S7900 OddForm Insertion Machine
Solution for Smart EMS THT PCB Assembly

An SMT line and a radial inserter take the easy part of a through-hole board. What is left over is small in count, awkward in shape, and expensive enough to keep four people at a bench.

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

Walk a through-hole department at the end of a shift and look at where the people are standing. They are almost never at the SMT line, which runs itself. They are rarely at the radial inserter, which is loaded with taped parts and cycles on its own. They are at the bench, in front of a rack of bins, picking connectors, relays, transformers and bulk capacitors out of bags and trays, checking which way round each one goes, and pushing the leads through holes by hand.

The count of parts at that bench is always smaller than the count of parts on the machine. That is what makes the bench invisible in a capacity review. A board with two hundred through-hole points may have a hundred and sixty of them on tape and forty in bags, and the forty consume more operator hours than the hundred and sixty. The cost model that counts points will point you at the wrong forty.

The reason is that packaging, not point count, drives manual time. A taped radial part arrives oriented, at a known pitch, in a sequence. A loose part arrives in a bag. The operator has to pick one, identify it, decide its orientation, dress or bend the leads if the body does not sit flat, insert it, and confirm it is seated. Every one of those is a separate motion, and several of them are a separate decision. That is why a bench with forty loose parts can cost more per board than a machine placing two hundred taped ones.

So the sizing question is not how many through-hole points your board has. It is how many of them arrive in a format that no machine in the line can currently feed. On the board analysed on this page — a real customer board, thirty-six through-hole lines — the split came out at twenty-six lines for an odd-form inserter, nine for an axial inserter, and one that stays manual. Sixty-nine percent of the through-hole work was sitting in a format the existing machines could not touch.

That ratio is the number to bring to a capital request. Not the machine's speed, not its footprint, not its accuracy — the proportion of your through-hole labour that is currently unreachable by any machine you own. If the answer is a third, an odd-form project is a capacity project. If the answer is five percent, it is not, and no amount of quotation detail will change that.

Getting the ratio takes one afternoon and no supplier. Take the current bill of materials. Delete everything that is surface-mount. For each remaining line, write down the supply format: loose bulk, ammo pack, tape, tray or tube. Then mark the lines whose format is loose bulk, tray or tube, because those are the ones a taped-only machine cannot take. The marked lines, divided by the total, is your ratio.

Do it for three boards, not one — a high runner, a low runner and the newest design. The newest design is the one that matters most, because it is the one whose layout is still open and whose packaging can still be influenced.

Send us the supply-format column for those three boards and we will tell you which lines an S7900 can take and which should stay where they are. It is a fifteen-minute answer, and it is a much better place to start than a machine quotation.

The decision that sets the price of an odd-form project is not which machine to buy. It is which lines go on which machine, and that decision is made before any quotation exists.

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

The scene: a quotation arrives for an odd-form insertion machine, sized against a whole through-hole board. The number is large, the case is thin, and the project stalls. Six months later the same board is reviewed again, this time line by line, and it turns out that a third of the lines on it should never have been on the odd-form machine in the first place — they are axial parts in ammo pack with a fixed lead span, and they belong on a machine that has been sitting in the plant all along.

This is the most common way an odd-form project is mis-scoped, and it is not a technical error. It is an omission: nobody wrote down the assignment rule before the quotation was requested, so the quotation answered a question nobody had asked.

The rule has three parts, and they are checked in order. First, is the component supplied in a format the odd-form platform can feed — loose bulk, tape, tray or tube? Second, does the part have a body that a vacuum nozzle or an air gripper can hold and a lead pattern that can be inserted without a fixture? Third, is it a component at all, or is it a wire assembly that is routed rather than placed? If the answer to the first two is yes and the third is no, the line belongs on the odd-form inserter.

Anything that fails the first test but is an ammo-pack axial part with a fixed lead span goes to the axial inserter. Anything that fails the third test stays manual. There is no fourth category, and the value of the rule is that it produces a decision rather than a discussion.

Applied to the customer board behind this page, the rule produced a clean split: twenty-six lines for the odd-form inserter, nine for the axial inserter, one that stays manual. The nine were four diodes and five resistors, all in ammo pack. The one manual line was a crimped ground harness — a wire assembly, not a feeder component. The twenty-six covered electrolytic and ceramic capacitors, noise-suppression capacitors, seven connectors, a spade terminal, a relay, a varistor, a sensing gauge, a transformer, a DIP IC and two trimmer potentiometers.

