A dead PLC rack produces a very specific kind of panic, because everyone's mind goes straight to the most expensive component: the processor, and with it the program, and with it the question of whether the last offline save was as recent as everyone hoped.
Take a breath. The single most common cause of a dark rack is the rack power supply — a component full of aged electrolytic capacitors doing hard thermal duty for a decade or three. Processors mostly don't die; supplies mostly do. As of August 2026, that ranking has not changed in the twenty-plus years these racks have been in service, and it is the reason the ten-minute check below exists.
The pattern to recognize
Rack supply failures rarely go from perfect to dead in one step. The classic progression:
- Faults when warm. The rack comes up fine in the morning, then faults mid-shift once the enclosure heats up. Aged capacitors lose capacity with temperature.
- Random processor faults / brownout behavior. Output ripple grows until logic browns out — which looks exactly like a "flaky CPU" and gets misdiagnosed as one constantly.
- The hard-down morning. Eventually it won't start at all. Dark rack, no LEDs anywhere.
If your "CPU problem" has a thermal or time-of-day pattern, suspect the supply first.
Why supplies are the wear item
The physics is not mysterious, and it is worth thirty seconds because it explains both the failure and the repair. A switching power supply lives and dies on its electrolytic capacitors: sealed cans of wet electrolyte doing filter duty at elevated temperature, around the clock, for decades. Heat dries them; equivalent series resistance climbs; ripple grows; regulation sags. The reliability literature published through IEEE has identified electrolytic capacitor aging as the dominant wear-out mechanism in power conversion for as long as the field has measured it, and EPRI work on industrial equipment reliability lands on the same component class.
The practical consequences cut both ways. It means the supply was always going to be the first thing in the rack to fail — no one mistreated anything. And it means the failure repairs cleanly and durably, because replacing every electrolytic with quality low-ESR parts rated for the thermal environment gives back a supply whose wear items are new. That is the whole logic of the flat-price repair.
It also means every other supply of the same vintage in the plant is on the same clock. The Department of Energy has documented for years how much of US manufacturing runs on control equipment decades past its design life; capacitors in those cabinets do not care that the strategy is "run to failure," they simply schedule the failure for you.
The 10-minute check
- Look at the supply's own LED (1746 and 1756 supplies have one). Dark or flickering with good incoming power = verdict nearly in.
- Verify incoming power at the supply terminals — confirm your 120/240VAC (or 24VDC) is actually present and stable. Rule out the upstream breaker, fuse, and terminal connections.
- Swap test if you have a spare. Racks make this easy: power down, swap the supply, power up. Five minutes to certainty.
- No spare? Measure the backplane. With appropriate care and the manual's pinout, sagging DC on the backplane with good AC in convicts the supply.
One warning: a supply that's sagging can also be dragged down by a shorted module. If a replacement supply immediately behaves the same way, pull modules one at a time to find the offender — that failed module is usually repairable too.
And one safety note that belongs in the procedure rather than the fine print: this is live-cabinet electrical work. Verifying incoming power happens under your plant's lockout/tagout discipline per OSHA 1910.147 where de-energized, and under the shock and arc-flash boundaries of NFPA 70E from the National Fire Protection Association where measurement requires the panel live. A dark rack is an inconvenience; a shortcut in a 480V cabinet is a different category of morning.
| What you observe | Most likely verdict | Next move |
|---|---|---|
| Supply LED dark, incoming power verified good | Supply failed | Repair path — $349 flat on 1746/1756 families |
| Rack faults mid-shift, fine in the morning | Supply capacitors aging | Repair before the hard-down morning picks itself |
| Random CPU faults, no pattern in the logic | Supply ripple browning out the rack | Test/swap the supply before condemning the CPU |
| Spare supply fixes it | Original supply confirmed | Repair the original, keep it as the spare |
| Replacement supply sags the same way | Shorted module dragging the bus | Pull modules one at a time; the offender is repairable |
| Supply verified good, rack still dark | Backplane or CPU territory | Test-and-verify before buying anything |
The economics are lopsided
- SLC 500 (1746-P1/P2/P4) and ControlLogix (1756-PA72/PB72/PA75) supplies: our flat repair is $349, including rectifier/switching-stage repair, a full recap, fan replacement, and a real load test at rated output. 24-month warranty.
- A used supply of the same vintage runs $60–$250 — carrying the same 20-year-old capacitors that just failed in yours. You're buying someone else's countdown.
- A new 1756 supply, where still sold, is many hundreds of dollars with lead time.
