Signal Conditioning for PC-Based Control
When you’re using a standard control solution such as a programmable automation controller (PAC), says Bob Nelson, business manager for controller I/O and software at Siemens, “the manufacturer takes care of the inner-workings of that controller, so the user doesn’t have to worry about them. When moving to a PC-based open-control environment, that openness gives great flexibility, but also puts more responsibility on the user to make sure the system can deliver the performance needed, and to avoid pitfalls of not tying the pieces together properly.”
So, why would any control engineer bother taking on this burden?
“As machines become more complex, control is no longer limited to traditional digital ON and OFF or slow speed analog signals,” said Swapnil Padhye, signal conditioning product marketing engineer for National Instruments. “Modern control applications demand acquiring data from complex sensors, such as accelerometers, and making decisions at higher speeds. PC-based systems offer advantages with their faster processors and availability of advanced feedback algorithms in the software.”
All control systems are equal
Nearly every automated system, from a clothes dryer to the 2 km (1.25 mile) diameter Tevatron at Fermilab in Geneva, IL shares the same basic architecture. It is a classic control loop, which starts when sensors collect information about the equipment it’s supposed to control, then analyzes that information with a programmable controller, and sends control signals to the equipment under control. The controller can only communicate over its data bus, so incoming signals have to be converted from their original form to digital “words” sent over the bus. This is a job for data acquisition (DAQ) electronics.
But DAQ electronics can’t always deal with any old signal that sensors put out. Different kinds of sensors put out enormously varied signals. Thermocouples, for example, put out millivolt-level signals that must be carefully terminated, and are a somewhat nonlinear function of the temperature they are supposed to sense. Resistance temperature detectors measure the same physical parameter, but require a precisely controlled excitation current. Its output signal is a voltage that may be two orders of magnitude larger than a thermocouple output at the same temperature. Thermistors are also resistance devices, but they have a negative temperature characteristic. Thus, putting a constant current through a thermistor produces a voltage that decreases with increasing temperature, instead of increasing.
The highly varied signals produced by sensors need to be matched to the DAQ electronics input. This is the job of the signal-conditioning modules. Since each sensor type has different signal conditioning demands, a wide variety of modules are available to meet those demands. In the end, the user selects sensors to meet measurement needs, and then selects signal conditioners to meet those sensor needs.
Finally, the system needs drive electronics to convert the output bus signals to an electrical signal with the appropriate form (ac, dc, TTL logic, etc.) and provide enough power to get the job done.
The difference between PC-based control and more traditional technology using programmable logic controllers and programmable automation controllers is mainly a matter of packaging. As William Bolton points out on page 3 of his book “Programmable Automation Controllers,” “A programmable logic controller (PLC) is a special form of microprocessor-based controller that uses a programmable memory to store instructions and to implement functions such as logic, sequencing, timing, counting and arithmetic in order to control machines and processes….” Personal computers (PCs), on the other hand, are general-purpose digital computers designed to adapt to a wide variety of applications.
The PLC or PAC vendor has taken on the responsibility for making hardware and software selections that fit the unit for its role in the control-system application. As Nelson points out, when a control engineer chooses to go the PC-based control route, responsibility for making those choices falls directly on his shoulders.
This fact has ramifications that can be good or bad, depending on the situation. It does, however, give the control engineer or system integrator a great deal of power to adapt and fine tune the control system for a specific application.
Signal conditioning basics
“Examples of signal conditioning,” says Padhye, “are amplification of small signals, attenuation of high or unsafe signals, filtering signals to reduce noise, isolating high voltage signals, providing excitation to passive transducers, or completing circuits for bridge-based sensors.”
Signal conditioning usually (although not always) deals with analog signals and uses analog electronics. The three parameters that characterize an analog signal are amplitude, frequency, and phase. The signal conditioning electronics is there to modify these parameters as they modify the sensor’s output signal to fit the analog-to-digital convertor (ADC) at the data acquisition system’s front end.
