Sensor Signal Conditioner
Pick a sensor, get its signal and where it goes wrong. Thermistors, RTDs, thermocouples, bridges and strain gauges each have a tab, and the last tab follows the signal through the amplifier, filter and ADC with a noise and error budget.
1 · NTC thermistor
Steinhart-Hart coefficients (optional)
Fill all three to override Beta. Blank uses the Beta model.
Fit Steinhart-Hart from three datasheet points
—
2 · Voltage divider
3 · Linearization and readback
Result
Resistance vs temperature
Divider output vs temperature
Linearization
| Method | Series resistor | Peak nonlinearity |
|---|
Peak nonlinearity is the largest gap between the divider output and a straight line through the two range ends, expressed in °C.
R-T table
| T (°C) | R (Ω) | Vout (V) |
|---|
1 · RTD
2 · Callendar-Van Dusen (IEC 60751)
R(T) = R0 (1 + A T + B T²) T ≥ 0 °C R(T) = R0 (1 + A T + B T² + C (T − 100) T³) T < 0 °C A = 3.9083e-3 B = −5.775e-7 C = −4.183e-12 α = 0.00385 Ω/Ω/°C
3 · Readback
Result
Resistance vs temperature
Measurement error
Lead-wire error and self-heating, in °C of reading error.
R-T table
| T (°C) | R (Ω) | V (mV) | Wire error (°C) | Self-heat (°C) |
|---|
IEC 60751 tolerance classes
Tolerance of the element itself, ± °C: AA 0.1 + 0.0017|t|, A 0.15 + 0.002|t|, B 0.3 + 0.005|t|, C 0.6 + 0.01|t|.
1 · Thermocouple type
2 · Temperature → voltage
3 · Voltage → temperature
Result
Voltage vs temperature
NIST ITS-90 reference function against a 0 °C cold junction.
Sensitivity (Seebeck coefficient)
1 · Bridge
+Vex
┌──┴──┐
R1 R3
├─ A ─┼─ B ─ → Vout = V(B) − V(A)
R2 R4 (sensor)
└──┬──┘
GND
Result
Bridge output vs R4 change
Sensitivity
Signal chain
Sensor
Bridge / divider
Amplifier
Filter
ADC
Result
Signal level through the chain
Half the sensor span, in mV, after each stage.
Noise budget
µV rms at the ADC input. Log scale.
Noise by stage
| Stage | Signal (mV) | Noise (µV rms) | SNR (dB) | Share of noise power |
|---|
Accuracy budget
Static errors referred to the amplifier input.
| Term | µV at input | % of span |
|---|
From a physical quantity to ADC counts
Every sensor here turns temperature or strain into a small change in resistance or voltage. The tool computes that change from the standard reference function for the sensor, then shows what limits it: linearity, lead wires, self-heating, cold-junction error, noise, and converter resolution.
Thermistors
The Beta model is a two-point fit and drifts away from the real curve outside the 25/85 °C pair. Steinhart-Hart with three coefficients follows it to a few millidegrees over a 100 °C span. The linearization table compares three series resistors: the geometric mean of the end resistances, the inflection formula Rs = R(Tc)(B − 2Tc)/(B + 2Tc) that puts the divider's inflection point at the centre, and the resistor that minimizes the peak error found by a direct search.
RTDs
The wire error uses the slope of the curve at each temperature, which falls from 0.39 Ω/°C at 0 °C to about 0.32 Ω/°C at 850 °C for a Pt100. The dissipation constant δ is the power that raises the element 1 °C above its surroundings and depends on the package and the medium it sits in. Moving air gives a much smaller value than a stirred liquid.
Thermocouples
A thermocouple measures a temperature difference. The voltage you read needs the cold-junction temperature added back, as a voltage and not as degrees. The tool inverts the NIST polynomial numerically, so it stays exact over the whole range instead of using the separate approximate inverse polynomials. The last line shows why a 1 °C error at the cold junction becomes nearly a 1 °C error at the hot junction when both sit at similar sensitivities.
Bridges and strain gauges
A quarter bridge is slightly nonlinear: at GF = 2 and 1000 µε the exact output is 0.1 % lower than the usual GF·ε/4 formula. Half and full bridges with gauges on opposing arms are linear.
Signal chain
Noise is referred to the ADC input so the terms add directly. A filter limits how much amplifier noise reaches the converter and also stops out-of-band signals aliasing. The tool warns when the filter cutoff is above half the sample rate, when the swing clips the ADC, and when it uses so little of the range that more gain would help.
Assumptions and limits
- Resistance and voltage models only. Thermal lag, radiation, stem conduction and sensor drift are not included.
- Thermocouple results hold inside each type's ITS-90 range. Type B is reported from 25 °C up, because its EMF is not monotonic near room temperature. Tolerances of real wire (for example ±1.5 °C or 0.4 % for type K, class 1 and 2) come on top.
- The thermocouple tab does not model the cold-junction sensor itself. Enter the temperature it reports.
- The chain treats half the sensor span as an RMS signal and a band-pass filter's centre frequency as its noise bandwidth. Flicker (1/f) noise and amplifier current noise are not included.
- Engineering estimate. Check the part datasheet before committing a design.
References
- NIST, ITS-90 Thermocouple Database, Standard Reference Database 60 (srdata.nist.gov/its90): reference functions and tables for types B, E, J, K, N, R, S, T. Coefficients were taken from the public-domain table in the Python package
thermocouples_reference0.20 and checked against NIST table values. - IEC 60751:2008, Industrial platinum resistance thermometers and platinum temperature sensors: Callendar-Van Dusen coefficients and tolerance classes.
- J. S. Steinhart, S. R. Hart, "Calibration curves for thermistors," Deep-Sea Research 15, 1968, 497–503.
- National Instruments, Measuring Strain with Strain Gages, and Hottinger Brüel & Kjær, An Introduction to Measurements using Strain Gages: bridge equations and quarter-bridge nonlinearity.
- Analog Devices, Linear Circuit Design Handbook (ed. H. Zumbahlen), chapters on sensors, noise and converters; Texas Instruments application notes on RTD, thermocouple and bridge front ends.
- H. Johnson, Thermal agitation of electricity in conductors, Phys. Rev. 32, 1928; H. Nyquist, Phys. Rev. 32, 1928.
Sensor reading off by a degree?
Kyros builds measurement front ends: sensor selection, excitation, cold-junction and lead compensation, low-noise amplification and the firmware that calibrates it in production.
