Industrial PC no power after a CMOS battery swap: what to check first

You swapped the coin cell on a vision PC or robot-controller motherboard and now nothing responds: no fans, no POST, no pendant. Before ordering a board, work the power-on path in order, because most boards in this state are not dead. The sequence that recovers them: touch the front-panel PWR_ON pins together for about one second, return the clr_cmos jumper to its run position, prove the supply outputs, then set the BIOS auto-start option the battery swap erased.

Each check below gives the measurement to take and the decision it hands you, for ATX-style cabinet PCs and for 24 V-input industrial machines. The habit that prevents the whole event, capturing your settings before the cell goes flat, is covered at the end, alongside the guide on saving your data when a CMOS battery runs low.

Why is the board completely dead after a battery swap?

Because the swap cleared the board's stored power state, and the board is now waiting for a power-button event it never gets: on ATX-style boards the button only pulls the PWR_ON pins to ground for under a second, so no button means no start.

The power button on these boards is not a switch in the power path; it is a momentary request. Pressing it bridges two pins on the front-panel header, the logic releases the supply, and everything spins up. Controller cabinets rarely wire that button to the door. They rely on a BIOS flag that starts the board as soon as AC returns, and the battery swap plus the clr_cmos clear erased that flag. The board now sits in its off state, healthy and silent, indefinitely. Waiting a day changes nothing: this is stored logic state, not a charge that creeps back. The repair is to hand the board the button event it is waiting for, then put the flag back.

How do you start the board without a working power button?

Short the two PWR_ON pins on the front-panel header for about one second: that touch is electrically identical to pressing the power button, and it starts most boards stuck in this state.

1. Find the board model, silkscreened between the expansion slots or near the memory, and get its manual from the maker's site. The front-panel header page draws the pin block and names the power-on pair, marked PWR_ON on many industrial boards.
2. Leave the cabinet powered so standby is present. Touch both PWR_ON pins together with a flat screwdriver or a jumper cap for about one second, then lift off. Fans should move and POST should begin within a second or two.
3. The moment it starts, set the BIOS auto-start option covered below, or you will repeat this after every outage.
4. If you use a jumper cap, take it off again: a button held about four seconds or longer is read as a shutdown request on ATX-style boards.

The same fault and the same one-second header touch on a robot-controller PC are walked through in this controller PC boot failure after a CMOS battery swap account, including the BIOS change that makes the fix permanent.

Can the clr_cmos jumper stop the board starting?

Yes, a board left with its clear-CMOS jumper in the clear position often refuses to power on or POST at all, and it stays that way however long you wait; put it back to run before anything else.

If the battery change involved a jumper clear, check where the cap sits. The typical three-pin block runs pins 1-2 for normal operation and pins 2-3 for clear, and the manual states the labelling for your board. A cap left across the clear pair holds the RTC circuit in reset, which looks exactly like a dead board with a fresh battery in it. With AC off, set the cap to clear for five to ten seconds, return it to normal, then start from the header. While you are there, measure the cell across its faces: a healthy CR2032 reads about 3.0 V, and below about 2.8 V the low-battery behaviour returns.

How do you prove the power supply before blaming the board?

Measure the outputs: an ATX-style supply in standby shows 5 V on the purple 5VSB wire and starts when PS_ON (pin 16, green) is grounded, and its 12 V and 5 V rails must stay within ±5 % once running.

Grounding PS_ON to any black wire issues the same request the header makes, straight at the 24-pin connector, and the supply fan and the board should come alive. Measure while it runs: 12 V between 11.4 V and 12.6 V, 5 V between 4.75 V and 5.25 V. The purple 5VSB wire should show about 5 V even before that; with mains applied and no standby reading, the fault is the supply or its input side, not the logic.

On 24 V-input industrial machines, measure at the board connector under load. A reading that collapses when the board tries to start points at the feed or the wiring. Check the inline fuse, and treat an open fuse as a symptom: after board work, a blown DC fuse usually means a standoff short or a pinched lead, not old age. Size a replacement from the running current:

A worked example: LED lighting at 1.6 A, a vision controller at 1.2 A and a camera input at 0.7 A on a 24 V feed gives I = 1.6 + 1.2 + 0.7 = 3.5 A, so P = 24 V × 3.5 A = 84 W. The 125 % continuous-load margin puts the fuse at 1.25 × 3.5 A = 4.4 A, so fit a 5 A slow-blow, which sits 43 % above running current and rides the LED inrush. The DC line fuse and noise filter guide takes that arithmetic further, including filtering.

