English Views: 0 Author: Elsa Long Publish Time: 2026-11-05 Origin: Site
Most mounter selections start with a speed figure and end with a BOM that does not fit, and the JUKI chip mounter BOM feeder nozzle selection method exists to prevent exactly that order of events. The machine is chosen last, because the machine is the answer rather than the question. Your printed circuit board assembly declares three things that a placement platform has to satisfy: the packages you place, the way those parts arrive at the machine, and the tool that picks each one up. Those three declarations become a BOM ledger, a feeder ledger and a nozzle ledger, and the platform is whatever satisfies all three without forcing a manual step. The machines that can do this sit inside the SMT pick-and-place platforms we build and supply, and the differences between them show up in the ledgers long before they show up in a rate column.
This guide works through the ledgers in the order that prevents expensive mistakes. It reads a BOM by package rather than by part count, converts the package mix into a feeder requirement in positions and tape widths, then converts the same mix into a nozzle requirement in orifice sizes and tool families. Only after those three are written down does it compare heads and rates, and only then does it produce a configuration. Every calculated figure is labeled as a hypothetical assumption, every published specification is attributed to the product page or to JUKI's own data, and nothing here claims a customer result that you cannot verify on your own board.
A bill of materials looks like a list of parts, and most buyers read it that way: count the lines, sort by size, ask for a machine that covers the biggest and the smallest. That reading hides the three things that actually constrain a placement platform. A BOM fixes which packages must be handled, how those packages arrive at the machine, and which tool picks each one up, and those are three different constraints with three different failure modes. A package that exceeds the component ceiling cannot be placed at all; a package that arrives on a tray consumes a different kind of position than one on tape; a package that needs a nozzle you do not stock turns into a setup stop on the first production run. Reading the BOM as a constraint set means writing those three answers down before you look at a machine, and it is the same discipline described in the wider JUKI machine selection method, applied here at the level of individual packages rather than machine families.
The order matters because the constraints are not equally negotiable. Package size is close to absolute: if the tallest part is 30 mm, a head with a 25 mm clearance will not place it, and no feeder or nozzle choice changes that. Feeder format is negotiable at a cost: a part that arrives on a tray can sometimes be re-reeled, and a part on a stick can sometimes be trayed, but each conversion adds handling and a failure point. Nozzle choice is the most negotiable of the three and the most expensive to get wrong in practice, because a nozzle that is close enough to work will work badly for months before anyone traces the defect rate back to it. Writing the three ledgers in that order, hardest constraint first, keeps the negotiation where it belongs.
The first number most people quote from a BOM is the placement count, and it is the least useful number for machine selection. A board with 8,000 placements and 45 unique lines behaves completely differently from a board with 8,000 placements and 220 unique lines, even though the cycle time arithmetic looks similar at first glance. The first board fits a small feeder bank and changes over in minutes; the second needs a bank large enough to hold most of those 220 lines, or a disciplined staging scheme, or a second machine. Rate is a function of placements; feeder capacity, changeover time and setup labour are functions of lines. Confusing the two is how a plant buys a fast machine and then discovers it spends its day swapping carts.
Count both, and count them per product family rather than per BOM. Take the last twelve months of production, list each family, and for each one record the number of unique lines, the number of placements, and the widest and tallest part. The unique line count drives the bank; the placement count drives the head; the widest and tallest parts drive the envelope and the head clearance. A family with 300 unique lines and 2,000 placements is a feeder problem with a modest rate requirement, and buying it a 100,000 CPH rotary machine solves the wrong half. A family with 40 lines and 30,000 placements is the opposite case, and no amount of feeder capacity will make a flexible head competitive on it.
The three ledgers can be written on one page each. The package ledger lists every unique footprint, its body size, its lead pitch or ball pitch, its height, and its mass class if it is heavy enough to need force control. The feeder ledger lists, for every line, the delivery format (tape, tray, stick, bulk), the tape width if it is on tape, and the number of positions that line will occupy in the bank. The nozzle ledger lists, for every line, the pickup area, the nozzle family that fits it, and whether that family is already in your stock. The three ledgers share one key, the line number, and that is what makes them useful: a line that appears in all three with a consistent answer is a line the machine can run, and a line that produces a contradiction is the line that will stop the line at 2 a.m.
Build them in a spreadsheet, not in a document, because the useful questions are all filters. Filter the package ledger by height to find the part that exceeds a candidate machine's clearance. Filter by body size to find the part that exceeds its component ceiling. Filter the feeder ledger by format to see how many tray positions you actually need. Filter the nozzle ledger by family to count how many tools a full changeover requires. Each filter produces a number that a supplier can answer with a yes or a no, which is a far better conversation than a general discussion about flexibility. The ledgers also survive the purchase: they become the setup sheet, the spare-parts list and the training document for the machine you eventually choose.
Contradictions between the ledgers are common and they are always resolved by the machine, not by the plan. A 40 mm connector that fits the component ceiling but sits 1 mm from a 25 mm-tall transformer may be unplaceable in sequence, because the head has to travel over parts already placed. A 0201 chip that needs a 0.25 mm orifice nozzle may share a feeder bank with a 56 mm tape that occupies eight positions, and the bank that looked generous on paper is suddenly full. A BGA that arrives on a tray in a matrix of 20 may need a tray stacker that consumes floor space the line does not have. None of these shows up in a speed comparison, and all of them show up in commissioning week.
