English Views: 0 Author: Elsa Long Publish Time: 2026-11-03 Origin: Site
Most JUKI buying mistakes are made before anyone looks at a machine. A team starts from a model number someone recommended, then works backwards to justify it, and the mismatch only appears in the third month of production when a 33.5 mm connector has no head that can place it or the feeder bank runs out of positions mid-changeover. Learning how to choose a JUKI pick and place machine properly means starting from seven measurable inputs — BOM, board and panel size, component range, feeder demand, throughput, accuracy and budget — and letting those inputs eliminate models until one configuration remains. That is the reverse of a catalog conversation, and it is the only sequence that survives contact with a real product mix. The models referenced here sit in the SMT pick-and-place machine lineup; every calculated figure below is labeled as a hypothetical assumption.
The framework has three parts. The first turns your product into a requirement sheet that a supplier can quote against, so no one has to guess what you actually build. The second tests that sheet against the four physical limits of any placement platform: what it can pick up, what it can hold, how many feeders it can carry, and how fast it can repeat the cycle at the accuracy you need. The third converts the survivor into money — capex, configuration cost and the five-year cost of owning the wrong envelope — and closes with a repeatable procedure and a worked example you can copy for your own numbers.
A placement machine is a set of limits dressed as a product. It can pick up components inside a size window, hold boards inside an envelope, carry a number of feeders, and repeat a cycle at a stated speed and accuracy. Nothing in that list is a preference, and nothing in it can be judged from a photograph. The reason so many evaluations stall is that the product has never been written down in the same units the machine is specified in. A BOM exists in the ERP, a board drawing exists in CAM, an output target exists in the sales forecast, and none of them have been converted into placement requirements that a supplier can answer. This chapter builds that conversion, and it is the step that determines whether the rest of the evaluation is a decision or a justification. The seven inputs below are ordered deliberately, because the early ones remove whole classes of machine before any price is discussed. That ordering is the practical core of how to choose a JUKI pick and place machine without spending a month on quotations you cannot compare.
Work through them in order and write down a number or a range for each, even when the answer is approximate. An approximate requirement sheet that says "largest component 33.5 mm square, board up to 1,150 × 520 mm panel, 129 unique BOM lines" is worth far more than a wish list that says "fast, flexible and reliable". The first can be quoted, tested and rejected on evidence. The second can only be argued about, and arguments are where schedules and budgets are lost. Keep the sheet in one page, date it, and treat it as a living document: the JUKI pick and place machine you shortlist today may need re-checking when the next product family arrives.
Input | What you collect | What it eliminates early |
|---|---|---|
BOM | Unique part count, package mix, quantity of 01005/0201 chips, largest body, tallest part | Heads and models that cannot cover the range in one program |
PCB size | Bare board, panel size, thickness, warpage, edge clearance | Models whose board envelope is too small or too narrow |
Component range | Smallest chip, largest connector, finest pitch, PoP or odd-form needs | Vision type, head type and nozzle family |
Feeder demand | Unique lines by tape width, tray and stick counts, expected changeover frequency | Feeder input count and cart strategy |
Throughput | Boards per month, placements per board, shifts, planned downtime | Number of machines and line balance |
Accuracy | Finest pitch, required tolerance and Cpk, inspection capability downstream | Model tier and vision specification |
Budget | Capex ceiling, floor space, manpower, five-year consumable and service cost | New versus used, single machine versus line |
Take the three highest-volume assemblies and the two most difficult ones, and build the sheet from those five rather than from the whole catalog. High-volume boards set the throughput and feeder requirements; difficult boards set the component range, accuracy and vision requirements. If you design for the average product you will under-specify both ends. For each board, record the placement count, the unique line count, the smallest and largest component, the finest pitch, the panel dimensions and the annual quantity. Then take the maximum of each column and the sum of the volumes. That gives you a requirement envelope with a real production weight behind it, and it takes an afternoon with the BOM export, not a week. Two further columns are worth adding: the number of 01005 or 0201 chips, because they drive nozzle and vision selection, and the tallest component, because it can limit which head can reach the board without collision.
Write the sheet so that a supplier cannot answer it with adjectives. Instead of "high mix", write "average 129 unique lines, 14 changeovers per week, batch size 400 boards". Instead of "accurate enough", write "0.4 mm pitch QFP at 100 percent of pins, target first-pass yield 99.5 percent". The test of a good requirement sheet is simple: two different suppliers reading it should arrive at the same shortlist. If they arrive at wildly different machines, the sheet is still too vague, and the comparison you are about to run will measure salesmanship rather than fit.
