Reciprocating Belt Feeder
for Smart EMS PCB Assembly

The cheapest part number on the quotation can be the most expensive one on the floor.

Full articles are in English. The summary above carries the argument in your language.

A purchasing team compares three quotations for the same odd-form connector and picks the cheapest unit price. The part arrives taped, on reels, at a good rate. Six weeks later the line is running at 78 percent of the rate the quotation assumed, and nobody can point at the number that went wrong, because the number that went wrong is not on the quotation at all.

The cost is not in the part. It is in the tape. Someone had to pay a supplier to tape it, someone has to receive and store the reels, someone has to change the reel mid-shift, and someone has to dispose of the carrier tape afterwards. Then the tape imposes its own constraint: the reel only holds so many parts, so the machine stops when the reel empties rather than when the board is finished. Every one of those costs lands on the floor, and none of them appear next to the unit price.

Do the comparison on a per-board basis instead. Take the part price, add the packaging surcharge the supplier charges for taping, add the handling minutes per reel multiplied by the fully loaded labour rate, and add the changeover stops per shift multiplied by the line rate. The number that comes out is the real cost, and it is frequently 15 to 30 percent above the unit price you started with.

That is where bulk feeding changes the arithmetic rather than the price. A bulk feeder accepts the part loose, so the taping surcharge disappears, the reel handling disappears, and the feeder refills from a box instead of stopping on an empty reel. On the S-MVF01 the component envelope runs from 3 x 3 x 3 mm to 20 x 20 x 20 mm and the unit is 850 x 140 x 260 mm, so it fits in a feeder bank where a bowl feeder would need floor space.

The habit to build is a quarterly review of which parts are still arriving taped. As the mix moves, a part that was high volume last year may now be a 40-board-per-week job, and it is still paying a taping surcharge and a reel changeover for the privilege. Move the low-volume odd-form parts to bulk, keep the genuinely high-volume ones on tape, and re-check the split every quarter.

Pick the three odd-form part numbers with the worst per-board cost and ask for a bulk-feeding trial on those three. Send a sample bag of each; the answer will be a feeder configuration or a straight no, and both are useful. Start with the sample bag, not the purchase order.

A bowl feeder is priced per part. That is fine until the part mix changes.

Full articles are in English. The summary above carries the argument in your language.

A vibratory bowl feeder is a beautiful piece of engineering. It is also, in most factories, a single-part machine with a footprint and a price to match, and that combination only works while the part it was built around stays in production.

The problem arrives with the third part number. The first bowl feeder was specified for the connector you were building in volume. The second was justified because the first could not be retooled quickly. By the third, you are paying for three bowl feeders, three footprints, three sets of tooling, and three sets of spare tooling, and each one sits idle whenever its part is not running. The tooling cost is real, but the idle footprint is what actually hurts, because floor space near the line is the scarcest thing you own.

Model the changeover cost rather than the capital cost. For each candidate part, write down the retooling time, the retooling cost, the annual volume, and the number of hours per year the feeder would sit idle. Any part that would idle more than half the year is a bad candidate for a dedicated machine, no matter how good the unit price looked.

The alternative is a feeder that is built around a component family instead of a single part. The S-MVF01 handles multi-pin odd-form components as a family and presents them through a selection track rather than a dedicated bowl, so a changeover is a track adjustment rather than a new feeder. It also occupies 850 x 140 x 260 mm at 12 kg, which is small enough to live in the machine's feeder bank instead of on the floor.

The recurring question to put on the engineering agenda is simple: which of our dedicated feeders is currently the wrong machine for its part? That question gets asked every time the mix shifts, and it is much cheaper to answer than to discover during a capacity review.

Take the part with the lowest annual volume that still has a dedicated feeder, and price a family-based feeder against the idle hours. If the idle hours do not justify it, keep the dedicated machine - and write down the reason, so the next review does not relitigate it.

Three physical tests you can run on a sample bag before anyone quotes a feeder.

Full articles are in English. The summary above carries the argument in your language.

Before anyone quotes a feeder, you can run three tests on a sample bag that will tell you whether bulk feeding is plausible. None of them need a machine, and all three take under an hour.

The reason this matters is that bulk feeding fails for physical reasons, not commercial ones. A part fails because the selection track cannot tell one orientation from another, because the leads bend when parts tumble together, or because the part is so tall and narrow that it will not lie down reliably. Those are properties of the part, and they are visible in a bag on a bench.

Test one is the orientation test. Empty the bag onto a tray, and try to sort the parts into piles by orientation using only a spatula. If a person cannot do it quickly by feel and by eye, a mechanical track will not do it either. Test two is the tumbling test. Shake the bag for a minute and inspect the leads for deformation. Test three is the aspect test: measure the length and the height of the part and divide them. The specification sheets for the S-MVF01 rule out anything under 1:1.2.

