Publish Time: 2026-10-28 Origin: Site
High-mix low-volume line planning · Published by I.C.T · For production, process and manufacturing engineers
High-mix low-volume production is a different engineering problem from mass production, and the Hanwha SMT line planned for it has to be planned by different rules. In high volume the line is tuned once and run for weeks on one board; in HMLV the line is re-configured several times a day, sometimes several times a shift, moving between dozens or hundreds of part numbers at mid-to-low batch sizes. The constraint stops being the machine's peak placement rate and becomes the sum of every changeover, every program switch, every feeder swap and every first-article check the plant has to perform to stay honest. This guide plans a Hanwha SMT line for exactly that reality, and it builds on the Hanwha SMT placement equipment range as the hardware baseline.
A Hanwha pick and place machine for high mix low volume is planned around five axes that mass-production planning tends to under-weight: feeder preparation, program management, component range, first-article verification and line balancing. The Hanwha SMT line for high-mix low-volume production is simply the hardware expression of that plan, and getting the plan right matters more than buying the fastest head on the spec sheet. Each axis is treated as a changeover cost, not a one-time setup, because in HMLV the same board is rarely run twice in a row. The plan below walks through how to standardise feeders so swaps are fast and error-free, how to manage programs across a large product list, how to match the component mix to the right placement head, how to gate every new part number with a disciplined first-article workflow, and how to balance the line so throughput holds up when the product mix keeps shifting. The aim is a line that loses less time to change than it gains from a fast machine, and that is the difference between a specification that looks good and a cell that actually ships product.
Table of Contents
High-mix low-volume is not a gentler version of high-volume; it is a different optimisation target with its own failure modes, and the Hanwha pick and place machine for high mix low volume has to be designed around change rather than around steady-state speed. In a high-volume plant the dominant cost is the time the line spends placing components, so the right move is to maximise placement rate and minimise idle. In an HMLV plant the dominant cost is the time the line spends not placing components — swapping feeders, loading programs, verifying first articles, re-balancing for a different board — and a machine that is 20% faster but 40% slower to change can destroy more throughput than it creates. The planning question therefore shifts from "how fast can one board run" to "how little time is lost moving between many boards".
This opening chapter sets the frame the rest of the guide works inside. The first subsection defines what HMLV actually means on the floor, with the order sizes and changeover frequencies that separate it from both prototyping and mass production. The second explains why the high-volume playbook — buy the fastest single mounter, run long campaigns — fails for HMLV, and the specific ways it fails. The third lays out the five planning axes the guide follows: feeder preparation, program management, component range, first-article verification and line balancing. The fourth maps those axes onto the Hanwha platform, because the choice between an SM-series and a DECAN-series line is partly a HMLV decision in itself.
On the floor, HMLV means a lot of different boards, each in modest quantity, with the line reconfigured often. A typical HMLV contract manufacturer runs batches of 20 to 500 boards per order, keeps 50 to 500 active part numbers in the system, and changes the product on a placement line somewhere between two and ten times per shift. Contrast that with high volume, where a single panel may run for a week, or with pure prototyping, where a board is built once and never repeated. HMLV sits in the demanding middle: the volume is real enough to need automation, but the variety is high enough that setup and changeover dominate the day. The economics are decided by the changeover, not the cycle.
The numbers make the priority concrete. If a line places at 25,000 components per hour but spends 45 minutes changing over between orders, a single changeover costs more placed volume than two hours of running. Run six changeovers a shift and the line is effectively down for half the shift no matter how fast the head moves. That is why HMLV planning treats feeder swaps, program loads and first-article checks as first-class costs to be engineered down, on the same footing as placement speed. A Hanwha SMT line for this environment is judged by its ability to absorb variety, and every later chapter in this guide is really a way to shrink the changeover tax.
The high-volume playbook fails for HMLV because it optimises the wrong variable. Its instinct is to buy the highest-CPH mounter money allows and to group production into long campaigns so the machine rarely stops. On an HMLV floor that instinct backfires twice. First, the fastest head is usually the most specialised head, meaning it handles the narrowest component window and needs the most feeder changes to cover a wide mix. Second, long campaigns are impossible when each order is small, so the "run it for a week" assumption never holds and the machine spends its life reconfiguring instead of placing. The speed you paid for is spent in the changeover you did not plan for.
The failures show up in predictable places. A line tuned for one board discovers that its feeder set is wrong for the next order and the swap eats the shift. A program pipeline built to load one golden dataset cannot version the dozens of BOMs the plant actually runs, so operators retype coordinates by hand and introduce errors. A first-article process designed for a quarterly new product introduction cannot keep up with daily new part numbers, so boards ship before they are verified. Line balancing done once for a single product becomes meaningless the moment the mix changes. Each of these is a high-volume habit colliding with an HMLV reality, and each is addressed head-on in the chapters that follow.
The guide is built around five axes, chosen because they are where HMLV lines win or lose the day. Feeder preparation covers how feeders are standardised, staged and calibrated so a changeover is a fast, error-free swap rather than a troubleshooting session. Program management covers how CAD, BOM and placement data are turned into versioned, repeatable programs across a large product list, so the right board builds the first time. Component range covers how the mix of package sizes and the board envelope are matched to the right placement head and nozzle set, so no part number is outside what the line can place. First-article verification covers the inspection gate that lets a new or changed part number run safely. Line balancing covers how placements are allocated across machines so throughput survives the shifting mix.
These five axes are not independent; they are a chain, and a weak link anywhere stalls the line. Feeder prep is worthless if programs are wrong, programs are worthless if the head cannot place the parts, the head is worthless if first articles are skipped, and all of it is worthless if the line is unbalanced so one machine is buried while another waits. The planning therefore has to be done as a system, not as five separate purchases. The reference configurations in the final chapter show how the five axes come together on real Hanwha hardware, but each axis is developed on its own first so the reasoning is visible and the trade-offs are explicit rather than hidden inside a vendor quote.
