2 – 4
Advanced Topics
gral and derivative one at a time. A simple
P
-controller generates a
control variable that is proportional to the error signal.
1
Increasing the
P
gain should cause the process output to respond
by moving closer to the setpoint. Generally, enough amplification
should be used so that the process output is brought reasonably close
to the setpoint. Too much gain, however, will cause the system to
oscillate. Start with a small
P
gain, and increase by factors of two
until the system begins to oscillate. Then back o
ff
in small amounts
until stability is recovered.
While it is possible to maintain stability with a simple
P
controller,
in general this will lead to a finite, non-zero
ε
. Increasing
P
will
tend to reduce the resulting
ε
, but too much proportional gain will
eventually lead to oscillations.
One way to eliminate this nonzero error problem is to include an
o
ff
set at the controller output. The SIM960
O
ff
set
parameter can be
turned on and adjusted to hold the process power at a level that main-
tains a smaller error. However, this is only a coarse improvement,
since the necessary power level may change with time.
Integral control provides an “automatic” way to dynamically adjust
the e
ff
ective o
ff
set to zero the error; in older controllers, integral
action was called “automatic reset” for this reason. Integral control
simply integrates the error signal with respect to time. Thus the
controller output changes until the error has been reduced to zero,
near which point the controller output slows and stops changing. If
the error drifts over time, the integrator responds by adjusting the
controller output to cancel the error. So it is much like having a
dynamic output o
ff
set constantly responding to system changes. As
with proportional gain, too much integral gain can cause oscillation.
Again, start with a small
I
gain and increase by factors of two until
oscillation begins, then back o
ff
until stability is recovered.
Though integral control is e
ff
ective at reducing the error, it is not as
e
ff
ective as proportional control at responding quickly to changes.
This is because the integrator needs time to build up a response. To
further enhance the response of the process to rapid changes, deriva-
tive control is often employed. Derivative control is proportional
to the rate of change of the error, so it is relatively unresponsive to
slow changes, but rapid changes to the system produce a significant
response. Derivative control reduces oscillations that can result from
step changes to a system.
1
During the tuning process, it is important to keep in mind that the di
ff
erential
input range of the SIM960 is
±
1.0 V. It is good practice to occasionally glance at
the
OVLD
indicator to ensure the input amplifier is not saturated.
SIM960 Analog PID Controller
