Publish Time: 2026-11-11 Origin: Site
Configuration planning guide · Published by I.C.T · For SMT process engineers, production owners and equipment buyers
Most conversations about the mounter platforms that accept these feeder and nozzle sets start with the placement rate and end with a purchase order, and the part in between — the configuration — is left to a spreadsheet filled in a week before installation. That order is backwards. The head decides how fast a part can be placed; the JUKI feeder nozzle option configuration decides whether the part is there to be placed at all, whether it is the right part, whether the machine can keep running while the next product is prepared, and how many minutes each changeover costs across a year. Two factories can own identical machines and post line efficiencies twenty points apart purely on this layer.
This guide treats configuration as a production decision with four ledgers: the slot budget (how many feeder positions the product mix actually needs), the bore budget (which nozzle family covers the component range without a changeover in the middle of a run), the option ledger (what each add-on buys and does not buy), and the evidence ledger (what a supplier must demonstrate before you sign). Every calculation below is written as an explicit assumption so it can be re-run with your own BOM and your own verified machine data; published specification tables are used as a starting point and flagged where they should be checked per model rather than trusted at family level.
A placement machine is sold as a head on a frame with a rated speed, but it only produces through three mechanical interfaces that are chosen separately: where the components come from, what picks them up, and which optional systems watch the process while it runs. The head cannot place a part that no feeder presents, cannot hold a part that no nozzle fits, and cannot recover time lost to a changeover that the option set does not shorten. That is why the same model can run at one site as a flexible workhorse and at another as a bottleneck, and why configuration deserves its own planning pass rather than a line item on an installation checklist. The rest of this section explains what those three interfaces control, how to price a bad choice in minutes and defective boards, and which four ledgers this guide uses to keep the decision honest.
The first interface is the feeder port: a slot with a bus connection, a mechanical stop and, on newer platforms, an identification reader. How many slots exist, how wide each is, and whether the slot can carry a double-lane feeder define the maximum component count a single setup can hold. The second is the nozzle holder on the head, where bore diameter, tip geometry and spring travel determine which component sizes can be picked without a mid-run change. The third is the option layer: tray changers, stick or vibratory supply, component verification, feeder identification and automatic tool change. None of the three appears in a headline speed figure, and all three can stop a line. A useful habit is to sketch the machine as a box with these three openings labelled, then write the product requirements against each opening instead of against the brochure. In a JUKI feeder nozzle option configuration these three openings are chosen together or not chosen at all, and teams that do this usually discover that their constraint is the slot budget or a nozzle family gap long before they discover anything about placement rate.
The cost of a configuration error rarely appears as a single dramatic failure. It appears as a stream of small stops: a feeder loading sequence that forces an operator to stop and re-splice, a nozzle that needs swapping between two component families in the same run, a tray part that has to be hand-placed because no tray option was ordered. As an illustration only — not a measured case — take a line that loses six minutes per shift to a feeder re-splice that a pre-set cart would have avoided, and eight minutes to a nozzle swap that a second holder would have avoided. Fourteen minutes on a 480-minute shift is roughly 3 percent of available time, which on a line running 20,000 placements per hour of effective output is around 12,000 placements per shift not produced. Add the boards that need rework because the wrong part was loaded from an unverified feeder, and the annual figure usually exceeds the price of the options that would have removed the cause. The arithmetic is deliberately simple; the point is that configuration costs are recurring and option prices are one-off.
Every decision in this guide lands in one of four ledgers. The slot budget counts feeder positions in a common unit so a product mix can be compared against a machine class. The bore budget counts nozzle families and the number of holders needed to cover the component range without an in-run change. The option ledger prices each add-on against the specific failure it removes, so an option that solves nothing measurable is not ordered by habit. The evidence ledger lists what the supplier must show on your boards before the configuration is accepted. Working in this order matters because the ledgers constrain each other: adding tray supply consumes slots and head time, a wider nozzle family can cost a holder position, and an identification option is worth more when the slot budget is tight and feeders are staged ahead of the run. When someone asks how to evaluate a JUKI feeder nozzle option configuration, the honest answer is to fill the four ledgers with your own numbers and see which one binds first.
