Flow Loop Reads High, Noise on the Signal, and a Hunting Valve
An orifice-plate flowmeter measures filtrate flow in a 150 mm line. The differential-pressure transmitter is ranged 0 to 100 kPa, corresponding to 0 to 150 t/h, and outputs 4–20 mA. The DCS analog input applies a square-root extraction to the normalized signal before displaying flow. During commissioning, the instrument technician zeroed and spanned the transmitter, then also enabled the square-root output jumper inside the transmitter; the factory default for this model is linear output. The DCS now displays 116 t/h. An independent clamp-on ultrasonic meter temporarily installed on the same line reads 90 t/h. The loop is also oscillating with a period of about 20 s and the control valve is visibly hunting. The cable shield is grounded at both the field junction box and the DCS cabinet. A 400 A VFD feeder runs in the same cable tray for about 30 m parallel to the signal cable.
1. Show by calculation that the double square-root extraction explains the reading error, and state the correction.
Read the worked answer
True flow of 90 t/h on a 0–150 t/h range is 90/150 = 0.600 of span. Because flow varies as the square root of differential pressure, DP is 0.600² = 0.36 of its range:
DP = 0.36 × 100 kPa = 36 kPa
Linear transmitter output: 4 + 16 × 0.36 = 9.76 mA
With the transmitter's square-root jumper enabled: 4 + 16 × √0.36 = 4 + 16 × 0.600 = 13.6 mA
At the DCS, the normalized value is (13.6 − 4)/16 = 0.600, and the DCS applies another square root: √0.600 = 0.7746 → 0.7746 × 150 = 116.2 t/h
This matches the observed 116 t/h almost exactly, confirming the diagnosis: the square root is being extracted twice, so the indicated flow is biased high and, critically, the error is flow-dependent (worst at mid-scale, zero at 0% and 100%).
Correction: set the transmitter output back to linear (disable the internal square root), verify the DCS extraction is enabled, and re-verify the loop at three points - for example 0, 50, and 100% of DP - using the ultrasonic meter or a calibrated DP source. Document the change and check whether any trend, totalizer, or ratio controller upstream was configured assuming the wrong signal.
2. Assess the shielding and grounding arrangement and specify how the signal wiring should be corrected.
Read the worked answer
The shield is grounded at both ends, which creates a ground loop. Any potential difference between the field junction box and the DCS cabinet - including the common-mode current generated by the nearby 400 A VFD - drives current along the shield. That current couples magnetically into the signal pair and appears as noise on the 4–20 mA loop.
Correct practice:
- Ground the shield at one point only, normally at the DCS or marshalling-panel end where the instrument is referenced. Leave the shield isolated at the field end (tape it back; do not connect to the transmitter housing unless the manufacturer's installation drawing requires a local RF ground through a capacitor).
- Carry the shield continuously through the junction box - do not create a pigtail or break it into a terminal block that interrupts the foil.
- Keep the signal pair twisted all the way to the termination, and separate it from the VFD feeder by at least 300 mm, or use a dedicated steel conduit or a separate tray with a grounded divider. Cross power cables at 90° where crossing is unavoidable.
- Confirm the transmitter is not locally grounded through its mounting (isolated thermowell or insulated fitting if needed) and that the DCS input is a true differential input.
Distinguish this from the VFD motor cable, where shielding is correctly grounded at both ends; the rule differs because that shield carries common-mode current intentionally. Also check whether loop power supply return is grounded at more than one point.
3. The valve still hunts with a 20 s period after the flow measurement is corrected. Describe the tests you would run to isolate the cause and the most likely findings.
Read the worked answer
A 20 s limit cycle in automatic, with a stable loop in manual, has four common causes: controller tuning, valve stiction or hysteresis, process dead time/gain, and a measurement dynamics issue. Test in a fixed order.
1. Manual step test. Put the controller in manual, step the output 5% and hold. Record valve position, flow, and time.
- If valve position lags then jumps, or if the same output gives different positions depending on direction, the cause is stiction/hysteresis - usually the actuator, positioner, or packing. Acceptance is typically <1% hysteresis. Fix the valve; do not tune around it.
- If the valve moves smoothly but flow takes many seconds to respond, the cause is dead time - detune the controller (wider proportional band, slower integral) or improve the measurement location.
2. Compute/observe loop dynamics. With an ultimate period Tu ≈ 20 s, the integral time should generally be on the order of Tu/2 to Tu and the proportional gain set from the observed process reaction. Excess integral action (too small a reset time) is the most common cause of a slow, sustained cycle of this length; the controller winds up and overshoots each half-cycle.
3. Check signal conditioning. Verify the DCS input filter, scan rate (a scan slower than the process time constant is not the issue here, but a very fast scan with a noisy signal is), and that the corrected square-root scaling did not leave a gain mismatch - the previous double square root distorted the apparent process gain, which may have led someone to tune aggressively.
4. Check for external forcing, such as the VFD drive on the same line modulating flow, or a downstream pressure controller fighting this loop (two controllers in conflict).
5. Retune last, using a conservative setting derived from the manual step test, and verify with a set-point change and a load disturbance.