Signal conditioning is the electronics between a sensor and the analog-to-digital converter that turn a sensor’s raw output into a clean signal the converter can measure accurately. It amplifies small signals, filters out noise and aliasing, supplies the excitation some sensors need, isolates the measurement from ground loops and hazardous voltages, and converts currents or bridge outputs into a voltage in the ADC’s range. Without it, many sensors either cannot be read at all or are read with errors far larger than the sensor itself.
This article is part of our complete guide to data acquisition (DAQ), which explains how sensors, conditioning, the ADC and software fit together. Here we go deeper on the conditioning stage: what each function does, what each common sensor needs, and how to specify it for a real system.
Why raw sensor signals need conditioning
A typical DAQ input measures ±10 V. Most sensors produce nothing like that:
- A Type K thermocouple produces about 41 µV per °C.
- A strain gauge bridge or load cell produces a few millivolts per volt of excitation, and nothing at all without that excitation.
- An IEPE accelerometer needs a constant current supply before it outputs anything.
- A process transmitter outputs a 4–20 mA current, not a voltage.
- A power or motor measurement may sit hundreds of volts above ground.
Worked example: why gain matters
On a ±10 V range, a 16-bit ADC divides 20 V into 65,536 steps, so one step (1 LSB) is about 305 µV. For a Type K thermocouple at 41 µV/°C, that is roughly 7.4 °C per step, useless for most temperature work. Amplify the signal by 100 so the effective range is ±100 mV, and one step becomes about 3 µV, or under 0.1 °C. The ADC did not change; conditioning made its resolution usable.
The main signal conditioning functions
| Function | What it does | Typically needed for |
|---|---|---|
| Amplification | Increases small signals to use the ADC’s full range and improve resolution | Thermocouples, bridges, microphones, low-level biological signals |
| Attenuation | Reduces large voltages into the input range | High-voltage and power measurements |
| Filtering | Removes noise and frequencies above half the sample rate (anti-aliasing) | Almost every analog measurement |
| Excitation | Supplies the voltage or current a passive sensor needs | Strain gauges, load cells, RTDs, IEPE sensors, LVDTs |
| Bridge completion | Adds precision resistors to form a full Wheatstone bridge | Quarter- and half-bridge strain gauges |
| Isolation | Breaks the electrical path between sensor and DAQ ground | Ground loops, high common-mode voltage, operator safety |
| Current-to-voltage conversion | A precision shunt turns a current into a voltage | 4–20 mA transmitters |
| Linearization and compensation | Corrects non-linear sensors and reference effects | Thermocouples (cold-junction compensation), thermistors, RTD leads |
| Protection | Limits overvoltage and transients before they reach the DAQ | Industrial and field wiring |
Signal conditioning by sensor type
The sensor decides most of the conditioning. This table covers the sensors we see most often in test and measurement systems:
| Sensor | Typical output | Conditioning needed | Watch out for |
|---|---|---|---|
| Thermocouple | Tens of µV per °C | High gain, cold-junction compensation, lowpass filtering, often isolation | CJC accuracy and open-thermocouple detection |
| RTD (e.g. Pt100) | About 0.385 Ω per °C | Precision excitation current, 3- or 4-wire connection, gain | Lead-wire resistance and self-heating |
| Thermistor | Large, non-linear resistance change | Excitation and linearization | Narrow useful range |
| Strain gauge | A few mV per V of excitation | Excitation, bridge completion, gain, shunt calibration, remote sense | Lead resistance, temperature drift, noise |
| Load cell or bridge pressure sensor | Typically 2–3 mV per V (full bridge) | Stable excitation, gain, filtering | Excitation drop over long cables |
| IEPE accelerometer or microphone | AC voltage riding on a DC bias | 2–20 mA constant current, AC coupling | Cable faults and the low-frequency limit of AC coupling |
| Charge-mode piezoelectric sensor | Electric charge (pC) | Charge amplifier | Cable movement noise; needs low-noise cable |
| LVDT | AC signal proportional to position | AC excitation and demodulation | Excitation frequency vs. measurement bandwidth |
| 4–20 mA transmitter | Current loop | Precision shunt (250 Ω gives 1–5 V), loop power | Ground loops when several transmitters share a supply |
| High voltage or current | Up to hundreds of volts, or amps | Attenuation, isolation, current transformers or Hall-effect sensors | Safety ratings and common-mode voltage |
Thermocouples have their own article, thermocouple data acquisition, because cold-junction compensation and accuracy budgets deserve more space than a table row.
Filtering and anti-aliasing
Filtering is the conditioning function people most often get wrong, because its most important job happens before the data exists. An anti-aliasing filter must be analog and sit in front of the ADC: once a high-frequency signal has been sampled too slowly, it appears as a false low-frequency signal that no software can remove. Set the filter just above the highest frequency you need and sample comfortably above twice that frequency; our article on high-speed data acquisition covers how to choose the sample rate.
- Butterworth filters have the flattest passband and suit most amplitude measurements.
- Bessel filters preserve waveform shape, which matters for pulses and transients.
- Elliptic filters roll off fastest but distort phase.
- Delta-sigma ADCs include steep digital anti-aliasing filters, which is why they are popular for vibration and acoustics, at the cost of some delay.
Separate from anti-aliasing, a lowpass filter well below the sample rate improves noise on slow signals such as temperature, and a notch or averaging over whole mains cycles rejects 50/60 Hz pickup.
