Publish Time: 2026-10-26 Origin: Site
Specification matching guide · Published by I.C.T · For SMT process, production and project engineers
Specification sheets reward the reader who compares numbers; production rewards the reader who reconciles them. A pick-and-place machine's rated CPH, its placement accuracy and its PCB size envelope are not three independent scores — they are three faces of one trade-off, and the right machine is the one whose trade-off matches the boards, the mix and the takt time the line actually runs. Rated at optimum, machines in the same family headline anywhere from 22,000 to 92,000 CPH; accuracy spans a range from ±40 to ±25 microns; and the same catalog stretches from a 460 by 250 mm dual-rail window to 1,200 mm long boards. The difference between a line that hits takt and a line that investigates itself is rarely any single number — it is whether the three were matched. This guide reads the specifications the way the schedule reads them: jointly, against real boards, in the order the Hanwha placement machine catalog itself is structured.
The method follows the logic of the Hanwha pick and place machine CPH accuracy PCB size trade-off itself: CPH first, because throughput is where optimism concentrates and where IPC-9850's measurement discipline matters most; accuracy second, because it is the specification that protects yield and cannot be retrofitted; PCB size third, because geometry is the constraint that quietly eliminates half the catalog before any commercial question is asked; and then the two layers no datasheet displays — product mix and takt time, which decide how much of any rated number the line will ever see. The closing sections translate the three matched specifications into line balancing: how the mounter's realistic rate is aligned with the printer and the reflow oven, when one machine becomes two or a dual-rail configuration, and how the whole analysis ends in a configuration that survives contact with production rather than one that merely survived the quotation.
Table of Contents
PCB size and panelization: geometry that constrains the choice
Line balancing: the mounter against printer, oven and schedule
Every specification on a mounter datasheet is easy to read alone, and that is precisely the trap. Rated CPH sells the machine, accuracy protects the yield, and PCB size defines the physical envelope — and each of the three, maximized in isolation, quietly taxes the other two. A buyer who selects on CPH alone buys a speed the boards may never see, because the rating was measured on geometry and components their products do not have. A buyer who selects on accuracy alone overpays for precision the BOM never exercises. A buyer who selects on board size alone discovers late that the only machines with the right window run fewer feeders than the census requires. The three specifications form a system, and the machine decision is sound only when the system is balanced — which is why the matching exercise in this guide reads them jointly, in a fixed order, against the production reality the line will face.
The section is built around four questions: what each specification actually governs, in investment terms; why the three correct each other and why the corrections always flow in predictable directions; where the product-mix layer sits — the variable that no datasheet shows but that decides how much of every rated number is real; and what matching method the rest of the guide follows. Answer those four and the reader holds the frame on which every later section hangs: CPH against takt, accuracy against the smallest package, geometry against the largest panel, and the whole set against the mix the factory actually ships.
Read as investment language, the three specifications manage three different risks. Rated CPH governs time: it decides, after the efficiency factor is applied, how many machines the schedule needs and how much floor, staffing and maintenance each of them carries. Accuracy governs yield: it sets the smallest package and the finest pitch the line can place at a first-pass yield worth shipping, and it is the one capability in the trio that cannot be upgraded in the field — no service visit converts a ±40 micron platform into a ±25 micron one. PCB size governs geometry: it fixes which machines the panels physically fit, which rail widths and support schemes they demand, and whether a dual-rail, wide-body or long-board tier enters the conversation at all.
The three risks also fail differently, and the failure asymmetry is worth internalizing before any datasheet is opened. A CPH shortfall announces itself slowly — overtime, weekend shifts, a capacity review a year after commissioning — and is usually repairable by adding a machine. A geometry mismatch announces itself immediately and embarrassingly: the panel does not fit, and the configuration conversation restarts. An accuracy shortfall is the quiet one: boards pass, ship, and come back, or fail at test in ways that cost multiples of the machine's price before anyone connects the losses to a specification that was skimmed. Matching the three specs is therefore not an exercise in finding the biggest numbers; it is an exercise in pricing three different failure modes against the same production plan.
The trade-offs run in consistent directions, which is what makes joint reading possible. Rated CPH is measured at the optimum: minimal travel, lightest components, full spindle utilization on a board shaped to flatter the machine. Every departure from that optimum — larger boards, heavier components, tray picks, fine-pitch vision time — spends rated speed, and the accounting is unforgiving: a 92,000 CPH flagship can run a mixed industrial board at a third of its rating without anything being wrong. Accuracy behaves as the counterweight: the platforms and configurations that hold the tightest tolerances spend time doing it, either in slower vision processing on hard rows or in architecture that trades raw cycle count for controlled placement. And PCB size taxes both: a long panel rides the conveyor longer per placement opportunity, and the machines with the widest windows are usually not the machines with the highest ratings.
There is a disciplined exception that proves the rule. The DECAN S2's 92,000 CPH alongside a ±25 to ±28 micron accuracy class shows what a platform generation buys — the magnetic-levitation dual gantry holds flagship speed and flagship precision at once, and that combination is exactly what the premium is paying for. But the exception does not repeal the corrections; it reprices them. The board-size envelope, the feeder count and the configuration cost still apply, and the buyer who reads the S2 as "fast and accurate, so geometry no longer matters" will rediscover the third specification the day the first 1,200 mm panel arrives. The correct habit is to let every number check the others: speed against the boards that must run fast, precision against the parts that must run right, geometry against the panels that must run at all.
Between the three rated specifications and the machine the factory should buy sit two variables the datasheet never prints: the product mix and the takt time. Product mix decides the efficiency factor — the fraction of rated CPH the line actually realizes — because every additional product family adds its own feeder map, its own tray stations, its own nozzle assignments and its own changeovers, and the machine spends rated speed on all of that overhead rather than on placements. A single-product line on chip-dense boards might hold 0.6 to 0.7 of rating; a high-mix job shop running boards with stage-vision rows and tray content can live at 0.3 to 0.5. The same rated machine, therefore, is a different production asset in the two factories, and no comparison that ignores the mix can rank machines honestly.
