Support field guide · 2026-10-16

LTE and 5G FWA signal levels for support agents

A support desk reference for fixed-wireless access (FWA): read the radio measurements, compare the symptom with the evidence, then choose a reversible next step.

Discuss a guided pilot

What the four measurements mean

ReadingUnitWhat the agent learns
RSRP — Reference Signal Received PowerdBmPower of the cell’s reference signal. A less negative value is stronger: −85 is stronger than −105.
RSRQ — Reference Signal Received QualitydBReference-signal quality relative to total received power. Read it alongside SINR and load symptoms.
RSSI — Received Signal Strength IndicatordBmTotal received power across the measured bandwidth, including interference and noise. Strong RSSI alone is not proof of a good link.
SINR — Signal to Interference plus Noise RatiodBUseful signal relative to interference and noise. Higher is better; a negative value means those unwanted components exceed the useful signal.

On 5G NR, record the exact field name: SS-RSRP, SS-RSRQ and SS-SINR use synchronization signals; CSI measurements use a different reference. Do not silently mix them with LTE readings. Check the CPE manual when its dashboard only says “RSRP”. Zyxel’s connection-status reference explains the reported fields; its 5G guide distinguishes NSA and SA.

LTE: good, fair and poor starting ranges

These are triage bands, not 3GPP service guarantees. “Good” combines the vendor’s good and excellent bands; “fair” means investigate. Boundary values are assigned explicitly below so adjacent bands never overlap.

LTE metricGoodFair / investigatePoor
RSRP (dBm)≥ −90≥ −100 and below −90below −100
RSRQ (dB)≥ −15above −20 and below −15≤ −20
SINR (dB)≥ 13above 0 and below 13≤ 0

5G NR: good, fair and poor starting ranges

Use the same conservative initial triage bands below, then qualify them against the CPE model, firmware and operator’s radio design. A 5G label does not justify a different universal pass mark. In NSA, capture both the LTE anchor and NR leg; in SA, capture the NR serving-cell readings. A missing NR reading is unknown, never zero or “good”.

5G NR metric (record SS or CSI label)GoodFair / investigatePoor
RSRP (dBm)≥ −90≥ −100 and below −90below −100
RSRQ (dB)≥ −15above −20 and below −15≤ −20
SINR (dB)≥ 13above 0 and below 13≤ 0

Sources for both tables: Teltonika’s 4G/5G recommendations provide the numerical bands. Zyxel’s 5G signal-quality guidance independently explains the metrics and warns about SINR below zero; it is not an identical grading scale. The three-band grouping and tie handling here are editorial choices. Neither table diagnoses congestion or promises a speed. RSSI has no standalone pass/fail row because bandwidth and unwanted power change its meaning.

Ask these questions before changing anything

  1. Where is the CPE? Indoor or outdoor, which room or wall, cabinet or window, orientation, antenna type and cable condition. Is the symptom new after a move or installation work?
  2. Which radio connection? Mobile operator, CPE model and firmware, LTE / NR NSA / NR SA / 3G, band (for example B3 or n78), channel number, channel bandwidth and any band or cell lock. Keep LTE and NR fields separate.
  3. When does it happen? All day or mainly in the evening? One subscriber or several in the same area? Record timestamps with timezone and repeat the same test during a quiet period.
  4. Which local path? Compare a controlled Ethernet test with Wi-Fi at the same time. Record upload, download, latency and loss to the operator’s approved test target. Note competing household traffic and any plan cap or fair-use limit.
  5. Does the serving cell change? Record the full serving-cell identifier and PCI together with band/channel. PCI alone is reused and does not uniquely identify a tower. Keep real subscriber/location data in the private support ticket.

Symptom → evidence → next action

Suspected causeEvidence to collectReversible first step / escalation
Indoor placement or obstructionWeak RSRP; repeat readings in several safe positionsTry a window and a different orientation, keeping the same test conditions. Coated glass can still attenuate radio signals. If there is a repeatable improvement outside, request a qualified outdoor-CPE installation; do not put an indoor unit in the weather.
Wrong band lockA forced band, missing aggregation or loss of the NSA LTE anchorRecord the old setting. Have an authorized operator test automatic selection, then a supported band if justified. Arrange local recovery before a change that could cut remote access.
Evening cell congestionSlow Ethernet tests at peak time; compare signal and throughput off-peak and with nearby affected linesStable radio readings do not rule out cell or backhaul load. Send timed comparisons to the mobile operator; stronger antennas alone cannot add scheduler capacity.
Ping-pong between cellsRepeated cell/band changes correlated with packet loss or interruptionsCompare stationary placement and antenna orientation. Escalate the timestamped sequence. A supported band/cell lock is a controlled experiment, not a permanent default.
3G fallbackAccess technology actually reports 3G; LTE/NR measurements absentVerify coverage, enabled modes, SIM/service provisioning and CPE band support with the operator. Do not apply LTE thresholds to RSCP or Ec/Io, and do not force LTE-only without a fallback/recovery plan.
Interference despite strong signalGood RSRP with weak SINR/RSRQCompare another safe position or an operator-approved band. Escalate persistent quality problems with measurements; do not diagnose interference from RSSI alone.

The placement examples in Zyxel’s FWA troubleshooting guide show why location testing is useful. Keep before/after evidence and change one thing at a time.

Worked ticket: strong power, poor evening service

Fictional training values, not a customer or live product screen. The serving cell is DEMO-CELL-A (fake ID), PCI 101 (invented), LTE B3. At 20:00 UTC: RSRP −86 dBm, RSRQ −18 dB, SINR 2 dB. At 10:00 UTC: RSRP −85 dBm, RSRQ −12 dB, SINR 16 dB. The CPE and wired test setup stayed in the same place.

The evening sample has good power but weaker quality. This is a reason to investigate time-dependent load or interference, not proof of either. Add timed throughput, latency/loss, plan limits and the operator’s cell-side evidence before assigning a cause. For a 5G NSA ticket, add a separate NR row; do not relabel the LTE sample “5G”.

Escalation checklist

  • Include model/firmware, access technology, bands/channels, timestamps, separate LTE/NR readings and serving-cell changes.
  • Add placement details, wired versus Wi-Fi comparison, peak/off-peak results and affected-service scope.
  • List every change, the before/after result and the restored setting. Agree who owns the next test.
  • Share identifiers only through the operator’s approved support channel. Never attach SIM credentials, passwords or a full unredacted configuration to a public thread.

Where ISPAgents fits

Planned: automatic cellular verdict. This page teaches a manual support workflow; it does not imply automatic cellular diagnosis, automatic band locking or a certified LTE/5G model list. A guided pilot can qualify a specific CPE model, firmware and the fields it exposes.

Read the TR-069 integration scope and CPE compatibility guide, then discuss a guided pilot. Confirm data availability and action boundaries before promising a rollout.

Frequently asked questions

Does good RSRP guarantee a fast connection?

No. RSRP measures reference-signal power. Interference, cell load, the service plan, backhaul and the local Wi-Fi or Ethernet path can still limit throughput.

Should an agent lock the CPE to a band?

Only as an operator-approved test with a recorded starting configuration and a way to restore automatic selection locally. A lock can remove a useful LTE anchor, carrier aggregation or fallback coverage.

Does ISPAgents automatically diagnose cellular signal problems?

The automatic cellular verdict is planned. This guide is a manual troubleshooting reference; discuss model and firmware qualification through a guided pilot.