Housings can protect outdoor instruments, which can take quite a beating from rain, snow, hail and falling ice. Such instruments can fail slowly over time unless enclosed in appropriate housings. You should configure the housing so it doesn't affect the sensor reading. For example, a housing for a temperature sensor shouldn't act as a heat sink, lowering the sensor's reading. Conversely, if a housing has fins to draw heat from an enclosed sensor during warm weather, you should mount the fins vertically. Otherwise, warm air won't be able to rise away from the housing.
DEALING WITH SIGNAL NOISE
Proper grounding of electrical signals is important. Signals often are referenced to a ground potential. Undesirable electrical ground loops occur when an extraneous current flows through the instrumentation wiring between two points that are supposed to be at the same voltage but aren't. The resulting electrical interference can cause random fluctuations in sensor output and even may damage the sensor. You must ground all instruments together at one master grounding point or to a grounding grid spread throughout the plant. Isolation techniques such as transformers and fiber optic communications can minimize grounding problems.
Electromagnetic and radio frequencies are common in plants that use walkie-talkies, pagers and wireless networks. Interference from these sources can adversely affect sensor signals. Interference also results whenever a current changes dramatically, such as when relay contacts engage or static voltage discharges, generating a spark.
Replacing electromechanical equipment with solid-state devices will eliminate arc-generated interference. Alternatively, simply relocating switch boxes and relays to instrument-free areas of the plant may suffice. Passive shielding of a source also is a solution.
PID controllers tuned to provide appreciable derivative action are particularly susceptible to the effects of measurement noise. They tend to react aggressively to every blip in the measurement signal to quickly suppress deviations from the setpoint. If a blip turns out to be nothing but noise, the controller will take unwarranted corrective actions and make matters worse.
Many modern digital instruments come with built-in digital signal processors and filters. These instruments replace complex analog components like oscillators, mixers and filters with mathematical operations executed inside a digital signal processor (DSP). Such a DSP (similar to a processor inside a personal computer but designed for specific "number crunching" applications) can perform complex operations at blazing speed. Recent advancements in DSP techniques can help greatly in separating extraneous noise from measurement signals.
Compared to analog hardware, a DSP offers more alternatives and much greater flexibility. These benefits lead to more effective methods of separating a real signal from process noise. Tangible advantages include: improved measurements in applications involving vibration, hydraulic noise and temperature fluctuation.
A DSP provides faster analog/digital conversion of a sensor signal, so it can handle a greater number of sample points in a given time than prior technologies. Digital filters with sharp drop-offs eliminate signal frequencies created by hydraulic and line noise that are outside the targeted measurement range. Advanced filtering techniques such as automatic filter adaptation and frequency weighting further give a processor further capabilities to accurately extract the signal from a potentially noisy process signal.
Powerful digital signal processing techniques for separating signal from noise include:
• Moving average. The user defines a band of measurement values. The DSP maintains a moving average of the incoming data values over a selectable window size. Values that fall outside the band are ignored and replaced by the average value (Figure 4). The instrument registers the number of error measurements. If the error total exceeds 50% of the window size, the measurement value is held. If the hold time setting is surpassed, the procedure resets, a new average value is calculated, and the window shifts accordingly.