Ask the expert: Interview with Sandi Kocjan
Digital signal conditioning for piezoresistive MEMS pressure sensors using Sensor Signal Processor (SSP)
Digital signal conditioning is a key enabler for turning the raw output of piezoresistive MEMS pressure sensors into accurate, stable, and intelligent sensor signals. Modern SSPs combine amplification, conversion, compensation, and digital interfacing in a compact solution, helping improve performance while simplifying calibration and system integration.
This questionnaire explores the key benefits, design considerations, and future potential of digital signal conditioning for next-generation pressure sensors.
All Sensors: From a sensor-performance and manufacturing perspective, what are the main advantages and trade-offs of digital signal conditioning with an SSP compared with conventional compensation using laser-trimmed passive resistor networks?
Mr. Kocjan: The main advantages of using an SSP are easier and more flexible calibration, compensation of sensor offset, sensitivity, temperature effects and non-linearity, and reduced dependence on laser-trimmed passive components. This can simplify manufacturing and improve calibration consistency. In contrast, laser-trimmed sensors require more sophisticated and expensive production equipment, precise trimming processes and additional verification steps, increasing manufacturing complexity and cost.
The main trade-offs with an SSP are the need for Analog/Digital conversion and digital processing, which can result in longer response time and higher current consumption compared with a simple analogue compensation solution. ADC (Analog-to-Digital Converter) resolution, noise and processing latency can also limit performance. Therefore, the choice is a balance between better calibration flexibility and simpler production versus response time, power consumption and digital processing limitations.
Mr. Kocjan: The optimal PGA gain and ADC input range should be selected based on a detailed understanding of the basic sensor output over the complete pressure and temperature range. Sensor offset, sensitivity, temperature drift, non-linearity and production tolerances must all be considered. The PGA gain should be set as high as possible to use most of the available ADC input range and maximise resolution, while still leaving sufficient margin to prevent ADC or PGA saturation at the maximum pressure, temperature extremes and sensor tolerances. Therefore, the gain selection is a compromise between maximum ADC utilisation and sufficient operating margin. Good characterisation of the sensor over pressure, temperature and production variation is essential for selecting the optimum setting.
"For low pressure ranges below 10mbar we can achieve total accuracy down to 0.25%FS."
All Sensors: What calibration model (mathematical correction) do you use for offset, sensitivity, nonlinearity, and temperature compensation, and how many pressure/temperature calibration points are necessary?
Mr. Kocjan: Typically, we use a multi-point calibration to compensate for sensor offset, sensitivity, nonlinearity and temperature effects. For our standard compensated temperature range, calibration is performed at three temperature points, typically 0 °C, 35 °C and 70 °C. At the reference temperature of 35 °C, the sensor is calibrated at three pressure points: 0, PN (nominal pressure) and PN/2. At temperature ends (0°C, 70°C) we normally take 2 calibration points per temperature (for offset and span). This results in seven calibration points in total. The calibration coefficients are then used by the SSP to calculate the corrected pressure output over the specified pressure and temperature range. The exact number of calibration points and the mathematical correction model can be adapted depending on the required accuracy, sensor characteristics and compensated temperature range.
All Sensors: Which temperature signal should be used for compensation: SSP temperature, MEMS bridge temperature, or an external temperature sensor?
Mr. Kocjan: If possible, the MEMS temperautre sensing element should be used for compensation, because it provides the most direct indication of the temperature experienced by the pressure-sensitive element. This is especially important when the sensor is in direct contact with the pressure medium, where the MEMS temperature can differ from the temperature measured by the SSP. Using the SSP temperature may introduce additional errors due to thermal gradients and different thermal time constants between the MEMS and SSP. An external temperature sensor can also be used, but its accuracy depends on its physical location and thermal coupling to the MEMS element. Therefore, the MEMS temperature is generally the preferred choice when possible.
All Sensors: What total accuracy or Total Error Band can we realistically achieve over the full pressure and temperature range, including MEMS, SSP, calibration, packaging, and aging?
Mr. Kocjan: The achievable Total Error Band mainly depends on the performance and stability of the MEMS sensing element itself. For higher pressure ranges, typically above 1 bar, a total accuracy in the range of 0.1%FS can realistically be achieved with proper calibration and temperature compensation. With higher accuracy demands we can perform additional »post calibration« procedures, which correct offset and span parameters @25°C as an additional process after the standard calibration procedure. With this we can minimise the error of the calibration system itself. For low pressure ranges below 10mbar we can achieve total accuracy down to 0.25%FS.
