SMT Feeder Repair Acceptance Test: 5-Step SOP (2026)

08:05 on a Tuesday in Suzhou. An EMS plant takes delivery of 30 Yamaha CL feeders back from a local repair shop. Every one of them is tagged “repaired.” The line supervisor mounts the first feeder on a YSM20 and jogs it. Inside five minutes the machine has thrown 11 pick errors, and the line is down for the second time that week.

This is not a rare story. Merex engineers have been called out to three near-identical incidents in the last 12 months, and the root cause is always the same: nobody ever defined what “repaired” means.

The distance between “it was repaired” and “it can go back on a production line” is where line-stop minutes, mispicks and 01005 defects are born. This guide closes that distance with a procedure you can actually run: five steps, measurable limits, a form you can print, and the blind spots that most repair shops miss.

By the end you will have:

• A five-step acceptance SOP for repaired feeders
• Brand-specific limits for pick error rate and placement deviation
• A printable acceptance sheet for the line and for purchasing
• The four blind spots that let bad feeders through

---

1. Why “Repaired” Is Not the Same as “Production-Ready”

A feeder is a miniature precision mechanism. Gear train, ratchet, tape clamp, spring, sensor, ID chip, cover, base plate, tape inlet, reel. Several hundred components in most models, all of them working against tolerances tight enough that a worn spring changes placement accuracy.

A repair normally replaces a subset of that: a tired gear set, a broken ratchet spring, a deformed cover. That work is usually done competently. What is usually missing is the verification that the reassembled unit still holds its tolerances once components are loaded on it.

Five things decide whether a feeder is production-ready:

• Gear mesh clearance, in the 0.05 mm range once backlash is accounted for
• Ratchet return spring force, which most manufacturers publish no specification for
• Stepper motor torque consistency, which depends on the winding and the magnets, not on the gearbox
• Sensor response thresholds, which drift as hours accumulate
• Tape clamp chamfer, which decides whether a component releases cleanly from the carrier tape

Many third-party shops close a repair order with a five-minute empty-tape test. Five minutes does not even cover one full indexing cycle at production speed, let alone a thermal cycle, and it never puts a component through the release point.

Merex field rule: a feeder that has not passed all five steps below should not go onto a primary line. It can go to a backup machine, a spare-parts shelf or a non-critical station.

---

2. The Five-Step Feeder Acceptance SOP

Step 1 – Visual and cosmetic inspection (about 5 minutes)

Purpose: reject the obvious failures before anything touches the machine.

Checklist:

• Model on the nameplate matches the paperwork. On Yamaha, confirm the CL, SS, CS or SL family. On Fuji, confirm the W8, W12, W24 or W44 size.
• No cracks or deformation in the cover, the metal tape-inlet piece or the base latch area.
• All fasteners present. No “one screw was missing, it still runs.”
• Sensor lens clean: no scratches, no oil, no fingerprints.
• ID chip seated correctly, not skewed, not damaged.
• Visible gear teeth free of chipping and heavy wear debris.

Blind spot #1: skipping the visual inspection because the technician is in a hurry. A skewed ID chip then surfaces as an identification failure that takes the whole machine down.

Step 2 – Mechanical dry run (10-15 minutes)

Purpose: verify mechanical transmission and sensor response with no components loaded.

Tool: a feeder calibration jig, or the placement machine’s manual jog mode.

1. Jog 50 index steps and listen. Gear mesh should be smooth, with no clatter and no catching. A repeated clunk means assembly backlash is out of tolerance, or a chipped tooth is still hiding in the gearbox.
2. Cycle the tape clamp release and lock 10 times. The feel should be smooth with even damping. Sticking at one position means a fatigued spring or a worn clamp guide.
3. Turn an empty reel 30 revolutions to check ratchet and pawl engagement. Not one slip is acceptable.
4. Block the sensor with a metal shim and confirm that the machine or the jig reports tape present and tape absent correctly.

Merex bench reference values for these four checks, for use when the OEM does not publish an equivalent figure:

Brand family	Step noise	Tape clamp release force	Ratchet return time
Yamaha CL / SS	55 dB or less	8-12 N	80 ms or less
Fuji NXT W8 / W12	52 dB or less	6-10 N	70 ms or less
Samsung SM / SMN	58 dB or less	9-13 N	90 ms or less
Panasonic CM	55 dB or less	8-11 N	85 ms or less

These are acceptance reference values from our own bench work, not manufacturer specifications. Where the OEM publishes a figure for your model, the OEM figure wins.

