Troubleshooting
Troubleshooting a Panel PC HMI Dark Screen After Reboot
Short answer
Treat the dark screen as a startup-chain fault and measure first: the 24 VDC feed at the HMI terminals under load, then the flashlight test. A faint image means the backlight path; no image anywhere with a good feed means board startup or graphics. The replaced SSD only rules out worn flash if the restored image itself was healthy.
On this page
- What the dark screen already tells you
- Measure the 24 VDC feed at the HMI terminals under load
- Worked example: the 24 VDC feed on a dark panel PC HMI at startup
- The flashlight test splits backlight failure from no video
- Splitting the panel from the board with an external display output
- Record the board alarm exactly
- What the replaced SSD does and does not prove
- The disabled graphics driver and heat
- Decisions by finding
- Questions asked
A panel PC HMI that freezes and then stays dark after a reboot has failed somewhere in the startup chain: the 24 VDC feed, internal power conversion, board startup, graphics output, the panel and its backlight, or the boot medium and operating system. Work that chain in order and the fault separates fast. Measure the input at the HMI terminals while the unit tries to start, not at the supply, then put a flashlight to the panel face. A faint image means the backlight path; no image with a good feed means board startup or graphics. The SSD you already replaced from an image narrows the storage question but does not close it, and a graphics driver disabled months ago is a condition to undo, not a fix to keep.
What the dark screen already tells you #
The earlier freeze and the later dark display are two observations of one machine, not one proven fault. A cursor that lags only across a serial link is a communication problem, the case covered in Why an HMI Freezes or Updates Slowly Over Serial; a cursor that freezes everywhere and is followed by a panel that never lights moves hardware ahead of any software diagnosis. Classify what you have before you open the cover.
| Observation | Where the fault lives | Discriminator |
|---|---|---|
| No board indication at power-up | Incoming feed or internal power conversion | Measure the 24 VDC feed at the HMI terminals under load |
| Board indication present, panel dark | Board startup, graphics, panel supply, backlight | Flashlight test, then a documented external display output |
| Image appears, then the system stops responding | Storage, operating system, driver, thermal | Drive health diagnostic and the Windows event log |
| Pointer freezes only after the unit warms up | Temperature-sensitive hardware | Inspect the cooling path and time the failure against temperature |
Expect a repeatable classification: either no electrical startup, board activity without a visible image, or an image followed by an operating-system stall. Each lands the fault in a different place.
Measure the 24 VDC feed at the HMI terminals under load #
The supply's label is a rating, not a measurement, and rapping the bezel proves nothing. The number that decides is the DC voltage present at the HMI input terminals while the unit is commanded to start.
- Read the printed input rating and range off the unit's nameplate. Do not substitute a rating remembered from another panel.
- With the meter on DC volts, measure across the power supply's output terminals while the panel tries to start, and write the number down.
- Move to the HMI input terminals and repeat in the same seconds. Wiring resistance, a loose terminal, or a collapsing supply only show up under load.
- A reading that holds at the supply but sags at the panel is a connection or conductor fault between them. A reading near zero at both points upstream to the branch breaker or feeder feeding the circuit.
- Measure with insulated probes and keep your other hand off the cabinet metal; isolate and lock out the feed before you reseat anything.
- With power isolated, reseat the supply connector and accessible board connectors and look for discoloration, bent contacts, and contamination. Let contact-cleaner residue evaporate fully before re-energizing, and remember that a cleaned terminal is not proven continuity: measure it again under load.
If the input sits inside its printed range under load and the board still repeats the same alarm state, the fault boundary has moved inside the panel.
Worked example: the 24 VDC feed on a dark panel PC HMI at startup #
Inputs, each with its source:
- Feed voltage class: 24 VDC, the standard industrial panel supply and the first thing to verify at the panel on this fault class.
- Accepted input range: 19.2 to 28.8 V, the range a typical 24 VDC panel nameplate prints. Read the one on your unit; its printed numbers govern.
- Reading at the supply output terminals during startup: 24.1 VDC, an example bench reading.
- Reading at the HMI input terminals in the same seconds, panel commanded to start: 18.9 VDC, same meter, same moments.
The rule in words: the reading that decides is the one at the panel terminals under load. It must sit inside the printed nameplate range, and the difference between the two readings, the drop, must stay under 5 percent of nominal, which is 1.2 V on a 24.0 V feed.
The arithmetic:
- Drop = 24.1 V − 18.9 V = 5.2 V
- Drop as a share of nominal = 5.2 V ÷ 24.0 V × 100 = 21.7 %
- Allowance = 0.05 × 24.0 V = 1.2 V, and 5.2 V exceeds it
- Lower limit check = 18.9 V is below the 19.2 V printed floor
The result: the supply holds 24.1 V, but the panel sees 18.9 V only while starting, losing 5.2 V, about 22 percent of nominal, between the two points and landing below the printed range.
The decision: the feed, not the panel, is the first fault. Isolate, reseat, and inspect terminals and any inline protector, verify the conductor size, and repeat the under-load reading. A drop above 1.2 V on a 24.0 V feed is a wiring fault whatever the panel does next, and nothing inside the panel is diagnosed until its terminals hold 19.2 to 28.8 V under load.
The flashlight test splits backlight failure from no video #
A dark LCD is not the same as a computer that is not running. The panel can lack backlight power while still producing an image, invisible at normal room light.
- Dim the room and press a flashlight against the screen face at a slight angle.
- Look for a faint, ghost-like version of the desktop or startup logo.