Takt time is the other half of the hidden layer, and it converts the schedule into the only throughput number that matters: seconds per panel. A line contracted to deliver a certain volume in a certain shift count has a takt, and the mounter's realistic rate — not its rating — must beat that takt with margin. The two hidden variables also interact: high mix lowers the efficiency factor, which raises the realistic rate required, which tightens the takt relationship, which can flip a one-machine answer into a two-machine answer without any rated figure changing at all. This is why the later sections treat mix and takt with the same rigor as the printed specifications — because in the arithmetic that actually decides the purchase, they are specifications, just unwritten ones.
The method runs in a fixed order, chosen so that each step filters the catalog before the next adds nuance. Rated CPH is converted to a realistic rate using the board's efficiency profile, and the result is compared against the takt the schedule sets — which usually resolves the machine count and eliminates the tiers that cannot reach it. Accuracy is matched to the smallest package and finest pitch the BOM and its roadmap contain, which selects the platform generation. PCB size and panelization are matched to the working envelope, which selects the machine tier and the rail configuration. Product mix then re-checks the efficiency factor and the changeover architecture, and line balancing aligns the mounter with the printer and oven on either side. Only after all five filters have run does configuration detail — feeders, trays, options — become worth pricing.
The order is not arbitrary; it follows cost of discovery. Geometry mismatches are cheapest to find on paper and most expensive to find on the floor, so they filter early even though they feel like a detail. Accuracy is nearly impossible to correct after purchase, so it filters before price negotiations create pressure to compromise. CPH is the most adjustable — machines can be added — but the efficiency factor must be honest or every downstream number inherits the error. Table 1 summarizes the frame: each specification, the decision it governs, and the failure mode that follows when it is read alone. The sections that follow work the frame column by column.
Specification | What it decides | Failure mode when read alone |
|---|---|---|
Rated CPH | Machine count, line speed, floor plan | Speed purchased that the product mix never releases |
Accuracy class | Smallest package, finest pitch, yield floor | Field losses that dwarf the machine price |
PCB size / panel | Machine tier, rail width, support scheme | Panels that do not fit, or a tier bought too wide |
Product mix (hidden) | Efficiency factor, changeover architecture | Realistic rate far below the rated headline |
Takt time (hidden) | Required realistic rate, machine count | A line that misses shipments at full rated speed |
CPH — components per hour — is the most quoted number on any mounter datasheet and the most misused in purchase meetings. The confusion is structural: the figure is true, but it is true of a measurement condition, not of the buyer's boards. Rated values are taken at the optimum, under the standardized discipline of IPC-9850, and every real product departs from that optimum in ways that spend speed at different rates. In the Hanwha pick and place machine CPH accuracy PCB size matching exercise, CPH is deliberately read first — not because it is the most important specification, but because it is the one most distorted by optimistic reading, and an honest throughput number has to exist before accuracy and geometry can be priced against it.
Four sub-sections work the throughput dimension. The first states what the rated figure actually measures and what it is for. The second builds the bridge from rated to realistic — the efficiency factor — and shows where the factor comes from on real boards. The third walks the Hanwha CPH ladder from 22,000 to 92,000 and reads it as a decision table rather than a leaderboard. The fourth defines takt time and uses it as the judge: the schedule, not the datasheet, ultimately decides how much machine the line needs.
A rated CPH figure is a capability statement: the throughput the machine achieves on a reference board designed to minimize travel, with the component mix weighted to the packages the head cycle handles fastest, and the vision load kept to the in-flight recognition that adds no dwell. Under IPC-9850, the measurement discipline the industry shares, that figure is repeatable and comparable across machines — which is exactly what it is for. Where buyers go wrong is not in doubting the number but in misusing it: a rating computed for comparison is read as a production forecast, and the gap between the two is then attributed to the machine rather than to the reading.
The correct use of rated CPH is therefore ranked comparison and envelope confirmation. It ranks machines within a class honestly, because every machine is measured under the same flattering conditions. It confirms that a candidate platform operates in the right throughput league for the application before deeper analysis is spent on it. What it does not do is size a line — sizing starts when the rating meets the buyer's own geometry and mix, which is the work of the next two sub-sections. Buyers who keep the rated figure in its lane find that datasheet numbers and floor results finally agree; buyers who treat the rating as a promise spend years in a reconciliation neither party can win.
The bridge from rated to realistic is one number with several inputs: the efficiency factor, defined as realized throughput divided by rated throughput. Chip-dense boards with small travel and high head utilization run 0.5 to 0.7 of rating on machines in this class. Mixed boards — stage-vision rows, tray content, connector-length verification — run 0.3 to 0.5. The inputs are mechanical and traceable: each placement's travel time on the actual board outline, the proportion of picks that route through slower stations, the vision time spent on hard components, the panel transport that interrupts placement, and the changeover overhead amortized across the batch the mix implies.
Two disciplines make the factor honest. First, it is computed per product family, not averaged across the portfolio at purchase time, because the family that dominates next quarter's schedule — not the portfolio average — sets the rate the line must hold. Second, it is verified, not asserted: a demonstration on the buyer's own panel, timed and counted, converts the factor from an assumption into a measurement, and every subsequent calculation inherits that measurement instead of a hope. With the factor fixed, the arithmetic is simple and public — realistic CPH equals rated CPH times the factor; required machines equal schedule placements per shift divided by realistic placements per machine-shift — and the 20 to 30 percent margin the line should hold over demand is applied where it belongs, on the required count, not inside the factor, where it would quietly hide capacity the schedule is borrowing.
