Observers or sensors for drive controls?

Electrical drives are the tireless driving force behind almost all the movement, dynamics, and convenience found in an industrialized society today. They cover a wide range of applications, from vacuum cleaners to elevators and electric vehicles, all the way to fully automated production lines. With the exception of very-simple applications, electrical drives usually are controlled, regulated and networked by powerful microcontrollers.

The greatest potential for innovation in modern drive engineering lies in the possibility of increasing system intelligence by replacing expensive sensor hardware with inexpensive “observer” software.

The increased intelligence of electrical drives occurs in conjunction with the monolithic fusing of power electronic and microelectronic components at the chip level, making it possible to install complete drive electronics as a power chip in the terminal box or plug connector of the motor.

Drive Engineering Overview
The ability to cover all four quadrants of the speed/torque plane is a prerogative of electrical drives. Controlled electrical drives with their excellent dynamic properties, high flexibility, and efficiency have become an important pillar in the automation industry.

The history of drive engineering began more than 140 years ago, when Werner von Siemens discovered the electrodynamic principle in 1866. It continued with the development of rotary and linear machines as well as transformers at the end of the 19th Century. The history of power converters did not start until the 20th Century with the development of mercury-arc rectifiers. The development of thyristors caused an enormous leap in the innovation of power converters when introduced in 1959. It now was possible to use these controlled-conductivity semiconductor valves to control rectifiers, inverters and converters without problems, and very efficiently. This process was supported by the introduction of integrated circuits such as operational amplifiers and gate modules.

A completely new era started in the mid-1970s when signal processing changed from analog and discrete signal processing to digital technology based on microprocessors. Such digital electrical drives can be used in all areas of advanced automation with connection to a master computer system by parallel or serial interfaces (fieldbuses). The advent of voltage-controlled power transistors, or insulated gate bipolar transistors (IGBTs), in the 1980s cleared the way for practical use of model-based control processes, such as the estimation of actual values with observers or the field-oriented control of three-phase machines with frequency-inverter supply. The power profile of electrical drive engineering increasingly will be governed by highly integrated and multi-functional power electronics in the form of intelligent power chips.

Model-Based Drive Control
t;The goals of model-oriented control theories concentrate on refining the control process and reducing the sensor workload. Two pioneering developments here were 1) the field-oriented control of three-phase machines by means of a mathematical model, and splitting up the motor current into a flux-forming and a torque-forming component, and 2) EMC control with IxR compensation for DC drives to eliminate the need for a speed sensor.

The replacement of expensive sensor hardware with inexpensive observation software is a milestone in drive engineering because it offers a tremendous potential for growth combined with considerable competitive advantages for users.

Figure 1 below compares the basic control modes in drive engineering. Based on the physical equation of motion, which states a change in speed can only be achieved using acceleration torque, the result is that the familiar cascade structure (nesting) is the optimum control structure for all electrical drives. This means the output of an external closed-loop speed controller indicates the setpoint for any secondary torque controller. As the torque in electrical drives is derived from the motor current, the torque controller will be a current controller that also protects the machine from overcurrents.

In a conventional controller design (Figure 1a), the equation systems will be transferred from the time range to the complex range using the La Place transformation. This makes it possible to issue statements concerning the stability of dynamic systems quickly and easily. The systems are, in turn, suitable for designing matching control equipment (P, PI, PID).

The design of state controllers and state observers (Figure 1b) takes place in the state space (time range), which makes it possible to evaluate electrical drives for controllability and observability. Compared with conventional drive controls, this means you don’t need to measure speed when integrating a state observer. A mathematical resource in the state space is the matrix operation and the pole setting.

If you implement cascade control in Figure 1a or state control in Figure 1b into a microprocessor system, both cases require an expensive analog interface (ADC) that has to operate at a higher scanning rate and bit-width due to the severe ripple of the electrical state variables. You eliminate this immense sensor expense completely when the actual value of the electrical state variable is determined by a binary observer (Figure 1c). Non-linear observer algorithms for voltage and current mean values can be derived using the Fourier transformation due to the periodic switching operation of the converter valves.

Observers Replace Sensors
Observers are mathematical models that calculate less-accessible quantities (speed, torque) online based on state variables that can be easily measured (voltages, currents, control states). This is why the process often is referred to as actual value estimation. The reasons for using such observers in electrical drives are mainly economical because expensive sensor hardware can be replaced by inexpensive observer software. The model equations of the observers do, of course, have to be very realistic to achieve satisfactory results. Figure 2 below shows in more detail the structure of actual value detection without sensors, in comparison to the conventional method using sensors. An observer cascade is the central element of drives without sensors. It consists of a state observer (machine model) for the mechanical state variables and a binary observer (converter model) for the electrical state variables. Depending on the drive machine and converter circuit in operation, feedback for this process is limited to two to three binary switching commands.

