>
>
2026-05-13
What kind of failures might these circuit boards experience after 15-20 years of use? How can they be diagnosed on-site? And which replacement part numbers should be looked up? The Fanuc 18i-MB has been running in the workshop since the early 2000s. Most of the controllers we see sent for repair today have been running continuously for over 15 years, and after handling dozens of units, the failure modes are fairly predictable. The power board is the first place to show signs of aging—electrolytic capacitors dry out, fans seize up, DC voltage rails begin to drift, while other parts of the CNC remain unaware. This article will discuss which components are prone to failure, what to check first, and which part numbers are applicable.
Inside the 18i-MB control unit (the A02B-0283-B series), the power supply board takes in 24 VDC from the external transformer and steps it down to the multiple lower DC rails the CNC needs internally — typically +5 V, ±15 V, +3.3 V, and battery-backed memory voltage. It feeds the main board, the FROM/SRAM cards, the FSSB card, the graphics card, and the LCD/MDI interface.
It's not a heavy-duty unit like the alpha or alpha-i PSM modules that power the servo drives — those are separate, much larger devices (A06B-6077, A06B-6087, A06B-6110 series). The power supply board inside the 18i-MB control is the small DC-DC converter that keeps the CNC brain alive. When it fails, you typically get either no display at all, intermittent boot failures, or random system alarms.
The power supply board on the 18i-MB is the internal DC-DC converter — separate from the alpha PSM that drives the servos.
The 18i-MB system as a whole carries an A02B-0283-B series designation — the third group of digits varies by hardware option (drive count, language pack, network options). A common full system part number you'll see in the field is A02B-0283-B503. Inside that controller, the power supply boards you'll encounter most often are from the A20B-8101 family. These are the boards that fail and need replacing.
| Part Number | Function | Typical Fitment |
|---|---|---|
A20B-8101-0285 |
Power supply module board | 18i-MB control unit, mid-late production |
A20B-8101-0011 |
Power supply module board | 18i-MB / 18i-TB common variant |
A20B-8101-0180 |
Power supply board | Earlier 18i series controllers |
A20B-8101-0191 |
Power supply board | Earlier 18i / 21i variants |
A20B-8100-0135 |
Main board (LCD-mounted) | 18i-A — often confused with power supply |
A02B-0283-B503 |
Complete 18i-MB control module | Full system part number, not the PCB alone |
Part numbers verified against the published Fanuc 16i/18i/21i hardware catalogue and current distributor listings (DNC Electronics, MRO Electric, CNC Electronics, Fanucworld). The 14th and later digits of the part number encode hardware revision — confirm the exact suffix on the failed board before ordering a replacement.
A20B-8101-0285/01A or similar. The slash and the characters after it are the hardware revision — different revisions are generally interchangeable, but if you're sourcing a refurbished board, asking for an exact revision match avoids surprises.The failures cluster into four buckets, in roughly this order of frequency:
This is by far the most common cause of failure on any Fanuc CNC power supply board built before about 2010. Aluminium electrolytic capacitors have a finite lifetime — typically rated at 5,000 to 10,000 hours at 105 °C, which translates to roughly 15–25 years at typical control cabinet temperatures (40–55 °C internal). The electrolyte slowly evaporates through the rubber seal. Capacitance drops, ESR rises, and the DC rail ripple becomes excessive.
You'll see symptoms like the controller booting intermittently, random alarms during heavy machining (when the servo load draws transient current), or the LCD flickering. On the bench, the failed capacitors are often visually obvious — bulged tops, leaked electrolyte residue around the base, or domed pressure relief vents on the larger units.
The controller cabinet fan isn't on the power supply board itself, but when it fails, the board takes the punishment. Internal temperatures climb from a normal 45–55 °C to 70 °C or higher. Capacitor lifetime halves for every 10 °C rise — so a controller that should have given another 5 years of service might fail in 6 months once cooling is compromised. Always check the cooling fans first when you see thermal-pattern damage on a board.
Less common, but it does happen. Through-hole connectors that carry the input 24 V or feed the heaviest output rails can develop hairline cracks at the solder joints after years of thermal cycling. The symptom is intermittent — the board works fine at room temperature, fails after the machine warms up, then recovers when cooled down. Reflow with fresh solder usually fixes it; reflowing without cleaning and using flux just spreads the problem.