Read the twenty-six by packaging rather than by function and the machine configuration falls out: seventeen lines loose bulk, so a bowl feeder; five taped, so a radial tape feeder; three in trays, so a tray feeder; two in tubes, so a tube feeder. That is a four-family feeder set, and it is the actual answer to the question the original quotation was trying to answer.

The split is also what makes the investment case honest. Quoting one machine against thirty-six lines inflates the scope. Quoting an odd-form machine against twenty-six lines and pairing it with an axial machine for the other nine produces a smaller number on each machine and a line that runs at two different cycle rates instead of one compromised one. The customer solution document behind this page proposes exactly that pairing.

Keep the rule in a spreadsheet and re-run it whenever a board changes. The assignment is not permanent — a part that moves from tray to tape changes category, and a design change can move a line from manual to machineable. A split that gets reviewed quarterly stays accurate; a split done once at project stage becomes the reason the next project is mis-scoped.

Send us your through-hole BOM with the supply-format column filled in and we will run the same rule against it and return the three-way split. Do that before you ask anyone for a price, including us.

Feeder stations are bought per supply format, not per component type. Read the same BOM the second way and the machine's price stops being a mystery.

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

Here is the meeting that goes wrong. Engineering lists fourteen odd-form components. Purchasing asks how many feeders that is. Someone says fourteen, because there are fourteen components, and the quotation comes back for fourteen stations. Then the machine is installed and it turns out that two of those components are in tubes, three are in trays, four are loose and five are taped — which is four feeder families, not fourteen of anything, and the station count on the quotation had been calculated against the wrong unit.

The unit of purchase on an odd-form machine is the feeder station, and a station is specific to a packaging format. A bowl feeder presents loose parts from a vibratory bowl. A tape feeder presents taped parts at a pitch. A tube feeder presents sticks of parts by gravity or push-rod. A tray feeder presents a matrix of parts in a pocketed tray. They occupy different footprints, they have different capacities, and they are not interchangeable. Counting components tells you nothing about how many of each you need.

So count the other way. Group your BOM by supply format and count the lines in each group. That count is the station requirement per family. On the customer board behind this page, thirty-six lines collapse to seventeen loose bulk, nine ammo pack, five tape, three tray and two tube — which is a much smaller and much more specific question than "thirty-six components".

Then check the count against what the platform documents per model. The S7900 catalog gives station counts model by model: a 205 bowl feeder takes 4 stations on the S7900-II, IV and VI-P and 8 on the S7900-VI; a tray feeder takes 1 on the II, IV and VI-P and 2 on the VI; the tape and tube families range from 4 to 20 depending on which feeder width and which model. The catalog also prints the caveat that matters most: when different feeder types are mixed, the quantity depends on the combination. Stations are not purely additive.

That caveat is the reason a feeder layout has to be drawn rather than summed. A bowl feeder and a tray feeder do not occupy the same width, and a machine configured for eight bowls will not accept the same number of trays. This is where a supplier's own experience with the combination is worth more than a specification table, and it is worth asking for a station layout drawing rather than a station count.

There is a second-order benefit to counting by format. Once the four families are separated, you can see which of them is under-used and whether a packaging change would remove a feeder. Moving a part from tray to tape, or from tube to tape, is a conversation with the component supplier — and it is usually cheaper than adding a station. On the board analysed here, three tray lines and two tube lines is a small enough number that a packaging conversation is worth having before the feeder order is placed.

Southern Machinery manufactures its own feeders rather than buying them in, which is why the platform documents all four families on one base and describes the stations as plug-and-use with no wiring between them. That matters for expansion: adding a station later is a feeder and a program line, not a re-engineering of the machine.

Do the count before the quotation. Group your BOM by packaging, count the lines per group, and ask for a station layout drawing against that count rather than a station total.

Send us the packaging column and the model you are considering and we will return the station layout, including which families can grow and by how much. That drawing is the document a capital request should carry, not the component count.

Floating components are diagnosed as a wave-soldering problem. The cause is at the insertion stage, and it has four documented remedies.

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

A board comes off the wave with a bridge under a connector, or a void where a transformer should be soldered, or an open on a relay pin. The wave is checked, the flux is checked, the preheat profile is checked, and eventually the board is run again with the same result. The parts are being inserted correctly. They are simply not being held down, and the solder wave lifts them.

The mechanism is worth stating plainly. Wave soldering pushes molten solder against the underside of the board with enough force to move anything that is not mechanically restrained. A lead that has been pushed through a hole and left straight has nothing holding the component against the top surface. The part lifts a fraction of a millimetre, solder enters the gap, and the joint that forms is not the joint the design intended.