Blind spot #2: checking that the feeder “turns” but never verifying the ratchet under load. The unit then indexes fine on the test bench and skips every few cycles on the line, which shows up as a rising defect rate rather than a clean alarm.

Step 3 – Electrical and sensor verification (10 minutes)

Purpose: confirm the feeder and the machine agree on what the feeder is, and that the signals are clean.

1. Install the feeder in the machine and confirm the machine identifies model, size and ID.
2. Watch the status feedback in the machine’s error window. OK, NG and Not Ready transitions must be real-time and accurate.
3. With tape loaded and then removed, confirm the material-present signal toggles correctly.
4. Read and write the ID chip where the platform supports it, and confirm the stored data matches the physical model.
5. Run for 30 minutes and watch for thermal drift alarms on the models that report them.

Blind spot #3: skipping this step because the repair shop’s test bench is not the production machine. A mismatched ID chip means the feeder feels fine while the placement program runs the wrong feeder parameters, and accuracy quietly degrades over a shift.

Step 4 – Loaded board run (30-60 minutes)

Purpose: this is the step that decides whether the feeder goes back on the line.

1. Load the actual component tape the line runs, not test tape.
2. Jog 200 index cycles in manual mode and count the pick errors. The pick error rate must be 0.5% or less. Tighter customer standards call for 0.2% or less.
3. Place 50 components and check placement position with AOI. Position deviation must stay within +/-0.1 mm for 0402 and larger.
4. Run 30 minutes at normal production cadence. No feeder-related alarm is acceptable in that window.
5. Pull the placement and SPI/AOI data and look for a drift trend, not just a pass or fail.

Acceptance limits at a glance:

• Pick error rate: 0.5% or less
• Placement deviation: +/-0.1 mm or better
• Feeder alarms during 30 minutes of production running: zero
• Tape tension: stable, with no slack and no overtension

Blind spot #4: the five-minute “no errors, so it passes” test. The feeder runs fine cold. Then it heats up, the lubricant thins, the gear set expands, and indexing accuracy walks out. This is exactly why price alone is a poor buying criterion for feeder parts.

Step 5 – Burn-in (optional, 2-4 hours)

This step is for high-reliability production: automotive electronics, medical devices, aerospace.

1. Run the feeder continuously for four hours on the machine.
2. Record pick error rate, position accuracy and temperature every hour.
3. Strip the unit after the run and inspect the gear set, ratchet and clamp for wear.
4. Compare key parameters against a new feeder of the same model.

Burn-in is too expensive to run on every unit, and most plants do not do it. The workable compromise is to pull one or two units from each new supplier, each repair shop and each in-house bench and burn them in. If the sample fails, the batch goes back.

---

3. What Changes by Feeder Brand

The acceptance sequence is the same for every brand. What differs is which part actually breaks, and that is where cross-brand acceptance mistakes get made.

Yamaha (CL / SS / CS / SL)

• Watch the gear material. The OEM uses POM; aftermarket replacements often use PA66+GF, which is harder but less tough in impact.
• The most common failure on a CL is the tape-inlet metal thickness. An aftermarket part that runs thick makes the component jam at the release point.
• Yamaha ID chips are encrypted, so an aftermarket chip has to be verified for compatibility, not assumed.
• Suggested limits: pick error 0.3% or less, position +/-0.08 mm or better.

Fuji (NXT / CP / IP)

• NXT puts a premium on ratchet spring force consistency, and that force is the number most often missing on a rebuilt unit.
• A repaired NXT gear set often indexes more noisily than new. Treat a clear step-noise increase as a finding, not as a character trait of the machine.
• NXT uses a proprietary communication protocol. In our own compatibility sampling, aftermarket ID chips passed on roughly 85% of the units tested, which is good enough to be dangerous: you cannot tell the failing 15% apart without testing.
• Suggested limits: pick error 0.2% or less, position +/-0.07 mm or better.