Amplitude characterizes a signal’s level. If it is a voltage signal, the amplitude specifies its peak voltage level; if it’s a current signal, then amplitude specifies the peak current; etc. For nominally dc signals, the amplitude is just the dc voltage, current, etc. Signal conditioning electronics modifies amplitude through amplifier gain and limits it by dynamic range.
Frequency for control signals usually translates into bandwidth. Most control signals are not pure tones at a single frequency. Even frequency-modulated signals, where the sensor’s output value is represented by a shift in the signal’s frequency, take up a significant amount of spectral real estate. Even nominally dc (zero frequency) signals include frequency components representing how the physical parameter being sensed varies over time. Signal conditioning electronics modifies control signal frequency content through filters.
|Signal conditioning electronics interface a nearly limitless variety of sensors to the data acquisition electronics needed to digitize the signals and pass them over the computer bus to the controller.|
Phase in control applications generally relates to temporal correlations between the signal and actions in other parts of the system. Phase shifts introduced in the signal conditioning electronics interact with phase shifts arising elsewhere in the control loop, and can lead to oscillations and other loop-dynamic phenomena.
In addition, there are issues with matching impedances, zero levels, and ground potentials.
“Control engineers would risk quite a bit by bringing field signals from sensors directly into a data acquisition card inside the PC,” says Bob Smith, vice-president for sales and marketing at Dataforth Corporation. “There’s always the potential for dangerous voltages to appear on those lines. There could be as much as 440 volts in factories. So we need input isolation, transient protection, and input voltage protection filtering in signal-conditioning modules.”
“Another huge problem a signal conditioner can eliminate is ground loops,” he continues. “Signal-conditioner isolation barriers break any type of ground loop that could occur.”
Another issue is linearization. Sensors use physical phenomena to produce electrical signals, such as resistance changes due to stretching of wires to measure strain. Generally, these phenomena are selected to be at least approximately linear. Demands of control applications, however, have come to exceed the linearity of most sensor phenomena. Measurements in most cases have to be corrected for second-order, or sometimes even third order, linearity errors.
“The advantage of doing linearization inside the signal conditioner,” says Smith, “is that you do not burden the PC with the time to either perform a mathematical computation such as a Taylor-series expansion, or go to a lookup table.”
“I think scaling units conversion is huge,” says Chuck Cimino marketing director for data acquisition products at Keithley Instruments. “If you’re in a controls application, what you want to see is temperature, pressure and flow, not millivolts. How do you correlate in your mind, millivolts to degrees Celsius? You want the system to do it for you.”
“Filters are always important,” he continues, “because many of the signals include frequency components that are beyond what you’re interested in. Proximity sensors, for example, can pick up vibrations.”
So, a general signal-conditioning subsystem will likely include:
Electrical isolation—typically by opto-isolators that convert electrical voltage levels to optical signals, then back again—to break ground loops and limit voltage transients;
One or more amplifiers to modify the sensor signal’s amplitude (including linearization), as well as to match impedances, zero levels, and ground potentials;
One or more filters to control the spectral response characteristics; and
Excitation sources as needed.
|Signal conditioner modules incorporate circuitry to provide the isolation, amplification, filtering and excitation needed by the particular sensors for which they are designed.|
“Today,” says Padhye, “devices are available with dedicated signal conditioning functionality, although newer products have started combining signal conditioning and data acquisition into a single device. Both form factors are commonly used, each providing benefits for different types of applications. Using front-end external signal conditioning and multiplexing hundreds of channels to a single data acquisition provides great cost benefits to large channel count systems, although with a limitation on maximum speeds achieved with this architecture.”
“What we do,” says Nelson, “is go through a PCI card to a Profibus network to reach our standard I/O sub-system.”
“Most systems using industrial PCs are mounted in some type of NEMA cabinet for protection,” says Dataforth’s Smith. “The signal conditioners are usually mounted on DIN rails in those cabinets.”