The Hanwha platform maps well to HMLV because its two families — the SM series and the DECAN series — span the range from flexible multi-function placement to high-throughput chip shooting, and a line can be composed from both to match a specific product mix. The SM471Plus is a dual-gantry chip shooter rated around 78,000 CPH with an optional dual-track conveyor, the SM481Plus extends IC reach, and the SM482Plus stretches to wide bodies and long 1,200 mm boards. The DECAN line restarts the ladder one generation tighter: the S1 is a single-gantry speed entry near 47,000 CPH, the F2 is a dual-gantry value platform near 80,000 CPH, the S2 is the 92,000 CPH flagship, and the L2 is a 56,000 CPH multi-function tier optimised for flexible placement. For HMLV the interesting machines are the flexible ones and the value-throughput ones, not the bare flagship.
The planning implication is that a HMLV line is usually composed, not bought as a single box. A flexible multi-function mounter such as the DECAN L2 or the SM481Plus carries the awkward packages and the low-volume oddities; a higher-throughput machine such as the DECAN F2 or the SM471Plus carries the bulk chip placing; and a fast single-gantry such as the DECAN S1 can be added to relieve a bottleneck when the mix skews toward volume. Because all of these share the same feeder form factor and the same programming environment, the composition can change as the product list changes — which is exactly the adaptability HMLV demands. The later chapters show how to size and balance such a composed line, and how to keep the feeder and program layers consistent across every machine in the cell.
Feeder preparation is the single largest, most repeatable cost on a HMLV Hanwha SMT line — the Hanwha pick and place machine for high mix low volume lives or dies by how fast it changes feeders — and it is also the most engineerable, which is why it leads the planning. Every changeover is, at the physical level, a set of feeders moving from a shelf to the machine and back, and the time and error rate of that move decide how often the line can switch products. A plant that treats feeders as a shared, ad-hoc pile will spend its life hunting for the right tape, mis-seating reels and discovering missing parts mid-run; a plant that standardises and stages feeders turns the same changeover into a predictable, minutes-long swap. The work in this chapter is to make feeders a managed system rather than a daily scramble.
Four moves structure the feeder plan. The first is set standardisation — deciding a fixed feeder assignment per part number so the same component always lives in the same slot, which removes a whole class of placement errors. The second is staging and pre-kitting, building the next order's feeder stack offline so the changeover is a swap, not a build. The third is calibration and error-proofing, so a feeder that goes on the machine is known good and a wrong reel is caught before it loads. The fourth is counting and budgeting, so the plant owns enough feeders of each type to cover the active product list plus a changeover queue without starving the line. The Hanwha Samsung SMT tape feeder range the line is built on supports all four moves when it is planned as inventory, not as an afterthought.
Set standardisation means giving every active part number a fixed feeder slot on the machine, so the same component always occupies the same station regardless of which order it appears in. Without standardisation, the same 0402 resistor might sit in slot 12 on Monday and slot 27 on Tuesday depending on who built the setup, and the placement program has to be edited to match — which is both slow and error-prone. With standardisation, the program references slot 12 for that resistor every time, the operator pulls the pre-kitted feeder for slot 12, and the machine, the program and the physical setup agree by construction. The standardisation is captured in a slot map that all programs and all kits reference, and it is the backbone of every faster move in this chapter.
The slot map is built from the union of components across the active product list, not from a single board, because HMLV is about the whole list, not one order. The high-runners — the 0402 and 0603 passives, the common SOT-23 and SOT-89 transistors, the usual QFP and QFN devices — earn permanent dedicated slots near the centre of the machine where changeover impact is lowest. The long-tail parts earn shared or overflow slots and may move between orders, but even they follow the same map so the program logic stays consistent. Standardisation also pays off in training and troubleshooting: an operator who knows slot 12 is always that resistor can spot a wrong reel at a glance, and the Hanwha Samsung SMT tape feeder range with its consistent mechanical interface makes a fixed slot assignment physically easy to hold across the cell.
Feeder type | Best for | Changeover behaviour on a HMLV line |
|---|---|---|
8 mm tape feeder | 0402–1206 passives, small SOT and QFN | Highest count; assigned by the slot map and pre-kitted as a unit |
Stick feeder | LEDs, axial parts, odd-form devices | Lower count; staged separately and verified before load |
Tray (waffle / stick tray) | Large QFPs, BGAs, bare ICs | Loaded at the machine; orientation checked at first article |
Vibration feeder | Very small or loose components | Specialised; presentation verified per part number |
Staging and pre-kitting move the changeover work off the clock and onto a bench, so swapping products is a fast exchange rather than a build-from-scratch. Instead of pulling reels one at a time while the line is stopped, the next order's complete feeder set is assembled, verified and labelled on a separate cart or table before the current run ends. When the changeover comes, the full set is swapped in as a unit, the old set is swapped out as a unit, and the line is back running while the removed set is being broken down and re-kitted elsewhere. The changeover becomes a logistics hand-off, not an engineering task, and its duration is bounded by how fast a cart can be pushed, not by how fast an operator can think.
The discipline that makes staging work is a two-bin or two-cart rhythm: while cart A feeds the live line, cart B is being built for the next order, and the roles alternate. The kit for an order is checked against the slot map and the BOM before it reaches the machine, so a missing or wrong reel is caught on the bench, not mid-run. Pre-kitting also enables parallel changeovers, where one operator stages while another runs, which is what lets a HMLV cell hit two to ten product switches per shift without the line collapsing into setup. The feeder inventory has to be large enough to support two or more complete sets at once, which is where the budgeting subsection below earns its keep — staging fails the moment there are not enough feeders to build the next kit while the current one is still mounted.
