample. L3 filtering is the standardised exponential moving average RRC applies to those L1 samples before evaluating events or sending a report.
L1 smoothing removes per-RE noise; L3 smoothing removes fast-fading swings so mobility decisions ride the slow, meaningful trend rather than instantaneous dips.
L3 uses the recursion Fn = (1−a)·Fn−1 + a·Mn, where Mn is the latest L1 sample, Fn the filtered output, and a = (1/2)^(k/4) with k set by filterCoefficient.
The filterCoefficient lives in the QuantityConfig IE of the RRC measurement configuration (RRCConnectionReconfiguration), and there is a separate coefficient for RSRP and for RSRQ. It takes values such as fc0, fc1, fc2, fc4 … fc19. A small k (e.g. fc0, giving a = 1) means "trust the newest sample fully" — fast reaction, more ping-pong. A large k (e.g. fc9 or higher) weights history heavily — slow, stable, but laggy. Tuning this coefficient is one of the main levers an operator has to trade handover responsiveness against stability.
The full pipeline: CRS REs → PHY estimate (L1 filtering) → L3 exponential filter (filterCoefficient) → event evaluation (A/B events) → MeasurementReport. The quantity in TS 36.214 is the L1 sample Mn; the value the network acts on is the filtered Fn.
One consequence worth remembering for interviews: the reporting periodicity and the filter are independent. Even with periodic reporting configured, each report carries the filtered quantity, and the entry/exit conditions of event-triggered reporting are tested against Fn, not the raw sample — which is why time-to-trigger and filterCoefficient together shape how twitchy or sluggish mobility feels.
⚠ Common pitfalls / gotchas
- Reading RSRP without RSRQ/SINR. A strong
RSRPat a dense-network cell edge can still be unusable — a co-channel neighbour of equal power tanksRS-SINR. Judge coverage byRSRP, but judge usability byRSRQ/RS-SINR. - Comparing RSSI across bandwidths.
RSSIscales withN, so an RSSI figure is meaningless unless you know how many RBs it was measured over. This is also whyRSRQcarries theNfactor. - Forgetting L3 filtering. A drive-test tool showing raw L1 swings will not match the value that fires events. Events test
Fn, so an aggressivefilterCoefficientor longtimeToTriggercan explain a "why didn't it hand over?" mystery. - Treating the report as raw dB. Reported quantities are integer indices into TS 36.133 tables (e.g.
RSRP0–97), not floats — off-by-one mapping errors are a classic log-parsing bug.
Reporting Ranges, and What Drives What
Put the quantities side by side and the division of labour is clear — one measures total power, one isolates signal, one measures quality against total power, one measures signal against interference:
| Quantity | Definition (TS 36.214) | What it tells you | Range (TS 36.133) |
|---|---|---|---|
RSSI | Total linear-average received power over N RBs on CRS-bearing OFDM symbols (signal + interference + noise) | How loud the whole band is; the denominator of RSRQ | Not reported alone (internal term); dBm |
RSRP | Linear average of the power of the CRS resource elements, per RE | True coverage / signal strength, bandwidth-independent | RSRP_00 … RSRP_97 (−140 to −44 dBm, 1 dB step) |
RSRQ | N × RSRP ÷ RSSI | Quality folding in load and interference | RSRQ_00 … RSRQ_34 (−19.5 to −3 dB, 0.5 dB step) |
RS-SINR | CRS RE power ÷ interference+noise power on those REs | Best throughput predictor; drives CQI/MCS | SINR_00 … (−23 to +40 dB, 0.5 dB step) |
The reported quantity is never a raw dBm/dB float — it is an integer index into a mapping table in TS 36.133. The two you must know cold:
| Reported value | Measured quantity | Step | Notes |
|---|---|---|---|
RSRP_00 | RSRP < −140 dBm | — | Bottom of range |
RSRP_01 | −140 ≤ RSRP < −139 dBm | 1 dB | RSRP index n maps to −141+n dBm |
RSRP_n | −141+n ≤ RSRP < −140+n dBm | 1 dB | |
RSRP_97 | RSRP ≥ −44 dBm | — | |
RSRQ_00 | RSRQ < −19.5 dB | — | Bottom of range |
RSRQ_n | −20+n×0.5 ≤ RSRQ < −19.5+n×0.5 dB | 0.5 dB | 0–34 base range |
RSRQ_34 | RSRQ ≥ −3 dB | — | Top of base range (Rel-13 extends beyond) |
So RSRP reports as an index 0–97 spanning −140 to −44 dBm in 1 dB steps, and RSRQ reports as an index 0–34 spanning −19.5 to −3 dB in 0.5 dB steps. The Rel-13 extended RSRQ pushes the top of the range higher (and adds negative-index extensions) with the same 0.5 dB granularity, so a very clean lightly-loaded cell can report better than the legacy −3 dB ceiling. Encoding as an index is what lets the whole measurement travel in a handful of bits inside a MeasurementReport.
How the numbers drive mobility
In RRC_IDLE, cell selection hinges on RSRP through the S-criterion: the cell-selection receive level Srxlev = Qrxlevmeas − (Qrxlevmin + Qrxlevminoffset) − Pcompensation must be positive, where Qrxlevmeas is the measured RSRP and Qrxlevmin is broadcast in SIB1. Coverage is a signal-strength question, so RSRP is the natural input. Cell reselection then ranks candidate cells by an RSRP-based R criterion (Rs/Rn with hysteresis Qhyst and offsets), and some deployments add an RSRQ threshold (threshServingLowQ, Rel-11+) so a UE reselects away from a cell that is strong but congested.
