Many applications involve environmental or structural measurements, such as temperature
and vibration, from sensors. These sensors, in turn, require signal conditioning before a data
acquisition device can effectively and accurately measure the signal. Signal conditioning is one
of the most important components of a data acquisition system because without optimizing
real-world signals for the digitizer in use, you cannot rely on the accuracy of the measurement.
Signal conditioning needs vary widely in functionality depending on your sensor, so no instrument
can provide all types of conditioning for all sensors. For example, thermocouples produce
very low-voltage signals, which require linearization, amplification, and filtering, while strain
gages and accelerometers need excitation. Other signals may need none of these but strongly
rely on isolation from high voltages. The key to a successful signal conditioning system
is to understand the circuitry you need to ensure an accurate measurement whatever your
channel mix.
This document covers the specific conditioning requirements you need for the most common
sensor types and discusses key considerations for developing and maintaining a conditioned
Most signals require some form of preparation before they can be digitized. Thermocouple
Filtering
Filters reject unwanted noise within a certain frequency range. Often, lowpass filters are used
to block out noise in electrical measurements, such as 50/60 Hz power. Another common
use for filtering is to prevent aliasing from high-frequency signals. This can be done by using
an anti-aliasing filter to attenuate signals above the Nyquist frequency. Anti-alias filters are a
form of lowpass filter characterized by a flat passband and fast roll-off. Because accelerometer
and microphone measurements are commonly analyzed in the frequency domain, anti-aliasing
filters are ideal for sound and vibration applications.
Isolation
Voltage signals well outside the range of the digitizer can damage the measurement system
and harm the operator. For that reason, isolation is usually required in conjunction with
attenuation to protect the system and the user from dangerous voltages or voltage spikes.
Isolation may also be needed when the sensor is on a different ground plane from the
measurement sensor, such as a thermocouple mounted on an engine.
Excitation
Excitation is required for many types of transducers. For example, strain gages, accelerometers,
thermistors, and RTDs require external voltage or current excitation. RTD and thermistor
measurements are made with a current source that converts the variation in resistance to
a measureable voltage. Accelerometers often have an integrated amplifier, which requires current
excitation provided by the measurement device. Strain gages, which are very-low-resistance
devices, are typically used in a Wheatstone bridge configuration with a voltage excitation source.
To achieve the best measurements, understanding the signal conditioning needs for each
measurement type is paramount. Based on the sensors you require to perform an application,
certain types of signal conditioning you need to consider certain types of signal conditioning
to ensure the best measurements possible. Table 1 provides a summary of signal conditioning
types for the different sensors and measurements.
Temperature Sensors
The most common sensors used to measure temperature are thermocouples, RTDs, and
thermistors. These sensors typically emit a low-output voltage measured in the millivolt range.
The output of these sensors is too small for measurement devices with a large input range to
measure accurately. For example, a typical signal range for a thermocouple is ± 80 mV. If you
have a 16-bit digitizer with a range of ±10 V, you can use only 0.8 percent of the range of the
ADC. To solve this problem, use amplification to increase the size of your output signal to match the range of the ADC.
As discussed previously, thermistors, RTDs, and thermocouples often output signals very close to 0 V; therefore, offset errors from the measurement device can be a large factor in overall
accuracy. Offset error is the deviation in measured temperature relative to the reference temperature. Many devices support a built-in autozero function that automatically measures the
internal offset before you acquire temperature data and compensates for offset error in the measurement device. If the measurement device does not support autozero, ensure
that the device is regularly calibrated and use the specification document to identify how offset error affects the overall accuracy.
Since temperature measurements are usually sampled at a slow rate, these measurements are susceptible to high- frequency noise. Lowpass filters are commonly used to eliminate high-frequency noise and 50 Hz and 60 Hz power line noise, which is prevalent in most laboratory or industrial environments.
