Menu
Application GuideSelection Guides17 min read

Why Soil Moisture Sensors Give Flat, Spiky or Frozen Readings

A soil moisture trace that stays flat, jumps to an implausible value or stops updating is easy to mislabel as a calibration problem. This field guide separates soil contact, power, wiring, logger and wireless faults with a diagnostic sequence and acceptance checklist.

Updated October 4, 2026How Equipvia researches
Agricultural water pumping station for sensor signal-path fault checks
Diagnose the signal path before the calibration menu: soil contact, power, cable, connector, logger scaling and wireless backhaul each produce characteristic symptoms. Irrigation pumping station photograph via Wikimedia Commons (CC BY-SA 2.0). View current BGT listing.

Disclosure: We may earn a commission if you purchase through links on this page, at no additional cost to you.

A soil moisture trace that stays flat, jumps to an implausible value or stops updating is easy to mislabel as a calibration problem. Calibration can matter, but it is often the wrong first intervention. A probe may be sitting in an air gap, a cable may have water ingress, a connector may have oxidised, the logger may be applying the wrong register or scaling, or a wireless node may be reporting an old value. Recalibrating a sound sensor will not repair any of those faults.

The useful diagnostic question is not "what percentage is correct?" until the signal path is working. First establish whether the sensor is responding to a controlled wetting or drying event, whether the raw output changes, and whether the value survives the trip from probe to dashboard. The sequence below is designed for field technicians and procurement teams who need to separate installation, electronics and data problems before replacing hardware. It also keeps absolute accuracy claims separate from the more modest evidence that a sensor responds and repeats.

Start with the symptom rather than the calibration menu

Record the symptom, the last known good time and what changed immediately before it. Rain, irrigation, fertiliser injection, trench work, a firmware update and a battery replacement are useful clues. Capture the raw register, voltage or current as well as the displayed percentage; a frozen dashboard can hide a changing electrical signal, while a bad scaling rule can make a healthy probe look dead.

Observed symptom First distinction to make Safe first check
Flat line No physical response, no power, or stale logger value Compare raw output and timestamp; apply a small, documented wetting change.
Constant zero or full scale Open circuit, wrong units, bad contact, or genuine extreme condition Inspect connector and insertion; compare with a second depth or reference point.
Sudden spikes Cable movement, EMI, intermittent contact, or a real wetting front Overlay the trace with irrigation, cable movement and logger events.
Slow drift Sensor ageing, salt or temperature effect, soil settling, or software scaling Check temperature, EC context, installation depth and raw counts before recalibration.
Wireless gaps Probe failure, node battery, radio path, gateway or cloud queue Read the local node and gateway status separately; do not infer probe failure from the dashboard alone.

The four failure layers

1 Soil contact and placement

Dielectric and capacitive probes measure a volume around the sensing element, not an abstract point. An air pocket, loose backfill, a probe tilted in a borehole or a probe placed against a stone can change that volume. The result may be a low value, a weak response to rain or a trace that looks plausible but is biased. Preferential flow can wet the cable trench or one side of a probe without representing the surrounding root zone. Arizona Cooperative Extension and UF IFAS both emphasise representative placement and good soil contact for field measurements [E01][E02].

Check depth, orientation, insertion method and the material immediately around the sensing element. Do not pull a buried probe repeatedly to "test" it; disturbance changes the very condition being measured. If the installation is suspect, compare a carefully installed spare at the same depth rather than averaging two uncertain readings.

2 Power, cable and connector integrity

A probe can be electrically alive but operationally unusable when supply voltage sags during a reading, a shield is terminated incorrectly or moisture reaches a connector. Long cables add resistance and can pick up interference. Corrosion, crushed cable, rodent damage and a loose terminal can create intermittent spikes that disappear when a technician touches the cable.

Measure the supply at the sensor while it is sampling, not only at the cabinet. Record the output with the cable still connected, then repeat with a known-good lead if the design permits. For RS485, check polarity, termination and address conflicts. For analogue voltage or current, confirm the input impedance and whether the logger expects 0–3 V, 0–5 V or 4–20 mA. Never assume that two products with the same connector use the same pinout.

3 Logger, register and scaling errors

A flat chart can be produced by a logger that is polling the wrong register, retaining the last successful value or rejecting a CRC response. A value that changes in raw counts but not on the dashboard points to conversion, unit or database logic. The reverse is also possible: a display may move because a software rule is filling missing data, even though the probe has stopped responding.

Save a short raw capture before changing the configuration. Confirm model, address, baud rate, parity, sampling interval, register map, byte order, signedness and scale factor. For a current-loop device, calculate the expected engineering range from the configured shunt and compare it with the logger's assumption. Keep the old configuration so a retest can be reproduced.