This is why the clinch exists. After insertion, a mechanism beneath the board bends the protruding leads over so the component cannot move. On the odd-form platform the clinch is a bottom-side unit, and it is documented as standard on one model in the family and optional on another rather than universal — which is the first thing to check against your configuration.

The catalog documents four bend patterns, and they are not cosmetic variations. Outward bend turns the leads away from each other. Inner bend turns them toward each other. Circular bend produces a curl with no fixed direction or angle. Two turns produces a specified number of rotations at a specified angle. Which one is correct depends on the pad geometry, the lead diameter, the hole clearance and how much board space the bent lead will occupy.

The lead-bending specification is equally specific: lead diameter under 1.0 mm and at least 2 mm of exposed lead, subject to final design. Both numbers matter. A lead thicker than 1.0 mm will not form reliably. A lead that protrudes less than 2 mm gives the bender nothing to work with, and the result is a partial bend that looks secure and is not.

There is a second consequence of clinching that changes the cycle time and therefore the capacity model. The catalog's per-component speed for the four-head S7900-IV-P is printed as 2 seconds with lead clinching, against 0.6 seconds for the six-head S7900-VI without it. That is a large difference and it is not a defect in either figure — it is the cost of the clinch operation, printed where you can see it. A capacity model that uses the fast number for a configuration that clinches will be wrong by a factor of three.

So the decision has three parts, and they should be made together: does this board need clinching at all, which bend pattern does each part need, and what does the clinch add to the cycle time. Answering only the first one produces a machine that either cannot hold its parts down or is slower than the business case assumed.

Build a clinch table for your board. One row per odd-form part, with lead diameter, exposed lead length, pad geometry, and the bend pattern you want. Where a lead is under 2 mm of exposure, that is a design change to raise on the layout rather than a machine setting to argue about.

Send us the odd-form part list with lead diameters and hole sizes and we will return a clinch recommendation per part, with the cycle-time consequence stated for each. That table is the one that makes the capacity model honest.

Three source documents describe the same machine family at three different speeds. None of them is wrong. Reading them together is the skill.

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

Every insertion machine is sold on a rate, and every production manager has watched a rate fail to appear. The pattern is consistent enough to be predictable: the machine runs at the quoted figure on the easiest board, the mix changes, and the realised rate settles somewhere below it. The engineer who wrote the business case was not exaggerating. They were reading one number out of a set of three.

On the S7900 the three numbers are visible in the source material, and they differ by more than a factor of three. The customer solution document gives 2,400 components per hour with conditions attached to the figure. The 2025 catalog gives speed per component, model by model, as 0.8 seconds on the two-head S7900-II, 0.7 on the four-head S7900-IV, 2 seconds on the four-head S7900-IV-P with lead clinching, and 0.6 on both six-head models. A third document, the Precision S7900 presentation, states a maximum of 4,700 components per hour.

None of these contradicts the others, because they are measuring different things. A per-component figure is the head's cycle time under defined conditions. A per-hour figure is that cycle time converted, sometimes with a utilisation assumption folded in. A maximum figure is the ceiling on the best possible combination of component sizes and angles. To compare them you have to convert them to the same basis, and the conversion is not trivial — which is exactly why quoting one number and hiding the others is a common supplier habit.

The catalog also prints a second, more useful number that is easy to miss. Its capacity section describes a real case — a board with two cavities, four insertions per machine, one variant and four pieces per machine — running at a cycle of roughly 11 seconds and a capacity of roughly 650 pieces per hour. That is a production figure on a real board, not a head figure, and it sits far below 2,400 per hour. The gap between 650 pieces per hour and 2,400 components per hour is not an error; it is the difference between counting boards and counting components, and between a real mix and a clean test.

So the discipline is to use three numbers, not one. The ceiling is what the manufacturer publishes and what the brochure says. The planning rate is what you expect across a realistic mix, including board transfer, feeder attention and the occasional stopped head, and it is what belongs in a capacity model. The measured rate is what your board actually runs at, on your line, with your operators — and it is the only number you should ever quote outside the project.

Most suppliers give you the ceiling and let you guess the planning rate. Southern Machinery's material is unusual in printing the ceiling, the per-component cycle and a real-board production case in the same document, which gives you the raw material to build the planning rate yourself rather than accepting someone else's.