In RRC_CONNECTED, handover is triggered by measurement events the network configures in ReportConfigEUTRA, each testing the filtered RSRP or RSRQ (chosen via triggerQuantity) against a threshold, with a hysteresis and timeToTrigger:
| Event | Entry condition (intuition) | Typical trigger quantity | Used for |
|---|---|---|---|
A1 | Serving becomes better than a threshold | RSRP/RSRQ | Cancel inter-freq/inter-RAT measurements (back in good coverage) |
A2 | Serving becomes worse than a threshold | RSRP/RSRQ | Start measuring neighbours; arm handover |
A3 | Neighbour becomes better than serving by an offset | RSRP (often) | Intra-/inter-frequency handover trigger |
A4 | Neighbour becomes better than a threshold | RSRP/RSRQ | Handover to a specific target (e.g. load balancing) |
A5 | Serving worse than threshold1 and neighbour better than threshold2 | RSRP/RSRQ | Coverage-driven handover away from a failing serving cell |
B1/B2 | Inter-RAT neighbour better than threshold (B2 also needs serving worse) | RSRP-equivalent per RAT | Handover/redirection to UTRAN/GERAN/other RAT |
The pattern is that A-events are E-UTRA (intra-LTE) and B-events are inter-RAT. Operators commonly use RSRP for coverage-driven events (A2/A3/A5) and switch triggerQuantity to RSRQ when they want the UE to flee a cell that is strong but overloaded — the exact scenario RSRP can't see. Because these events test the L3-filtered value, the same filterCoefficient and timeToTrigger from the previous section decide how quickly an event actually fires.
LTE ↔ NR: NR keeps the same three quantities but renames and re-anchors them — SS-RSRP/SS-RSRQ/SS-SINR measured on the SS/PBCH block, and CSI-RSRP/CSI-RSRQ/CSI-SINR measured on CSI-RS — because NR has no always-on CRS. NR measurements are also beam-level then cell-level: the UE measures per SSB beam, then consolidates the best beams (above absThreshSS-BlocksConsolidation) into a cell result. The A1–A5/B1/B2 event framework carries straight over, and NR adds A6 (for secondary-cell/CA neighbour comparison). L3 filtering with filterCoefficient works identically.
Q. Two cells report the same RSRP but very different RSRQ. What does that tell you?
A. Signal strength is equal, but one cell carries far more load/interference. Since RSRQ = N × RSRP / RSSI, equal RSRP with lower RSRQ means a higher RSSI — extra interference or traffic inflating total received power. You'd prefer the cell with the better RSRQ.
Q. Why is RSRP, not RSSI, used for cell selection?
A. RSRP is a clean per-RE signal power that is bandwidth-independent and free of interference, so it reflects true coverage. RSSI lumps in interference and noise and scales with bandwidth, so a strong RSSI could just mean a noisy environment — useless for judging whether a cell actually covers you.
Q. What is N in the RSRQ formula and why is it there?
A. N is the number of resource blocks over which RSSI is measured. RSRP is per-RE while RSSI is wideband, so multiplying RSRP by N scales it up to match the same bandwidth before dividing — without it the ratio would be meaningless.
Q. What is the difference between L1 and L3 measurements, and where does filterCoefficient apply?
A. L1 is the raw per-period PHY estimate; L3 is the exponential moving average RRC applies before events/reports, Fn = (1−a)Fn−1 + a·Mn. filterCoefficient (in QuantityConfig) sets a and thus how much the filter smooths — the network acts on Fn, not the raw sample.
Q. What are the reporting ranges for RSRP and RSRQ?
A. RSRP maps to indices 0–97 (−140 to −44 dBm, 1 dB step); RSRQ maps to indices 0–34 (−19.5 to −3 dB, 0.5 dB step), with Rel-13 extending the range. Both are reported as integer indices, not raw dB values.
Q. When would an operator trigger handover on RSRQ rather than RSRP?
A. When a cell is strong but congested. RSRP stays high under load, so a coverage-based event wouldn't fire; RSRQ drops as RSSI rises with interference/traffic, letting an A2/A5 event move the UE off the overloaded cell.
Summary
LTE's four measurement quantities form one chain. RSSI is total received power over N RBs on CRS-bearing symbols — signal plus interference plus noise. RSRP isolates just the CRS resource elements, giving a bandwidth-independent, per-RE signal-strength value in dBm. RSRQ = N×RSRP/RSSI folds load and interference back in as a dB quality ratio. RS-SINR compares CRS power to interference-plus-noise on those same REs and is the best throughput predictor.
None of these is acted on raw: the L1 PHY sample Mn is smoothed by the L3 filter (Fn, tuned by filterCoefficient) before events fire, and it travels as an integer index into TS 36.133 tables (RSRP 0–97, RSRQ 0–34). Coverage decisions — the S-criterion, A3/A5 handover — lean on RSRP; quality decisions in loaded networks lean on RSRQ or RS-SINR. Pick the right metric and mobility works; pick the wrong one and you either miss a needed handover or ping-pong on a fade.
Where measurements connect
These raw quantities are the inputs to mobility: the UE first has to find a cell to measure, the reference signals give it something clean to measure, and the network then configures which measurements report and when they trigger action.