Sensors of all types have come a long way in the last few years and tend to be more accurate and more stable than their predecessors. These sensors are not always simple to use. Designers of conditioning circuitry for these sensors often find development of this circuitry somewhat daunting. However, with a little basic knowledge and the help of a new on-line sensor design tool, much of the challenge of this process is alleviated.
While many types of sensors are available, the pressure sensor enjoys the most applications. Hence, this article will discuss the basic operation of the Wheatstone Bridge-based pressure sensor, and processing circuitry for interpreting the bridge sensor output, including offset and gain correction. In Part 2 of this article, we'll introduce and discuss an on-line tool that can be used to design the signal conditioning circuitry for these types of sensors.
Many pressure sensors use micro-electromechanical system (MEMS) technology and are comprised of four force-sensitive resistors connected in a Wheatstone bridge configuration.. When there is no pressure impressed upon the sensor, all resistors in the bridge are equal. When pressure is exerted upon the bridge, two resistors opposite each other will increase in value and the other two resistors will decrease in value. The amount of increase and decrease is equal to each other.
Unfortunately, things are not quite so simple because sensors have offset and gain errors. The offset error is the output that is present when no pressure is applied to the sensor. The gain error indicates how sensitive the sensor output is with respect to the external force applied to the sensor. A typical sensor might be specified for an excitation voltage of 5 V and have a nominal full-scale output of 20 mV/V. This means that with an excitation voltage of 5 V, the nominal full-scale output is:
20 mV/V × 5 V = 100 mV.
The offset voltage might be 2 mV, or 2% of full scale, and minimum and maximum full scale output voltages might be 50 mV and 150 mV, or ±50% of nominal full scale.
With two equal-value resistors in series, the voltage at the junction between those resistors is half the voltage across the two of them. If one resistor value increases by 1% and the other resistor value decreases by 1%. the voltage at the junction of the two resistors changes by 1%, However, if two of these resistor strings are placed in parallel and the lower resistor of one side and the upper resistor of the other side decrease by 1% and the other two resistors in the strings increase by 1%, the difference between the two “mid” points will change by 2% from zero difference. This configuration of two parallel branches is called a Wheatstone bridge, Figure 1 .
Figure 1: The Wheatstone bridge is driven with excitation voltage VEX and differential output voltage V.
Without knowing the offset and the exact relationship between sensor output voltage and pressure, we can only have a rough estimation of the pressure exerted upon the sensor. This means that a method of calibration is required to attain better accuracy.
Fortunately, offset and full-scale errors of a given sensor are fairly stable over time, so once a sensor is calibrated, the correction factors might not need to be changed for the life of that sensor, depending upon the accuracy requirements. Still, it is common to recalibrate the system at each power up.
The basic signal conditioning circuit consists of an instrumentation amplifier and an analog to digital converter (ADC). The instrumentation amplifier amplifies the small output voltage from the sensor to a level suitable for the ADC, which converts the amplified sensor output voltage to a digital word that is processed by a controller or DSP (Figure 2 ). The instrumentation amplifier is used to avoid loading of the bridge, which would change the sensor output voltage.
Figure 2: Basic pressure sensor-conditioning circuit.
(Click on image to enlarge)
The full-scale output of the sensor is the maximum input that will be seen at the amplifier input. When the sensor output is at full scale, the input to the ADC should be near its full-scale point, which is usually the ADC reference voltage, VREF . The required gain of the amplifier is (Equation 1 ):
where VREF is the ADC reference voltage and “Sensor FS” is the full-scale output of the sensor. The gain of the instrumentation amplifier, assuming perfect resistor matching, is (Equation 2 ):
As stated previously, there are two challenges with respect to the sensor that need to be addressed. The first challenge is that the sensor has an output offset. This can be adjusted with an appropriate voltage at VOFF of Figure 2. Alternatively, this offset may be negated in software after the sensor output is digitized. If the offset is handled in software, the value of VOFF becomes zero volts.
The problem with subtracting the offset in software is the placing of a limitation on the measurable sensor range. If the offset is positive, there is a limit on the maximum sensor output that can be measured because the amplified sensor output may reach the ADC full scale earlier than expected. If the offset is negative, very small sensor output levels may not be accurately measured since the ADC output code would not rise above a value of zero until the amplified offset value was overcome.