The range's rated figures form a ladder, and each rung exists because a class of production needs it. The SM485 holds the lowest headline at 22,000 CPH because its job is insertion: its four-plus-one force-controlled configuration buys pin-in-paste capability, not speed. The SM482Plus at 30,000 trades speed for the widest body and connector tolerance in the SM series; the SM481Plus at 40,000 widens the IC window further up the capability curve. The DECAN generation restarts the ladder a class tighter: the S1 at 47,000 as the single-gantry speed entry, the L2 at 56,000 as the multi-function and LED-optimized tier, the F2 at 80,000 as the dual-gantry value speed machine, and the S2 at 92,000 as the flagship where magnetic-levitation architecture holds both the top rate and the tightest accuracy. The SM471Plus, at 78,000 with its dual-gantry chip-shooter focus, remains the high-volume chip specialist of the SM platform.
Read as a leaderboard, the ladder invites the wrong purchase; read as a decision table, it filters cleanly. The question is never "which number is highest" but "which rung does the schedule's realistic rate require, after the efficiency factor, with margin" — and then, within the surviving rung, which architecture serves the accuracy and geometry rows that the next sections develop. Table 2 assembles the ladder with the context that makes the figures meaningful, and it is worth noting what the table already implies: several production problems in this range have two rungs that answer them, which is why the specification matching continues past CPH rather than ending on it. The DECAN S2 92,000 CPH flagship mounter anchors the top of the table, and the tiers below it each buy something the rung above traded away.
Machine | Rated CPH (optimum) | Architecture focus | Accuracy class |
|---|---|---|---|
SM485 | 22,000 | 4+1 force-controlled insertion head | Insertion-focused configuration |
SM482Plus | 30,000 | Wide-body tolerance, stage vision options | ±40 μm chip class; ±30 μm stage |
SM481Plus | 40,000 | Extended IC reach, stage vision options | ±40 μm chip class; ±30 μm stage |
DECAN S1 | 47,000 | Single-gantry speed entry | ±28 μm chip class |
DECAN L2 | 56,000 | Multi-function, LED-optimized recognition | ±28 μm chip class |
SM471Plus | 78,000 | Dual-gantry chip shooter, dual-rail option | ±40 μm chip class |
DECAN F2 | 80,000 | Dual-gantry value speed | ±28 μm chip class |
DECAN S2 | 92,000 | Magnetic-levitation dual gantry, modular rail | ±25–28 μm |
Takt time is the drumbeat the schedule assigns to production: available production time divided by required output, expressed for an SMT line as seconds per panel. It is the only throughput number that originates entirely on the buyer's side — from the shipment forecast, the shift pattern and the yield plan — and it is the standard against which every machine calculation in this guide is finally judged. A line whose takt allows 45 seconds per panel has a completely different mounter conversation from one that must hold 20, even if both run the same board outline, because the takt decides the realistic rate required, and the required rate decides the machine count and tier long before any datasheet column is consulted.
The practical discipline is to compute takt per product family and to design the line for the tightest family that will run in the planned configuration, because takt is not averaged — a line is only as fast as its binding product. The mounter's realistic rate is then compared against the takt with the family's efficiency factor already applied, and the comparison yields one of three outcomes: comfortable margin, meaning the tier can absorb growth; thin margin, meaning configuration effort (feeders, dual rail, panel strategy) must buy the difference; or negative margin, meaning the answer is a second machine or a higher rung, and no configuration wizardry on the first machine will change it. Takt also disciplines the rest of the line: the printer must present panels at the same beat and the reflow oven must carry them through at the same beat, which is the subject of the line-balancing section later in this guide. CPH capability, efficiency honesty and takt arithmetic together answer "how fast" — the next specification answers "how well."
Of the three specifications, accuracy is the only one the factory cannot walk back. A throughput shortfall is solved with a second machine or a shift pattern; a geometry surprise is solved with a different tier; but an accuracy shortfall announces itself as escapes, rework and customer returns, priced long after the invoice is approved and attributed everywhere except to the specification that was skimmed. Accuracy also sets the vocabulary for everything else in the matching exercise: the smallest package and the finest pitch in the BOM are not line items but votes, and they elect the platform generation before throughput or price are allowed a ballot. That is why, in the Hanwha pick and place machine CPH accuracy PCB size framework, accuracy is read immediately after the throughput reality is established — once the schedule has said how fast, the BOM says how well, and the two answers must come from the same machine.
The figure below shows the machines that carry this discipline in their production setting, and the section unpacks accuracy in four sub-sections: what the micron figures actually measure, and why they are system statements rather than mechanical ones; how the smallest package in the BOM — 01005 and below — maps to the accuracy classes across the SM and DECAN generations; where fine pitch turns accuracy into a vision question; and how to match the accuracy class to the product roadmap rather than to today's BOM alone.
A placement accuracy figure describes the distribution of placement error — translation in X and Y plus rotation in theta — under defined measurement conditions, and the industry standardizes those conditions under IPC-9850 in the same way it standardizes CPH. The figure is a system statement: it is delivered by the interaction of mechanical rigidity, the vision system's ability to find and center the part, thermal behavior over a shift, and the calibration regime the maintenance plan keeps. That is why two machines with similar mechanical tolerances can hold different accuracy classes, and why the figure moves when the configuration changes — the stage-vision options on the SM481Plus and SM482Plus tighten IC-class precision to around ±30 microns precisely because the stage camera resolves what flying vision, in motion, cannot.
The reading rule that keeps the figure honest is to match it against the requirement, not against another datasheet. A ±40 micron chip-class figure comfortably serves packages whose pad geometry leaves tens of microns of margin; a BOM that bottoms out at 01005 spends much more of that margin, and a roadmap that names finer packages spends it faster. The numbers across the Hanwha range are distinctive enough to be decision-grade: the SM platform holds ±40 microns in its chip class, the stage options step to ±30, and the DECAN generation holds ±28 on chips and ±25 on ICs, with the flagship tightening both. The buyer's task is not to admire the smaller numbers but to compute the margin the BOM and roadmap actually leave — the same arithmetic this section applies in its final sub-section, and the reason accuracy is never bought as a trophy specification.