The published JUKI SMT machine model list is a useful map but a poor decision tool. It shows that the family spans compact chip shooters, flexible modular mounters and high-speed rotary-head machines, and it gives a board envelope, a placement speed and a feeder count for each. What it cannot tell you is how those numbers behave on your board. A published optimum speed is measured under defined conditions on a reference pattern; your board has a different mix of chip sizes, a different number of feeder indexes per placement and a different vision load. A published feeder count is a maximum with a specific feeder type, and the type matters, because double-lane 8 mm feeders do not occupy the same space as 24 mm feeders or a tray bank.
Read the list in three passes. First, strike out every model whose board envelope cannot hold your largest panel, because no amount of programming fixes a board that does not fit. Second, strike out every model whose component range stops short of your largest body or tallest part. Third, compare what remains on feeder capacity and speed. What the list hides is equally important: changeover time, nozzle family coverage, software maturity, spare-part lead time and local service response. Those are not in the table, and on a five-year horizon they often cost more than the difference between two models on the same page.
The inputs are not independent, and the conflicts are predictable. Throughput pushes you toward a fast rotary head with a small component range and few feeder positions; component range pushes you toward a flexible head with a lower placement rate and a taller component clearance. Board size pushes you toward a wider machine, which costs floor space and money, while feeder demand pushes you toward more input positions, which costs width again. Budget pushes you toward one machine, while accuracy and range often argue for two. Recognising these trade-offs early prevents the most expensive mistake in the whole process, which is buying one machine that is mediocre at everything instead of two that are good at their jobs.
Resolve conflicts by ranking the inputs against the cost of being wrong. If a wrong board envelope means the machine cannot run your main product at all, that input is a hard constraint and everything else must bend around it. If a shortfall in feeder positions means an extra changeover per shift, it is a soft constraint with a calculable cost, and it may be worth accepting. Write the ranking down before you see prices, because after prices appear every constraint starts to look negotiable. A shortlist built this way is also easier to defend internally: each model either clears the hard constraints or it does not, and the soft ones are settled with numbers rather than opinions.
The BOM is the only document that describes your product in the same granularity as a placement program, which is why it deserves to be read as a machine requirement rather than as a purchasing list. Every unique line is a feeder position, a nozzle decision and a vision task. Every package type is a question about whether the head can pick it, carry it and place it without dropping or rotating it. Teams that read the BOM this way discover their real constraints in a week; teams that read it as a price list discover them in production. The distinction matters because component range is the input that most often disqualifies a model outright, and it does so silently: a machine that cannot place one connector still runs, it just sends that connector to a second process, a rework bench or a subcontractor, and the cost of that decision never appears in the quotation.
Read the BOM in four passes. First, sort by package and count the unique lines in each group. Second, find the extremes: the smallest chip, the largest body, the tallest part, the finest pitch. Third, mark the parts that need special handling, such as package-on-package stacks, connectors with locating pegs, or shielded components that vision cannot see through. Fourth, compare the totals with the machine you are considering. The output is a short list of parts that either fit or do not, and that list is what you take to a demonstration, not a generic sample board. A machine demonstration on the supplier's reference board proves very little; a demonstration on your five difficult lines proves almost everything.
Export the BOM with quantity, reference designator, package and manufacturer part number, then add two columns of your own: tape width and feeder type. Tape width is usually derivable from the package and the supplier's packing specification, and it is the number that decides how many feeder positions the program will consume. A board with 129 unique lines is not 129 feeder positions; it is a mix of 8 mm, 12 mm, 16 mm and wider tapes, plus trays and sticks, and the mix determines whether a 56-position bank is enough or whether you need 160. Add a third column for annual usage, because that tells you which parts will run out mid-shift and force a reel change at the worst moment.
The fourth column is the one most teams forget: whether the part is actually placed by this machine at all. Some BOM lines are intentionally hand-placed, some are through-hole and belong to the DIP process, and some are added as alternatives that will never be programmed together. Removing those lines before counting feeder positions can cut the requirement by ten to twenty percent, which is often the difference between two models. Do this work once, in a spreadsheet, and keep the spreadsheet as the base for every future quotation. It becomes the single source of truth for the project, and it stops the slow drift where each supplier quotes a slightly different version of your product.
Component range has three independent limits and they fail in different ways. The smallest chip limit decides whether 01005 or 0201 parts can be picked at all, and it is governed by nozzle availability and vision resolution rather than by the head alone. The largest body limit decides whether a connector or a shield can be carried without the nozzle losing vacuum or the part swinging during travel. The height limit decides whether the head can pass over a tall part already placed on the board, and it is the limit most often ignored until a collision or a skipped placement appears in the program. A machine specified at a 6 mm component height cannot place a 10 mm transformer, no matter what the speed figure says.
