Digital Control Systems: Sampling and Z-Transform

Modern technology, from drone stabilizers to automotive engine control units, relies on digital control systems. These systems bridge the physical, continuous world and the discrete, computational world of microprocessors. Understanding how to sample continuous signals, represent system dynamics algebraically, and reconstruct commands is fundamental to designing effective and stable digital controllers.

The Bridge Between Continuous and Discrete: Sampling

A digital control system's first task is to measure the continuous physical world. It does this through sampling, the process of capturing the value of a continuous-time signal at regular, discrete instants. This is performed by an Analog-to-Digital Converter (ADC). The time between samples is called the sampling period, $T_s$, and its inverse, $f_s=1/T_s$, is the sampling rate or sampling frequency.

The sampled signal, $x(k{T}_s)$—often abbreviated as $x(k)$—is a sequence of numbers representing snapshots of the original continuous signal $x(t)$. This conversion is non-reversible; information about the signal's behavior between sample points is lost. Therefore, the choice of $T_s$ is a critical design decision with profound implications for performance and stability, as will be discussed in the section on the Nyquist criterion.

From Samples to Commands: Hold Circuits

After a digital algorithm processes the sampled data, it must output a command that can act upon a continuous-time plant (e.g., a motor or valve). A Digital-to-Analog Converter (DAC) performs this task, but a simple DAC that outputs an impulse at each sample instant is impractical for physical systems. Instead, a hold circuit is used to reconstruct, or "hold," the output value between samples.

The most common type is the Zero-Order Hold (ZOH). A ZOH circuit takes a discrete value $u(k)$ and maintains that constant voltage or current until the next sample instant $u(k+1)$ arrives. The resulting output is a staircase approximation of the intended continuous control signal. While this introduces a phase lag and distorts high-frequency components, its mathematical simplicity and effectiveness make it the standard in most digital control implementations. The ZOH's effect must be accounted for in the overall system model.

The Z-Transform: Algebra for Difference Equations

Inside the digital processor, the controller is implemented as a difference equation. For example, a simple digital filter might be $y(k)=0.5y(k-1)+0.3u(k)$. Analyzing stability and frequency response directly from such time-domain equations is cumbersome. The z-transform provides the essential tool to simplify this analysis.

The z-transform of a discrete-time sequence $x(k)$ is defined as: $$X(z)=\sum _{k=0}^{\infty }x(k)z^{-k}$$ where $z$ is a complex variable. Its power lies in its property that a time shift corresponds to multiplication by $z^{-1}$. Applying the z-transform to a linear difference equation converts it into an algebraic equation in $z$. This allows us to derive the pulse transfer function, $G(z)=Y(z)/U(z)$, which is the discrete-time equivalent of the continuous Laplace-domain transfer function.

For instance, taking the z-transform of the difference equation $y(k)=ay(k-1)+bu(k)$ gives $Y(z)=a{z}^{-1}Y(z)+bU(z)$. Solving for the ratio $Y(z)/U(z)$ yields the transfer function $G(z)=b/(1-a{z}^{-1})=bz/(z-a)$. Stability is now easily assessed: for a causal system, all poles (roots of the denominator in $G(z)$) must lie inside the unit circle ($|z|<1$) in the complex z-plane.

Designing the Sample Rate: Nyquist and Aliasing

Selecting the sampling rate $f_s$ is a fundamental engineering trade-off. Sampling too slowly wastes available processor bandwidth and can lead to poor dynamic response. Sampling too fast increases computational load and hardware cost without tangible benefit. The absolute lower bound is governed by the Nyquist-Shannon sampling theorem.

The theorem states that to uniquely reconstruct a continuous signal from its samples, the sampling frequency must be greater than twice the highest frequency component present in the signal, $f_s>2f_{max}$. The frequency $f_N=f_s/2$ is called the Nyquist frequency.

Violating this criterion leads to aliasing, a phenomenon where high-frequency signal components "masquerade" as low-frequency components after sampling. Imagine a stroboscope (sampler) illuminating a spinning wheel. If the wheel rotates faster than half the strobe frequency, it can appear to be spinning slowly or even backwards—this is aliasing. In control systems, aliasing of high-frequency noise or unmodeled dynamics can corrupt the sampled signal and destabilize the controller. To prevent this, an anti-aliasing filter (a low-pass analog filter) is placed before the ADC to attenuate any signal components above the Nyquist frequency.

Common Pitfalls

1. Ignoring the Effects of the Hold Circuit: Treating the digital controller as a pure $G(z)$ without modeling the ZOH's effect is a common error. The ZOH introduces an effective delay of approximately $T_s/2$ seconds and alters the frequency response. A more accurate analysis models the continuous plant combined with the ZOH before deriving the discrete equivalent.
2. Sampling Based Only on System Bandwidth: While a good starting point is to sample at 10 to 30 times the system's closed-loop bandwidth, this alone doesn't guarantee avoidance of aliasing. High-frequency sensor noise or structural resonances can still alias into the control bandwidth if an adequate anti-aliasing filter is not designed and implemented.
3. Misapplying Continuous-Domain Intuition to the z-Plane: Stability regions differ. In the s-plane (Laplace), stability requires poles in the left-half plane. In the z-plane, it requires poles inside the unit circle. A pole at $z=-0.9$ is stable (inside the unit circle) but corresponds to an oscillatory response at the Nyquist frequency, which might be undesirable. Direct translation of s-domain designs requires careful discretization methods (e.g., Tustin's method, matched pole-zero).
4. Confusing the Z-Transform with the Discrete-Time Fourier Transform (DTFT): The z-transform is a more general tool. The DTFT is essentially the z-transform evaluated on the unit circle ($z=e^{j\omega T_s}$). It is used for frequency response analysis, while the z-transform in its full form is used for analyzing transient response, stability, and system structure.

Summary

• Digital control hinges on three core operations: sampling a continuous signal with an ADC, processing it with a difference equation, and reconstructing a continuous command using a Zero-Order Hold (ZOH) circuit in a DAC.
• The z-transform converts time-domain difference equations into algebraic transfer functions $G(z)$ in the complex z-domain, where stability is determined by poles lying inside the unit circle.
• The Nyquist-Shannon sampling theorem dictates that the sampling rate $f_s$ must exceed twice the highest frequency in the signal to prevent aliasing, which is mitigated in practice by anti-aliasing filters.
• Selecting $T_s$ is a critical design trade-off, and the distorting effects of the sample-and-hold process must be included in a faithful dynamic model of the complete digital control loop.