4 Wireless and backhaul failures

When a wireless system is involved, separate four timestamps: the sensor sample, the node transmission, the gateway receipt and the cloud display. A gateway can show a stale but neatly formatted value. Battery depletion, a changed antenna position, vegetation or a metal enclosure can reduce packet delivery without affecting the probe. Read the local node or service port where available, then inspect RSSI, retry count, queue depth and last-seen time.

Do not "fix" a wireless gap by increasing the sampling rate blindly. More transmissions can shorten battery life and create more collisions. First establish whether the missing data is local, radio or cloud-side. A wired comparison probe at the same depth is often the fastest way to avoid replacing a healthy sensor because of a network fault.

A field sequence that avoids false diagnosis

  1. Freeze the evidence. Export the raw series, displayed series, timestamps, configuration and alarms. Note irrigation, rainfall, fertiliser, maintenance and any physical disturbance.
  2. Check the physical installation. Confirm depth, orientation, soil contact, cable strain relief and that the sensing section is not sitting in a void or preferential-flow path.
  3. Check power under load. Measure supply at the probe during a reading. Inspect terminals, splices, shields, glands and connectors for corrosion or moisture.
  4. Test the shortest signal path. Read the probe locally or with a short known-good cable. For digital outputs, confirm the raw frame; for analogue outputs, measure voltage or current directly.
  5. Apply a small controlled change. Use a documented wetting or drying change that will not damage the crop. A response that appears at the probe but not the dashboard localises the fault downstream.
  6. Compare depth or a reference. A multi-depth probe or a second instrument can reveal whether the problem follows the soil layer, the cable, the channel or the individual sensor.
  7. Correct configuration before calibration. Fix register, scaling, units, address, power and radio settings. Re-run the controlled change after each material correction.
  8. Only then evaluate calibration. If the trace is repeatable but biased against a reference method, document the soil-specific relationship. Keep the raw output and the calibration version together.

When a replacement is justified

Replacement is justified when the sensor fails a repeatability check after installation and wiring are known good, when the raw output is outside the documented electrical range, when the sensing element or housing is physically damaged, or when drift remains after cleaning and a controlled comparison. A replacement is not justified merely because a generic percentage disagrees with a soil sample. That disagreement may be a calibration or measurand problem, not a failed probe.

Record the failed unit's serial or batch, installation depth, raw output, supply, temperature, EC context and the test method. This evidence supports a warranty claim and prevents a new unit from being installed into the same air gap or configuration error.

Sourcing candidates for a troubleshooting-led purchase

The two offers below serve different diagnostic roles. They are sourcing candidates, not ranked recommendations. Marketplace values are supplier or platform statements and must be confirmed for the exact variant in the quotation.

Candidate — BGT multi-depth RS485 probe

Candidate — Honde configurable capacitive probe

RFQ questions that prevent a troubleshooting dispute

  • What exact quantity is reported: relative moisture, volumetric water content, raw dielectric value, voltage, current or a supplier-defined index?
  • Which model, depth, output, cable length, firmware and wireless module are included in the quotation?
  • What are the raw output range, register map, byte order, scale factor, units and error response when communication fails?
  • What soil or substrate, temperature and EC range were used for the stated accuracy, and is the figure typical or guaranteed?
  • How is site-specific calibration uploaded, versioned and restored after a logger or firmware replacement?
  • What evidence accompanies each shipment: serial or batch record, test report, wiring diagram, installation guide and warranty procedure?

Acceptance checklist

Before accepting a deployment, save a commissioning packet containing the installed depth and coordinates, photographs of contact and cable routing, supply voltage under load, raw output, displayed value, logger configuration, gateway last-seen time, and at least one controlled wetting or drying response. Mark every statement as measured, supplier-stated or not yet verified. If the buyer needs VWC, add paired reference samples and report the tested range and error; do not convert a successful response check into an accuracy certificate.

The practical stopping rule is simple: once the signal path is healthy and repeatable, decide whether the remaining difference is acceptable for the buyer's decision. If it is not, move to a documented site-specific calibration or choose a sensor with evidence that better matches the soil and reporting requirement. This order keeps a wiring fault from being hidden by a new calibration curve.

Bottom line

Diagnose the signal path before touching the calibration menu. Establish that the probe responds, that the value survives the trip to the dashboard and that the trace is repeatable; only then decide whether accuracy work is needed. Soil contact, power, cable and connector faults, logger scaling and wireless backhaul each produce characteristic symptoms, and a short controlled wetting change localises most of them quickly. BGT is a wired multi-depth reference candidate for depth and installation checks; Honde is a configurable candidate for separating output, connector and logger problems. Both remain sourcing candidates until the quoted variant passes the buyer's field sequence.

Evidence and source notes

Share this article

Related research

View all guides
Cooling & Condensation·20 min read

How to Size a Chiller for a Rotary Evaporator

Size a rotary evaporator chiller by required condenser temperature and cooling capacity at that temperature—not by minimum temperature or flask size alone.

Updated September 19, 2026