Specifying a sensor
Defining sensor specification is an important first step either in selecting an existing sensor or developing a custom sensor. This tutorial contains a discussion of key parameters of sensor specification.
The sensitivity can be defined as a slope (the first derivative) of the noiseless sensor's response vs. the measurand (the measured property).
Consider an optical sensor which measures concentration of dissolved agent (solute) in a solution. Such a sensor would typically measures the power of a light beam attenuated by the solution. For simplicity, assume that the incident beam power is constant and known. The attenuated beam power (the measurand) can be measured with a limited precision due to the noise of the sensor electronics and the shot noise of the signal (light power). Thus, given a sensitivity, S, the smallest measurable change in the measurand, dM, is determined by the ratio of a sensor's response noise, dR, and the sensitivity:
The greater the sensitivity, the lower is the importance of the sensor's response noise. These concepts are illustrated in Figure 1.
Figure 1. Sensitivity, S, of a linear-response sensor is defined as a slope of the sensor response, R, vs. measurand, M. The smallest change, dM, in the measurand which the sensor is capable of detecting depends on the sensor response noise, dR, as shown in Equation 1. In a nonlinear-response sensor, the sensitivity is a local slope of the sensor response as a function of the measurand and changes with the measurand value.
Effective noise of the sensor's response generally decreases with time allotted for the measurement if noise reduction techniques are employed. Thus, the measurement time allowed by the application is an important parameter to consider.
The sensor's response noise is an important parameter in defining the sensor precision, i.e. the standard deviation of the measurand value measured by the sensor. However, that standard deviation may also depend on the measurand itself. Consider a particle counting sensor, which has a precision of 1 count. Even if suspension of particles were "homogeneous", i.e. on the average there were N particles per unit volume of the suspension throughout it, each measurement of the number of particles per unit volume would have nevertheless return a random number from a population with the mean value of N. Thus, the measurand itself may impose a precision limit.
Another important characteristics of a sensor is its accuracy, i.e. the deviation of an observed measurand value from the actual (true) value. The sensor accuracy is important when performing absolute
The measurand range is defined by the mimimum detectable measurand value and the maximum allowable measurand value. This range is determined by the sensing technique and the sensor design. If an application calls for a large dynamic range (where the adjective large is defined in the context of the sensing technique/sensor design) it may be necessary to use several techniques (sensor types), each best suited to operate in a sub-range of required measurand range.
One should also consider the sensor's signal range. The signal is a physical parameter which is measured by the sensor. For example, in a distance sensor based on the triangulation principle, the signal is the light power reflected from an object. The measurand is the distance of that object from the sensor, calculated from the sensor geometry and distribution of intensity of light reflected (say diffusely) by the object. Triangulation distance sensing is quite tolerant of the signal magnitude because it does not use the signal intensity to measure the distance. However, even in this case there is a singal range in which the sensor operates properly. If the surface of an object absorbs light, very low light power may be returned to the sensor and prevent it from measuring the distance reliably. Similarly, if the object's surface mirror-reflects light in the sensor wavelength range and is inclined at an angle which causes the sensing beam to miss the detector, the measurement may be impossible.
Finally consider a relationship between the measurand range and sensitivity. In a sensor with a high sensitivity, the signal reaches the upper limit of the measurand range at a value which is lower
than that in a sensor with a low sensitivity. In some cases it is possible to use a variable-sensitivity sensor to cover a large measurand range. Precision of such a sensor necessarily decreases with increasing the measurand range.
If a sensor is to be integrated into an applicationís hardware, the mass and volume (sometimes some critical dimensions specifically) of the sensor may need to be considered. It is sometimes overlooked that a complete sensor system may comprise several components: the sensor itself, a power supply, an amplifier, an
analog-to-digital converter, and a data logger or a data transmission device. Thus, a sensor system usually has a greater volume and mass than the sensor itself.
These conditions may include the ambient light intensity and spectrum, temperature, pressure, humidity, the presence of dust or other aerosols, vibration, and acceleration. Environmental conditions
dictate how robust the system needs to be. A sensor system intended for a laboratory may be pretty delicate. Other environments, for example, a factory floor, may require more robust sensors.
Consideration of operator skills is important in supervised applications of sensors. Both a minimum set of skills and a level of knowledge required to operate the sensor may need to be considered. If the sensor is to be based on technology that is new to your personnel, care must be taken to familiarize them with that technology. Failure to do so may prevent the proper operation of the sensor.
An entertaining example (in retrospect perhaps), is one of a telecom company. Its field technicians were very skilled in installing copper cables. However, when the company began installing first fiber-optics connections, only a few of these connections worked. No wonder, the technicians put liberal amounts of dark conductive grease at butt junctions of the fibers. This made it very difficult for light to pass from one to
another fiber resulting in a faulty connection.
Most modern sensor systems use a computer or a processor in one form or another. This makes interaction with a human operator or with other parts of your application much more efficient. Computerization of a sensor also enables one to collect statistically significant data about a product, production process, or a phenomenon one researches. These data can be used to fine-tune your process or to discover new facets of the phenomenon.
The software user interface in a computerized sensor system may include
The data management tools may include
In addition to the process control functions these tools may potentially provide your quality control system with quantitative means of correlating the presence of defects with process parameters.
Depending on the sensor project complexity, an overall investment required to complete a sensor selection or development may include several components. These investment components enable one to
If an off-the-shelf sensor has been identified, the decision whether to purchase it or develop a custom sensor may be dictated by the sensorís features and by the scale of your application. Off-the-shelf systems usually require a smaller investment than custom sensor systems. However, the former systems are designed with universality in mind. This implies optimization for a broad range of applications and inclusion of
features which may be unnecessary for your application. If the scale of your application justifies it, a custom sensor can be a more cost-effective choice than an off-the-shelf system.
Last modified: . Copyright 2000-2016 MJC Optical Technology. All rights reserved.