Random RTD jumps of about 2 °C that stop when a door-mounted operator panel is disconnected point at coupling through the panel's wiring. Disconnect its 24 V DC feed first: the programming cable keeps the panel alive at about 5 V DC, so a quiet channel pins the noise on the powered 24 V circuit. Then work through bonding, ferrites, an isolator and a separate supply.
RTD inputs that jump by about 2 °C whenever a door-mounted operator panel is powered are usually a wiring and noise problem, and the panel is the suspect because the symptom follows its power state. The panel can couple switching noise into the analog circuit through three routes: its DC supply leads, its serial communication cable, or the shared ground and signal common. Because the symptom looks identical for all three, you find the one at work by changing a single thing at a time and watching the same channel. The first change is the cheapest and the most informative: disconnect the panel's DC feed while its programming cable keeps the panel alive, then watch. This article gives the test order, the numbers from an installation where the sequence was run, and the decision each result points to.
Why do RTD readings jump when the door panel is connected?
Random jumps of about 2 °C that stop the moment the door panel is unplugged are coupled electrical noise, not a drifting sensor: a jump that small is under an ohm on a PT100 input, which is why tiny coupled currents are enough.
Randomness is the tell. A real process change tracks the batch and holds its new value; noise jumps up and down with no pattern and returns to the same reading once the disturbance stops. Watch the channel on a steady process, because a moving temperature looks like noise in a log.
The pattern across several cabinets says more than any single cabinet. In one installation where this fault appeared, ten cabinets each carried a 6-inch door panel, and only 3 of the 10 showed the jumps; the other 7 stayed quiet with the same panel model. Identical hardware behaving differently by cabinet puts the cause in the cabinet: power wiring, grounding, cable routing or the RTD card. That is the useful part, because cabinet wiring is fixable without buying anything.
Read your own symptom against this list:
- Jumps with the panel connected, quiet the moment it is unplugged. The panel or its wiring is in the coupling path. Its supply feed and its serial line get tested separately below.
- Some cabinets of a fleet show it, the rest never do. The difference is cabinet-level. Compare an affected cabinet against a quiet one rather than against the panel datasheet.
- A spare panel shows the same jumps. The cabinet interaction is the cause. Stop swapping hardware and run the test order.
- A bond from the DC common to earth changes nothing. Grounding is not the route in that cabinet; move on to the supply and serial tests.
- The value drifts with the process and never snaps back. That is a real temperature or a wiring fault. Reversed or miswired RTD leads show up as a lost sign on bipolar RTD temperatures, which has its own fix.
How do I capture a baseline before touching anything?
Log the affected RTD channel for at least 5 minutes with the panel connected and running in its normal configuration, and write down the jump size, the channel, the units, the direction of each jump and the interval between jumps.
Nothing changes while you record the baseline, because every later comparison is measured against it. Keep the process steady: a moving batch mimics noise. Photograph the panel wiring, the supply terminals and the RTD card terminations now, so every test state can be restored exactly.
Set each test window by the jump interval you logged. When jumps arrive every few seconds, hold each state for 5 minutes and you will see several of them. When they arrive every few minutes, hold each state for 15 minutes so the state contains several intervals. A state held shorter than one interval tells you nothing about a change.
Which change do I test first: the panel's supply feed?
Disconnect the panel's 24 V DC supply first and leave its programming cable connected: the cable feeds the panel about 5 V DC, the panel keeps running in a low-power mode, and a quiet RTD channel marks the powered supply circuit as the coupling path.
This one change keeps the panel mounted, alive and still talking over the serial cable, so it separates the panel's powered state from its mere presence. Quiet readings at reduced brightness put the disturbance on the supply wiring or in the panel's own electronics at full power, and the fix is supply-side. Jumps that continue from the cable put the coupling on the serial line or the grounding, and the serial tests come next.
De-energize the feed at its breaker or fused terminal and verify 0 V DC across the panel's supply terminals with a meter before opening any wiring; this is the only step in the sequence that puts hands on power conductors.
After the window, restore the feed and watch the jumps return. A symptom that clears on one change and comes back when that change is undone is a result you can act on. A supply that also feeds other loads shows up in this test too: backlight flicker when the panel polls the PLC is the same family of fault, a supply sagging under load, as covered in HMI backlight flicker when polling: check the power supply.
What does an example one-change test sequence look like?
Ten identical cabinets each carried a 6-inch door panel, and 3 of the 10 showed random jumps of about 2 °C on their RTD cards; the sequence below cleared the noise and brought it back with single changes.
Step 1, baseline with the panel on its rated supply. The RTD channel jumps about 2 °C at random. On a PT100 element the standard platinum coefficient is 0.385 Ω per °C (IEC 60751), so 2 °C corresponds to 0.385 × 2 = 0.77 Ω. With a typical card excitation of 1 mA, kept low to limit sensor self-heating, that is 0.77 Ω × 1 mA = 0.77 mV of disturbance at the card input. Numbers that small are why a coupled current of a few microamps can move the reading.
Step 2, bond the DC common to earth. No change. The jumps keep the same size and rate.
Step 3, panel supply feed removed, programming cable in place. The panel keeps running in its low-power mode from the cable's 5 V DC. The RTD channel holds steady.
Step 4, feed restored, alternate serial cable on the second port. The jumps return.
Step 5, spare panel installed on the original feed. The jumps persist.
Step 3 is the decisive result and step 4 is the one to re-run on its own: in this sequence the feed and the serial cable changed together in step 4, so the supply is the strong lead and the cable is the second variable, tested again later with the supply unchanged. The 7 quiet cabinets agree with that reading, because the panel model is the same everywhere and the cause has to live in the affected cabinets' wiring.