Samsung (SM / SMN)

• The tape clamp chamfer differs from Yamaha, and aftermarket replacements that reuse the Yamaha standard cause intermittent release problems. This is a part-number trap, not a wear issue.
• SM inlets wear at the mouth, and the assembly tolerance of a rebuilt inlet is where the difference shows up.
• ID chip compatibility is generally good, but we still see a read or write failure in roughly one unit in ten.
• Suggested limits: pick error 0.4% or less, position +/-0.1 mm or better.

Panasonic (CM / NPM)

• The feed motor’s torque consistency is the sensitive point, and aftermarket motors vary more than the rest of the unit.
• Sensor response thresholds drift after a part replacement, so the sensor step is not optional on a CM.
• Suggested limits: pick error 0.3% or less, position +/-0.08 mm or better.

---

4. The Acceptance Sheet and the Blind Spots

A form the line and the buyer can share

Repaired Feeder Acceptance Record

• Repair source / workshop: ________
• Feeder brand / model / size: ________
• Quantity / repair order no.: ________
• Acceptance date: ________

Step 1 – Visual

• □ Model matches paperwork □ Housing undamaged □ Fasteners complete □ Sensor lens clean □ ID chip seated correctly □ No chipped gear teeth

Step 2 – Mechanical dry run

• □ Smooth indexing, noise within limit □ Clamp release smooth □ No ratchet slip □ Sensor responds correctly

Step 3 – Electrical

• □ Identified by the machine □ Status feedback live □ Material sensor toggles □ ID chip read/write correct

Step 4 – Loaded board run

• □ Pick error rate within limit (actual: _%) □ Position deviation within limit (actual: _ mm) □ 30 minutes with no alarm

Step 5 – Burn-in (optional)

• □ 4 hours continuous run □ Error rate stable □ Temperature within range □ Stripped inspection clean

Result: □ Pass to line □ Downgrade to backup / non-critical line □ Reject and return

Inspector / date: ________

The four blind spots, in one place

1. “It turns” is not “the ratchet holds.” Ratchet engagement only proves itself loaded, with resistance, at real indexing speed.
2. Thermal drift is invisible in a cold test. Plan the hot test if your product tolerates nothing less than zero mispicks.
3. A mis-flashed or swapped ID chip is the hardest failure to spot, because the feeder still indexes. It quietly loads the wrong parameters into the placement program.
4. Sensor thresholds are not recalibrated after a part swap. Wrong material-present detection and index counting errors are among the quietest failures on a line.

Questions worth asking before you sign

• What is the warranty on the repair, and does it cover the parts replaced or the whole feeder?
• Do you test the repaired unit on a machine of the same brand and model, or only on a bench jig?
• Can you provide the pick error rate and position deviation from the last batch?
• Who signs the acceptance record, and does the record travel with the feeder?

---

5. Acceptance Is the Last Gate in Spare Parts Management

Back to Suzhou. When our engineers ran the full five-step acceptance on those 30 Yamaha feeders, 22 failed. Every pick error rate was above 3%. Six had mis-programmed ID chips, four had slipping ratchets, and three had chipped gear sets still hidden inside the housing.

Half a day of acceptance would have kept 22 bad feeders off a main line. The cost of that week, in rework, downtime and customer claims, was orders of magnitude larger than the cost of the test.

Acceptance is not paperwork. It is the last gate between a cheap part and a line that keeps running. Get it right and aftermarket parts, refurbished units and in-house rebuilds all become safe to use. Get it wrong and the cheapest part on the shelf is a scheduled downtime event.

Merex Technologies has spent more than 20 years supplying SMT spare parts and running feeder acceptance in-house. We repair and rebuild feeders for Yamaha, Fuji, Samsung and Panasonic platforms, and every unit leaves on a calibration jig with its acceptance record attached. Our CPK tester and calibration jig are supplied with the parts, so the verification happens on your line, on your own feeders, instead of on a promise in a supplier’s email.

If your plant buys repaired, refurbished or aftermarket feeders without a written acceptance procedure, we can help. Send us your feeder list and your current acceptance sheet. Our engineers will review the part numbers, suggest a mixed OEM, aftermarket and refurbished split, and run on-site acceptance training for your line and purchasing team.

Contact: https://parts-smt.com/contact/ | Online store: https://parts-smt.com/product/
Shenzhen Merex Technologies Co., Ltd. – in-house machining – 20+ years in SMT spare parts