The way to find them early is to run the contradiction check deliberately: for each line, ask which of the three ledgers constrains it most, and then ask what happens to the other two if that constraint is relaxed. If relaxing the feeder format forces a new nozzle, the line has two costs rather than one, and it belongs on the shortlist of parts to discuss with the supplier before the machine is specified. Plants that do this usually find three to eight awkward lines in a BOM of two hundred, and those lines are worth more attention than the other 192, because those are the ones that decide whether the machine runs unattended or needs an operator standing next to it.
Package families differ in more than size, and the differences map onto different machine features, which is why JUKI chip mounter BOM feeder nozzle selection starts by reading the BOM by footprint rather than by part number. A chip resistor asks for a small orifice and a fast head; a QFN asks for vision that can find a nearly invisible lead; a connector asks for a nozzle with a wide contact face and a head that can apply force without crushing the housing; a BGA asks for vision that can see spheres under a body and a placement force low enough not to smear paste. Reading a BOM by package family rather than by individual part number compresses two hundred lines into six or seven decisions, and each decision has a published answer on the machine specification. This chapter takes those families in order of how much they constrain the machine, from the chips that only ask for speed to the parts that can eliminate a machine from consideration.
The published figures used here come from the product listings on this site, which quote the same specification table for the RS and RX families: component height 25 mm on the RS models, 3–10.5 mm head-dependent on the RX-7R and 3 mm on the RX-8; largest component 74 mm square or 150 × 50 mm on the RS models, 25 mm square on the RX-7R and 5 mm square on the RX-8; smallest component 0201 metric on the RS models, 03015 on the RX-7R and 0201 on the RX-8. Those four rows, read together, tell you more about which parts of a BOM are compatible than any narrative description of flexibility.
A 0201 metric chip measures 0.25 × 0.125 mm, which is smaller than the registration error of a poorly maintained printer. Placing it is not the hard part; seeing it, holding it and not losing it to an air pulse is. The machine side of the answer is a small orifice nozzle, a vision system that can resolve the part against a dark background, and a head that can move at speed without shedding the part on acceleration. The RS models publish 0201 as their smallest component, and the RX-7R publishes 03015, which is smaller still, so both families can handle this class on paper. Whether they handle it in your plant depends on the feeder, the nozzle condition and the paste deposit, which is why the nozzle ledger belongs next to the package ledger rather than after the purchase.
The practical test is a run of two hundred boards with the smallest part in the BOM placed at production speed, inspected for missing and rotated parts. If the miss rate is dominated by pickup rather than placement, the nozzle is the suspect; if it is dominated by recognition, the vision lighting or the feeder tape is; if it is dominated by tombstoning after reflow, the paste deposit and the placement force are. That triage is worth doing before you blame the machine, and it is worth doing with the nozzle you intend to run in production rather than with the one that came in the tool kit. A plant that cannot pass this test on 0201 should specify the machine for 0402 and above and treat 0201 as a separate decision, because a machine bought for a capability the process cannot support is a machine bought for a brochure.
A 0.4 mm pitch QFN is a vision problem more than a placement problem. The lead is 0.2 mm wide, the body is often 5 mm square, and the part has no leads to grip, so the nozzle has to pick it by the body and the vision system has to find the body edge rather than the leads. The machine feature that matters is the ability to teach a body-edge recognition window and to place with a force low enough that the part settles into paste without squeezing it out from under the pads. The RS family's Takumi inline head is described on the product pages as adapting to different shapes with soft force control, which is the relevant capability here; the rotary heads in the RX family are optimized for rate on small chips and publish a lower component ceiling, so a board dominated by fine-pitch QFNs is not their strongest case.
The paste-side limit is the one buyers miss. A fine-pitch part placed on a deposit that is 20 % too tall will bridge regardless of the machine, and a part placed with 30 % too much force will have its paste displaced before reflow. Both look like placement defects in the yield report and neither is fixed by a better mounter. Before specifying a machine for fine pitch, measure the paste height and volume on twenty boards from the printer that will feed it, and confirm the printer can hold the tolerance the pitch requires. If it cannot, the mounter specification is not the constraint and buying a more accurate machine will not move the yield.
Connectors are where the families separate. The RS models publish a largest component of 74 mm square or 150 × 50 mm with a 25 mm height ceiling, which covers most board-edge connectors, USB housings, relay bodies and shield frames used in industrial and power products. The RX-7R publishes 25 mm square with a 3–10.5 mm head-dependent height, and the RX-8 publishes 5 mm square with a 3 mm height. Those ceilings are not marketing limits; they are the geometry the head and the vision camera can physically handle. A BOM whose largest part is a 30 mm connector eliminates the RX-8 and, depending on the head fitted, the RX-7R as well, no matter how attractive their rates look.
Measure the part as it sits on the board, including any metal tab, locating peg or housing overhang that extends beyond the footprint. The published ceiling refers to the body the machine sees, and a connector that is 24 mm square on the datasheet but sits on a 30 mm keep-out with a 6 mm overhanging latch may exceed the limit in the orientation the program requires. Where the part is close to the ceiling, the board-envelope comparison across the RS-1R, RS-1XL and RS-2 is worth reading for the panel geometry side of the same question, and the supplier should confirm the ceiling for the specific head and nozzle you intend to fit rather than the family maximum.
