What each measurement actually reports
Four measurements describe a radio link, and they are routinely confused with one another. They are not interchangeable, and a recommendation built on the wrong one is usually wrong. The distinction that matters most is between how much signal there is and how much of it is usable.
| Measure | What it reports | Range | Usable from |
|---|---|---|---|
| RSRP | Power received on the resource elements carrying the reference signal — signal strength alone, with interference excluded. | −44 to −140 dBm | ≥ −100 dBm |
| RSSI | Total wideband received power across the measured bandwidth — wanted signal, interference and noise together. | Context dependent | Interpret with RSRP |
| RSRQ | Reference signal quality — RSRP expressed against total received power, so it captures how loaded and how interfered the band is. | −3 to −19.5 dB | ≥ −10 dB |
| SINR | Wanted signal against interference plus noise. This is the measurement that determines modulation, and therefore throughput. | −20 to +30 dB | ≥ 10 dB |
RSRQ = (N × RSRP) / RSSI
N is the number of resource blocks across the measured bandwidth. Because RSSI includes everything the receiver hears, RSRQ falls when the band gets busy even if RSRP has not moved at all. A normal RSRP with a poor RSRQ is an interference or load signature, not a coverage one.
SINR = S / (I + N) ≈ RSRP / (RSSI − RSRP)
RSSI is wideband received power; RSRP is the reference signal only. The difference between them is everything that is not wanted signal. High RSRP with low SINR is the classic interference-limited cell.
Reading RSRP and SINR together
Neither measurement is a diagnosis on its own. Read as a pair they resolve almost every field observation into one of four situations, and each situation has a different remedy. This is the single most useful habit in drive test analysis.
| SINR good | SINR poor | |
|---|---|---|
| RSRP good | HealthyStrong signal, clean. Nothing to fix here — confirm the sample is representative and move on. | Interference limitedPlenty of signal, too many contributors. Look at overshoot, cell overlap, PCI conflicts and antenna tilt. |
| RSRP poor | Clean cell edgeWeak but uncontested. Acceptable at a boundary; a candidate for power, tilt or a new site if it is a service area. | Coverage holeNot enough signal and no dominant server. Needs a coverage answer — new site, repeater or indoor solution. |
Why the pair matters
Adding power to an interference-limited cell makes the problem worse, not better: every neighbour hears more of it too, and the SINR at the boundary does not improve. The matrix exists to stop that recommendation being made.
Path loss and propagation
Path loss is what separates transmit power from received power. In open space it follows a clean relationship with distance and frequency; in a real environment it does not, which is why empirical models exist.
PL = 20·log₁₀(d) + 20·log₁₀(f) + 32.45
d in kilometres, f in megahertz. Useful as a floor and a sanity check — real path loss is always higher. Note that doubling either distance or frequency costs about 6 dB.
PL = 69.55 + 26.16·log₁₀(f) − 13.82·log₁₀(h_b) − a(h_m) + (44.9 − 6.55·log₁₀(h_b))·log₁₀(d)
f in MHz, h_b base station height in metres, h_m mobile height in metres, d in kilometres, a(h_m) the mobile antenna correction factor. The model is empirical and band limited — check its validity range before trusting a result outside it.
The practical consequence of the frequency term is worth stating plainly: higher frequency means shorter range for the same link budget. An 1800 MHz cell covers roughly 60 per cent of the radius of a 900 MHz cell built to the same maximum path loss. This is the reason low band carries coverage and high band carries capacity, and it does not change with technology generation.
Thermal noise and receiver sensitivity
Every receiver has a noise floor set by physics and by its own electronics. Sensitivity is that floor plus whatever signal-to-noise ratio the chosen modulation needs.
N = −174 dBm/Hz + 10·log₁₀(BW_Hz) + NF
For a 10 MHz LTE carrier with a 7 dB noise figure: −174 + 70 + 7 = −97 dBm. Add the minimum SINR the modulation requires — around −5 dB for QPSK — and receiver sensitivity lands near −102 dBm.
The bandwidth term is why a wider carrier is not free. Going from 10 MHz to 20 MHz doubles the noise power, costing 3 dB of sensitivity and therefore some coverage — the capacity gain is real, but it is paid for at the cell edge.
Interference and cell identity
Once a network is dense, most quality problems are interference problems. These are the concepts that account for the majority of them.
| Concept | Definition | Field implication |
|---|---|---|
| Pilot pollution | Three or more cells received within about 6 dB of the best server, with no clear dominant. | Low SINR and unstable handover. Addressed with downtilt, power trim and a cleaner dominance plan. |
| Overshoot | A cell serving well beyond its intended footprint, usually from insufficient tilt or excessive height. | Interferes with distant cells and creates handover failures at long range. Visible as isolated far-field samples in drive data. |
| ICIC | Coordination of resource block allocation between neighbours over X2. | Reduces inter-cell interference specifically at the cell edge, where it is worst. |
| eICIC | Almost-blank subframes that protect a small cell from its overlaying macro. | Separates macro and small cell resources in time. Relevant wherever a heterogeneous layer is deployed. |
| PCI planning | Physical cell identity — 0 to 503 in LTE, 0 to 1007 in 5G NR. | Reuse must avoid collision and confusion. Mod 3 conflicts degrade the shared channel; plan the modulo relationships, not just the raw identity. |
Where this goes next
These measurements and models are the inputs to a link budget, which converts them into a maximum allowable path loss and then into a cell radius. That is the subject of the next article.