The smallest package in the BOM is the accuracy decision in its purest form, and it is also where the notation trap lives. Package codes come in inch and metric systems, and Hanwha datasheets write the window as 0402 (01005 inch) — meaning the 0.4 by 0.2 mm chip, the one the inch system calls 01005. Across the SM series, the flying-vision windows open at that package; across the DECAN generation, they extend a full class below, to 03015 — 0.3 by 0.15 mm. The accuracy logic follows the geometry: placing a smaller body spends the accuracy budget faster, because the same absolute error is a larger fraction of a smaller pad, and the vision system has fewer pixels of margin to work with.
The decision rule is binary in practice, and the earlier pillar on this range treats it as such. A BOM whose smallest specified package across the roadmap is 01005 or larger is inside both generations' envelopes, and accuracy then competes on the margin it leaves rather than on the class boundary. A BOM that names 03015 elects the DECAN platform, and the rest of the matching exercise proceeds within that platform's tiers. What the rule forbids is the comfortable middle: buying a platform whose window "closes enough" to the smallest package, on the assumption that production tolerance will stretch. It does not — yield on packages at the window's edge degrades exactly when the schedule is fullest, which is when the factory can least afford the investigation. The smallest package, read with its notation resolved and its roadmap trajectory noted, is the cleanest filter in the whole specification-matching exercise.
Between the micro-chip floor and the large-body ceiling sits the pitch range, and here accuracy stops being a mechanical property and becomes an optical one. A QFP with 0.3 to 0.4 mm lead pitch, a BGA with 0.4 to 0.5 mm ball pitch, a fine connector verified across its full length — each demands that the vision system not only find the part but characterize its leads or balls precisely enough to correct placement in real time. The Hanwha range answers with the two-camera architecture whose division of labor is laid out in the SM versus DECAN series specification comparison: flying vision handles the high-volume majority in motion, while stage vision — a precision camera with controlled lighting on the machine frame — resolves the rows whose geometry exceeds what a moving camera can confidently see.
For the specification-matching exercise, the pitch rows translate into a census rather than a philosophy. List every BOM line whose lead pitch, ball pitch or body reflectivity approaches the flying window's edge; if the list is empty, accuracy is served by the platform's flying system and the stage option is money better spent on feeders or margin. If the list runs to a handful of rows on every board, the stage-equipped configurations — or the DECAN tier's tighter native envelope — become the accuracy answer, and the cycle-time cost of routing the hard rows through the precision station is a small, known quantity on boards whose IC count is modest. If the list is long, the board itself is telling the buyer that the product belongs on the tighter platform, and the accuracy conversation has quietly merged with the throughput one. Either way, the census converts pitch from an adjective into a configuration line — which is what a specification is for.
The accuracy decision that ages well is bought for the roadmap, not the current BOM. Packages shrink on a schedule the product office controls but the equipment plan must anticipate: a line specified today at 0201 minimums will meet 01005 within a product cycle or two in most consumer and industrial segments, and a line specified at 01005 should at least be argued against a 03015 scenario before the contract is signed. The cost asymmetry makes the argument worth having: buying one accuracy class ahead of the roadmap costs a configuration delta at purchase, while re-platforming mid-life costs a second capital event, a second installation, and a yield investigation in between that no one budgets.
The matching rule is therefore a margin rule, computed from the roadmap's smallest package rather than today's. Hold enough accuracy margin that the next package step down still leaves comfortable pad geometry — on the SM platform, that usually means invoking the stage options when the roadmap tightens; on the DECAN platform, the tighter native class does the same work structurally. Document the assumption in the purchase file, the way the demonstration protocol in this series' selection framework documents capability: state the smallest package specified, the accuracy class purchased, and the margin held. Factories that keep this line in the file make platform reviews boring — which is the highest compliment a specification-matching discipline can earn, and the natural bridge to the next specification, where geometry takes its turn at constraining the choice.
Geometry is the most honest of the three specifications, because it does not negotiate. A panel either fits the machine's working envelope or it does not, and no demonstration, discount or datasheet adjective changes that. PCB size decides the rail width, the support scheme, the board-handling tier and, less visibly, the economics of every panel that rides through: a large board spends more transport time per placement opportunity, and a machine whose window is generous is rarely the machine whose headline rate is highest. That is why, in the Hanwha pick and place machine CPH accuracy PCB size matching order, geometry earns its own reading pass — it filters the catalog physically before the softer specifications are priced, and it interacts with CPH in both directions: bigger panels slow the rated machine, and the choice to panelize can speed the effective one.
Four sub-sections work the geometry dimension: the working envelope and what its numbers govern; panelization, and how the decision to panelize rewrites the throughput arithmetic; the long-board and dual-rail territory where the wide-body tier lives; and the cases where a panelization decision made on the design side changes the machine answer on the equipment side — the interaction that makes PCB size a conversation between engineering departments rather than a line on a quote.
The board-size specification describes the panel dimensions the machine's rails and transport accept, and it is read alongside the elements that make the figure real: rail adjustment range, support-pin field, the clamping scheme, and the clearance the head needs at the panel's edges and over its tallest components. Within the Hanwha range the envelope steps up through the tiers — the dual-rail option on the SM471Plus carries panels of 460 by 250 mm per lane, while the wide-body SM tier and the DECAN multi-function machines extend the reach for larger singles, and the long-board configurations cover panels up to the 1,200 mm class. The buyer's first geometry task is unglamorous: collect the largest panel every product family will run, add the clamping margin the transport requires, and compare the result against each candidate's envelope — in writing, per machine, before any other specification is discussed.
Two second-order geometry facts deserve the same attention. First, thickness and stiffness ride along with size: thin, flexible panels stress the support scheme more than large rigid ones, and a machine can accept the dimensions while struggling with the flex, which is a demonstration item rather than a datasheet one. Second, the envelope interacts with component height — tall parts near the rail edge stress head clearance in ways that only appear on the actual panel — and with fiducial placement, since the vision system needs reference marks it can reach. Factories that hand their machine supplier the true maximum panel, with its thickness range and its tallest components mapped, get envelope answers that hold; factories that hand over a nominal board size rediscover geometry during commissioning, when changes are most expensive.
