When one analog input goes dead and the others on the card read half their value or jump across the display, the shared path is at fault: the field wiring common, the card-to-base connection, or the base power supply. Record the readings, meter the loop current at the terminals, then meter the base supply while the system runs, and only then try a documented, compatible card.
When one channel on an analog input card stops reading and the others show about half their value or jump across the display, the fault usually sits in what the channels share: the field wiring and its common return, the card-to-base connection, or the base power supply. The card is the last thing to replace. This pattern tends to arrive after years of correct service, so it looks like sudden card failure, but the electrical trail usually leads to the base side. Work the order below: record the readings while the fault is live, meter the field signal at the card terminals, meter the base supply under load, and only then run a card swap that can test anything.
What each reading pattern points to
Decide from the pattern before removing anything. A fault that touches one channel lives in that channel's loop. A fault that touches several channels on one card lives in what they share.
| Reading pattern | What it points to | First check |
|---|---|---|
| One channel dead, the rest correct | That channel's transmitter, loop wiring or terminal | Meter that loop at the card terminals |
| Several channels sitting at about half value | A disturbed common return or a shared supply problem | Compare loop current at the transmitter and at the card |
| Several channels jumping in fast steps, for example sweeping from about 1 to 920 display units while the process holds steady | The base connection or the base supply | Meter the base supply while the system runs |
| Controller raw value steady, displayed value wrong | Display comms, tag selection or scaling | Trace from the raw count to the displayed value |
| Display freezes after a card swap | A swap that has not yet tested anything: model, configuration or seating | Restore, then verify model and seating from the hardware documentation |
Several channels moving together raises the priority of the shared wiring, the common return, the card-to-base connection and the base supply. The configured input type and the meter reading, not the displayed number, say whether an input is open, noisy, shorted or scaled wrong.
Record the fault before you touch anything
Capture the fault while it is present. For each channel write down the displayed value, whether it is stable or jumping, and whether the process itself is changing. If the program exposes them, compare the displayed value with the controller's raw input value and the scaled value in between. A steady raw value with a wandering display clears the acquisition and sends the trail through comms, tag selection or scaling; the way a scale value maps a raw count onto a display value covers that side.
Then build the timeline. Note when the readings went wrong against wiring work, electrical events and recent maintenance. In one job with this exact pattern, an operator had been wiring a 24 VDC modem, drew a spark and burnt a prox switch. The odd readings had started before that, so the spark and the readings were worked as two separate events.
Leave repeated power cycles out of the diagnosis. A cycle can clear a symptom for a while and hides which element changed. Run controlled tests, change one thing at a time, and write down what moved after each one.
Check the field signal at the card terminals
Most process transmitters send the 4 to 20 mA current signal standardized for electronic process instruments in ANSI/ISA-50.00.01: 4 mA at zero span, 20 mA at full span, linear in between. Read the transmitter's documentation for its active or passive loop type and wire the loop the way that documentation shows.
Hold the process at a known state and meter the loop current at the transmitter, then at the card's input terminals. The same current at both points clears the cable, the terminals and the route. A mismatch points at the wiring: a conductor loose under a screw, a shield drain wire touching a signal terminal, two loops commoned at a terminal block, or a cable run beside power conductors.
If the loop current is correct at the terminals and the controller's raw value is still wrong, the field side is cleared and the base side is next. Compact combination cards get the same order, signal first and card last; a walkthrough of erratic-reading checks for a compact analog combination card follows these same steps.
Follow the card's documented wiring diagram and isolation practice before moving any conductor. The loop is low energy; the cabinet around it is not.
Worked example: three of four channels erratic on a steady process
The inputs, and where each comes from:
- Four analog channels on one combination card, all seated in the same base: the channel count of the case, one channel dead and three erratic.
- Correct displayed value about 97 display units, meaning whatever scale the operator display is configured to show, on a process holding steady, with the program matching similar machines and 3 fault-free years behind it: the case values.
- Faulty readings: three channels sitting at about half the correct value, or jumping in fast steps between about 1 and 920 display units: the case values.
- Base supply history from a matching case: thermocouple inputs bounced while the 24 VAC base supply metered fine unloaded, and a replacement supply ended the fault.
The rule in words: a steady process holds its displayed value in a narrow band, so a reading that sweeps across many times the expected value is coming from the signal path. Count the channels that share the fault: one channel means its own loop, several channels on one card mean what they share, the base connection and the base supply.