Package family (typical) | How it usually arrives | Nozzle family that fits | Machine constraint it tests |
|---|---|---|---|
0201–0402 chip, resistor or capacitor | 8 mm tape | 501 or 502 (0.25–0.35 mm orifice) | Vision resolution and pickup stability |
0603–1206 chip, SOT-23, SOD | 8 mm tape | 503 (0.6 mm orifice) | Rate, not capability |
QFP, QFN, SOIC at 0.4–0.65 mm pitch | 12–24 mm tape or tray | 504–505 (1.0–1.7 mm orifice) | Body-edge vision and soft force |
BGA, CSP, LGA | Tray, sometimes 24–44 mm tape | 506 (3.2 mm orifice) | Ball recognition and force control |
Electrolytic, inductor, relay body | 16–44 mm tape or tray | 507–508 (5.0–8.0 mm orifice) | Component ceiling and height |
MELF diode | 8–12 mm tape | 511 (MELF-specific, 1.2 mm) | Round-body pickup without rolling |
Board-edge connector, USB, header | Tray or 44–88 mm tape | 517 (9.0 × 3.4 mm face) | Largest-component and height ceiling |
The table is a starting point, not a substitution for the nozzle list, and the orifice figures in it come from the part numbers published on the nozzle listing rather than from a generic rule. A part that sits between two families is decided by its pickup area and its mass, not by its name: a heavy inductor with a 6 mm body needs a larger orifice than a light QFN of the same footprint, because the vacuum has to hold the mass through acceleration. When a line is ambiguous, put both candidate nozzles in the ledger and resolve it on the machine with a pickup test rather than on paper with a guess.
Feeders are where a BOM turns into hardware, and this is the stage of JUKI chip mounter BOM feeder nozzle selection where most quotations go wrong, because the conversion is not intuitive. A line does not occupy one position; it occupies as many positions as its tape is wide, measured in 8 mm slots, unless the feeder is a double-lane type that carries two narrow tapes in the space of one. That single rule explains most of the surprise in a machine purchase. A BOM with 180 unique lines sounds like it needs 180 positions, and on 8 mm tape with double-lane feeders it may fit in far fewer, while a BOM with 90 lines that includes twenty 24 mm and ten 44 mm tapes may need more positions than the 180-line board. The feeder listing on this site covers the JUKI-compatible range from 8 mm through 88 mm, and the JUKI tape feeders from 8 mm to 88 mm are the physical units that decide how your bank is actually consumed.
The bank size that matters is not the family maximum. The RS models publish 112 feeder inputs, the RX-7R publishes 76 and the RX-8 publishes 56, and those are maximums under a stated feeder type rather than a promise about your mix. A bank of 112 positions loaded entirely with 8 mm double-lane feeders can hold more than 112 lines; the same bank loaded with 24 mm single-lane feeders holds a fraction of that. Two plants with identical machines and different tape mixes therefore report completely different changeover frequencies, and neither is wrong. This chapter converts a BOM into positions, then converts positions into a bank decision, and keeps the assumptions visible so you can redo the arithmetic with your own numbers.
Tape width is set by the pocket pitch and the component size, and it steps in standard widths: 8, 12, 16, 24, 32, 44, 56, 72 and 88 mm. In a bank measured in 8 mm slots, a single-lane feeder of width W consumes W/8 slots, so a 24 mm feeder takes three slots, a 44 mm feeder takes five and a half, and an 88 mm feeder takes eleven. Double-lane 8 mm feeders take one slot for two lines, which is why the same bank supports very different line counts depending on how many narrow parts a BOM contains. The arithmetic is crude but it is the arithmetic the machine uses, and doing it before the purchase converts a vague worry about capacity into a number you can compare with a quotation.
Width selection also has a floor set by the component, not by preference. A 12 mm part on 8 mm tape will not fit its pocket, and forcing a larger part onto narrower tape is a reliable way to create pickup errors that look like nozzle faults. The reverse, a small chip on wider tape, is wasteful but safe, and it is what happens when a plant standardizes on a narrower set of widths to simplify feeder stock. That trade is legitimate: fewer widths means fewer feeder types to buy, maintain and train on, at the cost of positions consumed. The right answer depends on how close your bank is to full, which is why the position arithmetic has to come before the standardization decision.
Take a hypothetical BOM of 120 unique lines and sort it by tape width. Suppose 78 lines are 8 mm, 18 are 12 mm, 12 are 16 mm, 8 are 24 mm and 4 are 44 mm. In single-lane terms that is 78 slots plus 27 plus 24 plus 24 plus 22, or 175 slots, which exceeds a 112-position bank. Load the 8 mm lines as double-lane feeders and the same BOM consumes 39 slots for those 78 lines plus 27 plus 24 plus 24 plus 22, or 136 slots, which still exceeds the bank. The conclusion is not that the machine is too small; it is that this BOM will not run in one pass on one bank, and the plant has to choose between a second bank, a staged setup, or a second machine.
Run the same arithmetic again with 105 lines instead of 120, and with 70 of them on 8 mm. Double-lane: 35 plus 20 plus 18 plus 18 plus 22, or 113 slots, which is one slot over a 112-position bank. That single slot is the kind of detail that decides a purchase, and it is invisible in any comparison that quotes feeder counts as a single number. Notice also how sensitive the result is to the 8 mm proportion: every two 8 mm lines converted from single to double lane frees a slot, so a BOM that is one slot over is often solvable by buying four more double-lane feeders rather than a larger machine.