The third number is the one you have to go and get. Do not accept a demonstration on a demonstrator's board. Send your own boards with your own component list and ask for the insertion count and the elapsed time to be reported separately, so you can compute the rate rather than be told it. Ask specifically for the run to include a component that requires an angle change and one that requires a span change, because those two events are what separate a ceiling from a plan.

Then write the capacity model with the planning rate, keep the ceiling in a footnote, and replace the planning rate with the measured rate after the first production week. Re-measure quarterly. A capacity model that gets re-measured is worth more than a faster machine bought on a worse assumption.

Send us your board and component list and we will return the three figures side by side for your case, with the assumptions printed against each one.

A 0-and-90-degree insertion limit is paid for in board area and jumper wires long before it is paid for at the machine.

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

Look at a through-hole layout drawn by someone who has spent years working with a machine that inserts at two angles. The constraint is visible even where it is not written down. Components sit in rows and columns. Where a signal path wants to travel diagonally, it travels in two orthogonal jumps instead. Where a part would naturally sit at an angle, it is rotated to the nearest right angle and a jumper wire is added to reach the point it needed to reach.

That is the real cost of a limited insertion angle, and it is not paid at the machine. It is paid in the layout: extra board area, extra jumper wires, extra insertion points, and extra hand assembly on the parts that could not be made to fit. A designer who is told the machine inserts only at 0 and 90 degrees will produce a board that respects that limit, and the board will be worse for it in ways nobody will ever attribute to the machine.

The odd-form platform removes the constraint. Insertion direction is documented across a full circle — the customer solution document gives 0 to 360 degrees with an adjustment resolution of ±0.1 degrees, and the catalog gives the same 0 to 360 degree range without the resolution figure. On a board with connectors that must face a particular edge, a transformer that sits where it fits, or a relay at an angle set by its footprint, the angle becomes a program field rather than a fixture.

The consequence is a design review that most teams have never run. Take your current family and count the odd-form parts whose natural angle is not a multiple of 90 degrees. Then count the jumper wires that exist purely to route around the constraint rather than to carry a signal. On a mature through-hole layout the second number is usually the larger of the two.

Run the test the other way as well. For each part that is currently rotated to a right angle, ask what would change if the angle were free: would the part move, would a jumper disappear, would a keep-out area shrink, would two parts that currently conflict stop conflicting. Those are layout simplifications, and each one removes an insertion point or a hand operation.

There is a second benefit that shows up on the bench rather than the board. Parts that fit at their natural angle are easier to hand-place on the lines that stay manual, because the operator is no longer fighting a layout that was rotated to suit a machine. And parts that move from manual to machine are the ones that carry the labour cost described in the first article on this page.

None of this requires a redesign of an existing product to be worth doing. It requires that the next layout, and the next revision of a current one, is drawn without the constraint — and that the design-for-insertion checklist asks about insertion angle explicitly rather than leaving it to the designer's habit.

Add one column to your design review: insertion angle, per part. If the column contains only 0 and 90, either your layouts are genuinely unconstrained or nobody has tested the constraint. Both are worth knowing.

Send us one board drawing and its odd-form part list with the insertion angles marked and we will tell you which parts could be simplified and what that does to the insertion count.

The board size a machine accepts is one number. Whether it can hold that board flat and square while heads work on it is a different question, and it is the one that decides the layout.

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

The scene: a long board is quoted against a machine whose maximum PCB length is comfortably larger. The machine is bought. On the first production run the far end of the board flexes under insertion force, the leads do not enter cleanly at the extremes, and the defect rate is highest at the two ends of the board. Nothing in the specification was wrong. The specification described an envelope, not a stiffness.

An insertion machine's board size figure is a maximum envelope. What it does not tell you is how the board is supported along its length, whether that support is adjustable, and what happens to a thin board in the middle of a long span. Those three things are what turn an envelope into a usable capability.

The catalog documents the support and transport arrangement in more detail than the envelope alone. The conveyor is a three-section design on every model in the family, with automatic width adjustment. Conveyor height is 750 mm ±20 on the two-head and four-head models and 900 mm ±20 on the belt-conveyor model and the six-head model. Maximum PCB weight is documented per model at 5 kg on four of the five and 2 kg on one. The catalog also requires a side width of at least 5 mm — the board needs a margin that the conveyor can grip.

That last requirement is the one that most often goes unread. A board whose edge is within a few millimetres of the outermost component cannot be gripped reliably, regardless of how large the envelope is. It is a layout constraint that belongs in the design review, not a machine limitation to be discovered at commissioning.

