Why did the panel stop the job, and why that code and not another one? Answer that before you pick up the phone. The code itself is a location. It tells you which watchdog tripped, not which part died. Two minutes of reading the history will usually separate a dirty strip from a dead board, and that difference is most of the repair bill. Pull the log, write the codes down in the order they happened, then run the checks below from free to expensive. Every step is something you can do at the machine with a flashlight, a lint free wipe, and a phone camera.
One rule before you start: change one thing per test. Two parts fitted at once and you have learned nothing, because the code that finally stops cannot be credited to either one. Keep a pen taped inside the cabinet door and date every entry you make. Six weeks of that page tells you more about this machine than any forum thread.
Start With the Error History, Not the Code on the Screen
The single code on the display is the last thing that happened, not the whole story. Most production machines keep a log behind a service or information menu. On Mimaki units you reach it from local mode through the function key into maintenance, then information, then error history. The UJF platform needs firmware 4.00 or later for that entry to exist. On the JV5 generation the information key goes straight in. The screen alternates the timestamp and the description, with the lowest index number being the most recent event.
Two more lists sit next to it and both are worth a look. The maintenance history records when cleaning actually ran, manual and automatic. The print mode history records resolution, pass count, scan direction, and job length for recent work. When a code only appears on long jobs at high pass counts, that history is the fastest way to prove the pattern. The log also rolls over when it fills, so a fault you shrugged off last month may already be gone. That is one more reason to shoot it while it is still on the screen.
Photograph the screen instead of writing it on your hand. Shoot the whole list, not just the top entry, then power down normally. Some machines wipe the log on a hard power cut, and a code you cannot quote is a code the service desk will not act on. Do this one thing even if you plan to keep printing: the log is the only record of how often the fault has already happened.
Why the Code Changes Every Time
A rotating set of codes is the most misread situation on the floor. Operators see encoder errors, then a printhead error, then a maintenance cartridge fault, and assume three parts are failing. They are usually watching one failure being reported from three different checkpoints.
The mechanism is simple once you see it. The machine cannot finish its homing or positioning routine, so it backs out and tries a different recovery path. Each path fails at a different test, and each test has its own error number. The carousel of codes is the recovery sequence, not a fault list. Clean the strip and the codes keep coming, because the strip was never the whole problem.
The tell is timing. If most of the entries cluster at power on, after a job ends, or whenever the carriage returns to the home position, treat it as one positioning or initialization fault and test the whole chain once. If a single code repeats at random moments mid job with the carriage already running, that code is probably real and local.
Here is the check worth doing before anything else. Power down through the normal shutdown so the carriage parks, then confirm the carriage slides freely by hand if your manual permits it. A stiff rail, a dragging ink tube bundle, or a wiper stuck above the platen will stall homing and generate exactly this kind of code rotation. Fix the drag, and the codes stop without a single part purchase.
Read the Code as a Location, Not a Sentence
Start with the words, not the digits. A lot of panels name the subsystem right on the line: a motor axis, a head data channel error, an ink end condition. Those are the easiest codes to work with because they already tell you which wire or line to follow. Write the word down before you look at the number.
Pure numeric codes carry structure. On several Epson wide format families the leading block of a long code points at the controller and memory area, the middle block at position sensors and servo parameters, and the tail at head driver temperature and interface faults. Do not carry that map to another brand. Each maker numbers its own way, and a code borrowed from a forum post about a different machine will send you to the wrong end of the printer.
Sort every code you collect into four buckets. Feedback covers anything about where the carriage or media is. Recognition covers chips, cartridge ID, and ink type mismatches. Fluid covers pumps, valves, tubing, and recovery. Heat covers heaters, thermistors, and warm up behavior. A fault that shows up twice in the same bucket means work that bucket's chain. Buying a part from a different bucket because the code looked similar is how a two hundred dollar repair becomes a two thousand dollar one.
Keep the bucket list on the wall behind the machine. After a few months it tells you which subsystem on this particular unit is the weak one, and that is the one worth keeping a spare for.
The Cheap Checks That Clear Most Codes
Work down this list in order. Stop when the code stops.
- Shut down through the menu, let the carriage park, then cut power. Forcing a carriage by hand while it is still locked can damage the belt, the carriage lock, the capping station, and the sensor bracket in one motion.
- Look at the encoder strip with a bright light before touching it. Ink mist and dried coating show as a dull band or a cloudy patch. Fingerprints show as a smear. Hold it against the light and look for a bright scratch line, which means the strip is finished and no amount of cleaning will help.
- Clean with a lint free wipe or a foam swab dampened with distilled water, or with the cleaner your manual names. Never use paper towels, cotton balls, shop rags, or compressed air. Fibers lodge in the sensor slot and air pushes debris deeper into it. Do not put a swab, a wipe, or a fingernail inside the sensor slot.
- Support the strip from behind with a clean dry finger and wipe along a short section with almost no pressure. Do not pull it, stretch it, or bend it. A stretched strip reads wrong even when it looks clean.
- Check that the strip is actually sitting back in the sensor slot when you are done. Ricoh's own knowledge base warns that a strip left outside the slot reports an encoder failure, and that is a code you created yourself.
- Reseat the carriage ribbon cable and every sensor connector you can reach without tools. A half inserted plug produces intermittent codes that look like a failing board and disappear after one push.
- Confirm whether a waste ink or maintenance box warning comes from a real level sensor or from a counter. If it is a counter, the box may be empty and the machine still stops.
Finish with a test that produces evidence. Print a small bidirectional alignment pattern and a color block on scrap media. If the defect shifts or clears after the strip cleaning, you have proved the feedback path was part of the problem. If nothing changes, you have eliminated the cheapest cause and earned the right to look further up the chain.
When the Code Is Telling the Truth
Use one question to decide between a part and a position. Move the suspect part to another slot, or fit a known good one, and watch where the code goes. A fault that follows the part is a failed part. A fault that stays in the original slot is upstream: ink supply, wiring, connector, or mechanism. That single test settles most arguments about whether the head, the damper, or the board is at fault, and it costs nothing but a few minutes.
Stop and call service when three conditions line up. Cleaning changed nothing, the same code returns at the same point every time, and a known good part still reports it. At that point you are looking at a board, a harness, or a mechanism that needs a trained hand, and another parts order only adds days.
Hand the technician a written sheet, not a description. List every code with its timestamp, what was on the machine when it happened, the media and print mode, the room temperature, which parts you already swapped, and what you cleaned. A tech with that sheet arrives with the right part. A tech without it arrives with a diagnostic meter and an hourly rate.
Count the cost in output, not in parts. Multiply your normal hourly output by the hours the machine is actually down, then add the media scrapped on the aborted roll, the expedited freight, and the call out fee. A code chased for two hours and a code chased for two days differ by a full production run, and that number is what makes the decision for you. The log you photographed at the start is what keeps that number small.