Feeder width (single lane) | Slots consumed in an 8 mm bank | Typical components | When it is the wrong choice |
|---|---|---|---|
8 mm, single lane | 1 | 0201–1206 chips, SOT-23, small diodes | When the bank is tight and double-lane is available |
8 mm, double lane | 1 for two lines | The same parts, high line count | When pocket pitch needs the full 8 mm width |
12 mm | 1.5 | SOD-123, MELF, small modules | Rarely — it is the usual step up from 8 mm |
16 mm | 2 | SOIC, small QFN, tantalum | When the part would fit a 12 mm pocket |
24 mm | 3 | QFP, larger modules, tray-to-tape conversions | When a tray feeder would be more economical |
32 mm | 4 | Large ICs, connectors on tape | When the part count per reel is low |
44 mm | 5.5 | Large connectors, power modules | When a tray is the original packaging |
56–88 mm | 7–11 | Very large bodies, shield frames | When a tray feeder or manual station is cheaper |
A feeder bank is not a fixed asset, it is a set of carts that can be prepared away from the machine. That fact changes the position arithmetic in a way that is easy to miss: if a BOM does not fit a 112-position bank but does fit two carts of 112 positions, the machine does not need a bigger bank, it needs a staging discipline and a cart exchange time short enough to absorb. Two pre-loaded carts, each carrying the lines for one product family, turn a changeover from a reload into a swap, and the swap time is measured in minutes rather than in the hours a full reload takes. The capital cost of the second cart is almost always lower than the capital cost of the larger machine, and the operational cost is the discipline of keeping the carts correct.
That discipline is the catch. A pre-loaded cart only works if the reel mapping is verified before the cart goes onto the machine, because a wrong reel on a prepared cart produces a wrong part on the first board of the run, and that defect is more expensive than the changeover it saved. The practical way to run it is a scan-and-verify step at cart build time, with the cart labelled by product family and the verification recorded. Plants that do this report changeovers that hold their time even when the operator changes, because the cart carries the knowledge rather than the person. Plants that skip it report prepared carts that nobody trusts, and the machine ends up being reloaded manually anyway.
Not every part can be taped, and the exceptions cluster in the same places across the industry: BGAs in matrix trays, connectors with fragile housings, modules that are too heavy for a pocket, and parts with a keep-out that a cover tape would violate. A tray feeder presents a matrix of pockets to the head and consumes space in a different currency, because it occupies bank space or floor space depending on the design, and it may need a stacker to hold several trays for a long run. A stick feeder gravity-feeds tubes and is cheap and compact, but its reliability depends on tube angle and part geometry, and a stick that jams at 3 a.m. costs more than the feeder saved.
The practical rule is to count tray and stick lines separately and treat them as a different budget. If a BOM has six tray lines, the question is not how many tape positions the machine has but how many tray positions it can present at once and whether a stacker is needed for the batch size. Six tray lines at 400 boards per batch may be fine with a single tray position and an operator reload; six tray lines at 4,000 boards per batch will not be. Write the batch size next to the tray line in the ledger, because the same six lines can be a minor inconvenience or the constraint that decides between two machines.
The parts that cause the most trouble in a placement program are rarely the smallest or the fastest. They are the awkward ones: a connector that has to be placed with 15 N of force and cannot be gripped by its body, a shield frame that is 22 mm tall and sits beside a part that is 24 mm tall, a transformer that arrives in a tray with 1 mm clearance between pockets, a daughterboard connector that has to go in last because everything else would block the nozzle path. Each of these is a legitimate SMT part, and each one can remove a machine from consideration even though the machine's rate figure is unaffected. This chapter covers the three categories that most often force the choice: tray-presented parts, odd-form parts that need force or a dedicated tool, and product families where the geometry rather than the part mix drives the decision.
The common thread is that these parts consume machine capability in ways the specification table compresses into single numbers. A 25 mm height ceiling and a 74 mm square component limit describe the envelope a part must fit inside, but they do not describe what happens when a tall part is placed early and a wide part has to travel over it. Nor do they describe the reliability of a nozzle that grips a connector by two rails rather than a flat face. Reading those rows as absolute limits is necessary but not sufficient; the sequence, the approach angle and the tool have to be checked for the specific parts, which is what makes the odd-form section of a BOM worth its own review.
A tape feeder is a fixed unit that occupies a known number of slots and presents parts at a known pitch. A tray feeder is a different kind of object: it presents a matrix, it may need to be indexed between picks, and its capacity is measured in trays rather than in slots. That difference breaks the position arithmetic that works for tape, because a tray line does not compete for tape slots in the same way and may instead compete for floor space or for a dedicated machine zone. A BOM with a dozen tray lines can therefore consume more of a machine's practical capacity than its line count suggests, and the machine with the larger tape bank is not automatically the better choice for it.
Two design details matter more than the tray count. The first is the tray pitch and pocket clearance, which decide whether the nozzle can reach a part without touching its neighbour; a tray of 20 mm modules at 21 mm pitch leaves 1 mm for the nozzle body, which may be less than the nozzle needs. The second is the reload behaviour: whether the machine pauses when a tray is exhausted, whether a second tray position lets it continue, and how long a reload takes. A tray line that pauses the machine every 40 boards is a different proposition from one that pauses every 400, and the difference is in the feeder design rather than in the placement head.
Odd-form parts are the ones where the head has to do something other than pick and drop. A connector with a press-fit body needs a controlled downward force; a shield frame needs a nozzle that contacts the rim rather than the centre; a tall electrolytic needs a head that can approach from a height the neighbouring parts allow. The RS family's published 25 mm height ceiling and 74 mm square component limit cover the geometry of most of these parts, and the Takumi inline head is described as adapting to different shapes with soft force control, which is the relevant capability for parts that must not be crushed. The RX family's lower ceilings, 3–10.5 mm on the RX-7R and 3 mm on the RX-8, make those machines a poor fit for a BOM whose odd-form content is tall.
