In signal processing, sampling is the acquisition of measured values of an analog signal at specific time intervals. The article shows what exactly is involved and what role sampling plays in sensor technology and for industrial use.
Sampling is embedded in the context of digital signal processing. This term covers all processing steps aimed at extracting information from a measured analog signal or preparing information for transmission for further processing. For digital signal processing to work, the analog signals must be converted so that they can be stored and processed. This conversion is also referred to as digitization.
Sampling in this context means the registration of measured values at so-called discrete points in time. In this way, a time-discrete signal is obtained from a time-continuous, i.e. uninterrupted, signal. Discrete-time means that the measurement is taken and documented at a specific time interval.
In other words, during sampling, the values of a signal are recorded at defined, usually regular intervals. The result is discrete values, also called sampled values.
Since an infinite number of values cannot be processed, the signal is quantized in time. In concrete terms, this means that the current value of the signal is recorded or sampled at discrete points in time. The intervals between the time points are usually always equidistant, i.e. of equal size. The measured values are also called sampled values. Secondly, the amplitude of the signal (the maximum deflection of the oscillation) must be recorded.
In order to explain digital signal processing in terms of system theory, the sampling as well as the signal reconstruction must be described mathematically "ideal". The mathematical description of the ideal sampling requires the definition of a so-called sampling function. For the sampling function a sequence of pulses is suitable due to the acquisition of the signal at discrete points in time.
In fact, such an ideal sampling function cannot be implemented. Reasons are the infinite steepness and infinitely short duration of pulses – thus, that the technical conditions do not allow the generation of ideal pulses. Therefore, the real sampling results from this fact.
The sampling rate or sampling frequency refers to the following. In digital signal processing, the term refers to the frequency with which an analog, continuous-time signal is sampled – that is, measured and converted into a discrete-time signal – in a given time.
The time interval within which the conversion takes place is the so-called sampling interval. If this time interval is constant, it is also referred to as the sampling frequency.
Sampling is always a periodic process. Therefore, it is measured in the unit Hertz (Hz), i.e. in periods per second. Because the constantly changing signal is only ever recorded at the time of the sampling process, a snapshot or instantaneous value is obtained each time. For this reason, the individual measurement results are called samples and are referred to as samples per second (SPS).
A constant sampling rate contributes to easier further processing of the signal. Despite the fact that the actual sampling only takes place in the context of digitization, the sampling rate is also referred to in further processing.
There is more than one sampling principle or method. However, most encoders use the photoelectric scanning method. This is contactless and therefore does not generate any wear. Photoelectric scanning can even detect graduation lines of a few micrometers. These generate output signals with extremely small signal periods.
The possible applications of sampling in the context of digital signal processing are almost endless. Exemplary application areas are the following:
Especially in industrial manufacturing, automation is advancing and signal processing brings great advantages in the context of electronic or physical measurements. Robotics is spearheading this development, and artificial intelligence is an essential component of Industry 4.0.
The reason for this is the rapid spread of cobots, i.e. collaborative robots. With their capabilities and thanks to the latest sensor technology, they are fundamentally changing production. At the same time, the new robots are much smaller and far less expensive than conventional robotic arms. Investments in this area are therefore also possible for medium-sized companies.
The sensor technology of the robots consists, for example, of camera systems and force or torque sensors. The robots thus form an interface to the environment and complement the work of humans, with whom they work side by side. They are also freely programmable and can therefore take on temporary tasks. This means they can be used for a certain time at one workstation, then reprogrammed and used again at another location.
Modern sensor technology is a key contributor to the development of increasingly sophisticated robotic systems. The main sensor technologies currently used in industry include tactile and proximity sensors as well as magnetic position sensors, gesture sensors, force and torque sensors, power management sensors and environmental sensors.
One example: Among the most widespread sensors in industry, but also in consumer electronics, are the so-called magnetic position sensors. All joints of industrial robots used in today’s industrial applications contain two or more position sensors. The reason: rotary movements or joint rotations are only possible with their help. Most robots have small, powerful DC motors that enable the movement of their joints and limbs. In order for these motors to be driven accurately, feedback on the current angle of rotation is necessary.
The magnetic position sensor reports the motor’s angle of rotation to the robot’s joint controller. The joint then moves to the favored position on the basis of the transmitted data and a closed control loop.
A robot joint requires two position sensors per axis. This means that an ankle joint that is to be capable of performing axial tilt and rotation motion requires four position sensors. Given the large number of joints in most robots, it is therefore no surprise that magnetic position sensors play such an important role in industrial robotics.
The flexibility of artificial intelligences thus results, among other things, from their intelligent sensors. This means they offer a wide range of application areas – which makes them very interesting for companies. On the one hand, they make a decisive contribution to increasing efficiency and accelerating manufacturing processes, and on the other, they bring enormous cost reductions.
It is also clear that the quality of the sensors increases the sensitivity of the robots. On the one hand, this makes AI more reliable when dealing with humans, and on the other, this ability helps to solve tasks that conventional robots fail at. For example, the tighter the tolerances of the components, the more complex the assembly. Here, the use of a robot that can scan the environment is helpful and ultimately increases the quality and speed of the process.
Sophisticated measurement technologies such as sensing as part of signal processing are what make sensor technology and robotic systems feasible in the first place. Only with the help of sensors can robots perceive their environment and become usable for industrial applications. Sensors enable spatial vision, which is used for object detection and collision avoidance. Many current service robots have 2D and 3D cameras that capture their working environment with high resolution. For example, even the smallest measurement tolerances on the workpiece can indicate the need for industrial spare parts.
