Why Does My Turbidity Meter Read High, Noisy, or Drift?

A turbidimeter measures one thing: the intensity of light scattered by suspended particles, which it converts to NTU or FTU through a calibration curve. Anything else that scatters light, blocks light, or changes the source intensity moves the reading exactly as real turbidity would. Bubbles, window fouling, sample color, stray light, and a stale calibration therefore account for most high, noisy, or drifting readings. The quantity that decides each case is the detector signal relative to a known reference. Compare a clean zero, a traceable standard, and a grab sample, and the fault separates into optics, sample, or calibration.

NTU Signal Patterns and What Produces Them

Classify the waveform before touching the instrument. Scatter from particles is continuous and tracks the process. Scatter from bubbles is intermittent and spiky. Fouling and source aging are monotonic, and they move in one direction between cleanings or calibrations.

Observed pattern	Physical cause	Deciding check
Random positive spikes on a steady baseline	Entrained air bubbles scattering light in the sample chamber	Degas or let the grab sample settle. If the spikes disappear, the problem is sample handling.
Slow upward creep between cleanings	Film or deposit on the optical windows adds fixed scatter	Clean the windows and re-read the same sample. A step down confirms fouling.
Slow downward creep over weeks or months	Light source intensity decay, typical of tungsten lamps as they age	Read a known standard. A low result on the standard points to source or calibration.
Reading low on colored samples	Absorption by dissolved color attenuates both incident and scattered light	Compare against a filtered or colorless sample of similar particle load.
Non-zero reading on turbidity-free water	Stray light, scratched cell, condensation, or fouled optics	Measure the zero in a clean cell. Wipe or replace the cell if the offset persists.
Reading flattens or falls as the process gets dirtier	Multiple scattering and absorption at high particle concentration	Dilute a grab sample and check whether the result scales with the dilution factor.
Offline analyzer and online analyzer disagree	Different optics, light source, or calibration reference	Calibrate both against the same standard and compare on a shared grab sample.

Scattered-Light Physics Behind the Reading

Light entering a turbid sample does two things. Particles scatter it in all directions, and particles plus dissolved color absorb part of it. The detector sees only the scattered fraction that reaches its viewing angle. Nephelometric instruments read the light scattered at right angles to the incident beam, which gives the unit its name. Some designs also read forward scatter or the transmitted beam.

Definitions of "nephelometric" versus "turbidity" geometry vary between references. Some descriptions assign nephelometry to forward scatter and call right-angle detection turbidity, which conflicts with the usual convention. Read the actual detector geometry from your analyzer's datasheet rather than inferring it from the unit label. The geometry matters because each angle responds differently to particle size, and two instruments with different geometries can disagree on the same water even when both are correctly calibrated.

Three physical effects set the limits of the measurement:

• Particle size and shape. Scatter intensity depends on particle size relative to the wavelength, not on mass alone. NTU tracks the concentration of suspended particles only for a stable particle population. If the size distribution changes, the NTU-to-mg/L relationship changes with it.
• Absorption. Dark or colored samples absorb light along the optical path, so the detector under-reads. The effect is strongest with broadband white sources.
• Concentration saturation. At high particle loads, light scatters more than once before it reaches the detector, and the right-angle signal stops rising in proportion to turbidity. Every instrument has a linear range. Above it, the reading is no longer a valid measurement.

NTU and FTU both reference formazin, a synthetic polymer suspension with reproducible scatter. The unit label identifies the traceability chain. It does not guarantee agreement between instruments with different optics.

Light Source and Detector Behavior

The optical chain has three parts: source, sample chamber, and detector. Each has its own failure mode, and each shows up in a different quantity.

A tungsten lamp emits broadband light. Its intensity falls with filament age, and its output shifts with supply voltage. Broadband light is also more sensitive to sample color. An LED source emits a narrow band, is more stable over its life, and is less affected by color if its wavelength sits outside the visible absorption of the sample. Because instruments with the two source types respond differently to the same sample, keep the source type consistent when you compare readings or transfer a compliance method.

Photodiodes cover most process ranges. Photomultiplier tubes (PMTs) give higher gain for very low turbidity, but their gain is sensitive to supply voltage and aging. At very low NTU values, the scattered-light signal is small compared with stray light and detector dark current, so the zero offset determines the lower limit of reliable measurement.

