Choosing the Right
Low-Pressure Sensor

Micromachined sensor chips have progressed to a point where pressure sensing elements are available in high volume. This one is a standard 1.5 mm by 1.5 mm die used in consumer applications. The sensors shown here are packaged in temperature-compensated, calibrated, and amplified hybrid designs and can be used as a turnkey solution in many low-pressure applications.

New low-pressure silicon sensors, offering response on the order of tenths of pounds per square inch, have made new products and applications possible. At the very lowest ranges, silicon sensors are now capable of measuring minute pressure changes in liquid levels or the pressure exerted during exhalation. These devices are on the leading edge of sensing technology and could replace mechanical devices in many applications. But how do you go about selecting the best one for your application? A good way to decide is to examine the difference between silicon and conventional components, the markets in which they excel, and the technical specs of your silicon options.

What, exactly, constitutes low pressure? Typical pressure ranges available in silicon sensors range from 0–0.15 psi F.S., and often up to 0–10,000 psi. Manufacturers' definitions of low pressure differ, based on their sensors' designs and their production process. A sensor that responds to pressures <5 psi typically requires different die topology and process techniques. For purposes of this discussion, therefore, anything < 5 psi will be considered low pressure.

The Silicon Difference
Micromachined silicon elements intended for low-pressure sensing are manufactured by a process akin to that used to make standard pressure range components, but with certain key differences. Standard-range elements incorporate a resistor (Wheatstone) bridge implanted in silicon that in turn is etched into a thin diaphragm. When stimulated by a voltage source pressure deflecting the diaphragm changes the resistor value and results in a change in output voltage. Low-pressure silicon sensors work similarly, but have distinctly different features, including bossed diaphragm structures and ~ 50% larger diaphragm areas for stress concentration. Thin, precisely etched diaphragm thickness and strategically placed resistor implants also greatly extend the sensor capability. In addition, conventional semiconductor batch processes make high-volume production at low cost a reality for silicon sensors.

Low-pressure silicon sensors are currently used in three key markets:

  HVAC. Low-pressure sensors play an integral part in heating, ventilation, and air conditioning systems. Among other tasks, they monitor venting and airflow, determine airflow volumes, detect loading caused by dirty filters, and control overall system pressure. These demanding applications require the ability to detect differential pressure changes on the order of 0.015 psi.

  Medical. Many medical applications would not be possible without reduced physical sensor size and increased ability to detect environmental changes and conditions. Designers have met these requirements with relatively miniature (1–2 mm) sensors that measure internal fluid pressures in the human body. These devices are typically catheter-mountable elements that can be inserted into such areas as the cranium, heart, or uterus for real-time monitoring during delicate surgical procedures, where disposability and low cost are musts. The lower pressures used in respirometers require sensor ranges under 0.5 psi. These tasks were previously accomplished by mechanical pressure switches whose characteristics tended to change over time.

  Automotive. Low cost and high reliability are essential to components that sense peripheral automobile conditions such as emissions, fuel vapor, and exhaust. Newer vehicles rely on silicon to monitor tire, manifold, and hydraulic brake pressures. The useful application range of silicon semiconductor sensors has been extended through creative packaging techniques that protect the element in harsh automotive environments.

Application Requirements and Sensor Types
Because each application has its own peculiarities, you must factor all aspects of the overall system into your selection process. Be sure to consider the pressure source input, the desired output, and all pertinent operating conditions. Mounting location and orientation, proximity to the pressure source, stress on leads, port and pressure linkage, and shock and vibration dictate package style and influence overall accuracy. In medical applications, for instance, the pressure-sensing die element must be mounted with no external packaging at all. Temperature fluctuations influencing span and offset may require more sophisticated sensors that are compensated and calibrated to negate the temperature coefficients of the materials. Environmental conditions such as moisture and contaminants govern the level of media protection required. The resulting tradeoffs in accuracy can be overcome by using microprocessor-based control to characterize and cancel undesired effects. System components (e.g., power sources, amplifiers, A/D converters, control circuitry) compatible with the sensor's analog output signal must be able to provide the desired resolution and accuracy of the overall design.

Each type of pressure sensor has material properties that change under applied pressure to produce quantifiable outputs. Keep your application requirements in mind as you consult the following list of commonly used detectors.

  Silicon Micromachined. These consist of a micromachined silicon diaphragm with implanted resistors that piezoresistively change value under the influence of pressure. Applications include medical ventilators and intrabody pressure measurements, and automotive air, vacuum, and vapor pressures.