Panelization is the design-side lever that rewrites the throughput arithmetic without touching the machine. Combining multiple units into one panel changes the ratio the mounter cares about: one transport event, one fiducial routine, one rail cycle now carries several boards' worth of placements, so the transport overhead per placement falls and the effective CPH on the product rises — often materially, because conveyor transport and board-handling pauses are among the quietest thieves of rated speed. Panelization also stabilizes thin boards through placement, and it moves work from the placement line's handling time into the depaneling process downstream, where it is usually cheaper.
The lever has limits that the matching exercise must respect. The panel's final size must still fit the working envelope, so panelization consumes geometry headroom the largest singles may need; depaneling quality and stress constrain how aggressively units can be packed, particularly near board edges; and the mix question returns — a product family that panels beautifully today may not share its panel with the roadmap's next variant. The disciplined approach prices both configurations: the same annual volume run as singles and as panels, each with its transport profile, its efficiency factor and its depaneling cost. When the panel route wins — and on high-volume, transport-bound products it usually does — the machine conversation shifts subtly: the required realistic rate falls, which can drop the answer one rung on the CPH ladder, and the geometry envelope must now serve the panel rather than the board. That two-way interaction is precisely why PCB size is matched rather than merely checked.
Three geometry scenarios occupy the top of the envelope, and each has a home in the range. Long boards — LED modules, power supplies, backlight assemblies in the 1,000 mm-plus class — route to the configurations built for them: the SM482Plus's wide-body platform accepts 1,200 mm boards and pairs that envelope with the large-body and long-connector tolerance the same products usually carry, which is why the SM482 Plus wide-body SMT mounter is the reference answer for industrial panels that most machines cannot physically hold. Dual-rail production answers a different geometry problem: when panels are small and volume is high, two lanes in one machine — as on the SM471Plus's 460 by 250 mm per-lane option — double the transport throughput without doubling the footprint, effectively buying CPH with geometry instead of with a second frame.
The wide-body and dual-rail routes also carry their own economics, and the matching exercise prices them honestly. Wide-body machines trade some rated speed for their envelope — the SM482Plus's 30,000 CPH against the SM471Plus's 78,000 is the clearest expression — so a line that needs both the long-board window and serious throughput is already hinting at either a mixed configuration or the DECAN multi-function tier's answer. Dual-rail configurations add mechanical complexity and changeover overhead per lane, and they pay off only when the product mix genuinely runs two lanes in parallel. The lesson across all three scenarios is the same: geometry at the extremes is not a nuisance around the specification but a decision-shaping variable in its own right, and the factories that treat it that way shortlist machines whose envelopes were designed for their products rather than machines whose brochures they liked.
The most valuable geometry insight in the matching exercise is that the causality runs both ways: just as the panel constrains the machine, the machine candidate can justify revisiting the panel. A design team that panels four small boards into a 400 by 300 mm carrier may find the resulting envelope forces the wide-body tier — while a two-by-two panel at 300 by 250 mm fits the high-speed tier the volume actually wants. Conversely, a factory already owning dual-rail machines may prefer two smaller panels per lane over one large panel, converting geometry the line already has into throughput it needs. These are not equipment concessions to design; they are the joint optimization that the specification-matching frame exists to surface, and the hour spent exploring them is the cheapest engineering on the project.
Making the interaction routine requires one artifact: a geometry matrix, maintained alongside the BOM census this series' component-configuration guide describes — every product family's single-unit size, its panel options, its largest panel and its thinnest board, checked against each shortlisted machine's envelope and support scheme. The matrix makes panelization a visible decision row in the purchase rather than an inherited assumption, and it protects the configuration across product cycles: when the roadmap adds a family with an awkward outline, the matrix shows which machine absorbs it and which panel redesign avoids the problem. Geometry, read this way, stops being the specification that silently eliminates candidates late and becomes the one that organizes the design-equipment conversation early — which is exactly the posture the next specification, product mix, also demands.
Product mix is the specification the datasheet never prints and the schedule always enforces. Every rated figure on the machine assumes a stable product: one feeder map, one tray plan, one nozzle assignment, one panel geometry. Real factories ship families, not products, and every family change spends rated capacity on overhead — feeder swaps, recipe calls, first-article checks, nozzle exchanges — that never appears in the CPH column. The efficiency factor introduced in the throughput section is where mix becomes arithmetic, and the changeover architecture is where mix becomes hardware. In the Hanwha pick and place machine CPH accuracy PCB size matching exercise, this is the layer that converts a correct-looking machine into a correctly-configured one: two factories can buy the same platform and run it at materially different realized rates, purely because one planned for its mix and the other discovered it.
The figure below shows the kind of line this section is sizing — a production environment where families rotate and the placement rate must survive the rotation. Three sub-sections follow: how product mix costs efficiency, and how to census it; the changeover penalty, and the docking and setup architecture that compresses it; and the feeder and tray capacity question under mix pressure, where the peak-concurrent requirement across families — not any single board's count — sizes the configuration.
Mix spends efficiency through three mechanisms, and each is measurable before the machine is bought. Routing overhead: a mix of boards sends different fractions of placements through slower paths — stage-vision rows, tray stations, long-connector verification — so the blended efficiency factor sits below the best family's, and the blend is set by the schedule's weighting, not by the average board. Transport overhead: families with different panel sizes and rail settings pay a setup at each transition, and thin-schedule transitions turn what looks like minutes into hours of weekly capacity. Utilization overhead: the head and nozzle plan that is optimal for family A is suboptimal for family B, and the machine runs at the compromise, not at either optimum.
The census that captures this is simple enough to run in a spreadsheet and valuable enough to run before demonstrations begin: for each product family, its placements per panel by component class, its panel geometry, its tray and fine-pitch fraction, and its scheduled batch size and changeover frequency per week. Weighted by the schedule, the census yields the blended efficiency factor the sizing arithmetic must use — and it yields something equally important: the identity of the binding family, the one whose combination of tight takt and low efficiency sets the machine requirement. Lines sized to the average family disappoint exactly when the binding family is scheduled; lines sized to the binder carry margin for everyone else. The census also feeds the next sub-section directly, because changeover frequency is one of its columns, and the penalty it produces is the next arithmetic to run.