Sequence is the part of odd-form work that no specification row captures. If a 22 mm transformer is placed before a 6 mm connector that sits 4 mm away, the head may not be able to reach the connector without the nozzle body contacting the transformer. The usual remedies are to place the tall parts last, to use a nozzle with a longer reach, or to move the placement to a second pass. Each has a cost: placing tall parts last can conflict with reflow profile or with adhesive cure, a longer nozzle reduces placement accuracy slightly, and a second pass adds handling. The one remedy that is not available is to assume the machine will find a way, because the machine executes the program it is given.
LED products invert the usual BOM shape. A strip or a driver board may have only 30 to 60 unique lines, but the placement count per board can be high because the same LED and resistor repeat along the length. That shape rewards feeder capacity per line and penalizes anything that interrupts a long run: the bank only needs a modest number of positions, but the machine needs to run for hours without a reel change, and the board may be long enough that transport rather than placement sets the cycle. A build such as the RS-1R-based LED placement configuration is aimed at exactly this shape, with the flexible head covering the LED body and the associated connectors rather than optimizing for chip rate alone.
The trap in LED work is assuming that a low line count means a simple machine. It often means the opposite: the placement count is high, the board is long, and the process is sensitive to placement force because an LED die does not tolerate a heavy touch. A plant running LED strips should test the LED part specifically, at production speed, and inspect for die damage and for tilt after reflow, rather than accepting a general statement about component range. The same board also tends to need a nozzle with a wide, flat contact face rather than a small orifice, which is another reason the nozzle ledger has to be built from the actual part rather than from the package name.
The nozzle is the cheapest component in the decision and the one most often chosen by habit. It is also the component that converts a correct machine choice into a stable process or a permanent defect rate. Orifice size sets how much vacuum reaches the part, which sets how reliably the part survives acceleration and how much it shifts on contact. Too large an orifice on a small part draws it into the tip and rotates it; too small an orifice on a heavy part drops it during travel. Both faults look like placement errors, both are intermittent, and both are usually blamed on the machine before anyone checks the tool. A nozzle ledger, built line by line from the pickup area and the mass of each part, removes that uncertainty before the machine is ordered.
The JUKI chip mounter BOM feeder nozzle selection work reaches its most granular point here, because the JUKI nozzle range is documented by part number and orifice diameter on this site's listing, and the numbers are specific enough to build the ledger from. The 500 series runs from a 0.2 mm orifice for the smallest chips to an 8.0 mm orifice for large bodies, with the 511 MELF type and the 517 connector type as distinct shapes rather than oversize circles. The 100 series covers a coarser range for larger parts. Reading that list as a menu of interchangeable tips is the mistake; reading it as a set of tools with published dimensions and intended part classes is the method. What follows converts a BOM into a family count, then into a setup cost, then into a specification.
The package name tells you the body size; the orifice needs the pickup area and the mass. A 0402 chip and a 0402-size LED can share a footprint but not a nozzle, because the LED's silicone lens needs a flat contact face and low force while the chip accepts a small circular tip. As a working rule, the orifice should be small enough that the part covers it completely and large enough that the vacuum holds the part through the acceleration of the head, which for most chip parts means an orifice in the same order of magnitude as the part's shorter dimension. That is why the 501 with a 0.25 mm orifice is the usual tip for 0201 and small 0402, the 502 at 0.35 mm covers 0402 and 0603, and the 503 at 0.6 mm covers 0603 through 1206.
Above that, the rule shifts from the part's footprint to its mass and its surface. A QFP with a 14 mm body is light for its size and picks well with a 1.0 mm orifice, while an inductor of the same footprint may need 1.7 mm or more because it is several times heavier. A BGA with a 27 mm body and a 1.0 mm ball pitch needs a tip that contacts the body without touching the balls, which is a shape question rather than an orifice question. The practical method is to test three candidate nozzles on the first ten parts of a new line, at production speed, and keep the one with zero misses across a hundred picks. That test costs an hour and settles a question that otherwise resurfaces as a yield investigation.
Three nozzle types deserve their own note because they are the ones that get substituted incorrectly. The 511 is the MELF type, published as a 2.0 mm face with a 1.2 mm orifice, shaped for a cylindrical body that will roll if picked with a circular tip on a flat face. A MELF diode picked with a standard 503 will place, then rotate during travel, then fail inspection at a rate that looks random until the tool is checked. The 517 is the connector type, published as a 9.0 × 3.4 mm face, and it exists because a board-edge connector has to be gripped by its housing across a wide span rather than sucked at a point; substituting a large circular nozzle risks tilting the part or marking the housing.
The 100 series, published from a 0.4 mm orifice at 101 up to a 3.2 mm orifice at 104, covers the coarser end: larger chip parts, simple SOT and SOD packages, and the general-purpose picks where a small orifice is unnecessary. Its presence in the range matters for a different reason: it shows that the choice is not strictly ordered by size, since a 104 with a 3.2 mm orifice overlaps the 506 at 3.2 mm in diameter while differing in shape and intended class. Two nozzles with the same orifice can behave differently on the same part, and the only way to know which is right is the pickup test rather than the dimension table.