The arithmetic:
- Half check: 97 / 2 = 48.5 display units, so a channel parked near 48 or 49 matches the half symptom exactly.
- Swing span: 920 - 1 = 919 display units on each jump.
- Swing against expected: 919 / 97 ≈ 9.5. The faulty channels sweep about 9.5 times the correct reading, in fast steps, on a process that is holding still.
- Share of channels: 3 erratic of 4 = 75% of the card's channels, all riding one base.
Result: half values and 9.5-times sweeps across 75% of one card's channels describe the shared acquisition path, and a start date that precedes the 24 VDC spark removes the spark from the list of triggers.
The decision, with its threshold: one faulty channel out of four sends you to that channel's loop first. Three or more erratic channels on one card send you to the base connection and the base supply first, and the card comes last. In this case, with loop currents correct at the terminals, the fault gets confirmed at the base supply under load, and the fix is the specified replacement supply or a base reseat per the hardware documentation.
Meter the base supply under load
A supply can sit inside its band on a meter with little current flowing and sag or oscillate once the card, its display and the I/O draw current together. That is the documented case behind this section: thermocouple inputs bouncing, a 24 VAC base supply that tested fine, and a new supply that ended the fault.
Meter at the base supply terminals while the controller and I/O are running, and keep the meter connected while the fault is showing. Compare what you see with the supply's own documentation for the installed hardware. A supply that also feeds solenoid or coil loads sags each time a coil picks up; that shared DC supply with solenoid coils that freezes a display is a fault of its own.
If the reading droops or wanders in step with the fault, fit the specified replacement supply and recheck every channel. If the supply holds its band while the fault shows, move to the base connection itself: seating, connectors and terminations, inspected and reinstalled per the hardware documentation.
Make the card swap a real test
In this case the replacement card froze the display while the original card kept producing wrong values. A swap that ends with a frozen display has tested nothing: model or configuration, seating, the base and the display path are all still in play.
A swap becomes a test when three conditions hold: the replacement is the exact documented model and configuration for the base and controller, it is seated per the manufacturer's procedure, and the controller and display run normally afterwards. A display that freezes has its own check order with power first, as covered in what to check when an HMI screen goes blank and freezes.
Change one element at a time and write down the result. With an approved known-good card available, put it in the suspect base; where the hardware design allows it, put the suspect card in a known-good base. If the symptom follows the card, the card is condemned. If it stays with the base, the base or its supply is condemned and the original card goes back to work.
Where this fault usually ends
Recommendation: with one channel dead and three erratic on one analog card, leave the card where it is. Record each channel's reading, meter every loop current at the card terminals against the transmitter, then meter the base supply at its terminals while the system runs and the fault is live. Fit the specified replacement supply if the reading droops or wanders with the fault. Reseat and inspect the base per the hardware documentation if the supply holds. Run a card swap only after both pass, with the exact documented model, one element at a time, and only accept the swap as a verdict when the display and controller keep running normally.
At the next scheduled shutdown, meter the base supply under load again and log the reading. A baseline taken while everything runs is what turns the next episode of half values and jumps into a one-hour job.
Questions
The odd readings started before the wiring spark. Does the spark still matter?
Treat the two as separate until the evidence links them. Date the first bad reading from the controller's data log or trend history, then compare that time with the wiring work. A spark that burnt a prox switch and a 24 VDC modem connection can still have damaged something nearby, but readings that predate the spark point at a cause that was already at work, most often the base connection or its supply.
Can a base power supply pass a meter check and still be the fault?
Yes. A supply can hold its band with little current flowing and sag or oscillate once the card, its display and the outputs draw current together. The documented case looked like this: thermocouple inputs bounced while a 24 VAC base supply metered fine, and a replacement supply ended the fault. If the fault will not hold still long enough to meter, run a recording meter on the base terminals for a shift and look for dips in step with the bad readings.
Should I add filtering when analog readings jump?
If the loop current meters steady at the card terminals and the raw value still moves, noise filtering is a legitimate last step: the card's configuration usually offers a filter or damping setting for exactly this. Apply it after the wiring and supply checks pass, since a filter smooths a real signal problem as readily as it smooths noise, and the fault stays in the wiring, waiting.
Planning information, not legal or engineering advice. Rules differ by place: check codes and permits with your local authority and your licensed trades.