DIN rail mounting started in Europe and, over the past several years, has become popular in the United States as well. Rails can be mounted on flat panels or walls. The modules are usually made of high-temperature plastic and have a special foot on them that snaps to the metal rails. The field wiring from the sensor and the computer wiring to the I/O board connect directly to the modules, so each module has a set of wires to the field and a set of wires to the DAQ input. “Marshaling of the wires is probably the messiest part of a DIN rail installation,” Smith reports.
Another popular signal-conditioner form factor, is what Smith calls “plug-in-the-panel” modules. These are small encapsulated plastic blocks about 1 inch by 1 inch by 1/2-inch thick. They have pins coming out of the bottom that interface to sockets on a circuit board called a back panel. Having 2, 4, 8, or 16 channels per back panel is typical. The computer-side wiring goes along printed traces on the back panel to a connector. To reach the DAQ input, a cable simply plugs into the connector on the back panel and directly into a connector on the DAQ board plugged into the PC.
A third robust form factor is the chassis, crate or mainframe type. The chassis provides computer-side signal wiring for modules plugged directly into slots, as well as cooling and power. Various standards, such as VXI, CompactPCI, and PXI, are available that essentially extend the computer bus architecture into a form suitable for instrument modules. Standards designated with an “XI” are based upon standard computer buses while providing additional features, such as triggering lines, that provide extra functionality needed by instrumentation systems. VXI, for example, is an extension of VMEbus, while PXI is an extension of PCIbus.
“The signal conditioners sit in there side-by-side,” says Smith. “They can be what I like to call ’slices’ where they’re not in an encapsulated module. They are circuit boards with little front panels that may have a power indicator or some test points available to the engineer.”
Typically, signal conditioning modules also provide connectors that make terminating field wiring fairly easy without removing the module from the chassis.
“Now there are two flavors of PXI,” Keithley’s Cimino points out, “and there are USB, FireWire, and now LXI which is an Ethernet standard for both instrument and data (ac). LXI is just some things added to the standard ethernet to address timing issues and other types of testing measurement programming issues.”
Surprisingly, the old General Purpose Instrument Bus (GPIB) still represents some 90% of networked instrumentation sales, according to Cimino. GPIB, also known as IEEE-488, was developed in the late 1960s by Hewlett-Packard as “HPIB.” It became an international standard in 1975 when formalized by the Institute of Electrical and Electronics Engineers (IEEE).
“In the instrument world,” Cimino says, “it’s been the mainstay. But in the controls world, I’d say it’s probably less than 5% or 10%.”
The process for choosing PC control system signal conditioning starts with the suite of sensors.
“The control system designer,” says National Instruments’ Padhye, “will have to know what types of signals the system are going to interact with before planning the strategy. Once the control system designer has an initial idea about the number and types of analog and digital channels, it is the signal conditioning functionality that converts signals to safe and measurable levels.”
Each sensor needs its own signal conditioner, and the characteristics of that sensor determine the signal conditioner. There are modules for type-J thermocouples, strain gauges and virtually any sensor type in virtually any form factor.
The choice of form factor for the system is largely subjective, although there are some situations where it is obvious. If you’re mixing PC-based control electronics with other hardware mounted on DIN rails, it makes sense to put the signal conditioning electronics on DIN rails, too. If you’ve got a Profibus network in place, Profibus would be an obvious choice. The form factor choice depends more on how you plan to bring the signals into the DAQ card and computer.
At the Large Hadron Collider being built on the French/Swiss border at the European Organization for Nuclear Research (commonly known as CERN), where intense radiation fields would fry any electronics mounted close to the beamline, choices are limited. CERN control system engineer Alessandro Masi chose to run long cables to signal conditioning modules mounted in PXI chassis as much as 800 m away.
 Bolton, W, Programmable Automation Controllers, 4th edition, Newnes, ISBN-10: 0750681128, (August 30, 2006)
|C.G. Masi is senior editor for Control Engineering. Reach him at firstname.lastname@example.org .|