Isolation, differential inputs and grounding
Noisy or drifting readings are more often a grounding problem than a sensor problem; our article on ground loops and galvanic isolation covers diagnosis in depth. When the sensor and the DAQ are grounded at different points, the difference in ground potential drives current through the measurement wiring and adds error, often at 50/60 Hz. Three tools deal with it, in increasing order of cost:
- Differential inputs measure the difference between two wires and reject voltage common to both, up to the input’s common-mode limit.
- Proper grounding and shielding: one ground reference per circuit, shields grounded at one end, and twisted pairs for low-level signals.
- Isolation breaks the conductive path entirely, so large common-mode voltages and ground loops cannot reach the measurement. It is also a safety requirement whenever sensors touch mains-level or hazardous voltages.
When comparing isolated hardware, check whether the rating is channel-to-earth only or also channel-to-channel, and whether it is a continuous working voltage or a short-term withstand test.
Where the conditioning lives
| Option | What it is | Good for | Trade-offs |
|---|---|---|---|
| Built into the DAQ module | Sensor-specific modules for thermocouples, bridges, IEPE and so on | Most test stands and lab systems | Module cost per channel; you choose the module per sensor type |
| External signal conditioners | DIN-rail or panel units that output a standard signal (0–10 V or 4–20 mA) | Industrial panels, PLC inputs, long cable runs | One unit per channel; adds accuracy and drift terms |
| Conditioning at the sensor | Amplified transducers, IEPE and smart sensors, sometimes with TEDS electronic data sheets | Simple wiring and noisy environments | Less flexible; the sensor fixes the range and bandwidth |
| Custom conditioning electronics | Purpose-designed boards matched to an unusual sensor or channel count | Specialized research and production test | Design and validation effort |
Note that “signal conditioner” often refers to the second row, a standalone DIN-rail module, while in a DAQ system the conditioning is usually built into the input modules. Both do the same job.
Worked example: conditioning a load cell
A 1,000 lb load cell rated at 2 mV/V, excited at 10 V, produces 20 mV at full load, or 20 µV per pound. On a plain ±10 V, 16-bit input, one step is about 305 µV, so the reading would move in steps of roughly 15 lb. A bridge input module with a ±25 mV range, built-in excitation and a 24-bit ADC resolves well under 0.1 lb, and remote sense wiring corrects the excitation lost in a long cable. The load cell is the same in both cases; only the conditioning changed.
How to specify signal conditioning
For each channel, write down:
- Sensor type, model and output range, including any rated output in mV/V.
- Excitation it needs: voltage, current or AC, and the cable length it must travel.
- Signal bandwidth, which sets the filter and the sample rate.
- Required accuracy and resolution in engineering units.
- Voltages present: common-mode voltage, nearby mains or motor drives, and whether isolation is required for safety.
- Environment: noise sources, temperature range and connector or cable constraints.
- Calibration: shunt calibration for bridges, TEDS, and how often channels must be verified.
Our article on choosing the right DAQ system turns this into a full requirements list for the rest of the system.
Common signal conditioning mistakes
- Connecting millivolt sensors to a ±10 V input and blaming the sensor for poor resolution.
- Relying on digital filtering to remove aliasing that only an analog filter can prevent.
- Using single-ended inputs for low-level signals in an electrically noisy plant.
- Grounding a cable shield at both ends and creating a ground loop.
- Ignoring excitation loss in long bridge cables instead of using remote sense.
- Forgetting that AC-coupled IEPE inputs cannot measure static or very slow signals.
- Treating isolation as optional when sensors touch hazardous voltages.
Systems Sciotex has built
Most of our data acquisition projects combine several sensor types, each with its own conditioning, in one synchronized system. Examples include a glove box tester that monitors component forces during automated cycling, a treadmill monitor that records physiological data from horses alongside treadmill speed and incline, and a stimulus and response system that applies voltage waveforms to biological samples while logging the tissue response.
Mixing sensor types in one system?
Sciotex designs custom data acquisition systems with the right conditioning for every channel, from thermocouples and load cells to IEPE and high-voltage inputs, built and supported in the US since 1997.
Frequently asked questions
Signal conditioning is the electronics between a sensor and an analog-to-digital converter that prepare the sensor’s output for accurate measurement, by amplifying, filtering, exciting, isolating or converting it.
Signal conditioning is the function. A signal conditioner is a device that performs it, often a standalone DIN-rail module that outputs a standard 0–10 V or 4–20 mA signal. In DAQ systems, conditioning is usually built into the input modules.
A strain gauge is a resistor that changes very slightly with strain. It needs an excitation voltage, usually bridge completion resistors, and high gain to turn a change of a few millivolts into a measurable signal.
Yes. Analog filtering, especially the anti-aliasing filter before the ADC, is one of the core conditioning functions, because aliased signals cannot be removed after sampling.
Use isolation when sensors and the DAQ are grounded at different points, when there is high common-mode voltage, or whenever sensors connect to hazardous voltages. Differential inputs are enough for many bench measurements.
Related articles
- What Is Data Acquisition (DAQ)? The Complete Guide
- High-Speed Data Acquisition: Sample Rates, Throughput and System Architecture
- Data Acquisition Software: Ready-to-Run, LabVIEW, Python or Custom?
- How to Choose a Data Acquisition System: A DAQ Requirements Checklist
- Sampling Rate, Nyquist and Anti-Aliasing Filters Explained
- Thermocouple Data Acquisition: Cold-Junction Compensation, Accuracy and Wiring
- Ground Loops and Galvanic Isolation: Causes, Diagnosis and Fixes