- A faint image identifies the backlight, its power supply, or the panel connection as the fault, with the computer quite possibly running; touch input may still be live, so treat the screen as active.
- No faint image, with a verified feed, keeps board startup, graphics, and internal power conversion in scope. Do not reinstall anything yet.
- Listen for storage and cooling-device activity only as an observation; sound is not proof of a successful boot.
Backlight failure that leaves a faint image is also how older operator terminals die on the bench, and the same face-of-panel procedure applies to dark-display backlight troubleshooting on an older operator panel.
Splitting the panel from the board with an external display output #
If the unit's documentation lists an external display output, connect a compatible display while the unit is powered down, then start it. Video on the external display with a dark integral panel isolates the fault to the internal display path: backlight, panel supply, or the panel connection. No output on either display, with the feed verified, keeps processor startup, graphics hardware, and power conversion in scope and points at board-level work.
If no output is documented for the exact unit, do not improvise adapters to guess; the flashlight result already carries the decision between panel path and board.
Record the board alarm exactly #
The alarm you found on the board under the cover is the fastest documented clue you have, but only if it survives your work. Photograph it before disconnecting anything, and copy every nearby identifier, the color, and whether the indication is steady or flashing, exactly as printed. Decode the pattern against the unit's own manual, not against a similar model. A reseated memory module and a cleaned terminal are reasonable hygiene, but neither answers the power question, and an alarm state that repeats with a verified feed confirms the fault is inside the enclosure.
What the replaced SSD does and does not prove #
Restoring an image to a new SSD and still getting a dark screen tells you two things. It lowers the odds that worn flash was the sole cause, and it does not validate the image: a clone reproduces corrupted system files, broken configuration, and disabled drivers along with everything else. A successful clone is not a health test.
- Identify the boot medium by its label and interface before drawing conclusions. The clicking-drive test applies to rotating hard disks only, where steady clicking indicates mechanical failure; an SSD is silent.
- Preserve the original medium and its image before any further write operations. Give it the discipline you give control logic and back up a PLC program the right way before a write touches it.
- Once video is back, read the Windows event log for disk, file-system, display-driver, thermal, and unexpected-shutdown events at the timestamps of the freezes, and run the drive manufacturer's health diagnostic on the result it reports.
- When testing software, distinguish your restored field image from a known-good recovery baseline for the exact unit: the first reproduces the existing state, the second tests whether that state caused the failure.
The disabled graphics driver and heat #
Two background conditions from this machine's history shape the endgame. A graphics driver disabled months ago is a diagnostic state, not a repair; the panel has been running without accelerated video, and the disabled driver removes your fallback path when startup fails. Once you can see video, install the driver package approved for the exact hardware and operating-system image, taken from the machine documentation or recovery package, then check Device Manager for unresolved display hardware.
Heat produces intermittent freezes before it produces a no-start. Inspect the processor and graphics heat-sink assemblies, the airflow path, and vents, and remove dust without driving it deeper into the unit. Replace thermal compound only if the heat sink comes off or the interface is visibly degraded, and restore uniform mounting pressure. Start from cold, note when video appears, and run the unit to normal operating temperature while watching for cursor delay or corruption. Video that appears cold and dies warm is a thermal fault: fix the cooling path before replacing boards.
Decisions by finding #
- Terminal reading under load below the printed lower limit, or feed drop above 1.2 V on a 24.0 V feed: fix the feed and wiring first; nothing inside the panel is diagnosed until then.
- Faint image on the flashlight test: backlight or panel-supply repair; leave storage, Windows, and drivers alone.
- Documented external video good, integral panel dark: internal display path; weigh repair against planning a replacement terminal from an older operator panel.
- No image on panel or external output with a verified feed: board startup, graphics, or power conversion; board-level repair or replacement.
- Image appears and then the system stalls: storage, operating system, driver, or thermal; event log and drive health before more parts.
Questions asked
How far can a 24 VDC panel feed sag before the reading counts as a fault?
Use the printed range on the nameplate; most 24 VDC industrial panels accept 19.2 to 28.8 V at the terminals. Any under-load reading below the printed lower limit is a feed fault, and so is a drop of more than 5 percent of nominal, 1.2 V on a 24.0 V feed, between the supply and the panel. Fix the wiring before opening the panel itself.
A flashlight shows a faint desktop on the dark panel. Is the computer dead?
No. A faint image means the panel is generating video and the fault sits in the backlight, its power supply, or the panel connection. The computer may have been running the whole time, and touch input can still be live. Repair the backlight path before touching Windows, the boot drive, or any driver.
Is a clicking boot drive proof that the drive is dead?
Only for a rotating hard disk, where steady clicking indicates mechanical failure. An SSD has no moving parts and gives no audible warning, so identify the medium by its label and interface before listening for anything. On an SSD, run the manufacturer's health diagnostic and read the Windows event log for disk errors instead.
Does replacing the SSD from an image prove the drive was not the problem?
It proves the old drive's flash was not the only fault only if the image itself was healthy. Cloning copies corrupted system files and broken configuration along with everything else, and a successful clone is not a health test. Compare against a known-good recovery baseline before spending more on storage.
The graphics driver was disabled months ago and the unit worked. Can it stay that way?
No. A disabled driver is a diagnostic state, not a repair, and it removes your fallback video path the next time startup fails. Once you can see video again, install the driver package approved for the exact hardware and operating-system image, then check Device Manager for unresolved display hardware before the panel goes back to service.