The DC-DC regulator ICs and their associated MOSFETs are the active components doing the actual stepping-down. They can fail outright (short circuit, no output) or partially (output voltage drifts low). On older boards, the +5 V rail dropping to 4.6–4.7 V is a common partial-failure mode — it's still high enough that the LEDs light up, but not high enough for reliable CMOS logic, so you get unpredictable behaviour.
| Failure Type | Frequency | Symptom Pattern | Fix |
|---|---|---|---|
| Electrolytic capacitor aging | Very common | Intermittent boot, ripple-related alarms, visible bulged caps | Recap with same-spec or upgraded long-life capacitors |
| Heat damage (fan related) | Common | Discoloration on board, premature cap failure | Replace fan + affected components; check ambient temp |
| Solder joint fatigue | Occasional | Intermittent, temperature-dependent failure | Inspect under magnification; reflow joints with proper flux |
| Regulator IC failure | Occasional | Specific rail out of spec or absent | Component-level replacement; needs schematics or board swap |
| MOSFET / IGBT short | Rare | Board drawing excessive input current, fuse blown | Component replacement; often easier to swap board |
Not always, but it's near the top of the suspect list once you've ruled out the obvious software causes. A controller that freezes mid-cycle ("dead halt", "system locked") can have several root causes, and in our repair workflow we walk through them in this order:
| Cause | How to spot it | Power supply link? |
|---|---|---|
| Power instability / line transients | Freezes correlate with shop loads switching on (large motors, welders nearby) | Yes — failing caps can't filter line noise effectively |
| Controller logic / watchdog faults | Reproducible at a specific G-code or operation; consistent alarm number | No — software/parameter issue |
| Program or parameter error | Same freeze on same program; cleared by editing the program | No |
| Electromagnetic interference (EMI) | Random freezes during high-current events; worse when nearby drives are loaded | Partial — see Section 7 on grounding |
| Overheating (thermal) | Freezes only after long runs; clears after the cabinet cools down | Yes — heat accelerates power supply degradation |
| Mechanical / sensor faults | Freezes triggered by axis events, with related servo alarms | No — separate servo system issue |
If the freeze is random, correlates with heavy machining load, or appears after the machine has been running for 30+ minutes — and especially if a controller power-cycle clears it — that's the signature of a tired power supply. Failing capacitors can't hold the rail up under transient current draw from the servo and spindle interfaces, the CPU sees a momentary brownout, and the system halts.
A useful field test: after a freeze, open the cabinet immediately and feel the power supply board. If it's noticeably hotter than the surrounding boards, or if you can smell anything (ammonia from leaked electrolyte, or the slight burnt-plastic smell of stressed components), the power supply is the prime suspect.
The 18i series has hundreds of alarm codes, but only a subset points directly at power supply problems. The codes below are the ones we see most often when the power supply board turns out to be the root cause. None of them is a guaranteed power supply fault on its own — they each have other possible causes — but if you're seeing two or more of them together, the power supply moves to the top of the suspect list.
| Alarm | Description | Power Supply Connection |
|---|---|---|
910 / 911 |
SRAM parity / DRAM parity error | Rail voltage drops cause memory corruption |
920 |
Servo alarm — watchdog or RAM parity | Often power-related if intermittent on startup |
930 |
CPU interrupt — undefined interrupt | Voltage instability can trigger spurious CPU interrupts |
701 |
Overheating — fan stopped | Direct: fan failure leads to power supply damage |
401 |
Servo amp not ready | If multiple axes affected simultaneously, check control PSU first |
414 |
Servo alarm — digital servo | Sometimes power-related; usually points to amplifier itself |
| No display, fans run | Controller appears dead | Classic power supply failure pattern — main board is starved |
| Controller boots, then drops out | Random reboot under machining load | Caps failing under transient load; rail collapses during heavy current draw |
Alarm code descriptions referenced from the official Fanuc 16i/18i/21i alarm code list. Power supply attribution based on field repair experience — these alarms have multiple possible causes and should be diagnosed in context.
About 70% of suspected power supply failures can be confirmed or ruled out with a multimeter and 10 minutes of careful checking. The other 30% need the board on the bench. Either way, do this in order:
Sometimes, yes. We've had cases where a "failing" power supply board turned out to be perfectly healthy — the real issue was electromagnetic interference or grounding faults that confused the symptoms. A few things worth checking before condemning the board:
Grounding integrity. The 18i-MB cabinet must have a clean, low-impedance ground bond — typically less than 100 mΩ from the controller earth terminal to the building ground. Loose grounds or ones that have corroded over the years let line noise couple into the controller via the chassis. The symptoms — random reboots, communication errors, intermittent SRAM/DRAM alarms — look exactly like a failing power supply. A 5-minute ground bond check with a milliohm meter can save hours of board-swapping.
Nearby high-current switching. If a contactor, VFD, or large motor switches close to the controller — especially on the same incoming power feed — voltage transients can propagate into the 24 V rail feeding the controller. The power supply board's input filter is supposed to handle this, but after 15+ years its filter capacitors are tired and the protection drops off. Adding a separate isolation transformer for the controller (or moving the offending drive to a different feed) is sometimes a more effective fix than replacing the board.