The switching state of all valves can be described by characteristic converter values. There are three variables for line-commutated converters: the control angle (α), the current flow angle (λ) and the overlap (u). This means the full impact of the intermittent flow and the commutation operation are taken into account. Converters without commutation, such as AC power controllers and three-phase power controllers, operate in intermittent flow only, which means control angle (α) and current flow angle (λ) are sufficient for an exact description of the converter. The group of self-commutated converter circuits such as choppers, inverters and frequency inverters, includes deletable power semiconductors, which means only conduct duration (λ) and overlapping (u) will occur. All three characteristic converter values correspond to electrical angles, with the control angle indicating the triggering delay compared to the natural triggering instance. The current flow angle indicates the conduct duration at which the valves work, and the overlap angle indicates the time period in which the commutation processes take place. As the switching operations in the converters are constantly repeating themselves, α, λ, and u are periodic variables that can be represented by binary, pulse-width-modulated signals (αB, λB, uB). Gaining information from binary signals is the basis for the theory behind binary observers. By simply counting the individual switching data, the associated angles already are available in digital form and can be used directly for calculating the electrical state variables in suitable observer algorithms. Another advantage of binary interfaces is they are computer-friendly, and can be electrically isolated using very simple methods. This is especially important for electrical drives because there always has to be electrical isolation between microelectronics and power semiconductors.

Motor Control Without Sensors
If you compare the formation and processing of the actual values in Figure 2 more closely, you’ll notice that the complex and expensive sensors and analog interfaces (ADC) of conventional motor controls have been replaced by simple and inexpensive comparators (opto-couplers) and counter function blocks in motor controls without sensors.

For a better assessment of the dynamic behavior and the accuracy of drives without sensors, Figure 3 shows a 2 kW DC drive as an example for recording the complete state estimation vector during acceleration. These observed actual values of speed, armature voltage, and armature current were fed simultaneously into the respective controllers as actual value replacements, so we’re looking at current and speed control without sensors.

The dynamics correspond to the known sequences of conventional DC drives with sensors. The linear increase in speed during the acceleration phase indicates the dynamic accuracy of the current observer. Closer examination shows minor deviations from the actual state variables. This deviation is mainly caused by the speed observer, where temperature-dependent winding resistors cause model errors of a few percent via the system matrix. These values can be reduced by adjusting the state observer for the speed to the operating temperature of the machine. The algorithms of current and voltage observers, on the other hand, don’t contain any temperature-dependent elements, which means that these observers are very accurate.

The observer-oriented current detection isn’t limited to DC drives; analog sensors and interfaces can be replaced by a respective observer cascade in AC and DC drives. This replacement is possible mainly because the power semiconductors in AC power controllers, direct inverters, inverters and frequency inverters are modulated in binary switching mode to achieve high efficiency ratings.

Observer + Sensor = Redundancy
Another interesting application option results from the demand for greater operational safety in the automation of processes and products. Observers and sensors can be connected in accordance with Figure 3 (below) to form a redundant drive control. Redundancy according to DIN 40042 means the functioning presence of more technical resources than necessary for the function (e.g., control). This setup increases the reliablity of the system because if any component fails (e.g., sensor) its functionality can be taken over by another component (e.g., observer) without affecting the system function. Troubleshooting can be scheduled to take place during periods when the entire system is subject to standard maintenance, so there will be no production stop. In these cases, the drive controllers are robust enough to cope with sensor failure.

The dynamic response of this type of low-cost redundancy function to the sudden loss of an actual speed value is shown in Figure 4 below. The speedometer break occurs during regenerative braking operation in the second quadrant from a high-speed setpoint to a low one. The redundant drive control switches to the estimated value of the observer, once it detects a problem with the actual value. Because the actual observer value at the time of the problem is almost identical to the speedometer signal, the transition is hardly noticeable in the speed curve. Once the speed has reached the low setpoint, the converter software starts switching to the first quadrant. In this case, note that the transition routine receives its information from observed values only.

As the drive operates without sensors after failure of the speedometer, there will be minor limitations in the accuracy and setting range. The decisive advantage of the observer-based redundancy is that it avoids dangerous system states and keeps up emergency operation.

Summary
Modern electrical drives are characterized by the successful integration of power electronics and microelectronics. Furthermore, they don’t use expensive sensors to determine actual values for control purposes, but instead use inexpensive observer software. As there is no need to detect, convert, and process analog values, very little hardware is needed to implement this intelligent control strategy. This means the solution is particularly suited to low-cost automation applications. The inaccuracy caused by the estimation of actual values by using observers lies in the 1% range, giving electrical drives without sensors a maximum control range of 1:100. This speed setting range is sufficient for basically all standard applications at home, at work, and in commercial settings.