Calibration and error-proofing keep a feeder that goes on the machine honest, because a mis-pitched, mis-registered or mis-labelled reel is the fastest way to scrap a run. Every feeder type has a defined pitch and a defined pick position, and Hanwha's 8 mm tape feeders, stick feeders and tray handlers each have to be set, verified and periodically re-calibrated so the head picks at the right spot every time. A feeder that drifts out of calibration places components a fraction of a millimetre off, which may pass a visual check and fail a first-article or a field return weeks later. Calibration is therefore not a commissioning task but a recurring one, tied to usage count and to any knock the feeder takes during a swap.
The error-proofing layer sits on top of calibration and stops wrong reels before they load. The simplest version is a barcode or RFID check at the machine: the feeder's ID and the slot it is being placed in are matched against the program's slot map, and a mismatch is rejected rather than silently accepted. A second layer is a visual and weight check at kitting time, where the reel label is confirmed against the BOM line. A third is the machine's own push-up and vacuum test that flags a feeder that cannot reliably present parts. Together these turn feeder handling from "trust the operator" into "the system confirms the operator", which is what lets a HMLV line change products repeatedly without accumulating placement errors. The same discipline is what the acceptability standards expect at first article, and it is far cheaper to enforce at the feeder than to discover at inspection.
Budgeting answers the practical question of how many feeders the plant actually needs, and the answer for HMLV is always more than the machine's slot count, because a healthy line keeps spare, kitted sets moving in parallel. A single machine may hold 80 to 120 feeders depending on the model and the head configuration, but if the plant only owns exactly that many, then every changeover requires stripping the machine to build the next kit, and staging collapses. The rule of thumb is to own enough feeders to cover the largest single order's distinct components, plus a second complete set for staging, plus a buffer for wear, loss and calibration cycles — often two to three times the slot count in active use across the cell.
The budget is driven by the component union across the active product list, not by any one board. Count the distinct tape widths and part types that appear across the 50 to 500 live part numbers, weight them by how many slots each needs at peak, and size the feeder pool to support two or more complete kits simultaneously. Reserve a slice of the budget for the long-tail packages — odd connectors, large ICs on trays, stick-fed parts — that appear rarely but must never block a changeover when they do. Treat feeders as capital that earns its keep by enabling changeovers, and track their utilisation the way you track machine uptime, because a feeder shortage is just as much a line stoppage as a broken gantry, only quieter and easier to blame on the operator.
Program management is the second axis, and for a HMLV Hanwha SMT line it is where the product list either becomes a repeatable, low-risk asset or a stack of one-off files that somebody has to rebuild every time. A high-volume plant may keep a handful of programs, each tuned over weeks; a HMLV plant keeps dozens to hundreds, and the difference between a smooth cell and a chaotic one is whether those programs are generated, versioned and reusable or hand-edited per order. The planning goal is to make every new part number a small delta on top of an existing, verified library rather than a fresh start, so the line builds the right board the first time, every time.
This chapter treats program management as a pipeline with four stages. The first stage is a single source of truth for CAD and BOM data, so the placement program is generated from verified data rather than retyped by hand. The second is versioning and the CAD-to-placement conversion, which keeps every program traceable to the revision it was built for. The third is panelisation and the trial-program workflow, which lets a new board be proven on a small run before volume. The fourth is off-line preparation, which keeps program work off the live machine so changeovers stay fast. The selection logic that decides which machine a program targets is developed in this series' Hanwha chip mounter selection framework, and the buying discipline that should precede the whole line is in the Hanwha mounter procurement checklist.
The single source of truth is the foundation, because every placement program is only as reliable as the CAD and BOM data it was generated from. In a HMLV plant the temptation is to keep programs as local files on the machine or on an engineer's laptop, edited per order, and that is where errors breed: a coordinate is nudged for one board, a centroid is corrected in a hurry, and the next person to build that board starts from the wrong baseline. The fix is to hold one authoritative component library and one authoritative BOM per part number, in a system of record that every program is generated from and nothing is edited against. The machine loads programs; it does not author them.
The data standard matters here. IPC-2581, the open XML-based product data exchange, lets the CAD centroid, the BOM and the stack-up move together as one validated package, so the placement program is built from the same data the bare-board fabricator and the stencil maker used. When the pick-and-place data, the BOM and the CAD agree by construction, a whole class of "the program does not match the board" errors disappears. For HMLV the payoff is compound interest: every new part number reuses the same verified library and the same generation step, so adding the 200th board is barely more work than adding the 20th, and the program for a repeated order is a known-good artifact rather than a fresh risk.
Versioning keeps each program tied to the board revision it was built for, which matters enormously when the same part number is revised and re-run over months or years. A HMLV plant will rebuild a board long after the original engineer has moved on, and without versioning the line is guessing which program matches which BOM. The pipeline therefore stores programs by part number and revision, regenerates them deterministically from the source-of-truth data, and keeps the previous version so a rollback is a lookup, not a re-invention. The CAD-to-placement conversion — centroid extraction, fiducial definition, nozzle and feeder assignment, placement order — runs as a repeatable recipe, not a manual art.
The pipeline is what makes the Hanwha chip mounter selection framework operational, because the selection decides which machine family a program targets, and the pipeline encodes that decision into the assignment step. A board heavy with fine-pitch QFNs routes to a flexible multi-function head; a board dominated by 0402 and 0603 chips routes to a high-throughput head; a mixed board splits across both. The program records that routing, so the next time the board runs, the line already knows which machine gets which placements. Versioning also supports the audit trail a customer or a quality system may demand: show the program, show the BOM it came from, show the revision, and the build is defensible end to end.
Panelisation and trial programs turn a risky new part number into a cheap, contained experiment before it touches volume. Most HMLV boards are run in panels — several boards per framed array with tooling strips and fiducials — and the panel definition, the fiducial strategy and the placement order all live in the program. A new board should first be run as a small trial panel: a handful of pieces built and fully inspected, the program corrected against what the trial revealed, and only then released to the full panel and the full batch. The trial is where the program meets the physical board, and almost every new program needs at least one correction when it does.