Quantity	Limit it reveals	Where to read it
Lamp or LED intensity / reference detector signal	Source aging and supply drift	Analyzer diagnostics menu or service screen (see the manual for the parameter name)
Reading on turbidity-free water	Stray light and window contamination floor	Zero check with a clean cell or flow-through chamber flush
Reading on a known standard	Calibration slope error	Verification step against a traceable reference
Upper linear range	Onset of multiple scattering	Instrument datasheet specification table
Signal variance at steady process	Bubble and particle-counting noise	Trend of raw reading before damping or averaging

Sample Preparation: Bubbles, Dilution, and Filtration

Air bubbles are the most common cause of false high readings, because a bubble scatters light the same way a particle does. In online analyzers, bubbles come from pressure drops, cavitation, and aerated process streams. In benchtop work, they come from shaking or pouring the sample. A bubble trap, positive back-pressure on the flow cell, or a short settling time before reading removes them. Heavy signal damping only hides the spikes and still raises the average.

Dilution brings a sample above the linear range back into it. Use turbidity-free water and back-calculate:

T_sample = T_measured × (V_sample + V_diluent) / V_sample

Dilution changes the particle environment, and some suspensions flocculate or dissolve when diluted. Run two dilution ratios. If the back-calculated values agree, the dilution is valid. If they do not, the result depends on the ratio and needs a different approach, such as an instrument with a wider range.

Filtration to remove large particles changes the quantity being measured, since it removes part of the scattering population. Filter only when your method defines turbidity on a filtered fraction. Otherwise, report unfiltered results.

Calibration and Verification Procedure

Calibrate against standard reference materials of known turbidity, and bracket the operating range. A calibration point far from the process value leaves the slope untested where it matters.

1. Record the current reading on the process or on a retained grab sample. This gives you a before/after reference.
2. Isolate the analyzer from the control loop. Put any dosing, filter-backwash, or alarm logic that uses the turbidity signal in manual or bypass.
3. Inspect and clean the optical windows or sample cell with a lint-free wipe and a cleaning agent compatible with the window material. Check for scratches, which cause a permanent stray-light offset.
4. Check the source intensity or reference-detector value in the diagnostics menu. Replace the lamp or LED module if it falls outside the manufacturer's limit.
5. Run a zero check with turbidity-free water, free of bubbles. Note the offset.
6. Present the calibration standards in the order the manual specifies. Invert gently to resuspend them without adding air, wipe the cell exterior, and index the cell to the same orientation every time.
7. Accept the calibration only if the instrument reports a valid slope. A slope rejection indicates a degraded standard, a dirty cell, or a failing source.
8. Verify with an independent standard that was not used in the calibration, ideally one close to the normal process value.
9. Return the analyzer to service, and release the control loop only after the reading has settled on the live process.

Confirming Stable, Traceable Readings After the Fix

A successful calibration shows the instrument reads standards correctly. It does not show the process reading is correct. Confirm both:

• Zero: turbidity-free water reads at or near zero, with no upward drift over a few minutes. Drift at zero indicates condensation, a warming cell, or slow bubble formation.
• Span: the independent verification standard reads within the manufacturer's stated accuracy.
• Process agreement: a grab sample taken at the analyzer inlet and read on a separately calibrated benchtop unit agrees with the online value. Allow for geometry and source differences between the instruments.
• Signal quality: the raw trend at steady process shows noise without periodic spikes. Periodic spikes that line up with pump cycles or valve actions indicate air entrainment.
• Fouling rate: log the reading before and after each cleaning. The step at cleaning measures the fouling contribution and sets the cleaning interval.

Recurring Failure Modes on Process Turbidimeters

• Signal and window fouling in the same direction. Deposits on the optics raise the reading slowly, so the trend resembles a gradual process change. A cleaning step-change in the history shows which one it is.
• Degraded standards. Formazin and other suspensions have limited shelf life and settle in the bottle. An expired or unmixed standard shifts the whole calibration. Check expiry dates and mixing practice before you blame the sensor.
• Cell orientation and fingerprints. Glass imperfections scatter light differently at different rotations. Index-mark the cells and handle them by the top.
• Temperature condensation. Cold samples in warm, humid rooms fog the outside of the cell and give a rising reading. Warm the sample or dry the cell exterior.
• Using NTU as a mass measurement. An NTU-to-concentration correlation holds only for the particle population it was built on. Re-derive it when the source water, coagulant, or process changes.
• Over-range operation. Readings past the linear range can fall as turbidity rises. A control loop acting on that signal will move in the wrong direction. Configure an over-range alarm from the datasheet limit.
• Heavy damping. Long averaging masks bubble noise and delays the response to real turbidity breakthrough. Fix the sample hydraulics first, then set damping to the process time constant.

FAQ