Run the same three tests against the published envelope. The S-MVF01 accepts components from 3 x 3 x 3 mm to 20 x 20 x 20 mm, and it explicitly excludes parts with no obvious physical feature that still need to be oriented, parts whose leads deform easily, and parts below the 1:1.2 aspect ratio. Those three exclusions are on the sheet because they are the three ways bulk feeding actually fails.

Make the test part of how new parts are introduced, not part of how feeders are bought. A ten-minute bag test at the design stage is worth more than a week of debugging after the feeder is installed, and it gives the design team a concrete reason to add a chamfer or a flat.

Pick one candidate part, run the three tests, and photograph the result. Send the photographs with the sample bag. A clear photograph of the part lying in three orientations will get you a faster and more honest answer than a specification sheet alone.

A short acceptance routine that stops a bulk conversion from failing on the line.

Full articles are in English. The summary above carries the argument in your language.

Cutting packaging out of an odd-form part is one of the highest-return changes available to an EMS line, and it is also one of the easiest to get wrong, because the failure usually shows up two weeks after the changeover as an intermittent jam rather than as an obvious mistake.

The failure pattern is consistent. Someone proves the feeder on a clean sample of parts, the changeover is approved, and then production starts using parts from a different batch. The new batch has a slightly different flash on the moulding, or the leads are a fraction straighter, and the selection track that worked perfectly now rejects one part in fifty. Nobody suspects the part, because the feeder was proven.

The answer is to accept the conversion formally, with six checks. First, the orientation test on three separate batches, not one. Second, the lead-deformation test after tumbling. Third, the aspect ratio measured on ten samples, not one. Fourth, a continuous run of at least 1000 feeds. Fifth, a jam log that records the rejection rate rather than just pass or fail. Sixth, a written changeover procedure that names the two vibration channels and the belt speed so the settings can be restored after a rebuild.

The published figures give you the targets for checks four and five. The S-MVF01 specification sheets state 1000 consecutive feeds without abnormality and 48 hours of continuous ageing without abnormality, with a feeding accuracy of plus or minus 1 mm. Those are the numbers to hold a trial against, and they are also the numbers to quote back to the supplier when a batch behaves differently.

Then keep the log. The rejection rate per batch is the single most useful number for spotting a supplier whose moulding has drifted, and it costs nothing to collect once the feeder is logging jams anyway.

Write the six checks into the changeover procedure for one part number, run it, and see what the rejection rate actually is. If it is stable across batches, extend the procedure to the next part. If it is not, you have found the problem before the customer did.

Multi-lane looks faster on paper and is harder to keep running on a real mix.

Full articles are in English. The summary above carries the argument in your language.

Multi-lane feeding looks like the obvious answer to throughput. Four channels instead of one should mean four times the parts per hour, and on a clean, single-part job it often does. The difficulty is that a real feeder bank rarely runs a single clean part.

The problem with multi-lane is that every lane has its own selection track, its own sensors and its own failure modes, and a jam in one lane can starve the pick position for the whole bank unless the lanes are managed separately. On terminals, which have a defined orientation and a defined lead position, the added complexity buys less than it costs, because the single-channel rate is already close to the machine's own pick rate.

Work out the required rate before you choose the lane count. Take the number of bulk components per board, divide by the machine cycle time, and compare that with the single-channel feeding rate. The S-MVF01 publishes a feeding interval of 1.5 to 2 seconds per feed on a single channel, which is 1800 to 2400 feeds per hour at full availability. If your machine takes longer than two seconds per bulk part, a second channel will not help you at all.

For terminal and connector work, that calculation usually lands on single channel. The S-MVF01 presents one part at a time at a vacuum-held pick position with a fibre-optic part-present signal, so the machine never reaches for a part that is not there. One clean signal is worth more on a terminal line than three lanes that each need their own tuning.

The number to watch over time is the ratio of bulk parts to machine cycle. If the ratio starts to approach one, that is the moment to look at lane count again - and not before, because a second lane that runs idle is a second lane that collects jams.

Count the bulk components on your current worst board and the machine cycle time, and divide one by the other. If the answer is above two seconds per part, specify single channel and spend the difference on a better selection track.

Two vibration channels, one belt speed, one vacuum. Everything else is a consequence.

Full articles are in English. The summary above carries the argument in your language.

A bulk belt feeder has four adjustments that matter and a dozen that look like they matter. Knowing which four to touch, and in which order, is most of what separates a feeder that runs for a year from one that is retuned every week.

