What a Sensor Actually Reports
Why this matters
The number on a controller screen is the end of a chain of five or six conversions, and a technician who treats it as the process quantity will replace a perfectly good sensing element because a scaling parameter somewhere downstream was set wrong. The reading is a report on the sensor's own state, translated repeatedly. Everything useful about diagnosing controls comes from knowing which link you are looking at when you look at a number.
This card is about the chain and what the displayed value excludes. A sibling card covers where the element is physically located and why that caps how good your control can be; another covers the signatures each link makes when it fails. Those are cited at the end rather than repeated here.
The chain between the quantity and the number
Six links, and every one of them can change the number with nothing changing at the process.
- The medium touches the element. Heat, pressure or motion has to get into the sensing element. This is a physical coupling with its own resistance and its own delay.
- The element changes a property. Resistance, voltage, capacitance, frequency. This is the only link that has any contact with reality.
- Conditioning turns that property into a standard signal. A bridge, an amplifier, a cold-junction compensation circuit, an analog-to-digital converter.
- Transmission carries the signal. A 4 to 20 milliamp current loop, a 0 to 10 volt signal, a raw resistance run back to the controller, or a digital message on a bus.
- The receiving input scales it. The controller applies a range, a curve type and an offset to turn the incoming signal into engineering units.
- The display renders it. Rounding, averaging, smoothing, a refresh rate, sometimes a units conversion.
Only link 2 knows anything. Links 3 through 6 are arithmetic performed on an assumption.
The reading is not the process quantity
It is the element's own state, and the element gets there through a physical coupling that takes time and never fully arrives.
A temperature element inside a stainless steel thermowell with an air gap around it is measuring the well's inner wall, not the fluid, and the well's inner wall is a compromise between the fluid and everything the well is conducting heat to, including the pipe and the air outside the insulation. Fill that gap with a conducting compound and the same element reads differently, immediately, with nothing having happened in the pipe. On a rising process the well reads low; on a falling process it reads high; on a steady process it reads close, which is why bench checks pass and field behavior does not.
The same applies outside temperature. A pressure transmitter behind a long capillary or a partly plugged port reports the pressure at its diaphragm, which is the process pressure filtered by that restriction. A current transformer reports the field it sees, which includes anything else routed through it.
The practical consequence: a reading that disagrees with the process is not automatically a bad sensor. It may be a correct sensor with a poor coupling, and swapping the element changes nothing.
The reading is not continuous in time
Almost every displayed value is filtered. Controllers smooth analog inputs so the screen does not flicker, and that filter is usually a rolling average or a first-order lag with a time window that is not shown to you anywhere on the screen.
If the display carries a 30 second rolling average, a genuine 5 second excursion that peaks 20 degrees above normal shows up as roughly a 3 degree bump, because a fifth of the window carried the excursion and the arithmetic is 20 times 5 divided by 30, about 3.3. That calculation assumes a flat-weighted rolling average over the window and no other filtering upstream. On a first-order filter the shape is different but the direction is identical: short events are attenuated, and the shorter they are the more they are attenuated.
This is why a display can sit calm through the exact transient that is tripping a protective device. The device sees the peak. You see the average. The device is not lying.
The reading is not more precise than its scaling
A 4 to 20 milliamp input scaled 0 to 250 degrees over its 16 milliamp span carries 15.625 degrees per milliamp. Read 12.4 milliamps at the controller input and the engineering value is 12.4 minus 4, times 250, divided by 16: 8.4 times 250 is 2100, divided by 16 is 131.25 degrees.
Now set the same input's range to 0 to 200 by mistake, which is a two-keystroke error on most controllers and is invisible on the screen afterward. The same 12.4 milliamps becomes 8.4 times 200 divided by 16, which is 105 degrees. A 26.25 degree error, with a perfect element, perfect wiring and a perfect loop current.
Resolution stacks on top of that. A 10 bit input has 1024 counts across its span. Over a 250 degree range that is about 0.24 degrees per count, which is finer than the sensor's own accuracy and does not matter. Over a 2500 degree range on the same 10 bit input it is about 2.4 degrees per count, and now the display's last digit is fiction. Check the span before you trust a tenth of a degree.
The reading is not proof the loop is using it
This is the link people skip. A controller can display an input it does not control from. Mixed-air temperature can be shown on the screen while the economizer logic runs off a different input, or off a calculated value, or off a network point that has gone stale and is holding its last received value with no indication.
A displayed value that has not moved at all across a real change in conditions is the tell. So is a value that reads plausibly but that the output ignores. Before you diagnose a control decision, confirm on the controller's own configuration which input the decision reads. If the sequence of operation names an input the controller is not configured to use, the sequence is documentation and the configuration is what runs.
Worked example: one milliamp reading, two answers
A discharge temperature transmitter feeds a controller that keeps tripping a high limit while the screen shows a comfortable value.
At the controller input terminals the loop reads 12.4 milliamps. The transmitter's label states a 0 to 250 degree range over 4 to 20 milliamps, so the true engineering value is 131.25 degrees, computed above. The controller's own input configuration shows a range of 0 to 200, so the screen shows 105 degrees. The gap is 26.25 degrees, and it is entirely in link 5.
Two things follow, and both matter. First, the limit is not nuisance-tripping: the process really is at about 131 degrees, and the limit is doing exactly what it exists to do. Second, nothing is wrong with the sensor, the wire or the transmitter, so a parts swap would have consumed the visit and returned the same symptom.
Because a protective device is in this story, the fix does not start with the scaling. It starts with why the limit opened, which the corrected reading has just answered: the process is genuinely running hotter than the limit's trip point. Correcting the range makes the screen honest, and then the actual work begins on whatever is driving the temperature up - airflow, staging, a stuck actuator. Correcting the range and moving on would leave a system running at its limit with a screen that now agrees it is.
Reading loop current means either breaking the loop to insert a meter in series or using a clamp-on milliamp meter around one conductor. Either way you are inside an enclosure that in most equipment also carries line-voltage terminals; the loop being 24 volts does not make the panel 24 volts. De-energize the line-voltage side and verify dead before you open it if you can. If troubleshooting genuinely requires the circuit live, that is the narrow allowance in 29 CFR 1910.333(a)(1), and it comes with a meter and leads rated for the circuit's category and available fault current, plus the shock and arc-flash protection the assessment calls for under NFPA 70E-2021.
How to verify you got this right
Work the chain in the order that eliminates the most links per measurement.
Compare the raw signal to the displayed value first. One measurement at the controller input, run through the transmitter's stated range, tells you whether links 5 and 6 are honest. If they are, everything downstream of the transmitter is cleared in one shot.
Then compare the raw signal to a reference at the element. This separates link 2 and link 3 from link 1. A sibling card covers doing that properly against a known condition at two points.
Confirm the controller reads the input you think it does. Force the input to change by a known amount and watch which displayed values and which outputs respond. A point that does not move did not participate.
Record the range parameters before you change anything. The most common way this diagnosis goes wrong is a tech who corrects a range, does not write down the old value, and cannot later tell whether the range was wrong from commissioning or was changed by someone chasing the same symptom last season.
References
- 29 CFR 1910.333(a)(1) - live parts to be de-energized before work, and the narrow conditions permitting energized troubleshooting
- NFPA 70E-2021, 120.5 - process for establishing and verifying an electrically safe work condition
- Manufacturer documentation for the transmitter's stated range and output characteristic, and for the controller's input scaling and filter settings
- See related: Where a Sensor Is Reporting From and Why It Matters; The Sensor Failure Modes and What Each One Looks Like; How to Verify a Sensor Against the Real Quantity