For obsolete families this gets starker: SLC 500 is in Rockwell's long wind-down, and Siemens S7-300 hit type-discontinuation in October 2025. Repair is how those racks stay alive. The wider version of that calculation — downtime cost against lead time against sticker price — is the subject of the repair-versus-replace math guide, and it favors the bench more lopsidedly here than almost anywhere else, because the failed component is cheap to fix and the thing it protects is irreplaceable.
Add the labor arithmetic and it gets worse for the replacement path: with skilled maintenance labor running well north of fifty dollars an hour fully loaded on Bureau of Labor Statistics figures, the hours spent sourcing, questioning and commissioning an unknown used supply routinely cost more than the flat repair of the known one.
The full scope, family coverage and turnaround for this work live on the PLC repair page, and the same recap logic applies to standalone units via industrial power supply repair. Standard bench time is 3–5 business days; rush service exists for the mornings when the rack picked badly.
And if it really is the CPU
Straight talk, because this is where some shops get sold false hope: most dead PLC CPUs have no economic component-level repair path. They're proprietary silicon and memory. What's honestly available:
- Test-and-verify — we confirm whether the CPU is actually dead or the victim of the supply/backplane (it's the victim surprisingly often).
- Battery and storage care on families where that's the real issue.
- A working used CPU + your program backup as the recovery path when it truly is gone.
Which is the last point worth repeating to every maintenance team: the $349 supply repair is routine — if the program exists offline. Verify today that your ACD/RSS/project files are current and reachable — the NIST guidance on operational technology says the same thing about configuration backups at book length, and CISA advisories on industrial control resilience repeat it for a different threat model. Where your program actually lives, family by family, is covered in will my program survive the repair. The rack will eventually pick a morning to test you on it.
When that morning comes, start a repair online with the symptom and the family, or request a quote if you want a technician's read first — the evaluation costs nothing either way.
Frequently asked questions
My PLC rack is completely dark. Is the processor dead?
Probably not — the rack power supply is by far the most common cause of a fully dark rack, and processors are the victim of a failed supply far more often than they are the culprit. Check the supply's own LED against verified incoming power, or swap in a known-good supply if you have one; five minutes usually answers it. Condemning the CPU is the last step of this diagnosis, not the first.
What does it cost to repair a PLC rack power supply?
As of August 2026, our flat price on the common Allen-Bradley families — SLC 500 1746-P1/P2/P4 and ControlLogix 1756-PA72/PB72/PA75 — is $349, which covers rectifier and switching-stage repair, a complete recap with quality low-ESR capacitors, fan replacement and a genuine load test at rated output, with a 24-month warranty. That is against used units of the same vintage that carry the same aged capacitors, and new units that are several times the price where they still exist.
Why does my rack fault only in the afternoon?
Because aged electrolytic capacitors lose capacity as they warm up. A supply that is marginal at morning cabinet temperature falls below regulation once the enclosure heats through the shift, and the rack browns out or faults. A thermal or time-of-day pattern in "CPU problems" is one of the most reliable fingerprints of a dying supply, and it is the stage at which the repair is a scheduled event rather than a production stoppage.
Could a bad module be killing my power supply instead?
Yes, and it is the trap in the swap test: a shorted I/O or comm module drags the backplane down and makes a healthy replacement supply look as dead as the original. If a known-good supply misbehaves the same way, pull modules one at a time until the bus recovers — the module you remove last is the offender, and it is usually repairable at board level too.
Is a used replacement supply a reasonable fix?
As a get-running-today measure, sometimes; as the fix, no. A used supply of the same generation carries the same twenty-year-old electrolytics that just failed in yours, so you are installing someone else's countdown. The durable pattern is the one running plants converge on: repair the failed unit — which returns it with new wear items — and keep the used spare on the shelf for the next event.
Does this apply to Siemens, Mitsubishi and other rack families too?
Yes — the architecture is universal even where the part numbers differ. S7-300 and S7-400 rack supplies, Mitsubishi Q-series, GE 90-30: every one of them is a switching supply full of electrolytics feeding a backplane, every one follows the same warm-fault-first progression, and every one is cheaper to recap than the processor it protects is to replace. The check sequence transfers unchanged; only the LED locations move.
What should I do before the CPU ever dies?
Verify today that your offline project files — ACD, RSS, whatever your family uses — are current, dated and reachable by someone other than the one person who knows where they are. A dead CPU with a good backup is a recoverable event; a dead CPU holding the only copy of the program is a crisis. The supply repair protects the rack; only the backup protects the logic.
Have this exact problem?
Start a repair or get a free technician review — we'll tell you honestly whether it's worth fixing.