External signals to the controller. The 4–20 mA analog inputs, encoder feedback lines, and DI/DO signals routed into the 18i-MB can carry induced noise if the cabling isn't properly shielded or grounded at one end only. Repeated noise spikes stress the input protection circuits on the power supply and main boards, accelerating failure.
Related component aging. When the controller is 15+ years old, the power supply board isn't the only thing tired. The backup battery (memory backup) usually needs replacement every 2–3 years and is often overdue. The cabinet fan bearings dry out. Connector contacts oxidise. We've seen jobs where replacing the power supply alone didn't fix the problem — the new board went back into the same overheating, vibrating, dirty cabinet and failed within 12 months. Address the environment, not just the board.
Three realistic paths. Each makes sense in a different situation.
| Option | When it makes sense | Typical lead time | Watch out for |
|---|---|---|---|
| New genuine board | High-value machine, long expected remaining service life, downtime cost is high | Often 2–6 weeks (Fanuc is winding down 18i-MB production) | Genuine new stock is increasingly hard to source. Watch for counterfeit parts sold as "new" |
| Refurbished board | Standard machine, moderate downtime tolerance, cost-sensitive | 1–2 weeks, sometimes overnight | Verify the refurbisher tests under load, not just power-on. Insist on at least 90-day warranty |
| Component-level repair (recap) | You can spare the board for 5–10 days, machine still has years of life ahead | 5–10 working days | Quality of work varies hugely. Choose a shop that uses 105 °C / long-life capacitors, not bargain caps |
Counterfeits and "Chinese clone" boards have appeared more often as the 18i-MB ages out and genuine inventory dries up. The reliable indicators:
Part number label. Genuine labels are crisp, high-resolution, and the typography is consistent across the whole label. The barcode scans cleanly and the format matches Fanuc's documented coding scheme. Counterfeits often have slightly off-spec labels — wrong font, blurry barcode, or "made in" markings that don't match Fanuc's actual manufacturing locations.
PCB silkscreen quality. Genuine boards have sharp, evenly applied silkscreen. The Fanuc logo and revision markings are positioned consistently. Counterfeits often have rough silkscreen, off-centre printing, or visible artifacts from a lower-quality printing process.
Component markings. The ICs on a genuine Fanuc board carry markings from major manufacturers (Toshiba, Mitsubishi, NEC, Renesas, etc.) with date codes consistent with the board revision. If you see relabeled ICs, sanded surfaces under the labels, or date codes wildly inconsistent with each other, you're looking at remanufactured or counterfeit work.
Test before installing. Whatever you buy, power it up on a test rig before committing to a production install. A bench test with proper loading exposes most problems within 30 minutes.
Four things, ranked by how much trouble they save:
1. Clean the cabinet filters and check the fans every 6 months. This is the single most cost-effective maintenance task. Clogged filters and slow fans drive cabinet temperatures up by 10–15 °C, which roughly halves the remaining capacitor life. A 5-minute filter clean and a quick fan check (spin freely, no bearing noise) prevents most thermal failures.
2. Log and trend cabinet temperatures. If your maintenance system can log the cabinet temperature alarm (or you install a simple data logger), a slow upward trend over months is an early warning that filters are loading up or a fan is degrading — well before a 701 overheat alarm shuts the machine down.
3. Plan a preventive recap at 15–18 years. If your 18i-MB is approaching this age and is on a critical machine, scheduling a planned capacitor replacement during a maintenance window is far cheaper than the emergency call after it fails on a Friday afternoon. Budget around 1 day of downtime plus the recap labour.
4. Keep one tested spare on the shelf for each critical machine. A refurbished, bench-tested power supply board costs a fraction of a single day of unplanned downtime on a busy CNC. For a small fleet, having one shared spare per controller type is realistic and worth it.
The 18i-MB power supply board isn't a complicated piece of hardware, but it's the part of the controller that ages most predictably. Capacitors dry out, fans wear down, and intermittent symptoms slowly turn into hard failures. Most of the failures we see can be diagnosed in under an hour with basic tools, and most can be addressed by either a quality recap or a properly tested refurbished board.
What you want to avoid is the panic order at 2 AM on a production night. Knowing the part number on your specific controller, having a vetted source ready, and keeping a tested spare on the shelf for high-value machines is far cheaper than the alternative.
If you need a replacement Fanuc 18i-MB power supply board, share the exact part number from your existing board (including the revision suffix), plus the complete A02B-0283-B system designation. We supply genuine Fanuc PCBs and tested refurbished units with documented warranty terms.
যে কোনো সময় আমাদের সাথে যোগাযোগ করুন