A compact model such as the RX-6B compact JUKI mounter is positioned for exactly this everyday mix: small and large components in the same run, a feeder count that supports a normal BOM, vision for placement verification and a narrow body that fits a crowded floor. That is a useful shape of machine for a plant whose boards carry mostly chips plus a handful of larger parts. What it is not is a universal answer. If your BOM includes a 33.5 mm connector and a 12 mm tall part in the same program, the height and body limits must be checked line by line, not inferred from the family name. The same discipline applies to any model: find the three numbers, then find the BOM lines that sit outside them.
You can test a candidate machine against your BOM in about an hour without a demonstration. Sort the BOM by body size descending and take the top twenty parts. For each, note the body dimensions, the height, the weight class and whether the supplier's packing is tape, tray or stick. Then mark each against the published range of the candidate: inside, borderline, or outside. Borderline cases are the ones to ask about specifically, because published ranges are usually stated for a nominal condition, and a part at the extreme of the range may need a special nozzle, a slower placement mode or a vision parameter change that affects cycle time.
Repeat the exercise for the small end with the ten smallest parts, including the ones with the finest pitch. A part that is technically inside the range but needs a dedicated nozzle is not free: nozzles are consumables, they wear, and every additional nozzle family adds a setup step and a spare-part line. By the end of the hour you should have three lists: parts that clearly fit, parts that fit with conditions, and parts that do not fit. The third list is decisive. If it contains a part that appears on every board you build, the machine is wrong for you regardless of price, and no negotiation will fix it.
Feeders are where the BOM becomes physical, and they are the most common source of disappointment in a placement purchase. Two machines with the same nominal speed can have wildly different practical output because one has enough positions to keep the whole BOM loaded and the other forces a mid-shift changeover. The published feeder count is a maximum under a specific feeder type, and it interacts with tape width, cart design and how many products you run per week. This chapter converts the BOM into a feeder plan, then shows how the plan behaves when the mix changes. It is also where the difference between a chip shooter and a flexible mounter becomes visible in money rather than in adjectives. If feeder capacity is where your project would fail, settle it before any other part of how to choose a JUKI pick and place machine is discussed, because a shortfall here cannot be programmed away.
A feeder plan has three outputs: the number of positions required for the current BOM, the number required for the worst-case BOM in the next two years, and the number that can stay loaded between changeovers. Those three numbers rarely agree, and the gap between them is the decision. If you size for the current BOM only, every new product introduction becomes a setup project. If you size for the worst case, you buy positions you will not use for most of the year. The practical answer is usually to size for the worst realistic BOM and to use carts so that a changeover swaps a prepared bank rather than reloading reel by reel at the machine.
The count starts from unique lines, not from placements, and it ends with positions, not with reels. Group the BOM by tape width, then apply the lane rule of the feeder type you intend to use. An 8 mm double-lane feeder carries two 8 mm tapes in the width of one conventional feeder position, which is why a 160-position machine can hold far more than 160 small-tape lines. Wider tapes occupy one position each, sometimes more. Trays occupy a tray bank rather than tape positions, and sticks need a stick feeder that consumes a position and often a dedicated nozzle. The table below is a hypothetical example built on a 129-line BOM, and it shows why the raw line count is misleading.
Feeder group | BOM lines (hypothetical) | Positions consumed | What it constrains |
|---|---|---|---|
8 mm tape, double-lane | 96 | 48 | Main bank capacity; the widest lever on model choice |
12 mm tape | 18 | 18 | Bank width and cart layout |
16 mm tape | 8 | 8 | Bank width; affects lane sharing |
24 mm and wider | 4 | 4 to 6 | Consumes multiple positions on some banks |
Tray and stick | 3 | 1 tray bank plus 1 stick slot | Head reach and nozzle family |
Total | 129 | 78 plus tray | Compare with 56, 112, 160 or 240 positions |
Feeder type changes the arithmetic more than most buyers expect. Electronic feeders index by motor and report their status to the machine, which means the program can detect a missing or empty feeder before it causes a miss. Mechanical feeders are simpler and cheaper per unit but rely on the machine's indexing and give less feedback. The practical consequence is not only reliability; it is setup time and changeover discipline. With electronic feeders and a prepared cart, a changeover can be a mechanical swap plus a program recall. Without that, the same changeover becomes a manual reload with a verification step for every reel, which is where the twenty-minute job turns into a two-hour job.