Thermocouple
Thermocouples have specific signal conditioning requirements. Because cold junctions are
formed by the connection of the thermocouple to wires or terminals of the data acquisition
device, they generate voltages that add to your net measurement. For example, in the system
shown in Figure 4, instead of measuring AB, which is desired, the actual measurement is
AB+AC+BC. The additional voltages generated by the extra junctions are cold-junction error.
To eliminate this error, the known temperatures of AC and BC are subtracted from the total
measurement to obtain the true temperature. This adjustment is known as cold junction
compensation (CJC). Most thermocouple measurement devices include built-in CJC and
automatic scaling in software. If the data acquisition device does not have built-in CJC, the
temperature must be measured externally to account for this difference in software.
Although a CJC helps account for errors induced from cold junctions, the CJC itself
and how it is implemented can cause errors as well. The overall CJC error includes the error from the CJC sensor, the error from the device measuring the CJC sensor, and the gradient between the cold junction and the CJC sensor. The temperature gradient between the cold junction and the
CJC sensor is the largest factor. Placing the CJCs as close as possible to the thermocouple terminals helps reduce this type of CJC error. To reduce errors from the CJC sensor, use an accurate temperature sensor such as an RTD, a thermistor, or an IC temperature sensor designed for the temperature range the cold junctions will be subjected to. To reduce errors from the measurement device, invest in a device that offers the accuracy specifications you need for the application, calibrate the device as required, and use the device only within the conditions specified by the manufacturer.
Another source of noise that can affect thermocouples is directly mounting or soldering them to conductive materials or submersing them in water. When a thermocouple is connected to a conductive material, it is susceptible to common-mode noise and ground loops. Isolation helps prevent ground loops and can significantly improve the rejection of common mode noise. Conductive materials with large common-mode voltages require isolation to effectively measure large common-
mode voltages.
Thermistors and RTDs
Thermistors and RTDs are active temperature sensors that require voltage or current excitation.
It is important to note that sending a large excitation current results in self-heating, which
affects the accuracy of your measurement. If you cannot dissipate this extra heat, consider
lowering the excitation current. When using RTDs or thermistors, be sure to implement
amplification and lowpass filters as discussed earlier to help eliminate the effect of noise.
Strain Gage
Strain gage measurement involves sensing extremely small changes in resistance. The proper
selection and use of the bridge and signal conditioning are required for reliable measurements.
The three main types of strain gages are quarter, half, and full bridge. The name refers to how
many legs of the Wheatstone bridge are made up of actively sensing strain gages. Therefore,
you need bridge-completion circuitry for quarter- and half-bridge strain gages. Typically signal
conditioning circuitry strain gages are designed for half-bridge completion networks. If you
are using a quarter-bridge sensor, you need a third resistor commonly referred to as the quarter-bridge completion resistor. Similar to temperature sensors, most strain gages require amplification because they have relatively low output levels (less than 100 mV) which makes them vulnerable to noise. Using lowpass filters can help remove noise from unwanted high-frequency components.
Strain gages require voltage excitation levels between 2.5 V and 10 V. The change in output voltage for a given level of strain increases in direct proportion to the excitation voltage. Although a higher voltage excitation generates a proportionately higher output voltage and thus improves signal-to-noise ratio, the higher voltage can also cause errors because of self-heating. Self-heating changes a strain gage’s resistivity and sensitivity, affects the adhesive’s ability to transfer strain, and introduces temperature effects between lead wires and the foil gage. This has a large impact on measurements when the structure does not provide good heat dissipation, such as plastic. You can reduce self-
heating by either selecting a strain gage with a larger surface area for better heat dissipation
or reducing the excitation level.
If the strain gage circuit is far from the signal conditioning circuitry and excitation source,
the resistance of long lead wires and small gage wires can result in a lower excitation voltage
delivered to the bridge. Compensate for this error by using remote sense. Remote sense
measures the amount of excitation actually delivered to the sensor and regulates
the excitation supply through negative feedback to compensate for lead losses and deliver the
needed voltage at the bridge.