Changeover is where mix becomes visible on the clock. A placement-machine changeover stacks several activities: the feeder map swap, the tray-station reload, the nozzle-exchange to the new recipe, the program call and verification, and the first-article confirmation before the line is trusted again. Done at the rail, feeder by feeder, the session consumes a meaningful fraction of a shift; done with the docking-cart architecture the platforms share — feeders loaded and verified on carts offline, carts swapped in as a unit, recipes tying cart contents to programs — the same changeover compresses toward minutes. For high-mix operations, that compression is not a convenience; it is capacity, because every hour of changeover is an hour the takt arithmetic assumed was producing.
The high-mix penalty is therefore priced as a configuration decision, not endured as a fact of life. The cart count determines how many setups can be staged in parallel — a machine with more feeder positions than staged carts cannot change over faster than its carts allow; the electric-feeder fraction determines how quickly fine-pitch rows are verified and presented; and the software's recipe management determines how much of the map survives between runs of the same family. Factories running many families per day should treat these items as primary selection rows rather than option-line items, and the demonstration protocol in the closing section should include one timed changeover between the buyer's own products. The pattern echoes the whole guide's method: the mix the datasheet never shows becomes real only when it is converted into configuration the machine actually carries — and the same conversion applies to the last hidden specification, the feeder and tray capacity under mix pressure.
Feeder capacity reads differently under mix than under single-product production. The platform standard — 120 positions of 8 mm tape across the main configurations — is generous for one board and merely adequate for a portfolio, because the governing number is the peak concurrent requirement: the feeder load demanded when the families that change over fastest must run back to back, which typically runs 30 to 50 percent above any single board's count, before the width corrections that make 8 mm and 24 mm positions non-fungible. A mix-heavy factory that sizes to a single board's census discovers the shortfall mid-run, in the shape of feeder swaps on live machines — the most expensive interruption a placement line experiences.
The configuration answer follows the census, and it is one of the places where the platform tiers genuinely differ in character. The multi-function tier is built for exactly this pressure: the DECAN L2 flexible multi-function placer pairs its 56,000 CPH with the wide recognition and LED-optimized capability that mixed portfolios need, and its configuration accepts the feeder, tray and specialty-station load that a rotating family schedule generates. On the SM side the same load is answered with cart count and electric-feeder fraction rather than a different machine. The rule across both platforms is identical and worth stating as the section's close: capacity under mix is planned from the peak, verified in a timed changeover, and purchased as configuration — because the alternative, discovering the peak in production, is the one specification failure that announces itself on the shipping schedule rather than in a yield report.
Matching the three specifications produces a machine; matching the machine to the line produces a production system. The placement stage is almost always the SMT line's pacing constraint — it has the most moving parts, the longest cycle and the highest sensitivity to mix — which means the printer upstream and the reflow oven downstream are sized to the mounter's beat, not the other way around. Line balancing is therefore the final examination of the whole matching exercise: a line can hold a correct CPH choice, a correct accuracy class and a correct envelope, and still miss takt daily because the balance between the stages was never computed. This is the closing layer of the Hanwha pick and place machine CPH accuracy PCB size framework, where the specifications stop being machine properties and start being scheduling facts.
Four sub-sections complete the layer: the cycle-time alignment between mounter, printer and oven, and the communication that keeps it honest; the one-machine-versus-two decision, with the division-of-labor patterns that make two machines outperform one; the dual-rail and modular routes that buy capacity through architecture; and the buffering that lets a balanced line stay balanced when the mix moves. The section's premise throughout is that balance is computed, not hoped for — and that the computation uses the same realistic-rate arithmetic the earlier sections established.
The alignment starts with a single table: each stage's cycle time against the target takt, per product family. The printer's cycle — print, release, transport — must fit inside the takt with its own margin; the oven's conveyor speed must carry the panel through its profile within the same beat; and the mounter's realistic rate, converted to seconds per panel, must beat the takt with the 20 to 30 percent margin the schedule deserves. In practice the mounter defines the beat and the neighbors adapt: stencil printing is configured for the panel throughput the mounter implies, and oven length and zone count are chosen so the profile completes within the same cadence. The communication layer keeps the alignment live — the SMEMA interface and the platforms' line-control handshake pass board-present and machine-ready signals between stages, so a pause anywhere propagates as a controlled stop rather than a pile-up.
The balancing arithmetic also prices the transitions between stages honestly. Board handling — loaders, conveyors, buffers — adds seconds that belong in the takt calculation, not around it; double-sided assembly doubles the placement load and therefore the mounter requirement per finished board; and the panelization decision from the geometry section reappears here, because the panel that optimized mounter efficiency also sets the print pattern and the oven load per transport event. Lines that balance cleanly share a habit: the cycle-time table is maintained per family and revisited when the schedule shifts, because a balance that held for last year's portfolio can quietly fail on the current one. The table is also the earliest warning system the line has — a family whose placement seconds per panel drift toward the takt ceiling announces itself there, months before it announces itself in missed shipments.
When the realistic-rate arithmetic lands near a tier boundary, the decision becomes architectural: one machine at a higher rung, or two machines dividing the work. Both routes reach the same takt, but they fail differently, and the choice is a production-strategy question rather than a price question. A single flagship keeps the line simple — one recipe set, one maintenance regime, one changeover per family — at the price of concentrating the plant's entire placement capacity in one frame. Two machines divide the risk and the work: a high-speed chip shooter absorbing the passive majority while a multi-function machine handles ICs, trays and connectors, each optimized for its share, with the division usually recovering more effective throughput than the sum of their ratings suggests, because each head plan runs closer to its own optimum.