The double-lane option deserves its own calculation. If your BOM is dominated by 8 mm tape, double-lane feeders can nearly double the effective capacity of a bank, and that can move you from a large machine to a mid-size one. But the benefit only exists if the tapes are actually 8 mm and if the program can address both lanes without conflict. A high-speed chip shooter in the family, such as the RX-8 high-speed JUKI mounter, is published with a comparatively small feeder input count and a small maximum component size, which is a deliberate design trade: it is built to place enormous numbers of small chips from a limited set of lines, not to carry a 240-line flexible BOM. That is a perfectly good machine in the right line, and the wrong machine if feeder capacity is your binding constraint.
Decide early whether feeder positions will be dedicated or shared. A dedicated layout keeps the highest-runner parts permanently loaded and treats the remaining positions as a changeover zone; it minimises setup for the main product and costs capacity for the tail. A shared layout reloads almost everything per product, which is fine for long runs and painful for high mix. Most plants end up with a hybrid: a fixed block for the parts that appear on every board, and a cart-based block for product-specific lines. The hybrid works because it puts the recurring setup work where it repeats least, and because it lets a prepared cart be built while the machine is still producing.
Then count the real changeover cost, not the theoretical one. Time each step once: unload the previous cart, load the prepared cart, recall the program, verify feeder addresses, run the first-article check. Multiply by the number of changeovers per week, then by the loaded hourly cost of the line. That number belongs in the selection, because it is comparable across models in a way that "flexibility" is not. A machine with more positions and a better cart interface can pay for its premium in a high-mix plant within a year, and it can be pure overspend in a plant that runs one product for three weeks at a time.
Board size looks like the simplest input on the sheet and behaves like the most brutal one, because it is a hard constraint with no workaround. A machine either accepts your panel or it does not, and if it does not, the entire evaluation is over regardless of speed, accuracy or price. The complication is that the number buyers quote is usually the bare board, while the number the machine sees is the panel, including rails, fiducials and the clearance the clamp needs. A board that measures 480 × 300 mm can arrive at the machine as a 520 × 300 mm panel with rails, and that difference has eliminated more shortlists than any other single figure. This chapter separates the three dimensions that matter, then looks at what the envelope does not tell you about actually holding the board flat.
The envelope also has a lower bound, which is easy to forget. Very small boards, especially thin ones, need panelization to run at all, and the panel then has to fit the machine along with its rails. If your product range includes both a 40 × 40 mm module and a 1,100 mm LED strip, you are asking one machine for two very different geometries, and the honest answer may be that the strip runs on a large platform while the module runs on a smaller one or on a panel. Mapping the extremes early turns a vague "we need flexibility" into a concrete decision about how many platforms the plant actually needs.
Record four dimensions for each product: bare board length and width, panel length and width with rails, board thickness, and the keep-out zone at each edge. The panel figure is what you compare against the machine envelope, and it should be the largest panel you would realistically run, not the largest you can imagine. Then add the process requirements that consume space: fiducial clearance, clamp position, and whether the machine needs an edge to grip or can support the board from below. A panel that fits the envelope but has no usable clamp edge is still a problem, and it is usually discovered at the trial run rather than at the quotation stage.
Thickness and warpage belong in the same conversation. Thin boards deflect under clamping force and under the pressure of placement, which affects placement height and therefore accuracy. The support system, whether pins, a vacuum plate or a magnetic pallet, has to match the board, and the machine's specification should be read together with the support accessories you intend to buy. A machine quoted without the right support hardware is not a complete quote, and the missing items often appear as an unpleasant surprise in the installation week.
Long boards are their own category. LED lighting panels, industrial control strips and some appliance boards exceed a metre, and they need either a long-bed machine or a segment-placement strategy where the board indexes through the machine and is placed in sections. Both approaches work, and they cost differently: a long-bed machine costs floor space and money up front, while segment placement costs cycle time and programming complexity. Decide which cost you prefer before you shortlist, because the two paths lead to different models with almost no overlap.
A flexible platform such as the KE-3020VA large-board JUKI placement platform illustrates how the envelope changes the role of a machine. With a published board range from 330 × 250 mm up to 1,210 × 560 mm, it can hold panels that smaller chip shooters cannot accept, and its published feature set includes package-on-package placement, which is a different job from high-volume chip placement. That combination of large board and flexible component handling is worth real money in a plant building control boards, power electronics or lighting products, and it is the wrong purchase in a plant that only runs small panels at high volume. The envelope is not a specification to maximise; it is a specification to match.
Two numbers with the same product are not the same envelope. The usable area inside a machine is reduced by the clamp mechanism, the conveyor rails and the clearance the head needs at each end of travel. A machine published at a given maximum board size is stating a best case, typically for a board with the right edge conditions, and the usable area for your panel may be smaller. Ask for the clamp positions and the rail adjustment range, then draw your panel inside those numbers. It is a five-minute check that prevents a very expensive misunderstanding.