When a strain gage is installed and connected to a Wheatstone bridge, it is unlikely that zero
volts are read when no strain is applied. Strain gage imperfections, lead wire resistance, and
prestrained installation condition generate some nonzero initial voltage offset. In this case
perform offset nulling or null calibration in hardware or software to compensate for the inherent
bridge imbalance. In software, take an initial measurement before applying strain and use this
initial voltage in the strain calculations to calculate strain offset. Simple and fast, this method requires no manual adjustments. The disadvantage of software compensation with traditional measurements is a loss in effective measurement range due to large offsets. Another method uses hardware to balance the bridge. Measure the initial strain and then fine-tune a potentiometer as a leg of the Wheatstone bridge to physically adjust the output of the bridge to zero. By varying the resistance of the potentiometer, you can control the level of the bridge output and set the initial output to 0 V.
Load, Pressure, and Torque
The most common tool for measuring load, pressure,and torque is a full-bridge strain-gage-based sensor. In a full-bridge setup, all four of the Wheatstone bridge legs are actually strain gages; therefore, you do not need additional resistors or bridge-completion circuitry. Load, pressure, and torque sensors can output low or high voltages, depending on the excitation requirements of the sensor. Typically powered by a measurement device, low-excitation-level sensors output millivolt-
and volt-range signals, but high-excitation-level sensors require higher external power sources
to operate and output ±5 V, ±10 V, or 4–20 mA. Since this is a full-bridge strain measurement,
the signal conditioning discussed previously for strain measurements such as remote sensing
and shunt calibration also applies.
Accelerometers and Microphones
Sound and vibration measurements are closely related. Accelerometers and microphones both
measure oscillations but in different media; therefore, the theory of sound and vibration
measurements and their necessary signal conditioning techniques are similar. The type of signal
conditioning implemented for accelerometers and microphones depends on whether they have
built-in amplifiers. Because the charge produced by an accelerometer is very small, the electrical
signal produced by the sensor is susceptible to noise, and you must use sensitive electronics
to amplify and condition the signal.
Integrated Electronic Piezoelectric (IEPE) sensors integrate the charge amplifier or voltage amplifier close to the sensor to ensure better noise immunity and more convenient packaging. This signal conditioning provides a constant current source to power the circuitry inside the sensors. Since
piezoelectric accelerometers are high- impedance sources, you must design a charge-sensitive amplifier with low noise, high input impedance, and low output impedance. Like accelerometers, microphones can be powered externally or internally. Externally polarized condenser microphones require 200 V from an external power supply. Prepolarized condenser microphones are powered by IEPE preamplifiers that require a constant current source.
When IEPE signal conditioning is enabled, a DC voltage offset is generated equal to the
product of the excitation current and sensor impedance. The signal acquired from the sensor
consists of both AC and DC components, where the DC component offsets the AC component
from zero. This can lower the resolution of the measurement because the signal amplification
is limited by the range of the ADC. You can solve this problem by implementing AC coupling.
Also known as capacitive coupling, AC coupling uses a capacitor in series with the signal to
filter out the DC component from a signal.
LVDTs
Linear variable differential transformers (LVDTs) and rotary variable differential transformers
(RVDTs) are popular sensors for measuring position. LVDTs operate like transformers and
consist of a stationary coil assembly and moveable core. An LVDT measures displacement by
associating a specific signal value for any given position of the core. Signal conditioning
circuitry is essential for the proper operation of an LVDT.
You must generate a sinusoidal signal to provide excitation for the primary coil. This signal
is typically between 400 Hz and 10 kHz, and the frequency of the signal should be at least
10 times greater than the highest expected frequency of the core motion. You should apply the
same sine wave used for excitation to demodulate the secondary output signal. You should
also include a lowpass filter to remove high-frequency ripple. The resulting output is a DC
voltage that is proportional to the core displacement.