When the supply itself is the casualty, size the replacement from the label, not habit. The enclosed power supplies range covers the common cabinet voltages:

On a 48 V chain, start with the unit below. The LRS-75-48 prints 48 V and 76.8 W on its label. Dividing the two, 76.8 W ÷ 48 V = 1.6 A of continuous output; budget real load to about 1.3 A and the margin holds near 20 %.

Could the board have shorted on a standoff while it was out?

Yes, and it is the failure that actually kills boards: one extra or misplaced standoff under the PCB shorts VRM circuitry to the chassis, the supply folds back or never releases, and the board never wakes.

To reach the coin cell you took the board out or flexed it in the chassis, and on the way back a standoff in the wrong hole puts metal under a live trace. Bench-test before condemning anything: board on a non-conductive surface, only the CPU, one memory module and the supply connected, started from the header. If it runs stripped but dies in the case, the mounting is the fault. Refit with hardware only under the board's own mounting holes, and look for the screw that rolled under the PCB. Inspect the coin-cell holder too: a cracked clip from the swap leaves the cell floating, so the board boots but loses settings again, which is the failure that sends people in circles.

Why must you start it by hand after every power loss?

Because the BIOS option that made the controller start on its own, the restore-after-AC-loss setting, lived in the CMOS the clear wiped; set it back to Power On and the cell recovers itself after every outage.

The auto-start behaviour is one setting in the power or advanced menu, worded by board as restore on AC power loss, power on after power failure or similar, usually with Power Off, Last State and Power On as choices. For an unattended vision cell, set Power On: the controller restarts itself after every mains blip, and nobody walks the floor with a screwdriver. Last State suits machines that must resume only if they were running, but inspection cells that run the night shift belong on Power On.

What must you reset and re-check once it boots?

The clock, the drive mode and the camera links: BIOS resets to a default date and often to a different SATA mode, and cameras that enumerated before can sit undetected until the link is re-initialized, so work all three before declaring victory.

• Time and date: a reset clock breaks time-stamped logs and trips software whose license checks the date.
• Drive mode: a cleared BIOS often defaults the SATA mode away from what the operating system was installed with, and the board then reports no boot device. Set it back before reinstalling anything.
• Boot order: put the inspection system drive first.
• Camera links: if cameras that came up before now sit dark, run the camera power and detection checks before touching optics or software. On robot cells, the program that opens the camera at startup usually needs to re-initialize it after a controller reboot, as covered in camera connection after a power cycle.

If the board's own network ports died with it, a PoE card puts the machine-vision cameras back on a single add-in slot from the frame grabbers range:

Which fix matches your symptom?

Match the row and apply the fix: a dark standby rail is the input side, a lit standby rail that ignores the header is the jumper or a standoff, and a board that starts then dies about four seconds later has a held button.

How do you keep the next battery swap from stopping the line?

Photograph the BIOS screens and change the cell on a schedule, before the battery measures below about 2.8 V: settings you can re-type in minutes turn a surprise outage into a planned swap, and a CR2032 typically lasts three to five years in a cool cabinet.

• Capture boot order, SATA mode, the AC-loss option and the fan profile before any battery work, so the rebuild is typing, not archaeology.
• Keep one sealed CR2032 per cell on the spares shelf; change it at three to five years, sooner in a hot cabinet.
• Leave the AC-loss option on Power On and let the vision software auto-start with the system, so a flat cell costs settings, not production.
• After any clear: jumper to run, cell at about 3.0 V, header touch to start.
• Measure the 24 V feed and the fuse before condemning a board.

For the display-side machines, the industrial HMI CMOS battery failure walkthrough covers the same diagnosis where the cell backs an operator panel. And if the board stays dead with good rails and a clean bench test, the motherboard is the casualty: send the board model and your rail readings through the contact page and we will help you plan the rebuild.