The second-machine route also opens the value tier in a way the single-machine route cannot. A pairing that places the DECAN F2 high-throughput value placement machine on the chip-dense stream and a multi-function unit on the precision stream often prices below a single flagship while beating it on blended effective rate — the F2's 80,000 CPH doing the volume work the efficiency factor punishes least. The honest comparison prices both routes over the same horizon: acquisition, floor space, staffing, maintenance and changeover load per family, each expressed against the takt table. Whichever route wins, the decision is documented the same way as the specification matches — with the arithmetic shown — because the one-machine-versus-two question will return with the next product family, and a documented balance is the difference between a decision and a debate.
Between one machine and two lies the architectural middle ground, and it deserves its own arithmetic. Dual-rail configurations — the SM471Plus's two-lane option among them — run two panels in parallel inside one frame, doubling transport throughput for small panels and letting two families share a machine without sharing its feeder map. Modular line strategies extend the same idea across the line: placement capacity added in units, each sized to the takt of the moment, so growth is bought in steps rather than in a single oversized frame. Both routes trade some simplicity for scalability, and both are priced honestly only when the mix census is on the table — dual rail pays when small panels and parallel families dominate, and pays badly when large singles and long boards define the portfolio.
The future-capacity question is where these architectural choices earn their keep. A line bought exactly to today's takt converts every roadmap addition into a re-platforming debate; a line bought with an architectural expansion path — a spare lane, floor and utilities staged for a second machine, a printer and oven rated for the next step up — converts the same addition into a purchase order. The project-level method for planning this whole-line growth, laid out in this series' Hanwha mounter selection framework, treats the expansion path as a selection row in its own right: state the volume the line must absorb before the next capital event, then verify that the chosen architecture — rails, floor, utilities, software licensing — actually reaches it. Capacity that is designed for is cheap; capacity that is discovered to be missing is a second project.
A balanced line is a snapshot; mix makes it a movie. Families rotate, and with them the stage that binds: a chip-dense family loads the mounter, a large-panel family loads the printer's setup and the oven's profile, a tray-heavy family slows the placement stage in a way no other family does. Buffering is what turns those shifts from stoppages into absorption: buffer conveyors between placement and reflow hold panels when the downstream stage momentarily lags, unload stations decouple the line's tail from operator rhythms, and the placement machines' own lane management absorbs the head-of-line variation that dual-rail and multi-family running produce. The buffer arithmetic follows the same realism discipline as everything else — size the buffer for the transition moments the mix census predicts, not for averages that never occur.
The last balance that must survive is the human and procedural one. Recipes, setup carts and maintenance windows are scheduled against the same takt as the panels, and a line whose changeovers collide with its maintenance plan loses both. Factories that keep their balance over years share three habits: the cycle-time table is a living document reviewed per schedule change; the buffer and handling plan is re-checked whenever a family with new geometry joins the portfolio; and the whole-line takt is re-verified at every major recipe addition, using the same realistic-rate arithmetic the purchase was justified with. Balance, maintained this way, is not a commissioning-week achievement but an operating discipline — and it is the discipline that finally converts three matched specifications into shipments that leave the dock on schedule.
The matching exercise ends where every sound equipment decision ends: in a configuration whose reasoning survives review. CPH has been converted from a rated figure to a realistic rate and checked against takt; accuracy has been matched to the BOM's smallest package and its roadmap; PCB size and panelization have fixed the envelope and the tier; mix has been censused into an efficiency factor and a changeover architecture; and the line balance has been computed stage by stage. What remains is to compress that chain into a form a reviewer can audit — the sequence, the cross-check and the demonstration — so that the machine that arrives is the machine the analysis named, and the analysis remains on file when the next family or the next capacity review arrives. This closing discipline is what separates the Hanwha pick and place machine CPH accuracy PCB size method from datasheet comparison: the numbers do not just select a machine, they build a record.
Three sub-sections assemble the close: the matching sequence restated as a step-by-step procedure, each step with its filter and its output; the specification cross-check table that compresses the whole exercise onto one page; and the demonstration protocol that converts the matched configuration from analysis into evidence, using the buyer's own boards, parts and schedule.
Run in order, the sequence filters the catalog five times before configuration is priced. Step one computes takt per product family and identifies the binding family — the output is a required realistic rate, in placements per hour, per family. Step two applies the efficiency factor from the mix census to each candidate's rating, and eliminates every machine whose realistic rate misses the requirement with margin — the output is the surviving CPH rung or rungs, and the machine count. Step three matches accuracy: the smallest package and finest pitch across the roadmap select the platform generation and the camera configuration, and any candidate failing the smallest-package test is eliminated regardless of its speed. Step four matches geometry: the largest panel, the thinnest board and the panelization plan select the envelope, the rail scheme and the tier. Step five sizes the mix architecture — feeder positions at peak-concurrent, tray stations, cart count, electric-feeder fraction — against the shortlist that survived the first four filters.
The sequence's power is its auditability: every elimination names the specification that caused it, and the final configuration carries its reasoning with it. It also connects cleanly to the component-level detail this series' BOM-driven component configuration guide develops — the census of packages, feeders and trays that steps five consumes is built exactly as that guide prescribes. Factories that run the sequence in writing report the same experience: the catalog narrows to one or two candidates before any commercial factor is discussed, and the negotiation that follows is about configuration and support rather than about whether the machine fits — which is where a purchase conversation should be by the time a number is written on it.
The whole method compresses onto one page, and the page is the artifact that travels to the supplier meeting, the budget review and the capacity discussion a year later. Each row pairs a specification with the question the analysis must answer and the form the answer must take; a purchase is ready to sign when every row has an answer written against it, sourced from the analysis rather than from the quotation. The table below is that page — six rows, in decision order, with the demonstration row at the end closing the loop between analysis and evidence.