The workflow keeps the risk contained because the trial panel is small, the inspection is intense, and the correction is made to the program in the library, not to a one-off file. A common correction is a fiducial that the vision system reads unreliably, a component whose centroid was off, or a nozzle assignment that placed a part too slowly; all of these are fixed once, in the versioned program, and every future run benefits. For HMLV, where new part numbers arrive constantly, the trial-program workflow is the only way to keep first-pass yield acceptable without slowing the line to a crawl — the trial runs off-line and in parallel with production, and the corrected program is ready before the volume order needs it.
Off-line preparation is the discipline that keeps program work off the live machine, because a placement machine is far too expensive to use as a programming terminal. In a HMLV cell the next order's program is generated, verified and simulated on a separate station — using the same slot map, component library and machine model — while the current order is still running. When the changeover arrives, the machine loads an already-finished program; it does not wait for an engineer to type coordinates. The line's idle time is bounded by the load, not by the prep, and that is what lets a HMLV cell absorb two to ten product switches per shift.
Off-line prep also de-risks the changeover, because the program is validated before it touches the board. The verification step checks the program against the slot map, confirms every component has a valid feeder and nozzle assignment, and simulates the placement sequence to catch collisions or unbalanced head loading. The Hanwha mounter procurement checklist should include the off-line programming station as a line item, because the cell is only as fast as its slowest link, and a missing programming seat is a silent bottleneck. Plants that skip off-line prep discover it the hard way: the machine sits idle while an engineer edits a program on the HMI, and the throughput the hardware promised evaporates into setup.
Component range is the third axis, and it decides which Hanwha machine — or which combination of machines — the Hanwha pick and place machine for high mix low volume is built from, because no single placement head places every package well. A high-volume line can dedicate a machine to one narrow component window; a HMLV line has to place 01005 passives, large QFPs, connectors, LEDs and odd-form parts across the same product list, and the head selection is the engineering that makes that possible without constant re-tooling. The planning question is not "what is the fastest head" but "what head mix covers the component union at acceptable speed and accuracy".
Four subsections develop the head plan. The first matches the component mix to head type, because flexible multi-function heads and high-throughput chip shooters are different tools for different parts. The second plans nozzles and spindles for mixed packages, since the nozzle set is what actually touches the components. The third covers wide-body and long-board handling, which a HMLV line meets whenever a customer brings an oversized or 1,200 mm board. The fourth chooses the mounter tier for the mix, pulling the previous points into a concrete machine decision. Across all four, the aim is a line whose head configuration is wide enough to place the whole product list and focused enough to do it quickly — and that balance is the heart of HMLV planning.
The component mix decides the head type because placement heads trade flexibility for speed in predictable ways. A high-throughput chip shooter places small passives and simple ICs astonishingly fast but struggles with large, tall or odd packages; a flexible multi-function head places a far wider window — large QFPs, connectors, LEDs, through-hole-led parts on a hybrid line — but more slowly per component. A HMLV product list usually contains both, so the line needs both kinds of head, allocated by package rather than by board. The art is to keep the fast head doing the high-count simple parts and the flexible head doing the low-count awkward parts, so neither is the bottleneck for the wrong reason.
This is where the DECAN L2 earns its place on a HMLV line. The DECAN L2 multi-function flexible placer is built for exactly the mixed, awkward component window that HMLV throws at a line — it carries the connectors, the LEDs, the larger QFNs and the low-volume oddities that a pure chip shooter would choke on, while a faster neighbour machine handles the passive bulk. Pairing the L2's flexibility with a throughput machine means the line covers the whole component union without forcing every part through the slow head or every part through the narrow one. The head type is chosen from the component list, not from a speed contest, and the L2 is the flexible anchor of a balanced HMLV cell.
Component class | Recommended head | Planning note for HMLV |
|---|---|---|
01005–0603 passives, small SOT | High-throughput chip shooter | Highest placement count; concentrate on the fast head |
QFP / QFN, 0.4–0.5 mm pitch | Flexible multi-function head | Accuracy-critical; route to the tightest-placement-class machine |
Connectors, LEDs, odd-form | Flexible multi-function head | Low count, high variety; the flexible anchor carries these |
Large BGA / fine-pitch IC | Flexible head with tray feed | Verify at first article; allocate to one consistent machine |
Nozzles and spindles are the part of head selection that actually touches components, and they are planned from the package list, not assumed. Every component has a pick-and-place envelope — a minimum nozzle size, a maximum part height, a centre-of-gravity the vacuum must hold — and a HMLV line's nozzle set has to cover the union of those envelopes across all active part numbers. Too few nozzle types and the line cannot place some parts; too narrow a spindle assignment and the head spends time swapping nozzles mid-run, which is a hidden changeover cost that hurts HMLV more than high volume because the mix changes constantly.
The planning allocates nozzles by frequency: the common 0402, 0603 and standard QFP sizes earn dedicated spindles that never change, while the rare large or odd packages earn shared spindles swapped only when that part appears. The DECAN F2 high-throughput value mounter is a strong anchor for the high-count side of this plan, because its dual-gantry value architecture places the bulk chip volume fast while leaving the flexible head to manage the nozzles that the mix needs less often. The nozzle plan is captured in the same slot map and component library the feeders use, so a program knows which nozzle each part needs and the changeover only swaps what the next order actually requires. Good nozzle planning is invisible — the line just places everything — and that invisibility is the goal.
Wide-body and long-board handling is the edge case that a HMLV planner ignores at their peril, because the moment a customer brings a 1,200 mm board or an oversized panel, a line sized only for standard 50 × 50 to 510 × 460 mm boards simply cannot build it. HMLV contract manufacturers live on variety, and variety includes board size, so the line's conveyor, clamping and board-support strategy has to cover the long tail of larger formats. The Hanwha platform handles this with wide-body models and dual-track options: the SM482Plus stretches to long boards, and several models offer a dual-track conveyor that processes two boards or panels in parallel.