The customer solution document adds a second set of figures for the line rather than the machine: a three-phase ESD conveyor at 930 ±30 mm, a board range stated on one page as L50 × W50 to L490 × W420 and on another as L60 × W60 to L490 × W420, and two SMEMA interfaces for line handshaking. The two board-minimum figures are printed on different pages of the same document and are shown side by side on this page rather than reconciled. Where your own board sits relative to that range is worth confirming against a current drawing rather than a proposal document.

The practical work is to model the board, not just its outline. For each board you intend to run, record the length, the width, the thickness, the weight, the distance from each edge to the nearest component, and whether the board has a stiffener or a breakaway rail. Then check all six against the machine's figures and against the conveyor's grip requirement. A board that passes the envelope test and fails the edge test will still fail in production.

Two of those fields interact in a way worth calling out. Board thickness and insertion force are related: the catalog documents a thickness range of 0.6 to 2.0 mm, and one model in the family applies up to 98 N of insertion force against a 2 kg maximum board weight. A thin, long board in that configuration is a deflection problem waiting to happen, and the answer is either a support fixture, a different model, or a layout change — not a machine setting.

Where a board genuinely needs support beyond what the conveyor provides, the platform documents a carrier option with its own maximum size per model, ranging from 350 × 500 mm to 500 × 550 mm. A carrier is how you run a board that is outside the normal envelope or too flexible to run bare, and it is a per-product investment rather than a machine feature.

Send us the six fields for your longest and thinnest board and we will tell you whether it runs bare, whether it needs a carrier, and which model in the family fits it. That check takes less time than a commissioning visit.

The same machine is described three times in the source material, and the numbers do not match. That is normal, and the way you handle it decides whether the project survives an audit.

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

Anyone who has assembled a capital request knows the moment. Three documents about the same machine are open on the screen, and they disagree. One says the maximum board is 490 × 420 mm, another says 450 × 400, a third says 480 × 390. One says four vision cameras, another says one plus two. One says the machine weighs 1,300 kg, another says 1,300 to 1,800 depending on model. The instinct is to pick the most recent, or the most authoritative-looking, and move on.

That instinct is what fails an audit six months later. If the figure you wrote into the specification turns out to be the optimistic one, the machine does not meet the requirement and the project is exposed. If it turns out to be the pessimistic one, you may have excluded a machine that would have worked. Either way, the problem is not the disagreement between the documents — it is that the disagreement was resolved silently.

The discipline is simple to state and takes discipline to follow: when two sources disagree, print both, attribute each to its document, and never average them. An average of two real figures is a third figure that exists nowhere and that nobody can defend. Printing both makes the disagreement a known open point, which is a manageable state. Hiding it makes it a latent defect.

Applied to this machine family, the disagreements are extensive and they have a pattern. The customer solution document describes a configuration being proposed for one plant, with six heads, a specific conveyor height and a particular board range. The 2025 catalog describes the whole product family, model by model, with five distinct configurations. The Precision presentation describes the platform's capability in general terms. None of them is describing the same thing, which is why they do not agree — and that is exactly why a single reconciled number would be misleading.

So attribute by scope before you attribute by date. A figure that belongs to a five-model family table is not comparable to a figure that belongs to one proposed configuration. A maximum in a marketing presentation is not comparable to a per-component cycle in a specification table. Sort the sources by what they are describing, and most of the apparent contradictions resolve into different questions.

Where they genuinely conflict — where two documents describe the same scope and give different values — the conflict is information. A minimum board size printed as 50 × 50 mm on one page and 60 × 60 mm on another page of the same document tells you the document has not been internally reconciled. That is worth knowing before you rely on either figure, and it is worth raising with the supplier rather than resolving yourself.

Then mark what is unconfirmed. Three labels cover almost everything: verified material, meaning a figure printed in a document you can cite; concept illustration, meaning an explanation or a diagram rather than a measurement; and to be confirmed, meaning a value that is either absent or contradicted and that must be settled before it enters a contract. A specification where every line carries one of those three labels is auditable. A specification where they are mixed is a liability.

This page follows that rule throughout. Where the sources disagree, both figures are printed with their document named. Where a value is absent, it is shown as not stated rather than filled in. Where a figure belongs to a different model in the family, the model is named. The disclosure table further up this page exists for exactly this reason, and it is deliberately long.

Adopt the same three labels in your own capital request. The document will be shorter than you expect and considerably harder to attack — and when a supplier quotes a number that does not match your table, you will know which of you is looking at which document.

Send us the specification you are working to and we will mark it against the source documents, line by line, and tell you which figures are confirmed and which need to be settled before order.