CMOS battery low on a vision or industrial PC: save your data first

The controller sat unused for a season, and now it greets you with a "CMOS battery state low" message, an invitation to run the setup utility, and a keypress before it will boot. On an industrial PC or vision controller that warning is not housekeeping: the cell that keeps the clock also keeps BIOS settings, and on older controls it backs the battery RAM that holds machine and vision parameters. The order of work decides whether this is an afternoon or a week of re-teaching. What follows is the order that protects the data first: regain setup access, copy everything, then replace the cell, plus the battery arrangements you will actually find on the board and the repair-or-replace call for soldered RTC modules.

The decision: back up first, then touch hardware

Commit to one order: leave the machine powered, copy every parameter, job and calibration file, then replace the battery. A low-cell warning means the backup for the real-time clock and its settings RAM is at end of life. Makers of encapsulated RTC modules quote retention on the order of 10 years at 25 °C, and a machine that sat off for months has spent that reserve; the warning is the last polite notice you get.

While mains power is applied, the clock circuit runs from the supply and the cell is not being drained, so the window for a safe copy is now, with power on. Every cold start is a fresh coin toss on what still decodes.

What dies with the cell: date and time, often to a default such as 01-01-1980, boot order, the halt-on-keyboard-error behavior, onboard port settings, and, on a control that keeps parameters in battery-backed RAM, the machine parameters and compensation tables. The first group takes ten minutes to re-type. The second is why the backup comes first; the same discipline applies before any control work, as in configuration backup before troubleshooting, and it is what saves you when a camera configuration is lost after a job load.

Step 1: regain keyboard control in setup

The trap: F1 gets you into setup, then nothing moves. Three causes, in order of likelihood.

First, the front-panel socket is wired for the machine's own membrane keypad, not a full keyboard. A handful of conductors carries the keypad, so a standard keyboard plugged there delivers little beyond the one key that got you in. Second, if you are using a USB keyboard, the settings reset switched USB keyboard support off, so it stays dead until setup says otherwise. Third, the navigation is unobvious: on setup screens of this era, Tab jumps between the major sections, the arrow keys move within a section, and Alt-H lists the navigation keys.

The fix that works immediately: plug a PS/2 keyboard directly into the motherboard, with a passive adapter if the socket is the older large DIN, and bypass the front-panel wiring altogether. Then set the date and time, set halt-on-errors to ignore keyboard errors so the machine boots whether or not it detects the front keypad, and restore the boot order. USB on industrial controllers has its own quirks once the OS runs; see USB UVC camera support on an industrial controller. The same failure on panel PCs is walked through in this industrial HMI CMOS battery diagnosis.

Step 2: copy everything while it still boots

Copy in this order: machine parameters, vision jobs, camera calibration files, camera user sets, saved as described in saving camera settings across power cycles with a user set, then a full drive image of the controller. On DOS-era controls the operating system and the control programs live on the same drive, and a rebuild without that image means re-creating them from a printed copy.

For a typical 2 GB legacy drive, the copy time by path is:

The arithmetic makes the decision: image the drive over the network or by pulling it, and use removable media only for the parameter set as an emergency copy. Keep the machine powered the whole time. Field reports are consistent on this point: controls that booted fine several times after the warning, then came up blank.

On camera systems, keep known-good cable in the service kit for the bench copy, from data cables for Ethernet and USB cameras:

Cable length and bandwidth limits on Ethernet camera links are covered in 5GBASE-T camera interface: bandwidth, cables and distance.

Step 3: find the battery, because it is not always a coin cell

An empty, silkscreened "Battery" footprint does not mean the board has no battery; it means the backup moved. Boards of this era use three arrangements:

Coin cell: swap the CR2032 for the same type only. Where the holder is reachable, many technicians fit the new cell with the system powered so the settings never see a gap; the how-and-whether is laid out in this write-up of replacing the CMOS cell with the system powered.