Dual-track deserves specific attention in HMLV, because it is one of the few hardware features that directly shrinks changeover impact. With two lanes, the line can be loading one board while placing the other, or running two different products at once, which is a structural win for a plant that changes products often. The cost is some complexity in balancing the two lanes, but for a HMLV cell the trade is usually worth it. The planning decision is to size at least one machine in the cell for the largest board the product list realistically contains, and to reserve the dual-track option where batch sizes and mix support parallel flow — that way a long-board order is a normal job, not an exception that breaks the line's assumptions.
Choosing the mounter tier is the synthesis of the previous three subsections: it turns the component union, the nozzle plan and the board envelope into a concrete machine list. For most HMLV cells the answer is a composed line, not a single flagship, because no one machine is both the most flexible and the fastest. A typical HMLV cell pairs a flexible multi-function mounter for the awkward window with one or two higher-throughput machines for the chip bulk, and may add a fast single-gantry machine to relieve a bottleneck when volume spikes. The tier is chosen from the mix, and the mix is taken from the real product list, not from a vendor's headline CPH.
The practical method is to chart the active part numbers by component window and by batch size, then assign each cluster to a machine tier that places it well. The flexible window goes to a multi-function head, the chip bulk goes to a value-throughput head, and any fine-pitch or high-accuracy cluster gets the machine with the tightest placement class. Because the Hanwha families share a feeder form factor and a programming environment, the resulting line is re-composable as the product list shifts — add a machine, rebalance the slot map, and the cell adapts. The next two chapters show how that line is verified at first article and balanced for throughput, and the final chapter turns the whole plan into reference configurations built on these same tiers.
First-article inspection is the fourth axis and the quality gate that lets a HMLV line run many part numbers without shipping defects, because in HMLV the board you are building today may never be built again the same way, so every new or changed part number has to be proven before it reaches volume. A high-volume plant can amortise a long qualification across a million boards; a HMLV plant qualifies constantly, and the first-article workflow is the repeatable process that makes that affordable. It is the moment the program, the feeders, the stencil and the board meet for the first time, and the moment most hidden errors surface if they are going to.
The chapter builds the workflow in four steps. The first treats the first-article build as a risk gate, not a formality, where the line proves it can build the board at all. The second applies the IPC-A-610 acceptability standard and a concrete inspection checklist, so "good" is defined before the board is judged. The third covers measurement — placement accuracy and solder-joint quality — which turns the subjective into the quantitative. The fourth closes the first article and releases to volume only when the evidence is in. The verification discipline extends beyond the board to the machine itself: the same evidence-based mindset is what the factory and site acceptance checklist for SMT lines applies when a Hanwha machine is accepted, and the two checklists share a philosophy — measure, don't assume.
The first-article build is the risk gate because it is the first time the entire chain — program, feeders, stencil, placement, reflow — operates on the real board, and any weakness in that chain shows up here first. Treat it as a deliberate, contained experiment: build a small quantity, often a single trial panel, and inspect it intensely rather than running the full batch and hoping. The gate's job is to answer one question before volume — can this line build this board to spec — and to answer it with evidence, not optimism. A first article that passes unlocks the order; one that fails is cheap to fix at five pieces and ruinously expensive at five hundred.
The gate is especially important in HMLV because the product list turns over constantly and the line rarely benefits from repetition on a single board. A high-volume plant learns a board over a long run; a HMLV plant gets one first article and then may not see the board again for months, so that first verification has to be thorough enough to stand in for experience. The first-article build therefore exercises every placement the program specifies, on the actual feeders and nozzles assigned, using the actual stencil and reflow profile, so what is verified is the real process and not an idealised one. Skipping or rushing this gate is the most common way a HMLV line ships a defect on a board it has never run before, and the cost lands on the customer and the return rate.
IPC-A-610 is the acceptability standard that defines what "good" means before the first article is judged, and using it turns inspection from opinion into a shared, auditable criterion. The first-article checklist derived from it covers the things that actually fail: component presence and correct identity at every coordinate, polarity and orientation, solder-joint shape and fill, tombstoning and skewing on passives, bridging and insufficient solder on fine-pitch leads, and any mechanical damage from handling. Defining the criteria up front means the inspector knows what to look for and the build knows what it will be measured against, which removes the ambiguity that causes disputes and escapes.
The checklist is run on the trial panel with both automated and manual inspection — SPI and AOI catch solder-volume and placement defects at speed, while a manual optical and sometimes X-ray check confirms the joints the machines cannot fully judge, especially on bottom-termination and hidden-lead parts. Each finding is recorded against the program coordinate it relates to, so a defect points back to a specific placement, feeder or stencil aperture rather than to a vague "something was wrong". That traceability is what makes the first article useful: it does not just say the board passed or failed, it says why, and the why is what the next subsection's measurement and the final subsection's closure act on. The same evidence-based habit underpins the broader acceptance work on the line.
Measurement is what turns the first-article inspection from a visual impression into quantified evidence, and the two numbers that matter most are placement accuracy and solder-joint quality. Placement accuracy is measured by comparing the actual component position to the program coordinate — typically expressed as a mean offset and a spread in microns — and checked against the placement class the machine was quoted for, often in the ±25 to ±40 μm range for modern Hanwha heads. A first article that places within that window on the real board confirms the program, the fiducials and the machine are all aligned; one that drifts outside it flags a calibration, fiducial or program error to fix before volume.