Support and clamping also decide whether accuracy is real or theoretical. A board that flexes by 0.2 mm under placement pressure can move a placement outside a fine-pitch tolerance that the machine itself can hold. This is why acceptance testing should include your thinnest, most flexible board, not only the convenient rigid one. The conveyor interface matters here too: if the machine must hand boards to the next process, the transfer height and the handshake described by SMEMA have to match the rest of the line, or the operators will spend their shifts clearing transfer faults that have nothing to do with placement quality.
Throughput is the input that buyers think they understand best and measure worst. The published number is a maximum placement rate under defined conditions, and the number that matters is the rate at which your boards come out good at the end of the line. Between those two figures sit feeder indexing, vision tasks, board transfer, nozzle changes, component mix and the reality that a program rarely presents the head with its favourite pattern. Accuracy sits alongside throughput rather than after it, because the two are traded against each other in every head design: a rotary head that places a hundred thousand small chips per hour is not the same tool as an inline head that places a large connector within a tight tolerance. This chapter builds a defensible capacity number and then shows where accuracy enters the calculation.
The method is deliberately conservative. Start from the placements per board, multiply by the monthly board requirement, and divide by the time the machine is genuinely available after changeovers and planned stops. That gives the average rate the machine must sustain. Then compare it with a realistic rate derived from the published figure, derated for your mix. If the answer is close to the published number, the plan is fragile; if it has thirty to forty percent headroom, the plan can survive a bad week. The worked numbers below are hypothetical and exist to show the arithmetic, not to recommend a model.
Manufacturers publish two kinds of speed figure, and the difference between them is the single most useful piece of skepticism a buyer can carry. The optimum figure is a peak rate, usually achieved with a favourable component mix and the best head configuration. The IPC-9850 figure is measured under a standardized characterization method, which makes it closer to a comparable production rate across models. When a product page publishes both, the ratio between them tells you how optimistic the peak number is, and that ratio varies by head type. The table below uses figures published on the supplier's product pages for three JUKI models; treat them as published claims to be confirmed against the current datasheet, not as measured results.
Model | Optimum CPH (published) | IPC-9850 CPH (published) | Ratio | Board envelope (published) |
|---|---|---|---|---|
KE-3020VA | 20,900 | 9,400 | About 45% | 330 × 250 to 1,210 × 560 mm |
FX-3RA | 90,000 (0.040 s per chip) | 66,000 | About 73% | 410 × 360 mm, XL 610 × 560 mm |
RX-8 | 100,000 | Not published on the page reviewed | Not calculable | 510 × 450 mm, approximate |
Take a hypothetical plant that must ship 25,000 boards a month, with 260 placements per board, running 22 days on two shifts of 7.5 productive hours. The monthly placement demand is 25,000 × 260, which is 6.5 million placements. Available time is 22 × 2 × 7.5 × 60, which is 19,800 minutes. Dividing demand by time gives 328 placements per minute, or about 19,700 CPH that the machine must sustain on average, including every stop. That is the target the machine has to beat with margin, not equal.
Now derate a candidate. A high-speed modular model documented with an IPC-9850 figure of 66,000 CPH, such as the FX-3RA high-speed modular mounter, has a large apparent margin against 19,700. Even at sixty percent of its published IPC rate, it delivers about 39,600 CPH, which is double the requirement. A flexible model published at 9,400 CPH would need three machines to reach the same demand, which is not a criticism of the machine: it is a different tool, built for board size, component range and package-on-package work rather than for chip volume. The calculation is what separates the two roles, and it is why "which is faster" is the wrong first question.
Accuracy specifications are stated as a tolerance, often at a stated confidence such as plus or minus 0.05 mm at three sigma, and that form matters. A tolerance without a confidence level cannot be compared across models, and a tolerance measured on a favourable package does not describe your 0.4 mm pitch connector. What you actually need is repeatability at the component that decides your yield: the finest pitch part, the smallest chip, or the largest connector with the least self-aligning capability during reflow. Ask which component and which vision mode the figure refers to, and then test that component on your board.
Cpk turns the tolerance into a process statement, and it is the number to write into the acceptance criteria. A machine that places within tolerance on average but with wide scatter will fail a Cpk requirement even though the average looks good. In practice this is where the vision system, the nozzle condition, the board support and the clamp all meet. A useful acceptance test places the same fine-pitch component at the four corners and the centre of the panel, then measures the result at each position; edge-of-travel behaviour is where a machine's real accuracy budget shows.