Encapsulated module: the DS1287/DS12887 family packs the clock, calendar, settings RAM and a sealed lithium cell into one module. Socketed versions pull straight out; note the orientation notch before lifting. Soldered versions are a different job, covered next. For module-level diagnosis, this dead DS12887 CMOS battery troubleshooting guide goes into detail.

Off-board cell: the 3.6 V lithium on a daughter or motion card backs that card's memory, not the motherboard's. Treat the two as independent circuits: replace the one that raised the warning, then cold-test both.

Soldered module: repair the board or replace it

Board repair shops that do this weekly report the same failure mode: lifting a soldered RTC module off a multilayer board breaks inner traces often enough, hot-air station or not, that a cell job becomes a board repair. Seating against the outermost slot leaves no working angle for the iron.

The decision:

• Socketed module, or soldered with clear access and a board you can afford to have down for an afternoon: replace the cell or module. Bench job.
• Soldered, cramped, and the machine earns money every hour it runs: swap the SBC or controller. Image the drive first, in every case.
• Middle path: leave the module in place and graft external lithium cells behind a series Schottky diode, a BAT54 being the usual choice, so the board's supply cannot back-feed the cells. Follow a published replacement guide for the module family instead of improvising the connection.

A controller swap is also the moment to give the rebuild clean power: an enclosed unit from power supplies, sized with margin, following the DC line fuse and noise filter practice for a 24 V vision system. At the camera end, settle USB versus external 5 V camera power while the wiring is open. For a 48 V rebuild:

The LRS-75-48 is an enclosed supply rated 48 V, 76.8 W.

Re-enter settings and re-align the camera

After the swap, re-enter the BIOS first: date and time, halt-on-errors, boot order, USB keyboard support. Then reload the control's parameters, vision jobs and camera user sets from your copies. If the camera never moved, calibration holds. If anything on the stand moved, put the camera back at its marked working distance before re-teaching; the 3D camera working distance and positioning guide covers the geometry. Mark the stand column and base with a pencil line before loosening anything.

Where re-teaching is routine, camera stands with X-Y travel turn alignment from an hour into minutes:

The LS-MNT-0220 is a 60 mm manual X-Y plus rotation stage.

Prove the fix before you close the cabinet

Two tests, in this order:

1. Set the clock, save, then switch off at the mains, not the front switch, since standby rails can hold memory, and leave it ten minutes. On power-up, settings and time must be exactly as left. A failure means the new cell, its holder contact or the module seat is bad; fix it before going further.
2. Keep the machine unpowered from Friday to Monday and repeat the check. A healthy cell holds for years; a marginal connection shows up as drift within days.

Then check the record-keeping: a control that ran with a dead clock has image files and logs dated to the default year. Anything that sorts or ages records by date needs a one-time correction, or the audit trail stays wrong.

When the usual advice is wrong

• "Reset the clock and keep going." Right for an office desktop, wrong for a control that keeps parameters in battery RAM. The gamble loses exactly when you have no copy.
• "The holder is empty, so there is no battery." Look for the encapsulated module or the daughter-card cell first.
• "The USB keyboard has failed." It is the settings reset switching USB keyboard support off, not the keyboard.
• "It has booted ten times since the warning." End-of-life cells are erratic; that is what end of life means. Act on the warning, not on the failure.

Checklist: CMOS battery low, in order

1. Leave the machine powered; do not cycle it to test.
2. Photograph every setup screen.
3. Plug a PS/2 keyboard into the motherboard; navigate with Tab, arrows and Alt-H.
4. Set date and time, halt-on-errors and boot order; save.
5. Copy parameters, jobs, calibration and camera user sets; then image the drive.
6. Open the copies on another machine before you power down.
7. Lock out, open the cage, identify which of the three battery arrangements you have.
8. Replace the cell or module, or make the board-swap call for a soldered module against the slot.
9. Re-enter the BIOS, reload parameters and jobs, confirm camera alignment.
10. Run the ten-minute cold test, then the weekend test.

If the swap route is yours, the system configurator steps through camera, light, controller and cabling choices, the systems topic page collects related practice, and you can send us the controller details to have a replacement controller and supply matched to your cameras.