Solder-joint quality is measured through the reflow result: SPI verifies solder-paste volume and registration before reflow, and post-reflow inspection verifies fillet shape, wetting and absence of defects like voids, bridges or heads-up pillows. For fine-pitch and bottom-termination parts, X-ray measurement of hidden joints is part of the first article, because those defects cannot be seen optically and will not announce themselves until field failure. The measurement step is where HMLV planning meets the machine's real capability: a line planned around a tight placement class has to prove that class on the first article, on the customer's board, or the planning assumption was wrong and the line needs rebalancing. Good measurement makes that discovery cheap and early.
Closing the first article is the disciplined step that releases a part number to volume only when the evidence is complete, and it is the difference between a verification that protects the line and one that merely documents a problem after the fact. Closure means the inspection checklist is fully answered, the measurements are within the defined limits, every finding is traced to a cause and a correction, and the corrected program and stencil are updated in the library before the next build. Only then is the part number marked released, and only then does the volume order run. A first article that found issues but was not corrected in the source artifacts would simply repeat the defect on the full batch.
The closure record also becomes a reusable asset for the HMLV line, because the next time that part number runs — perhaps months later — the released program, the verified feeder set and the inspection baseline are already known good, and the changeover is a reload, not a re-qualification. This is the same evidence-first discipline the factory and site acceptance checklist for SMT lines applies when a Hanwha machine itself is accepted, and the two reinforce each other: a verified machine and a verified first article together make a HMLV run defensible end to end. The planning payoff is that qualification, done properly once, pays back on every repeat of the board, which is exactly how a HMLV plant keeps yield high across a turning-over product list.
Line balancing is the fifth axis and the one that converts the other four into actual shipped volume, because a Hanwha pick and place machine for high mix low volume is only as fast as its slowest machine, and the slowest machine changes every time the product changes. In high volume, balance is computed once for a single board and stays put for weeks; in HMLV, balance has to be recomputed per order, or at least per product family, so the placement load is spread evenly across the cell instead of piling onto one machine while another waits. The planning goal is a line whose machines share the work well enough that adding a product does not require re-engineering the cell.
This chapter covers four balancing moves. The first sets takt time and the real, achievable CPH of each machine, because quoted CPH is not placed CPH and balancing on the wrong number breaks the line. The second allocates placements across a dual-machine line so neither machine is starved or buried. The third uses buffers and dual-track to absorb changeovers without stopping the flow. The fourth brings in a fast single-gantry machine — the DECAN S1 high-speed single-gantry mounter — to relieve a bottleneck when the mix skews toward volume. Balancing is where HMLV planning pays off in throughput, and it is also where a poorly planned cell quietly loses every gain the fast machines promised.
Takt time is the cadence the line must hold to meet demand — the available production time divided by the number of boards required — and it is the yardstick every machine is balanced against. The trap is to balance against quoted CPH, the headline number from the datasheet, when the number that matters is real CPH: the placed-component rate after fiducial reads, nozzle changes, board transfer, and the inevitable small stops are accounted for. A Hanwha head quoted at 80,000 CPH may place closer to 55,000 to 65,000 CPH in a real HMLV mix with frequent fiducial reads and varied packages, and balancing on the 80,000 figure overloads that machine by a third.
The planning method is to measure or conservatively estimate real CPH per machine for the actual component mix, then divide the board's total placements by takt time to get the required placed-CPH, and confirm the cell's real CPH sum exceeds it with headroom. The headroom matters more in HMLV than in high volume, because the mix shifts and a machine that is balanced exactly to takt will fall behind the moment a heavier board appears. Building in a buffer — planning the cell to roughly 15 to 25 percent above takt — is what keeps the line meeting commitment when the product list turns over. Real CPH, not quoted CPH, is the only honest input to that calculation, and the first-article measurement from the previous chapter is part of how you confirm it.
Allocation is the core act of balancing: deciding which placements each machine in the cell performs so the load is even. On a two-machine HMLV line the natural split is by package and by count — the flexible multi-function machine takes the awkward, low-count parts (connectors, LEDs, large QFNs), and the high-throughput machine takes the chip bulk (0402, 0603, small ICs). The split is tuned so both machines finish their share in roughly the same time, which is the definition of balance. A line where the flexible machine finishes in three minutes and the throughput machine in nine is unbalanced and effectively capped at the nine-minute machine's rate.
The allocation is captured in the program and the slot map, so it is repeatable per part number rather than re-argued per order. For a turning-over HMLV product list, the allocation should be parameterised: define the split rule by component class, then let the generation pipeline apply it to each new BOM, so a new board is balanced automatically instead of by hand. Where a board's mix is unusual — heavy on fine-pitch, say — the rule is overridden for that board and recorded, so the exception is explicit. Good allocation means the cell's throughput is the sum of balanced machines, and the slowest machine is slow by design, not by accident of an unbalanced split that nobody noticed.
Machine in cell | Role | Component class | Placements / board | Cycle impact |
|---|---|---|---|---|
Flexible multi-function (L2-class) | Awkward window | Connectors, LEDs, large QFN | 180 | ~3.2 min |
Value-throughput (F2-class) | Chip bulk | 0402–0603, small IC | 620 | ~3.4 min |
Fast single-gantry (S1-class) | Relief valve | High-count chips on volume runs | offloads 300 | relieves F2 |
Buffers and dual-track are the structural tools that let a HMLV line keep flowing through changeovers instead of stopping for them. A small buffer — a board loader, a conveyor staging area or simply a few panels held between machines — lets one machine keep placing while the other changes over, so a changeover on machine B does not force machine A to idle. Without that buffer the line is a rigid chain in which the slowest event stops everything, and HMLV's frequent changeovers then stop the whole cell repeatedly through the shift.
Dual-track conveyor is the stronger version of the same idea: with two lanes, the line can load and place two boards in parallel, or run two different products at once, which structurally reduces the per-board changeover penalty. Combined with a buffer, dual-track lets a HMLV cell absorb a product switch on one lane while the other lane keeps producing, turning a full stop into a local slowdown. The trade is some balancing and programming complexity, but for a plant that changes products several times a shift the complexity pays for itself in uptime. The planning choice is to size at least one machine in the cell with dual-track and to design the board flow so a buffer exists between the placement machines and the downstream reflow and inspection, because that buffer is what converts changeover frequency into a manageable cost.