# | Specification | Question the analysis answers | Answer required |
|---|---|---|---|
1 | Takt time | Seconds per panel does the schedule allow, per family? | Required realistic rate; binding family named |
2 | CPH (realized) | Rated × efficiency vs requirement, with margin? | Surviving tier and machine count, +20–30% |
3 | Accuracy | Smallest package and finest pitch on the roadmap? | Platform generation and camera set, margins shown |
4 | PCB size / panel | Largest panel, thinnest board, panel plan? | Envelope, rail scheme, tier — fits in writing |
5 | Product mix | Peak feeder demand, tray load, changeover tempo? | Feeder/tray/cart configuration sized to the peak |
6 | Line balance | Do printer, mounter and oven share one beat? | Cycle-time table per family; buffer plan |
The demonstration converts the cross-check table from analysis into evidence, and the buyer writes the script. Bring the panels and components the table's rows name: the largest panel with its real thickness and support needs, the smallest package from a production reel, the finest-pitch IC on its actual packaging, and — if the mix census says changeover matters — the feeder map of the second family for a timed changeover. Require each surviving candidate to run the script on the configuration quoted, at production cadence, with results counted rather than narrated: placements per hour realized on the buyer's panel, yield on the smallest package, the envelope check on the largest, and the changeover timed end to end. A demonstration run this way tests the rows that decided the purchase; a demonstration run on the supplier's demo board tests nothing the buyer ships.
The record then lives where the analysis lives. Measured results are filed against the table rows they verify, the delivered configuration is checked at commissioning against the demonstrated one, and the file becomes the plant's standing baseline for the next capacity review or product family. The habit compounds across purchases: factories that keep demonstration records build a private, comparative database of how platforms behave on their own products — an asset no datasheet replaces and no competitor can copy. That is the natural end of the specification-matching exercise: not a machine chosen, but a method institutionalized, one that turns every future specification sheet from a brochure into a questionnaire the factory already knows how to answer.
Match your CPH, accuracy and PCB-size specs with I.C.T
If the matching exercise in this guide has raised questions your own numbers can answer, our engineers run this exact sequence with customers every week. Send your panel geometries, your product families with their takt requirements, and the smallest packages and finest pitches on your roadmap, and we will map them across the SM471Plus, SM481Plus, SM482Plus and the DECAN S1, F2, S2 and L2 — returning a configuration proposal with the realistic-rate arithmetic, the accuracy margins and the line-balance table shown, not just a model number. Whether the conclusion is one machine, a dual-rail configuration or a two-machine division of labor, it arrives as a completed cross-check table with the demonstration script attached, so the decision your team signs is the one your boards verified. Browse the range in the Hanwha placement machine catalog, or share your schedule and panel drawings to schedule a demonstration on the exact specifications that will decide your purchase — your takt, your smallest package and your largest panel, run and measured before any commitment is requested.
Rated CPH is measured at the optimum — minimal travel, favorable components, full head utilization — under the IPC-9850 discipline, and it is intended for comparing machines, not for scheduling production. Realistic throughput is the rated figure multiplied by an efficiency factor that reflects your boards: 0.5 to 0.7 for chip-dense products, 0.3 to 0.5 for mixed boards with stage-vision rows and tray content. Size the line on the realistic number, verified on your own panel, and hold 20 to 30 percent margin above the schedule.
A 01005 chip (0.4 by 0.2 mm) is within reach of both Hanwha generations: the SM series' flying-vision windows open at it with chip accuracy around ±40 microns, while the DECAN platform holds around ±28 microns. Both can place the package; the margin differs. If 01005 is your roadmap floor rather than today's floor, the tighter DECAN class buys comfort; either way, verify the specific package on the specific feeder in a scripted demonstration before signing, because capability at the window's edge is proven, not assumed.
Check the envelope against both extremes: the wide-body SM482Plus accepts 1,200 mm boards, and small panels run on any machine in the range, so physically the answer is yes. The trade-off is throughput — the wide-body tier trades rated speed for its envelope. If the small-panel volume is high and the LED volume is modest, a dual-rail high-speed machine plus the wide-body tier, or a mixed configuration, often beats one machine trying to serve both regimes; the cycle-time table per family makes the comparison honest.
Panelizing several units into one panel amortizes each transport event, fiducial routine and rail cycle across more placements, so the effective rate on the product rises and the required machine rate falls — sometimes by a full rung on the CPH ladder. The panel must still fit the working envelope, and depaneling quality constrains how aggressively you pack. Price both configurations with their transport profiles; on transport-bound, high-volume products, panelization is usually the cheapest throughput upgrade available, because it is purchased in the design office rather than on the equipment budget.
Compute it, do not assume it. Build the mix census: per product family, placements by component class, panel geometry, tray and fine-pitch fractions, batch sizes and changeover frequency. Weighted by the schedule, the census yields a blended factor — typically 0.5 to 0.7 for chip-dense single-family running, 0.3 to 0.5 for high-mix boards with hard rows. Size to the binding family, not the average, and verify the factor in a timed demonstration on your own panel, so every downstream calculation inherits a measurement instead of a hope.
When the realistic-rate arithmetic lands near a tier boundary and the gap cannot be closed with configuration. Two machines also win strategically: a high-speed chip shooter for the passive majority plus a multi-function machine for ICs, trays and connectors lets each head plan run near its own optimum, and the pairing often prices below a single flagship while beating it on blended effective rate. Decide with the cycle-time table over a five-year horizon — acquisition, floor, staffing, maintenance and changeover per family — and document the arithmetic, because the question returns with every new product family.
Takt equals available production time divided by required output, expressed as seconds per panel: take the shift minutes actually available for production, subtract planned downtime, and divide by the panels the schedule demands in that time. Compute it per product family, because takt is never averaged — the line runs at its binding family's beat. The mounter's realistic rate must beat the tightest takt with 20 to 30 percent margin, and the printer and oven are then configured to the same beat, which is the line balancing this guide's sixth section describes.
Mostly, with a structural exception. Tighter accuracy costs time when it is delivered by processing — stage-vision rows route slower than flying vision — and by architecture that trades cycle count for controlled placement. The exception is platform generation: the DECAN S2's magnetic-levitation dual gantry holds 92,000 CPH alongside a ±25 to ±28 micron class, showing that a new architecture can raise both at once. The premium buys exactly that combination — but the board-size envelope, feeder count and configuration cost still apply, so accuracy and speed are matched jointly with geometry, never alone.