Nozzle series | Published orifice / face | Parts it is normally specified for | Failure mode when substituted |
|---|---|---|---|
501 | 0.25 mm (alternate 0.2 mm) | 0201 and the smallest 0402 chips | Part drawn into the tip and rotated |
502 | 0.35–0.4 mm | 0402 and 0603 chips, small diodes | Marginal hold on fast moves |
503 | 0.6 mm | 0603–1206 chips, SOT-23, SOD-123 | Over-large orifice on small parts |
504 | 1.0 mm | SOIC, small QFN, 0.5 mm pitch QFP | Insufficient hold on heavy bodies |
505 | 1.7 mm | Larger QFP, QFN, medium modules | Contact on neighbouring tall parts |
506 | 3.2 mm | BGA, CSP, large IC bodies | Ball contact and paste smear |
507 / 508 | 5.0 mm / 8.0 mm | Electrolytics, inductors, relays, shields | Nozzle body interfering with the part |
511 | 2.0 mm face, 1.2 mm orifice | MELF diodes and cylindrical bodies | Rolling during travel, random rotation |
517 | 9.0 × 3.4 mm face | Board-edge connectors, headers, USB | Tilt, housing marks, misalignment |
101–104 | 0.4–3.2 mm orifice | Coarser chips, SOT/SOD, general picks | Shape mismatch rather than size |
The number that matters for production is not how many nozzles you own but how many families a changeover requires. A BOM that runs on three families changes over with three heads or three pockets; a BOM that needs seven families changes over with a tray of tools, a lookup table and an operator who has to get the mapping right every time. That difference shows up as minutes per changeover and as defect risk, and it is invisible in a machine specification. Count the families in the nozzle ledger and treat anything above four as a setup project rather than a detail, because a nozzle mapping error is one of the few mistakes that produces a defect on the very first board of a run and is easy to attribute to the machine.
Two habits reduce the family count without compromising the process. The first is to standardize the chip range on two nozzles rather than four, accepting a slightly larger orifice on the smallest parts if the pickup test passes, which frees two family changes. The second is to group parts by nozzle family in the program and in the feeder layout, so that a changeover that switches families does it once rather than interleaving. Both habits require the ledger to be accurate, and both are documented by the original JUKI nozzle part numbers rather than by generic descriptions. The original JUKI nozzles with published orifice sizes are the reference list this ledger should be built from, because a substitute with an unstated orifice makes the whole calculation unverifiable.
The spare-parts question has a simple answer once the family count is known: two of each family in use, plus one of each special shape. Two rather than one because a single spare fails at the worst moment and leaves the line down while a replacement is ordered, and because a nozzle that has been dropped is not always obviously damaged until it produces a defect. The special shapes, the MELF type and the connector type, are worth an extra unit each because they are used less often and therefore tend to be missing when needed. That list is small enough to hold in a drawer and large enough to cover a normal year.
Nozzle condition is a maintenance item, not a consumable. A nozzle that has picked ten million parts has a worn tip, a partially blocked orifice and a vacuum path that no longer holds the smallest parts reliably; it will still place 0603 chips perfectly and will fail on 0201 in a way that looks like a feeder problem. Inspect the tips under magnification at a fixed interval, measure the vacuum at the tip rather than at the pump, and retire a nozzle on evidence rather than on age. The interval that works in most plants is monthly for the high-use families and quarterly for the special shapes, and the record belongs next to the nozzle ledger so that the family count and the spare count stay consistent.
With three ledgers written, the head choice becomes a filtering exercise rather than a preference. The head has to reach the parts the package ledger allows, hold them with the nozzles the nozzle ledger requires, and pick them from the feeders the feeder ledger describes fast enough to meet the capacity target. Those are three separate tests and a head can pass two of them and fail the third. The published data on this site separates the heads clearly: the Takumi inline head on the RS-1R and RS-1XL, the parallel rotary P16S head on the RX-7R and the P20 rotary head on the RX-8, with component ceilings and rates that follow from those architectures. This chapter turns the ledgers into a head decision and then into a capacity number you can defend.
The capacity number is the last thing to calculate, not the first, because it is the only one that depends on the other three being right. A rate figure quoted under favourable conditions tells you how fast the head can move when nothing goes wrong; the number you need is how many boards per hour the line produces when the feeder bank is loaded, the nozzles are correct and the changeovers are real. That number can be estimated from the published rate with a stated efficiency and a stated overhead, and the estimate should carry both assumptions in writing so that nobody later mistakes it for a guarantee. It also helps to know which rate you are reading: an optimum figure describes the head under favourable conditions, while a figure characterized under IPC-9850 describes the same equipment through a standardized measurement procedure, which is why the two are not interchangeable and why a supplier should tell you which one a quotation uses.
The Takumi inline head is the flexible choice. It is published on the RS-1R at 47,000 CPH optimum with a 74 mm square component ceiling, 25 mm height and 0201 as the smallest part, and on the RS-1XL at 42,000 CPH with the same component range and a wider board envelope. Its value is range: it can place a 0201 chip and a 30 mm connector in the same program, with soft force control for parts that must not be crushed. The parallel rotary head on the RX-7R is published at 75,000 CPH with a 25 mm square component ceiling and a 3–10.5 mm height range, and the P20 rotary head on the RX-8 at 100,000 CPH with a 5 mm square ceiling and 3 mm height. Those ceilings are the price of the rate.
The filtering rule that follows is simple enough to apply without a spreadsheet. If the package ledger contains any part larger than 25 mm square or taller than 10.5 mm, the RX-7R is out; if it contains anything larger than 5 mm square or taller than 3 mm, the RX-8 is out. If either machine survives that filter, the next question is whether the placement count justifies its rate, which is a volume question rather than a capability question. In most mixed-product plants the filter eliminates the rotary machines immediately, and the real comparison happens between the RS models on envelope and feeder capacity, which is examined in the article on how the three JUKI architectures divide the work.
