Smart temperature sensors combine a sensing element with electronics that condition or encode its output for use by a controller or other system. In industrial automation, selecting among smart temperature sensors is not simply a matter of choosing the highest stated accuracy. The measurement objective, operating environment, temperature range, required response, and long-term stability all influence which sensor architecture is appropriate.
A useful starting point is to distinguish absolute temperature measurement from monitoring temperature changes or differences. Those tasks can place very different demands on the sensor and its signal-processing chain.
Match the sensor to the measurement task
Calibration and trimming contribute to the cost of highly accurate temperature sensors, and accuracy may need to be checked over time. Thermal cycling and mechanical loading can affect whether a system continues to meet its accuracy requirements. In many applications, however, repeatable measurements over a defined period matter more than closeness to a standard reference.
- Tracking change: When monitoring a process over time, resolution and stability may be more important than absolute accuracy. A system may need to detect small temperature shifts even if the exact temperature is not critical.
- Measuring a temperature difference: Thermal and infrared sensing systems can use thermopiles to measure a difference between junctions. An additional temperature measurement may be needed to account for temperature-dependent effects in the sensing system.
- Measuring absolute temperature: The sensor must be suited to the required operating range, and the system must account for the sensing element’s characteristics and interface requirements.
These distinctions help engineers avoid paying for a level of absolute accuracy that an application does not need, while ensuring that resolution, repeatability, and drift are not overlooked.
Compare sensing elements and operating ranges
Common options include bipolar junction transistors (BJTs), thermocouples, platinum resistance elements, and thermistors. Their strengths differ. BJTs are attractive for integration with electronics and offer high sensitivity over a limited temperature range. Thermocouples cover very broad ranges, but measure the temperature difference between a measurement junction and a reference junction; the reference temperature must also be measured to determine an absolute temperature. Platinum elements are used across a broad range, while thermistors provide high sensitivity over a more limited range but have strong nonlinearity.
Integrated electronics also impose a temperature limit: a smart sensor’s usable range cannot exceed what its electronic circuitry can tolerate. Where temperatures exceed that limit, a discrete sensing element paired with a suitable interface may be more appropriate. This separation can also suit applications with demanding stability or packaging requirements.
For integrated sensors, BJTs are useful because they can be fabricated alongside analog and digital circuits in CMOS technology. Their base-emitter voltage changes with temperature, while the difference between the base-emitter voltages of two transistors biased at different current densities is proportional to absolute temperature (PTAT). Combining these signals allows circuitry to shape the temperature response and compensate for some nonlinearity.
How CMOS techniques help manage measurement error
CMOS is widely used for integrated circuits, but component mismatch can create errors in current ratios and amplifier offset. CMOS circuits can also exhibit low-frequency, or 1/f, noise. Two techniques described for precision sensor designs address these limitations:
- Dynamic element matching (DEM) changes the positions or roles of matched components over a sequence of measurement states. Averaging across the sequence reduces systematic errors associated with component mismatch.
- Chopping modulates the sensor signal so that amplifier offset and low-frequency noise can be separated from the signal and reduced through filtering.
In the described BJT-based CMOS design, vertical PNP transistors are used as sensing elements, and DEM and chopping are applied to limit error sources. The design also uses one-point wafer-level trimming to correct process variation. These are design strategies, not substitutes for checking that a complete sensor system meets its application requirements.
Reading a duty-cycle-modulated temperature output
A duty-cycle-modulated (DCM) sensor encodes temperature in the proportion of each output period for which the signal is high. The output is a square wave, so a microcontroller can measure the high time and total period with a timer, then calculate the duty cycle. This approach moves some conversion and processing into the controller rather than requiring a full analog-to-digital converter in the sensor.
For sensors that use DEM and chopping across a repeating sequence of eight periods, processing should account for the complete sequence. The duty cycle can be calculated for each period and then averaged, or the high and total times can be summed before division. The first method provides the better accuracy described for this design; summing first is simpler but can leave a larger residual error because periods receive different weights. For slower measurements, averaging corresponding periods across multiple complete sequences can reduce the number of divisions while retaining nearly the same accuracy as period-by-period averaging.
A DCM output can also be used in a simple analog circuit. Filtering the square wave produces an average voltage that varies with duty cycle and therefore with temperature. Comparing that voltage with an adjustable set point can provide thermostat control without a microcontroller. The choice of readout method should account for how the averaging treats the repeating measurement sequence.
Translate specifications into system requirements
The source describes a BJT-based CMOS sensor with a DCM output and a stated operating range of -45°C to 130°C. Its reported accuracy is better than 0.1°C from -20°C to 60°C and 0.3°C across the full range. Reported resolution is about 0.17 mK with a one-second measurement time and about 3 mK at 1.8 ms. These figures illustrate that accuracy and resolution describe different aspects of performance: a sensor can resolve small changes without having equally small absolute error.
Before choosing a device or architecture, define the required range, absolute accuracy, resolution, response time, supply conditions, and acceptable long-term drift. Also consider packaging and mechanical stress, which can affect BJT-based sensor readings. For a DCM sensor, include the controller timer and the required averaging window in the design; the sensor output alone does not determine the quality of the system measurement.
Conclusion
Effective industrial temperature measurement begins with the measurement objective, not a single headline specification. Select the sensing element for the required range and environment, distinguish resolution from accuracy, and design the interface and processing around the sensor’s output. For smart temperature sensors using DCM, correct averaging is part of the measurement architecture. Treating the sensor, package, electronics, and controller as one system helps engineers make practical choices about precision, response, stability, and integration.
Access Official Technical PDF