The usual failure is tuning in the wrong order. An engineer sees parts backing up at the pick position and turns up the belt speed. The belt now delivers parts faster than the selection track can orient them, so more parts are rejected, the return track fills, and the whole thing looks like a track problem. Two hours later the belt speed is back where it started and nothing has been learned.

The order is: selection first, transport second, presentation third. Set channel A on the vibration controller, which drives the selection box, until parts are being oriented and rejected cleanly. Then set channel B, which drives the feeding track, until accepted parts advance smoothly to the front. Then set belt speed on the speed controller, clockwise to increase, so the box is refilled fast enough to keep the selection box fed but not so fast that it floods. Finally check the vacuum, which holds the part at the front of the track.

The S-MVF01 exposes exactly these controls. The manual describes channel A as the voltage frequency of the material selection box and channel B as the voltage frequency of the material feeding track, with a separate speed controller for the belt motor and a vacuum generator at the front of the track. Five air-blowing points handle the parts that have fallen out of the flow, so a jam clears itself instead of stopping the feeder.

Then leave it alone, and record the settings. The single biggest cause of a feeder that mysteriously stops working after a rebuild is that nobody wrote down the four numbers. Put them on a card inside the electrical control area and the next rebuild takes ten minutes instead of a shift.

Find the four controls on your feeder, write their current values on a label, and stick it inside the panel. That one action prevents more downtime than any amount of retuning.

The manual's own intervals, turned into a checklist a shift leader can actually own.

Full articles are in English. The summary above carries the argument in your language.

Most bulk feeder downtime is not caused by a failure. It is caused by a wear item that was never inspected, and by the fact that the inspection intervals live in a manual that nobody reads on the shift.

The consequence is predictable. The belt starts slipping and the parts stop travelling. The operator increases the belt speed, which wears the belt faster. By the time anyone looks at the tensioner, the belt has been running slack for a week and the drive gears have taken the load instead. What began as a five-minute adjustment ends as a belt and gear replacement.

The manual gives three intervals that are worth turning into a shift checklist. Lubricate the drive gears after three months of operation. Check the belt tensioner whenever the motor turns but the belt does not - the fix is to tighten it. Wipe the belt with alcohol whenever the belt turns but the material does not travel with it, because that symptom is surface contamination rather than tension. Add a fourth: check the gap between the armature and the coil on both straight-vibration units, which the manual specifies should stay under 1 mm.

On the S-MVF01 those are all reachable without dismantling the feeder. The manual documents the drive gear, the belt tensioner and the belt as separate serviceable items in the conveyor module, and it documents removing the tensioning screws to take the belt module off entirely. The straight-vibration units expose their coil, armature and spring blade for inspection.

Build the checklist around the two sensors as well, because they are what the feeder uses to protect itself. The front track sensor stops the front straight vibrator when material is detected, and switches the vacuum off after a minute if the part is not taken. The rear track sensor stops the rear vibrator and the belt when the track is full. If either sensor sees no material for more than a minute, the feeder alarms and stops the belt and the blowing. A sensor that has drifted produces symptoms that look like a mechanical fault.

Take the three manual intervals, add the armature gap, and write them onto one card for the shift leader. Then log every jam for a month and see which of the four actually fires - the log will tell you which interval to shorten.

Which numbers to insist on before you approve the spend, and which ones to ignore.

Full articles are in English. The summary above carries the argument in your language.

Feeder payback calculations go wrong in the same way every time. Someone puts a machine price on one side and an assumed labour saving on the other, and the resulting number is either so good that nobody believes it or so vague that nobody acts on it.

The weakness is that the labour saving is usually asserted rather than measured. The manual handling time per odd-form component is the number that drives the whole calculation, and it is almost never measured. It is guessed, and it is guessed low, because the person guessing is thinking about the motion of placing the part rather than the walk to the bin, the opening of the bag, the orientation check, and the walk back.

Measure four things on the line before you build the model. Time the manual operation over a full board, not a single component, so the walking and the bag handling are included. Count the bulk components per board. Record the machine cycle time. Then write down the machine investment and the fully loaded monthly wage, and put all five into a model that shows its own arithmetic.

That is exactly what the calculator on this page does. It takes the feeding interval, the component count, the boards per day, the production days, the manual handling time, the operators and the wage, and it produces machine time per board, handling hours released per month, labour released per year, and a payback period. It refuses to print a payback when the machine time is not shorter than the manual time, because a negative saving is not a saving.

Then revisit it. The first calculation uses today's component mix and today's wage. Both move. A payback model that is re-run every six months with the current mix is a management tool; one that is run once to justify a purchase is a formality.

Time the manual operation on one board this week and put the real numbers into the calculator. If the answer is a payback you can defend, you have your business case. If it is not, you have saved yourself an approval meeting.