Page 6 - untitled

ctllp05044Mistakes.qxd 4/18/2005 11:29 AM Page 6

THE FOUR BIGGEST MISTAKES IN INSTRUMENTATION ... A

>> Mistake #3: an inch. Whether those flaws are should be applied in the final
Generating gibberish appreciable or not will depend on stages of the project, just before
the amplitude of the original noise loop tuning.
N Nooiissee signal, the inertia of the rollers, and 4
Ground loops are not the only the tuning of the controller. Mistake #4:
source of noise that can distort a PID controllers tuned to provide Quitting too soon
sensor’s readings. Radio frequency appreciable derivative action are Even when the data filters are in
interference (RFI) is even more particularly susceptible to the place and the last loop has been
common in plants that use walkie- effects of measurement noise. They tuned, the project isn’t over. There
talkies, pagers, and wireless net- tend to react aggressively to every are some commonly neglected
works extensively. RFI also results blip in the measurement signal to chores that should continue as
whenever a current changes, such quickly suppress deviations from long as the instrumentation system
as when an electromechanical con- the setpoint. If a blip turns out to is in place.
tact or a static discharge generates be nothing but noise, the controller
a spark. will take unwarranted corrective C Caalliibbrraattiioonn
The sources of RFI noise must actions and make matters worse. Most instrumentation engineers
be eliminated or at least kept away know that a sensor must be cali-
from the plant’s instrumentation if at F Fiilltteerriinngg brated in order to associate a
all possible. Replacing electro- Unfortunately, it is not always pos- numerical value with the electrical
mechanical equipment with solid- sible to eliminate noise sources signal coming out of the transmitter.
state devices will eliminate arc-gen- altogether. It is often necessary to Yet all too often, the instruments are
erated RFI. Or, it may be sufficient filter the raw sensor data by aver- calibrated just once during installa-
to simply relocate switch boxes aging several samples together or tion then left to operate unattended
and relays to instrument-free areas by ignoring any changes less than for years.
of the plant. If all else fails, it may some small percentage. Many digi- The result is an insidious prob-
be possible to passively shield the tal instruments, like ABB’s FSM lem known as drift. A sensor’s out-
source of the interference or the 4000 flowmeter, come equipped put tends to creep higher and high-
instruments being subjected to it. with built-in filters. er (or lower and lower), even if the
3
Ignoring the problem is not an However, it is a mistake to think measured variable hasn’t changed.
option, especially when the source that number crunching alone can Deposition on the sensing sur-
of the noise is ordinary house cur- fix all measurement noise prob- faces, corrosion in the wiring, and
rent. At 60 Hz, house current oscil- lems. Filtering tends to increase the long term wear on moving parts
lates slowly enough to have an time required to detect a change in can all cause an instrument to
appreciable effect on some the measured value and can even begin generating artificially high (or
processes. introduce spurious information into low) readings. As a result, the con-
Consider the steel rolling appli- the signal. Worse still, it can mask troller will gradually increase or
cation again. A 60 Hz noise super- the actual behavior of the process decrease its control efforts to com-
imposed on the output of the thick- if it is overdone. pensate for a non-existent error.
ness gauge will pass through the It is generally more cost-effec- Analog instruments are particu-
controller and induce a 60 Hz oscil- tive in the long run to install sen- larly susceptible to drift, much like
lation in the roller pressure. If the sors correctly and minimize the old FM radios. The slightest nudge
sheet exits the rollers with a veloci- sources of interference than to rely on the dial could cause the radio to
ty of six feet per second, those strictly on mathematics to separate lose its signal. With modern digital
oscillations will appear as bumps in the data from the noise. When con- radios, the one true frequency for
the sheet appearing every tenth of structing a control loop, data filters each station is digitally encoded at