The measurement provides the window into the process. Smart transmitters have made the view clearer. Installation and operating condition effects are compensated for allowing the transmitter to have an installed accuracy close to the sensor capability. The main remaining considerations are filtering, noise, rangeability (turndown), threshold sensitivity, resolution, and wireless update rates.
Filters from signal processing and transmitter damping adjustments have an effect similar to the signal filters in controllers. See the Control Talk Blog "Effect of Controller Dynamics." The equation for the integrated error shows the effect of a measurement filter for given settings of controller gain and reset time. There are additional effects in terms of an increase in loop dead time and deception if the measurement filter becomes the largest time constant in the loop. On slide 12 of the ISA New Orleans short course “Effective Use of Measurements, Valves, and Variable Speed Drives” a simple equation is offered to estimate the attenuating effect of time constant. The equation can be used as the basis for setting transmitter damping settings to keep noise less than default trigger level of wireless transmitters. The equation can also provide an estimate of the actual process variable amplitude from the filtered amplitude.
There are many sources of measurement lags and pure delays. Since we live in a digital world and are entering into wireless control, we need to consider the additional dead time from discontinuous updates (periodic updates rather than continuous updates from analog devices with inline sensors). The practitioner needs to know when these are a concern.
The same approach is used to estimate the implied dead time from the existing tuning settings. If the increase in dead time from measurement dynamics causes the total dead time to become greater than the implied dead time, the controller will become oscillatory unless retuned. The deterioration in performance from the required slower tuning settings can be estimated from the equation for the integrated error.
For at-line analyzers, the total dead time from the sample transportation delay, analyzer cycle time, and multiplex time have a major effect on loop performance because of the large dead time introduced. The loop may become dead time dominant where the dead time is much larger than the process time constant.
As a process has progressively more non-self-regulation, a measurement time constant (lag) has a greater effect especially when this lag becomes large enough to be considered a secondary time constant. Thus, measurement lags are more detrimental in integrating than self-regulating processes and are more detrimental in runaway than integrating processes. For runaway processes, the window of allowable controller gains can close making the process unstable for all tuning settings.
Resolution and threshold sensitivity limits can create a dead time or cause a kick in the controller output from derivative action. If the rate of change of the process variable is extremely slow (e.g. temperature on large volumes), the dead time and kick can be large because such processes tend to have a large PID gain and a rate time setting. Even worse is the triggering of a false update from noise.
The wide range and 12 bit A/D of thermocouple input cards in 1980s vintage DCS and sadly in some modern day HVAC systems have a resolution limit of about 0.25 degF causing deterioration in the signal to noise ratio and kicking in the output so severe that derivative action cannot be used and the controller gain had to be decreased, further degrading temperature control. Fortunately, this problem is mostly history but may offer some guidance for wireless transmitter settings.
A small judiciously set threshold sensitivity limit can have a beneficial effect when properly used to screen out noise. The addition of this limit prevents limit cycles caused by resolution limits, deadband, and stick-slip by preventing noise triggering an update of the enhanced PID developed for wireless. The limit can also enable a higher controller gain by preventing noise from being amplified by the PID gain causing dither (high frequency oscillations) of the valve or variable speed drive. For near-integrating, integrating, and runaway processes, noise can be the primary constraint as to how high a controller gain can be used.