The fast single-gantry tier exists to relieve a bottleneck, and on a HMLV line that relief is often the difference between meeting takt and missing it. When the product mix skews toward volume — a run of several hundred boards of a chip-heavy design — the high-throughput machine can become the constraint even after balancing, because the bulk placing dominates. Adding a DECAN S1 high-speed single-gantry mounter to the cell offloads the highest-count simple components, freeing the flexible and value machines to focus on the parts only they can place, and the cell's real CPH rises without rebalancing the whole architecture.
The S1 is a good HMLV relief valve precisely because it is a single-gantry speed entry rather than a flagship: it is cheaper to add than a top-tier machine, it shares the Hanwha feeder form factor and programming environment so it drops into the existing cell without a new logistics layer, and it slots into the allocation rule as "chip bulk overflow". The planning insight is to design the cell with a placeholder for this relief tier from the start, so adding it later is a configuration change, not a redesign. A HMLV line that plans for the bottleneck before it appears can relieve it in an afternoon; one that did not plan for it discovers the bottleneck only after a customer order exposes it, and then spends weeks re-architecting the cell under pressure.
This chapter turns the five axes into concrete reference configurations, so the planning has a starting point rather than an open-ended blank page. The right size depends on the product list: how many active part numbers, how wide the component window, how fine the pitch, and what batch sizes dominate. A plant with 50 active variants and mostly standard packages needs a different cell from one with 500 variants and regular fine-pitch work, and the two reference cells below bracket the usual HMLV range. The point is not to copy a configuration blindly but to see how the feeder, program, component, first-article and balancing decisions land on real Hanwha hardware.
Four subsections make the references concrete. The first is a 1–2 mounter cell for 50 to 200 variants, the entry point most HMLV contract manufacturers start from. The second is a 3–4 mounter line for 200-plus variants and fine pitch, where the component window and the volume both justify more machines. The third weighs new versus used and the budget entry point, because a HMLV line is often a capital-constrained purchase and the used market is a legitimate option when verified. The fourth is the acceptance and handover checklist before go-live, the step that turns a purchased cell into a working, measured asset. The budget entry in particular is worth a close look: the SCM1-D pre-owned entry SMT machine can anchor a small HMLV cell at a fraction of new-machine cost when its condition is properly verified.
The 1–2 mounter cell is the right starting point for a plant with 50 to 200 active part numbers, mostly standard SMD packages and batch sizes in the tens to low hundreds. The simplest version is a single flexible multi-function mounter — a DECAN L2-class machine — that covers the whole component window on its own; this is the lowest-cost way to enter HMLV and it suits a plant whose volume per order is modest and whose mix is wide but not extreme. The limitation is throughput: a single head places everything serially, so as volume grows the cell hits its real-CPH ceiling and changeovers start to dominate the day.
The two-mounter version adds a second machine to split the load — typically a flexible head for the awkward window paired with a value-throughput head for the chip bulk — and this is where most serious HMLV cells land. The pair covers the component union, balances the placement load, and absorbs changeovers far better than a single machine, because the second machine provides the redundancy that lets one side reconfigure while the other runs. For a plant starting out, a pragmatic path is to open with the SCM1-D pre-owned entry SMT machine as the flexible anchor and add a throughput machine as volume justifies, which keeps the initial capital low while preserving the composed-line architecture the earlier chapters described.
The 3–4 mounter line is for a plant with 200-plus active variants, a wide component window that includes fine-pitch QFNs and bottom-termination parts, and enough volume that a two-machine cell would bottleneck. The composition typically pairs a flexible multi-function machine for the awkward window, one or two value-throughput machines for the chip bulk, and a fast single-gantry machine as the relief valve for chip-heavy runs. Three to four machines let the allocation rule spread placements finely, so no single head becomes the constraint and the line holds takt across a broad mix.
Fine pitch is the specific driver for the larger cell, because tight-pitch placement concentrates on the most accurate head and would overload a smaller cell. With three or four machines, the fine-pitch cluster routes to the tightest-placement-class head, the bulk routes to the throughput heads, and the relief valve handles overflow, so the line balances even when one order is fine-pitch-heavy and the next is chip-heavy. The cost is more capital and more balancing complexity, but for a plant at this scale the alternative — an under-powered two-machine cell missing takt daily — costs more in late orders and expedites. The four-axis planning (feeder pool, program pipeline, component range, first-article gate) scales linearly with machine count, so the cell built this way stays manageable as it grows.
New versus used is a real decision for a HMLV line, because the budget often does not stretch to a full new cell and the used market offers capable Hanwha-class machines at a fraction of the cost — but only when their condition is verified, not assumed. A new machine ships with a known build, a current configuration and a warranty, so its planning risk is low; a used machine ships with a history, and its real accuracy, calibration and remaining component life have to be proven before it is trusted with production. The discipline that governs that proof is laid out in this series' used versus new Hanwha mounter inspection guide, and it applies directly to the HMLV cell decision.
The SCM1-D is the natural budget entry: the SCM1-D pre-owned entry SMT machine can anchor a small HMLV cell at low capital cost, leaving budget for the feeder pool, the programming station and the inspection the line actually depends on. The planning rule is to treat the used machine's purchase price as only part of the cost — budget for verification, for any head or feeder refurbishment, and for a contingency — and to weigh that total against a new machine's lower risk. For many HMLV startups the right answer is a verified used flexible anchor plus one new throughput machine, which captures most of the capability at a manageable price while keeping the risk concentrated in a single, well-inspected asset rather than spread across an all-used cell.