Feeder positions and rate interact through changeover, not through placement. A machine with 112 positions and a BOM of 120 lines will change over between products whether it runs at 42,000 or 75,000 CPH, and the changeover time is the same in both cases. That is why a faster machine with a smaller bank can be the slower choice in high mix: it spends more of the shift being reloaded. The 76 positions on the RX-7R and the 56 on the RX-8 are adequate for a BOM of 40 lines and marginal for a BOM of 120, while the 112 positions on the RS models fit the wider BOM but at a lower rate. This is the trade the earlier arithmetic should have already exposed; if it has not, the feeder ledger is not finished.
There is a second-order effect worth stating: changeover is not only reload time, it is also verification time. Every line that changes needs a part verification, a first-article check or at least a reel scan, and that cost scales with the number of lines changed rather than with the number of boards. A plant running 30 changeovers a week on a 120-line BOM spends more time verifying than a plant running 5 changeovers a week on the same BOM, and the second plant can afford a smaller bank because it stages the difference. The bank size decision is therefore inseparable from the changeover frequency, and the changeover frequency is a scheduling decision as much as a machine decision.
Take a hypothetical product with 6,800 placements per board, 96 unique lines, a largest part of 22 mm square and a tallest part of 12 mm, and a requirement of 900 boards per month on a single shift. The package ledger eliminates the RX-8 on both the 5 mm square ceiling and the 3 mm height, and eliminates the RX-7R on the 10.5 mm height, so the comparison is between the RS-1R at 47,000 CPH and the RS-1XL at 42,000 CPH. Apply a 65 % efficiency factor for indexing, vision and feeder advance, then 85 % for availability and changeover: the RS-1R yields roughly 25,900 effective CPH and the RS-1XL roughly 23,200. At 6,800 placements per board that is about 3.8 boards per minute against 3.4, or 228 against 205 boards per hour before changeover.
At 900 boards per month the difference is under four hours of production, which is not a reason to choose either machine. The RS-1XL's wider 650 × 560 mm envelope is a reason only if the panel needs it, and the 112-position bank is the same on both, so the decision falls to board geometry. Run the same numbers at 8,000 boards per month and the four hours becomes thirty-five, which is still small against a changeover-heavy schedule but large enough to matter if the machine is the line constraint. That is the honest shape of the calculation: rate differences inside one head family rarely decide a purchase, and the ledgers usually do.
Planning item (hypothetical) | RS-1R | RS-1XL | Basis |
|---|---|---|---|
Published optimum rate | 47,000 CPH | 42,000 CPH | Product page, optimum conditions |
After 65 % process efficiency | 30,550 CPH | 27,300 CPH | Assumption for indexing, vision, feeder advance |
After 85 % availability and changeover | 25,968 CPH | 23,205 CPH | Assumption for stops, setup and ramp |
Boards per hour at 6,800 placements | 229 | 205 | Derived from the two assumptions above |
Feeder positions available | 112 | 112 | Product page maximum |
Envelope that decides the choice | 650 × 370 mm | 650 × 560 mm | Product page, rails included |
If the requirement instead includes a low-cost single-platform entry, the entry-level six-head chip placer build is listed on the same specification basis and can be compared with the same two assumptions, which is the only fair way to compare machines at different price points. Note what the table does not contain: a claim about your yield, your uptime or your labour. Those depend on your process, and any supplier who offers them as numbers without seeing your line is guessing. The table contains rate arithmetic and published limits, and it is deliberately small so that you can redo it with your own efficiency factor.
Everything up to this point is analysis; the purchase needs a document. The worksheet is that document, and its purpose is to make every assumption visible to both sides so that a quotation can be compared with another quotation on the same basis. Most quotation comparisons fail because two suppliers answer different questions: one quotes a base machine with a nominal bank, the other quotes a configured machine with tray feeders, an extra head and a nozzle set, and the totals are not comparable. A worksheet with the three ledgers attached removes that ambiguity, because both suppliers are answering the same seven rows. It also gives you a record to hold a supplier to at acceptance, which is worth more than the elegance of the form.
The sign-off test is the second half of the document, and it is the part that converts a specification into evidence. A machine that meets its specification on a demonstration board can still fail on your board, because your board has the awkward parts, the tight spacing and the paste height that the demonstration board was designed to avoid. The test should therefore use your boards, your BOM, your nozzles and your feeders, at production speed, for long enough to expose the intermittent faults. What follows is the worksheet, then the test, then the three cases where the honest conclusion is that one machine cannot cover the BOM and the configuration has to change shape.
Row one is the largest panel measured with rails, clamps and any carrier included, because that number rather than the bare board decides the envelope. Row two is the tallest component on each board, measured from the board surface to the highest point. Row three is the largest component by body area, including any overhanging latch or tab. Row four is the unique line count per product family, which drives the bank. Row five is the placement count per board, which drives the head. Row six is the feeder mix by width, which turns the line count into positions. Row seven is the nozzle family count, which turns the package mix into a setup cost and a spare-parts list.
Those seven rows are deliberately all facts you already have or can measure in an afternoon. None of them requires a supplier, and all of them are answerable from the BOM, the CAD data and a pair of calipers. That is the point: the JUKI chip mounter BOM feeder nozzle selection work should be finished before the first supplier conversation, so that the conversation is about confirming limits rather than about discovering requirements. A worksheet with all seven rows filled also protects against the most common procurement failure, which is a machine specified around a demonstration board and delivered against a production BOM that was never shown to anyone.
The test has three stages and each one answers a different question. Stage one is a static check: load the actual feeders, fit the actual nozzles, and run the program at low speed for one board to confirm that every line picks, recognises and places without a manual intervention. That stage finds the geometry problems, the nozzle mismatches and the feeder positions that do not index. Stage two is a production-speed run of at least two hundred boards, inspected for placement defects by type, which is where the intermittent pickup and recognition faults appear. Stage three is a changeover: switch the machine from one product family to another, timed, with the operators who will actually do it.