  Electronic. These include strain gauges and variable capacitance sensors that are not made of micromachined silicon. The strain gauges use the coupling effect of a deflected thick/thin film, foil, or bonded foil diaphragm. The variable capacitance components interpret a capacitive change and convert it into a signal. Strain gauges are used in tactile and mechanical pressure applications.

  Vacuum. Vacuum sensors measure gas pressure indirectly by using resistive changes in a heated wire as one leg of a balanced bridge. They are used primarily to sense vacuum levels attained in pumped vacuum chamber environments such as scanning electron microscopes and processing equipment.

  Piezoelectric. These sensors are made of polymers, crystals, ceramics, and films that generate electric polarization when mechanical force is applied. They are best suited for applications with a dynamic pressure source rather than for continuous static events such as barometric pressure. They work well in rugged environments.

  Variable Reluctance. Variable reluctance sensors incorporate an inductive half-bridge configuration in the form of a deflecting sensing diaphragm between two coils. The resulting change in inductance/impedance creates a varying AC signal. Typical applications include low-pressure HVAC.

Understanding the Specs
Once you define your design requirements, prioritize them to determine the most important selection criteria and judge which technology is most suitable for your job. How do you decide which criteria are more important than others? System specification requirements generally have certain aspects that cannot be violated. A prime example is the pressure operating range, which usually commands the highest priority. With low-pressure micromachined devices now available in volume, you can consider a wide response range in situations where it was previously too expensive. Along with pressure range and sensitivity, common criteria to rank are physical size, cost, accuracy, temperature performance, reliability, long-term stability, and media compatibility.

Accuracy and temperature performance often prove the most complex factors to manage. Each manufacturer has a slightly different way of describing accuracy specifications, which must be converted into comparable units to make a fair comparison possible. (Don't get caught by misleading "specsmanship"!) To help you make sense of the performance specs, the following lists explain the most commonly used definitions of accuracy, as well as some notable variations for each. The first group assumes a reference temperature of 25°C.

  Zero/Offset. The value of the output voltage at databook excitation conditions at 25°C with zero pressure applied. It is typically expressed as 0 ±mV. Due to ratiometricity, a higher supply voltage or current will result in a higher offset.

  Pressure Hysteresis of Zero. A measure of the repeatability of the zero when the sensor is subjected to one or more full-scale pressure cycles. Units are expressed as percent of full-scale output. Different sensor manufacturers may use different numbers of pressure cycles and various full-scale ranges when calculating the pressure hysteresis of zero.

  Span Pressure Hysteresis. A measure of the repeatability of the output span when the sensor is subjected to one or more pressure cycles. The value usually expresses the worst-case variation as a percent of full-scale output.

  Span Temperature Hysteresis. A measure of the repeatability of the span when the sensor is subjected to temperature cycling. In other words, the sensor is subjected to minimum and maximum operating temperatures. The difference in the span reading after cycling is a measure of temperature hysteresis of span. This parameter is specified in the percent of full-scale output. Thus, the full-scale pressure reading normalizes the worst-case variation in span over the full temperature cycle.

  Sensitivity. The ratio of the output signal change vs. the change in pressure. Units vary widely depending on supplier, but the value is usually expressed in mV/V or I/psi. Sensitivity is a key performance criterion defining the system's resolution.

  Long-Term Drift. A measure of the change in span output or zero behaviors over time. It is typically expressed in mV.

  Span Nonlinearity. Micromachined silicon sensors tend to have proportionally less and less output gain as pressure is increased. A pressure transfer

Figure 1. For calculating linearity, the best fit straight line error provides a measure of average error where the straight line is shifted for equal errors above and below the measured pressure. Terminal base linearity is the end point linearity from zero to full-scale pressure measured at the mid span.

line therefore shows less output for high pressure than a true linear extrapolation would indicate. There are two basic methods for calculating linearity. A best fit straight line (BFSL) error provides a measure of average error where the straight line is shifted for equal errors above and below the measured pressure (see Figure 1). A terminal base linearity is defined as the end point linearity from zero to full-scale pressure measured at the mid span. Terminal base nonlinearity is typically 2 x BFSL nonlinearity.

A second class of specifications describes temperature-related accuracy. All sensors have some shift in zero in addition to nonrepeatability when subjected to temperature variations. The bridge resistance typically changes by about +3000 ppm/°C (mV/V/°C for others) above room temperature. However, below room temperature (usually below –20°C), the resistance may actually show a zero slope and then invert. Under these conditions, the assumption that the resistor changes linearly with temperature is invalid and a second-order correction term must be incorporated into the sensor model. The result is that all sensors have both a linear and a second-order error. The best that can be done is to minimize the error, dependent on the acceptance criteria used by a specific manufacturer. Each manufacturer may trim parts slightly differently to achieve the optimum conformance to its criteria.