Acceptance and handover is the final step that converts a purchased or built cell into a measured, trusted asset, and it is where the first-article discipline of chapter five meets the machine-level verification the line deserves. Before the cell goes live, every machine is accepted against its specification — model, configuration, placement accuracy and board envelope — using the same evidence-based mindset, and every program, feeder set and first article is confirmed on the actual hardware. The handover checklist records what was verified, by whom, and against what standard, so the cell enters production with a known-good baseline rather than a hope.
The checklist is short but non-negotiable: confirm each machine's exact model and configuration; verify placement accuracy on a trial board within the quoted class; confirm the feeder pool covers the active product list with staging sets; confirm the program pipeline generates from the source-of-truth data; run and close a first article for each launch part number; and balance the cell against takt with real CPH. The used versus new Hanwha mounter inspection guide supplies the machine-level verification detail, especially for any pre-owned anchor such as the SCM1-D. A cell that passes this checklist is not merely installed; it is proven, and the proof is what lets a HMLV plant change products confidently, knowing the line will place what the program says it will.
Plan your Hanwha SMT line for high-mix low-volume production
A HMLV line is planned as a system, not bought as a single fast machine, and the five axes covered in this guide — feeder preparation, program management, component range, first-article verification and line balancing — are the levers that decide whether your Hanwha pick and place machine for high mix low volume absorbs variety or drowns in changeovers. The practical next step is to map your real product list before you commit capital: count the active part numbers, chart their component windows and batch sizes, and size the feeder pool, the off-line programming pipeline and the head mix against that measured data rather than against a vendor headline CPH. If you are weighing a new build against a verified pre-owned machine, or scoping a one-to-two mounter cell against a three-to-four mounter line, the trade-offs are concrete and the wrong call is expensive, so it is worth doing the maths on paper first. The Hanwha platform supports either path because its families share a feeder form factor and a programming environment, which means the cell you build today can be re-composed as your product list changes tomorrow, and the line you size to the mix you actually run stays a working asset long after the invoice is paid.
The practical limit is set by changeover time, not by the machine's placement rate. A well-planned cell with staged feeder sets, off-line program prep and a disciplined first-article workflow typically manages two to ten product switches per shift, which across a shift can cover dozens to a few hundred distinct part numbers in a day depending on batch size. The number rises when changeovers are engineered down — pre-kitted feeders, parallel programming, dual-track flow — and falls when the line rebuilds setups by hand. The planning lever is therefore changeover cost: shrink it and the part-number capacity rises without buying a faster head.
There is no single best machine; the right answer is a composed cell. A flexible multi-function mounter such as the DECAN L2 carries the awkward, low-count component window, while a value-throughput machine such as the DECAN F2 carries the chip bulk, and a fast single-gantry DECAN S1 can be added to relieve bottlenecks on chip-heavy runs. For the smallest HMLV operations a single flexible head may suffice, while larger mixes need three or four machines. The choice follows the component union and batch sizes of your real product list, not a speed contest between individual models — in short, the best Hanwha pick and place machine for high mix low volume is the cell, not the box.
A target changeover for HMLV is measured in minutes, often five to fifteen, not in the half-hour-plus that hand-built setups incur. The cuts come from feeder set standardisation (same component, same slot, every order), pre-kitting the next order's feeders on a separate cart, off-line program preparation so the machine only loads finished programs, and error-proofed feeder loading that catches wrong reels before they mount. Buffers and dual-track further hide changeovers behind continuing production. Each of these is covered in the feeder, program and balancing chapters and together they are what make frequent product switches affordable.
Not strictly, but it helps materially once you change products several times a shift. Dual-track lets the line load and place two boards or two products in parallel, so a changeover on one lane does not stop the other, and it provides the structural buffer that keeps flow alive through swaps. For a plant with modest changeover frequency a single-track cell with good staging may be enough; for a plant at two-to-ten switches per shift, dual-track on at least one machine is usually worth the balancing complexity it adds, because it converts a full stop into a local slowdown.
Make first-article inspection a repeatable, off-line workflow rather than an ad-hoc check. Build a small trial panel first, inspect it against an IPC-A-610-based checklist, measure placement accuracy and solder joints, trace every finding to a program or feeder cause, correct the source artifact, and only then release the part number to volume. Because the corrected program and feeder set are stored as released assets, the next run of that board reloads a known-good configuration instead of re-qualifying from scratch. The discipline scales with the product list because the work is done once per part number and reused on every repeat.
Either is viable, and the decision rests on verified condition versus risk. A new machine ships with a known build and warranty, so planning risk is low; a used machine must have its accuracy, calibration and component life proven before production trust. A common HMLV strategy is a verified pre-owned flexible anchor such as the SCM1-D paired with one new throughput machine, capturing most capability at lower capital while concentrating risk in a single inspected asset. Budget for verification and any refurbishment either way, and apply the used-versus-new inspection discipline to any pre-owned machine before it enters the cell.
Plan for more than the machine's slot count, because a healthy HMLV line keeps spare kitted sets moving in parallel. A practical rule is enough feeders to cover the largest single order's distinct components, plus a second complete set for staging, plus a buffer for wear, loss and calibration cycles — often two to three times the slot count in active use across the cell. The budget is driven by the component union across the active product list, not by any one board, and a feeder shortage is a line stoppage as real as a broken gantry, only quieter. Treat feeders as capital that earns its keep by enabling changeovers.
Before installation, gather the active part-number list with their BOMs, the component windows and package types, the board-size envelope including any long or wide boards, the batch-size distribution, and the required takt time from your demand. That data sizes the feeder pool, the programming pipeline, the head mix and the balancing rules. Holding it as a single source of truth — ideally in an IPC-2581-aligned format so CAD, BOM and placement stay consistent — lets every program be generated rather than retyped, which is the foundation of low-error HMLV production and the prerequisite for the first-article and balancing work this guide describes.