The final performance also depends on the SSP accuracy, calibration method, packaging-induced stress, mechanical assembly and long-term aging. These factors should therefore be considered together when specifying the final accuracy. With a well-characterised MEMS element, suitable SSP and multi-point calibration, the SSP itself normally does not have to be the limiting factor. The achievable value should ultimately be confirmed by testing the complete sensor over the specified pressure and temperature range.
All Sensors: What are the main reasons to choose one SSP architecture over another, in terms of accuracy, bandwidth, latency, calibration effort, and production cost?
Mr. Kocjan: The main criteria for choosing an SSP architecture are its performance characteristics, simplicity of use, and proven reliability. Accuracy, available bandwidth, conversion and processing latency, calibration capabilities and the required number of external components should all be considered. Production cost and the effort required for calibration and programming are also important, especially for high-volume applications. In practice, past good experience with a specific SSP and its reliability in similar applications can be an important factor. A well-known and proven device can reduce development time, calibration effort and production risks. Therefore, the best SSP is not necessarily the one with the highest technical specification, but the one that provides the required performance with a simple, reliable and cost-effective implementation.
Mr. Kocjan: The realistic effective resolution (Effective number of bits, ENOB) is typically lower than the nominal ADC resolution due to ADC noise, sensor noise, electrical interference, offset and other non-ideal effects. As a practical guideline, the effective resolution can be approximately 1–2 bits lower than the nominal ADC resolution, although this depends on the selected bandwidth, PGA gain, sampling rate and digital filtering. The actual pressure noise should therefore be evaluated at the required measurement bandwidth using the complete sensor and SSP signal chain, rather than considering the ADC specification alone. In practice, the final pressure noise and effective resolution are determined by the combination of sensor noise, analogue front-end performance, ADC performance and digital filtering.
All Sensors: How do you select the sampling rate and digital filter bandwidth to achieve the required compromise between noise, response time, and latency?
Mr. Kocjan: The sampling rate and digital filter bandwidth are mainly selected based on measurement and test results. The required pressure response time and bandwidth are first defined, and the sampling rate is then selected sufficiently high to capture the required pressure dynamics. The digital filter bandwidth is adjusted to achieve the required compromise between noise reduction and response time. A lower filter bandwidth generally reduces noise but increases response time and latency, while a higher bandwidth provides faster response but allows more noise to pass through. The final settings should therefore be verified by testing the complete sensor and SSP system under realistic operating conditions, including both static noise and dynamic pressure changes.
All Sensors: What dominates the final pressure noise: MEMS Johnson noise, PGA noise, ADC noise, reference/excitation noise or digital quantisation?
Mr. Kocjan: The MEMS sensor itself is typically the main contributor to the final pressure noise, particularly its piezoresistive bridge noise. The remaining noise comes from the PGA, ADC, reference or excitation supply, and digital processing. The contribution of the PGA and ADC becomes more important when the sensor output signal is very low. Reference or excitation noise can also have a significant effect if the bridge supply is not sufficiently stable. Digital quantisation noise is normally a smaller contributor when sufficient ADC resolution is available. Therefore, in a well-designed SSP system, the main limitation is often the noise of the MEMS pressure element, while the SSP and signal-conditioning chain should be designed so that their added noise remains significantly below the sensor noise.
“The best SSP is the one that provides the required performance with a simple, reliable and cost-effective implementation.”
All Sensors: Which diagnostic and self-monitoring functions does the SSP provide, and which sensor or signal-chain failures can it reliably detect during operation?
Mr. Kocjan: Typical SSP diagnostic functions include signal-loss detection, input over-range/under-range detection, supply-voltage monitoring, memory/communication checks, and detection of certain sensor or signal-chain faults. These functions can identify open or abnormal sensor signals and other electrical failures during operation. The exact diagnostic coverage depends on the specific SSP architecture and available diagnostic features.
All Sensors: Looking ahead, what developments or new capabilities would you like to see in future SSP generations to improve sensor performance, simplify calibration, or reduce overall system cost?
Mr. Kocjan: Actually, all of these aspects should be improved in future SSP generations. We would like to see higher accuracy, higher effective resolution, wider bandwidth, lower latency, simpler calibration and lower overall system cost. It would also be beneficial if SSPs could include some basic microcontroller functionality, such as simple logic, signal processing, diagnostics, communication handling or configurable outputs. This could reduce the need for a separate microcontroller and additional external components, simplifying the overall sensor design and reducing system cost. At the same time, the SSP should remain easy to configure and use. Overall, future SSPs should provide better performance, greater functionality, easier calibration and simpler, more cost-effective sensor solutions.
All Sensors: Thank you very much for the interview, Mr. Kocjan.