A fast mounter in front of a slow printer produces a queue, not output. Line balance means the placement section is sized against the real cycle time of printing, inspection and reflow, and against the transfer behaviour between them. If the printer needs 22 seconds per board and the placement section needs 12, the line runs at 22 seconds and the extra placement speed is wasted capital. If placement needs 30 seconds, the printer waits and the whole line inherits the bottleneck. Compute the cycle time of each station on your board, then decide whether the placement section needs one machine or two, and whether a buffer is the cheaper answer.
This is also where a second machine earns its place for reasons other than speed. Splitting the BOM between a chip-focused machine and a flexible machine often reduces total cycle time more than buying one faster machine, because each head runs the component mix it was designed for. It also gives the line a fallback: when one platform is down for maintenance, the other still produces part of the BOM. The trade is floor space, capital and additional programming, and the calculation above is what makes that trade visible instead of theoretical.
Price is the last input, not the first, and it should be treated as an envelope rather than a number. The quotation covers a machine, but the project covers feeders, nozzles, carts, support tooling, programming software, installation, training, spare parts and the floor space the machine occupies for five years. Two offers with the same headline price can differ by a third once the configuration is completed, and the difference is usually in exactly the items that were left out to make the price look good. Building a cost envelope means writing down every line you will actually pay for, then comparing offers on the completed list. It also means deciding what the wrong envelope would cost, because that number is often larger than the price difference that started the argument.
The envelope has a second purpose: it makes the new-versus-used question answerable. A used platform with a full feeder set, a documented service history and local support can be excellent value, and a used platform with missing feeders, unknown nozzle condition and no local parts channel can cost more than new within two years. Neither outcome is decided by the asking price. It is decided by the completeness of the configuration and by what you can verify before payment, which is why the budget chapter belongs after the range, envelope, feeder and capacity chapters rather than before them.
The first line is the platform itself: the base machine with the heads, vision and board handling it needs to meet your requirement sheet. The second is feeders, and this is where budgets are usually broken, because a machine quoted with a token feeder set needs another thirty to sixty feeders to run your BOM. The third is nozzles and the consumable set, including the spare family for the small end and the large end of your component range. The fourth is the invisible infrastructure: carts, support pins and plates, programming and line software, installation, training, freight, and the compressed air and power connections, each of which has a real number attached.
Ask for each offer to be broken into these four lines with quantities, then compare line by line rather than in total. A lower total with a thinner feeder set is not a cheaper offer, it is a smaller scope. The same applies to the nozzle family: a machine that needs five nozzle types to cover your BOM carries a different consumable burden from one that needs three, and the difference recurs every year. When the four lines are explicit, negotiation becomes a conversation about scope instead of a discount contest, and the final comparison reflects what will actually be installed on your floor.
What changes between new, used and refurbished is not mainly the machine's capability; it is the verifiability of its condition. A new machine comes with a known configuration, a warranty and a documented installation. A used machine comes with a history you have to reconstruct: hours run, maintenance records, head and nozzle condition, software version, and whether the feeder set is complete and matched. A refurbished machine sits between the two, depending on who did the work and what the report actually covers. In all three cases the selection framework is identical, which is a useful discipline: the envelope, feeder plan and capacity check do not care whether the platform is new.
What does change is the risk profile and therefore the acceptance test. For a used machine, add a condition inspection with measurable outputs: placement accuracy on a test coupon at several board positions, feeder indexing accuracy across the bank, vacuum performance at each nozzle station, and a full backup of the software and programs. Ask for the maintenance log and the spare-parts list, and check whether consumables are still available. If the platform is discontinued, the relevant question is not whether it still runs but whether it will still run in year four, when a head cable or a vision board fails.
The table below is a hypothetical comparison of two configurations for the same requirement, built only to show the shape of the arithmetic. Option A is one flexible platform sized for range and board envelope; Option B is two platforms, one chip-focused and one flexible, splitting the BOM. All figures are assumptions for illustration and are not quotations, and they exclude tax, freight and building works. The point is not which option wins; it is that the decision changes once consumables, changeover labour and the cost of a capacity shortfall are included.
Cost line (hypothetical) | Option A: one flexible platform | Option B: two-platform split |
|---|---|---|
Platform and configuration, year 1 | Lower capital, one installation | Higher capital, two installations |
Feeder and cart set, year 1 | Must cover the whole BOM on one bank | Each bank covers part of the BOM; some duplication |
Consumables, years 1 to 5 | One nozzle family, one wear profile | Two nozzle families, two wear profiles |
Changeover labour, years 1 to 5 | Higher if positions are short | Lower per product, more programs to maintain |
Capacity risk | Single point of failure | Partial production continues during a stop |
Floor space and manpower | Lower | Higher |
The first mistake is cutting feeder positions, because the saving is visible and the cost is a recurring changeover that nobody books against the project. The second is cutting the nozzle family, which shows up as hand-placed parts and reduced first-pass yield. The third is cutting the support tooling and the acceptance test, which shows up as accuracy complaints in month two that cannot be traced to a single cause. Each of these saves a small amount at purchase and creates an open-ended cost afterwards, and each is easy to avoid by putting the item in the requirement sheet instead of the negotiation.