The range of output-voltage values possible for the sensor full-scale output is the second problem. For example, a sensor with a nominal full-scale output voltage of 100 mV might have a specification that indicates the possibility for this full scale output to be as low as 50 mV and as high as 150 mV.
If the full-scale sensor output is less than nominal, the full-scale range of the ADC is not used. If the full-scale sensor output is greater than nominal, the ADC output will reach the ADC full-scale output value before the sensor output reaches its full scale. Furthermore, if there is drift in the sensor output or in the amplifier itself, there could be some uncertainty and inaccuracy in the readings.
Fortunately, there is little if any time drift in today's sensors, and careful choice of amplifiers can minimize amplifier drift. As a result, the circuit gain can be adjusted once during manufacturing and/or at system power up.
One way of doing this is to use a digital to analog converter (DAC) to adjust the VREF , the ADC reference voltage, to compensate for the sensor full-scale error, and another DAC to adjust the VOFF of Figure 2 to compensate for offset errors. A dual DAC, such as the DACxx2S085 from National Semiconductor, where the “xx” can be 08, 10 or 12 for the DAC resolution, would be a good choice for this application. Alternatively, these errors may be calibrated out in software after the sensor output is digitized.
The best solution to solve these two problems is adjusting the offset and gain errors during manufacturing and during software calibration whenever the system is powered up. This method allows for minimal error correction in software and preserves the largest amount of available dynamic range from the ADC.
The third problem is that a single-ended input ADC generally requires its input to be driven very close to zero volts to produce an output code of zero. The problem arises because the amplifier used to drive the ADC input can not produce an output that is lower than 50mV or so. This is common even if the amplifier is referred to as having rail-to-rail output capability.
While the inability of the circuit to provide a minimum ADC output code of zero is not a problem for some applications, it is for others. For those applications with a requirement for a minimum code of zero, the alternatives include:
- A negative supply for the amplifier driving a single-ended input ADC.
- Use a single-ended ADC with both positive and negative reference voltage that can be set to a value higher than device ground and offset the ADC input voltage accordingly.
- Bias the ground of the ADC to about 100 mV.
- Offset the ADC input and lose some codes at the ADC output, adjusting with software.
- Use a differential input ADC.
The disadvantage of using a negative supply for the amplifier driving the ADC is that a negative voltage source may not be available in the system, and providing one for this single amplifier is not seen as a viable alternative. However, a simple solution is available in the LM2787, a switched-capacitor voltage inverter from National Semiconductor Corp.
All ADCs have a positive and a negative reference. The difference between these two reference voltages is what is defined as “the reference voltage” of the ADC. The negative reference and positive reference voltages define the input minimum and maximum voltages, respectively. Unfortunately, many ADCs available today have their negative reference voltage internally defined as device ground. This has usually been done as a compromise to fit the ADC in a smaller package with less external pins.
Raising the ground of the ADC is generally not seen as a trivial task. Furthermore, biasing it too high would present problems of output interfacing because the logic low of the device will be a little higher than the ground bias amount. Nevertheless, doing this has the same effect as defining the ADC negative reference voltage to be a low value, perhaps 70 mV to 100 mV.
The addition of an offset to the ADC, with appropriate adjustment for the ADC full-scale input value is a viable approach, but decreases the dynamic range used by the ADC. This amounts to providing a positive VOFF in Figure 2, reducing the gain of the amplifier so that the ADC input does not exceed the ADC reference voltage, and doing a software adjustment on the ADC output code.
The use of a differential input ADC is the best alternative to getting an ADC output code of zero and maintaining circuit linearity over the entire range of input voltages at the ADC input, without the use of a negative voltage in the system. In this approach, the output of the differential amplifier is fed to the differential inputs of the ADC without a differential to single-ended amplifier stage. This is a simple, yet rather elegant, and very effective approach to the problem.
In Part 2 , we will introduce and discuss a new on-line tool that can be used to design the signal conditioning circuitry for these types of sensors.
About the Author
Nicholas “Nick” Gray is a Staff Applications Engineer with National Semiconductor's Data Conversion Products Group. Nick's career includes video circuit design and over 30 years as an applications engineer, mostly for data converter products. He received his BSEE degree from Gonzaga University, Spokane, Washington in 1965 and has done graduate work at California State University, San Jose, California.