The acceptance criteria should be written before the test and expressed in defects per million placements by class, not as a single yield number, because a single number hides which class is failing and cannot be acted on. Agree the criteria, the sample size, the inspection method and who signs. If the supplier cannot meet a criterion on your board, the discussion is still useful: it identifies whether the gap is a tool, a feeder, a program or the machine, and any of those can be fixed before shipment rather than after. A machine that passes all three stages on your product is a machine you can put into production without a second commissioning project, which is worth more than the days the test costs.
Some BOMs are simply too wide for one platform, and the worksheet will show it before the purchase rather than after. The usual signs are a line count that exceeds the bank by a wide margin, a package mix that spans a 0201 chip and a 30 mm connector with several tray lines, or a product family whose placement count is high enough to justify a dedicated high-rate machine. In those cases the right answer is a split: one flexible platform for the mixed work and one rate-focused platform for the volume work, or a dedicated build configured for the dominant product. A build such as the four-head configuration prepared for LED driver boards is an example of the second approach: the machine is shaped around one product family rather than asked to cover all of them.
The split has costs that belong in the comparison: floor space, a second program structure, a second spare-parts list, additional training and a second set of acceptance tests. It also has benefits that belong in the same comparison: each family runs on a machine matched to it, the nozzle families per machine fall, and the line keeps producing when one platform is down for maintenance. Cost the split against a single oversized platform including those items, and be honest about the changeover time that a single platform will spend serving both families. In most plants with a genuinely wide BOM, the split wins on throughput and loses on capital, and the decision turns on how much of the volume is in the dominant family.
Send the ledgers, not the brochure request
If you send us the seven worksheet rows, your feeder mix by width and your nozzle family list, we will answer with a configuration rather than a catalogue page: which platform clears your package ceilings, how many positions your tape and tray lines consume, which nozzle families your BOM requires and how many spares are worth holding. We will mark which figures come from the product pages on this site, which come from JUKI's published specification and which need confirmation on the current datasheet, and we will show the efficiency assumption next to the capacity number so you can defend it internally. Where a line is awkward, we will say so and name the workaround rather than leaving it for commissioning week. And if the honest answer is that your BOM needs two platforms or a dedicated build instead of one flexible machine, the review will say that, because a configuration that fails in production costs both of us more than a lost order.
The package mix, not the placement count. Sort the BOM by footprint and record the largest body size, the tallest part and the smallest part, because those three numbers eliminate machines before any rate comparison is useful. Then count unique lines per product family, since that number drives the feeder bank, and finally list the delivery format of every line. A BOM that opens with 8,000 placements and hides a 30 mm connector will point you at the wrong machine, and the package mix is what prevents that. The placement count comes last, because it only matters once a machine can physically build the board.
It depends on tape width, and the arithmetic is simple. A single-lane feeder consumes its width divided by 8 mm in slots, so a 24 mm feeder takes three slots and a 44 mm feeder takes five and a half. A BOM of 120 lines that is mostly 8 mm can fit a 112-position bank if the narrow lines run as double-lane feeders; a BOM of 120 lines with twenty 24 mm and ten 44 mm tapes will not fit even with double-lane loading. Do the calculation line by line rather than estimating, and note the tray lines separately, because they compete for a different kind of space.
No, and aiming for one nozzle per part is the wrong target. Aim for the fewest families that pass the pickup test, which for a typical mixed BOM is three to five: a small chip family, a mid-size family, a large-body family and the two special shapes, MELF and connector. Two parts in the same family can share a nozzle if their pickup areas and masses are close. The number to watch is the family count per changeover, not the part count, because every additional family adds a tool change, a lookup step and a chance of a mapping error on the first board of a run.
Only if the connectors fit its published ceilings. The RX-7R publishes a 25 mm square component limit and a 3–10.5 mm height range, and the RX-8 publishes 5 mm square and 3 mm. A board-edge USB or a 30 mm header exceeds those limits, so the connectors have to be placed somewhere else, which means a second process, a manual station or a flexible machine in the line. That is not a defect in the rotary design; it is the trade that buys its rate. The question to answer is what the excluded parts cost per board, and whether the saved placement time exceeds that cost.
Look at the defect signature. A part that is missing after pickup, or that arrives rotated by roughly the same angle each time, points at the nozzle or the vacuum. A part that is offset consistently in one direction points at vision or at the feeder pick position. A part that tombstones after reflow points at the paste deposit or the placement force. The quick test is to swap the suspect nozzle for a known-good one on the same line and run a hundred boards; if the defect follows the nozzle, the nozzle is the cause. If it does not, the tool was never the problem.
It depends on batch size and on the part. Re-reeling is cheap per reel but adds a handling step, a potential moisture and damage exposure, and a supplier relationship for every part you convert. A tray feeder costs more up front and occupies space, but it preserves the original packaging and works for parts that cannot be taped at all, such as BGAs in matrix trays or connectors with fragile housings. Compare the loaded cost per thousand placements rather than the price of the feeder, and remember that a tray line which pauses the machine every forty boards carries a cost that does not appear on either quotation.
Your boards, your feeders, your nozzles and your program, run at production speed. Three stages: a low-speed single board to prove every line picks and places, a production-speed run of at least two hundred boards inspected by defect class, and a timed changeover between two product families performed by your operators. Write the criteria before the test, express them as defects per million placements by class rather than as one yield number, and agree who signs. A machine that passes those three stages on your product is ready for production; a machine that passes only on a demonstration board is not yet tested.