The fourth mistake is buying capacity you cannot feed. A machine with large headroom needs work to keep it busy, and idle capacity still carries depreciation, floor space and maintenance. The right target is enough margin to absorb a bad week and a new product introduction, not enough to run a second plant. If the capacity calculation shows a comfortable margin, the next question is whether the money is better spent on feeders, on a second nozzle family or on the printer and inspection equipment that determine whether the fast placement section can actually run at speed.
The framework only pays off if it becomes a procedure the team can repeat, with outputs at each step and a decision at the end. The procedure below has six steps and takes about three weeks of part-time work, most of it in the first week. It ends with an RFQ that a supplier can answer precisely and an acceptance test that protects the investment. The reason to formalize it is not bureaucracy: it is that equipment decisions made without a procedure are usually made by whoever speaks last, and the cost of that appears years later when the plant tries to introduce a product the machine cannot build.
Run the steps in order and do not skip the ones that feel like paperwork. The requirement sheet is what makes the shortlist defensible; the envelope filter is what prevents a wasted demonstration; the feeder plan is what makes the price comparison meaningful; the capacity check is what stops you buying the wrong role of machine; the acceptance criteria are what convert a promise into a test. Each step produces a document, and the six documents together are the project file that survives staff changes, supplier changes and the next product introduction.
Step one is the requirement sheet from the first chapter: five representative boards, the seven inputs, and the extremes recorded rather than averages. Step two applies the envelope filter, removing every model that cannot hold your largest panel or your tallest component, using the published figures and confirming the clamp and rail conditions by question rather than assumption. Step three builds the feeder plan: unique lines grouped by tape width, converted to positions with the lane rule of the feeder type you intend to buy, plus the tray and stick requirement. The output of step three is a single number for positions, and a comparison against the models that survived step two.
These three steps eliminate most of the market in a week, and they do it without a single sales meeting. That is the point. A shortlist of two or three models, each with a written reason for surviving, is a much stronger position than a shortlist of six compiled from recommendations. Keep the eliminated models and the reason for each in the file as well, because the reason often changes when the product mix changes, and it is useful to know whether a model was excluded by a hard constraint or by a preference.
Step four runs the capacity calculation for each survivor: placements per board, boards per month, available minutes, required sustained rate, and the derated rate from the published figure. If a model clears the requirement with thirty to forty percent headroom, it passes. If it needs a second machine to clear it, that is a different project and should be priced as one. Step five is the line trial, run on your boards and your five difficult lines, with the printer and the downstream process behaving as they will in production, because a placement demonstration in isolation hides the transfer and support problems that cause most of the first-month pain.
Step six writes the acceptance criteria before the order, not after the installation. Include placement accuracy at the corners and the centre of your panel, feeder indexing across the full bank, a successful program recall after a simulated changeover, a first-article run with your inspection criteria, and documentation of the software backup and the spare-parts list. Add the training scope and the response-time commitment for service. These items cost nothing to write and are difficult to negotiate later, which is precisely why they belong in the RFQ rather than in a conversation after the machine has arrived.
Return to the hypothetical plant from the capacity chapter: 25,000 boards a month, 260 placements per board, two shifts, 22 days, 7.5 productive hours per shift. The required sustained rate is about 19,700 CPH. The BOM has 129 unique lines, converting to about 78 feeder positions plus a tray bank. The largest panel is 1,150 × 520 mm, the largest component is a 33.5 mm square connector, the tallest part is 8 mm, and the smallest chip is 0201. Finest pitch is 0.4 mm, and the plant wants first-pass yield at or above 99 percent with placement Cpk at or above 1.33 on the fine-pitch part.
The envelope filter removes every model that cannot hold 1,150 × 520 mm, which is a short list on its own. The feeder plan removes models with fewer than about 80 positions once the tray requirement is counted. The component range removes models that stop below a 33.5 mm body or a 6 mm height. What survives is a flexible platform sized for the envelope and the range, with a documented capacity figure that clears 19,700 CPH at a realistic derating. The high-speed chip shooter that looked attractive on the speed table is excluded by the feeder and component limits, which is exactly the kind of conclusion the framework is meant to produce before money is committed.