  Temperature Coefficient of Zero (Three Methods). This is not well defined, and the exact specification varies by manufacturer. Some use a simple line from the reference temperature (25°C) to the two extremes and specify that the error must be less than that butterfly pattern over the range of interest. A second definition of the temperature coefficient of zero is that the temperature error will be less than a given percentage of full scale anywhere within the temperature range. This is one of the least critical measures of performance, but it is currently the most widely used. The third approach is to compute the BFSL determined by the three data points.

Figure 2. Three techniques can be used to determine the temperature coefficient of zero. A simple line from the reference temperature to the two extremes provides the most information. Full-scale error provides the least information on device performance but is nonetheless the most wiedly used.

Of the three techniques, the first provides the maximum information, while the third (BFSL error) provides the user with limited accuracy. The second approach provides the least information on device performance. Figure 2 graphically compares the three approaches.

  Temperature Hysteresis of Zero. This reading is a measure of the repeatability of the zero when the sensor is subjected to one or more temperature cycles. The difference in the zero reading after cycling is a measure of temperature hysteresis and is usually specified as a percent of full-scale output. Thus, the full-scale pressure reading normalizes the worst-case variation in offset over the full temperature cycle.

  Temperature Compensation. This method of canceling the effects of various temperature coefficients uses complex algorithms to determine the trim value of thick film resistors. Systems operating under various temperature extremes will require some form of compensation either through system electronics or preset by the sensor manufacturer.

The combination of all error components can then be used to describe the sensor error's contribution to system performance. The following two definitions are most commonly used to describe overall sensor accuracy:

Worst-Case Error. A sum of all of the relevant individual errors, where:

   error (worst case) = E1 + E2 + E3 . . . En

Most Probable Error. Defined as the square root of the individual errors summed and squared, where:

   error (most probable) = (E1² + E2² + E3² + . . . En²)

RMS is independent of the direction of the error.

Chart Your Choices
Now that you know the choices available, you've made a set of a selection criteria, and you have a list of design constraints, you are ready to use them to make a decision. A simple spreadsheet can help as a tool for comparison. Often, many of the technologies fall out right away, and your comparison comes down to just a couple of manufacturers. As an example, suppose you are building a medical respiration device that must activate other electronics at a pressure switch point. Electronic switches can take care of this task, but the key design parameters might be pressure range, size, cost, and long-term stability (see Table 1).

TABLE 1

Evaluating the Choices
Critical Design Constraints	Example	Rank Switch	Rank Silicon Sensors
Pressure Range	0-0.2 psig	1 - Actuates at 0.2 psi	1 - Sensor range is 0-0.3 psi
Sensitivity	None required	2 - Will switch in desired range	1 - 5mV/V/psi; more controlable
Physical dimensions	Can be 0.3 in.2	4 - Exceeds target by 10%	1 - Meets spec
Target cost	<$2.00	5 - Exceeds target	1 - Meets target; additional switch circuitry required
Accuracy	< 1% Overall accuracy	3 - Actuation point not as consistent and controllable	2 - Better; more consistent accuracy
Temperature Performance	Must operate from 0°C-45°C	1 - Meets requirements	2 - Meets requirement; compensation required
Long-term stability (reliability)	< 0.1% F.S. drift over 10 yr.	4 - Requires mechanical recalibration periodically	2 - <0.2% over 5 yr.
Media compatibility	Air	5 - Works with air	5 - Works with air
Performance	Must switch when pressure reaches 0.015 psi ± 10%	3 - Mechanical wear; may require recalibration	3 - Element does not wear out, but needs additional circuitry to do the switch function
Totals		31	18

As the general comparison in Table 1 suggests, the switch and low-pressure sensor can perform comparably, but each has its own advantages and disadvantages. What is important here is that this selection process was not possible before the introduction of low-pressure micromachined silicon sensors. As you evaluate new designs, use both mechanical as well as silicon considerations. The advent of more environmentally harsh conditions, coupled with increasingly short product launch windows, gives designers the potential for new and untapped markets.

Rick Hagen is an Applications Engineer, EXAR Corp., 48720 Kato Rd., Fremont, CA 94538; 510-668-7367, fax 510-668-7025, rickh@exar.com

Home | Contact Us | Advertise © 2009 Questex Media Group, Inc.. All rights reserved. Reproduction in whole or in part is prohibited. Please send any technical comments or questions to our webmaster.