Will bonding the DC common to earth fix it?
Treat a bond from the DC common to earth as a design decision, not a first remedy: in the example cabinet it left the jumps unchanged, and every extra bond is a candidate path for a ground loop between equipment that is already earthed.
A ground loop disturbs an analog signal by adding or subtracting small voltages: two parts of the system sit at slightly different earth potentials, and current finds a path through the signal common or a shield. Every added bond is a new candidate path, which is why a quick bonding change can leave the symptom in place or move it somewhere else in the cabinet.
So inspect before you add. Trace where the panel chassis, the PLC, the RTD card, the supply common and the cabinet earth actually terminate, and compare the affected cabinet against a quiet one. Correct loose or unintended bonds according to the cabinet's grounding design and leave the designed bonding arrangement as found. Faults in this supply-and-ground family show up in other ways too, such as an HMI panel that needs a periodic power cycle.
Do ferrite cores and an RS232 isolator come next?
Snap a ferrite core over the panel's serial cable and one over its incoming supply pair, and if jumps persist after that, put an RS232 opto-isolator between the panel and the PLC to break the conductive path through the serial line.
A ferrite core is the cheapest hardware step, and it is the first countermeasure in vendor guidance on preventing sensor and PLC malfunctions near inverters, where the same coupling mechanism walks switching noise into analog wiring. Clamp the cores as close to the panel end as practical. The core adds high-frequency impedance that dissipates common-mode current before it reaches the analog circuit.
The RS232 opto-isolator answers a different route: it breaks the galvanic path through the serial cable, so a potential difference between the panel and the PLC stops driving current through the signal line. It is the right tool when the power-state test has cleared the supply side and the jumps remain whenever the panel communicates. On the supply route it will leave the symptom in place, which is why the step 3 test decides which of the two you buy first.
When does a separate or linear supply fix it?
When the noise tracks the panel's supply, give the panel its own 24 V DC source: a separate supply for the panel cleared the symptom in one case, and a linear supply in place of a switching one cleared it in another.
Separating supplies breaks the route by which the panel's switching load modulates a rail the analog circuitry references, while both supplies keep the same cabinet earth, so no new loop is created. Both versions of this fix kept the panel at its rated supply in the field cases behind the example.
Size the new supply from the panel's documentation rather than from the old supply's rating: read the required voltage range and the current draw from the panel's label or manual and size above that figure. A linear supply costs heat and space and produces far less switching ripple, which suits a small door-panel load well. Change the supply alone, hold the same test window as the baseline, and compare. Reading a symptom that appears in one supply state and disappears in another is a repeatable method, used again in an HMI that locks up on AC power but runs on 24 V DC.
What is the decision for each test result?
If the RTD channel goes quiet when the panel drops from its rated supply to the 5 V DC of its programming cable, the fix is supply-side; every other result in the sequence points to its own fix, listed below.
- Quiet at 5 V DC from the cable, jumps with the feed restored. Supply route. Fit a separate 24 V DC supply sized from the panel label, add ferrite cores to its supply leads, and step up to a linear supply if jumps persist.
- Jumps in every powered state, quiet only when the panel is unplugged. Serial route. Ferrites on the serial cable first, then an RS232 opto-isolator between panel and PLC.
- The bond to earth changed nothing. Grounding is not the route in that cabinet. Leave the designed bonding as found and spend the time on the supply and serial tests.
- A spare panel shows the same jumps. Cabinet interaction. Stop replacing hardware and re-check the routing: separate the panel's supply and serial wiring from the RTD leads and repair any damaged shield termination.
- The value never snaps back in any state. That is a sensor or card question, not noise. Check the element, the lead configuration and the card configuration on their own.
Recommendation: Disconnect the panel's supply feed before anything else, with the programming cable left in place so the panel stays alive, and hold the state for the observation window you set in the baseline step. If the jumps stop, fit a separate supply for the panel, sized from its label current, with ferrite cores on its supply leads. If the jumps persist, fit ferrites on the serial cable and then an RS232 opto-isolator. Run the panel from its rated supply in production and keep the low-power state for diagnosis.
Questions
Does the same test order apply to analog inputs other than RTDs?
Yes. Thermocouple cards, 4-20 mA transmitter inputs and millivolt signals all couple noise through wiring, supply commons and shields, so the one-change order holds. Convert the symptom into the signal's own units first: note the jump in displayed engineering units and in raw counts or mA if the card exposes them, so every test state is compared on the same scale.
Where should a ferrite core sit on the cable?
As close to the noisy end as practical, which for this fault means at the panel end of both the serial cable and the incoming DC supply pair. Snap-on cores clamp over the assembled cable without rewiring, one per cable. Where the cable is long enough, pass it through the core twice; each additional pass raises the impedance the core adds.
Can a door-mounted panel with a poor earth cause the same jumps?
It can contribute. A panel bolted to a painted door often reaches earth only through the hinge, and paint is a poor conductor, so the chassis can sit above cabinet earth potential. Add a braided bonding strap across the door hinge or run an earth lead to the panel's earth terminal, then repeat the baseline. The cabinets in this article had earthed panels and still showed the noise, so treat bonding as one check in the sequence, not the whole answer.
What if the jumps only appear while a variable frequency drive is running?
Then the drive is a likely source and the panel is the receiver. Inverter-driven motors conduct and radiate switching noise into nearby analog wiring. Vendor guidance on preventing sensor and PLC malfunctions near inverters recommends ferrite cores on sensor and communication cables, physical separation from motor wiring, and a check of the drive's output cable and filters. Re-run the power-state test with the drive stopped and with it running to confirm the source.
Planning information, not legal or engineering advice. Rules differ by place: check codes and permits with your local authority and your licensed trades.