The RFQ should restate the requirement sheet, not the model you have in mind. Include the five representative boards with placement counts and line counts, the panel dimensions, the component extremes, the feeder plan by tape width, the required sustained rate with the efficiency assumption stated, the accuracy and Cpk targets, and the acceptance test you intend to run. Then ask for the configuration to be quoted in the four cost lines, with feeder quantities and nozzle families listed explicitly, and ask for the changeover time for a representative product swap.
Add three questions that reveal more than the price. Ask which components in your BOM are outside the published range or need a special nozzle, because the answer tests whether the supplier read the sheet. Ask what the realistic rate is for your mix rather than the optimum figure, and on what basis. Ask for the spare-parts list with lead times for the head, the vision module and the feeder interface. The answers to these three questions will tell you more about the next five years than any discount you can negotiate. That is how to choose a JUKI pick and place machine with evidence rather than enthusiasm, and it is the difference between a machine that fits the plant and one the plant has to fit around.
Turn the requirement sheet into a configured JUKI line
Send us the five boards, the BOM summary and the panel dimensions, and we will map them against the JUKI range, quote the feeder and nozzle scope explicitly, and run the capacity arithmetic with you. You get a configuration that matches your envelope, your feeder plan and your output target, with the assumptions written down rather than implied. The review starts from the same sheet this article describes: unique lines by tape width, largest and smallest component, tallest part, panel size with rails, placements per board and the monthly board requirement. From those inputs we can show which models clear the hard constraints, which clear them only with a special nozzle or a slower mode, and where a second platform would beat a larger one. If a specification on your shortlist cannot be confirmed against the current datasheet, we will say so instead of quoting a number we cannot support, and if the honest answer is that your board mix fits a different family, that is what you will hear.
About three weeks of part-time work if the product data is available. The first week produces the requirement sheet, the envelope filter and the feeder plan, which is where most of the value is created. The second week runs the capacity arithmetic and requests configurations against the requirement sheet rather than against a model number. The third week handles the line trial and writes the acceptance criteria. Projects that take three months usually spend the extra time waiting for quotations that were requested before the requirement sheet existed, which then have to be re-quoted once the requirements become clear.
It depends on whether your BOM has two different jobs in it. If the board is mostly chips with a handful of larger parts, one flexible platform usually covers the range and the capacity. If the board combines a large chip count with large connectors, fine pitch or package-on-package work, splitting the BOM between a chip-focused machine and a flexible machine often gives a shorter cycle time and a useful fallback during maintenance. Run the capacity calculation for both options and include floor space, programming effort and the cost of a single point of failure, then compare the totals rather than the machine prices.
Sometimes, and the published range is the place to check rather than the family name. A platform documented at 0402 down to 01005 and up to a 33.5 mm square body covers a very wide span, but it may need different nozzles for the two ends and a slower placement mode for the extremes. The practical questions are whether the nozzle family is available locally, how many nozzle changes the program requires, and whether the component height limit allows the head to pass over parts already placed. A machine that covers the range with three nozzle types is a different proposition from one that needs six.
Use it as an upper bound and plan from a derated figure. When a product page publishes both an optimum rate and an IPC-9850 rate, the second is the more defensible basis because it comes from a standardized characterization method, and the ratio between the two shows how optimistic the peak number is. On the pages reviewed for this article the ratio differed substantially between models, which is exactly why comparing optimum figures across models can mislead. Derate further for your own mix, then require thirty to forty percent headroom over the calculated demand so that a bad week does not become a missed shipment.
Judge both with the same framework and change only the verification. The envelope, feeder plan and capacity check do not care whether the platform is new. What changes is what you must prove before payment: placement accuracy on a test coupon at several board positions, feeder indexing across the bank, vacuum performance at each nozzle station, a complete software backup, and a spare-parts list with lead times. A used machine with a complete feeder set, a documented history and a local parts channel can be excellent value; one with missing feeders and an unknown head condition can cost more than new within two years.
Plan for the largest panel you would realistically accept as an order in the next three years, not the largest you can imagine, and check the clamp and rail conditions for that panel rather than only the published maximum. Oversizing the envelope costs floor space and capital every month; undersizing it closes the door on a product family permanently. A practical compromise is to size the envelope for the realistic maximum and size the feeder bank with some spare positions, because feeder capacity is cheaper to add later than board handling capability.
Cover the four limits you selected on. Place your fine-pitch component at the corners and the centre of your largest panel and measure the result at each position. Run the full feeder bank and confirm indexing and address verification after a simulated changeover. Recall a program and produce a first article that passes your inspection criteria. Confirm the software backup, the spare-parts list and the training scope are delivered. Add the response-time commitment for service in writing. These tests take a day and they convert the specifications you bought into evidence you can hold.