Feeders are the part of a mounter that operators touch most and planners model least. They arrive in a width ladder, they consume machine slots in proportion to the tape they carry, and they determine how much of the component list can be staged at one time. The width ladder matters because a 24 mm tape does not use 24 mm of a slot bank — it uses three 8 mm positions, and a 44 mm tape uses six. Once a product list is converted into that common unit, the comparison against a machine's feeder capacity becomes arithmetic instead of guesswork, and the difference between machine classes stops looking like a rounding error. This section covers the ladder, the conversion, the class limits that set the ceiling, and the operational details — splices, low-reel warnings and staged carts — that decide whether the plan survives a real shift.
The published feeder range for JUKI machines on the parts side runs from a narrow 4 mm class through 8 mm, 8 mm long-body, 12, 16, 24, 32, 44, 56, 72 and 88 mm, and the tape feeder widths listed for JUKI machines are the ones to match against a BOM rather than a generic assumption. Each width exists because tape pocket pitch and carrier width scale together: an 8 mm tape carries the small chip families at 2 or 4 mm pitch, while 16 mm and 24 mm tapes carry larger discretes and small ICs with wider pockets. Width is not a free choice. A feeder narrower than the tape will not index reliably, and a feeder much wider than necessary wastes slots that another part could use. In practice most banks are dominated by 8 mm and 12 mm positions with a long tail of wider tapes, and it is that tail which silently consumes the budget.
Tape width | Typical content | Slot cost in 8 mm units | Planning note |
|---|---|---|---|
4 mm class | Very small chips on narrow carrier tape | 1 | Often the first width to run out at high mix; check availability before promising it |
8 mm | 0201 to 0603 chips, small diodes, LEDs | 1 | Backbone of the bank; order spares because these wear fastest |
8 mm long body | Same tape, longer reel or deeper pocket | 1 to 2 | Confirm whether it needs two positions on your slot bank |
12 mm | Larger chips, SOT packages, small inductors | 2 | Second most common width; a 12 mm tape never fits an 8 mm slot |
16 mm | SOP, SOIC, medium inductors | 2 | Counts double even though it looks modest on the BOM |
24 mm | Larger ICs, connectors, big capacitors | 3 | Where the slot budget starts to disappear |
32 mm | Wide body ICs, shielded parts | 4 | Plan these on the outer end of the bank for access |
44 mm | Large connectors, modules | 6 | Two of these equal twelve 8 mm parts off the list |
56 / 72 / 88 mm | Very large or irregular taped parts | 7 / 9 / 11 | Often better served by tray or stick supply instead |
The conversion is one line of arithmetic: for each feeder wider than 8 mm, divide the width by 8 and round up, then sum. A bank holding 20 feeders at 8 mm, 10 at 12 mm, 6 at 16 mm, 4 at 24 mm, 2 at 32 mm and 1 at 44 mm consumes 20 + 20 + 12 + 12 + 8 + 6 = 78 positions in 8 mm units. That number, not the count of feeders, is what has to fit the machine. The distinction matters when two products are compared: a board with 60 part numbers on 8 mm tape and a board with 34 part numbers spread across wide tapes can need the same bank, and only the converted figure shows it. Run the conversion per product, then take the maximum across the family as the base load and add a staging reserve, typically ten to fifteen percent, for active reels that must sit on the cart during a changeover. The result is a specification requirement you can hand to a supplier, and it is checkable in a meeting rather than discovered during the first production run.
Width | Feeder count (assumed) | Positions each | Total positions |
|---|---|---|---|
8 mm | 20 | 1 | 20 |
12 mm | 10 | 2 | 20 |
16 mm | 6 | 2 | 12 |
24 mm | 4 | 3 | 12 |
32 mm | 2 | 4 | 8 |
44 mm | 1 | 6 | 6 |
Total | 43 feeders | — | 78 positions |
Feeder capacity is published per model, and the spread across a single family is larger than most buyers expect. On the product pages for the platforms referenced in this guide, the maximum feeder inputs are listed as 112 for the flexible single-gantry models, around 76 for the parallel rotary class and around 56 for the fastest rotary head — figures taken from a family table, so confirm them against the datasheet for the exact model and lane configuration you are ordering. The practical consequence is that the fastest head can be the most constrained bank. A product needing the 78 positions calculated above fits the flexible class with room to stage, fits the parallel rotary class with almost nothing spare, and does not fit the fastest class at all without splitting the component list across two machines or moving wide parts to tray supply. That single comparison often reorders a shortlist, and it is invisible in a speed table.
Class in the family table | Max feeder inputs (published) | Head type | What the bank limit implies |
|---|---|---|---|
Flexible single gantry | 112 | Inline head | Wide component range and staging space; slower per placement |
Wide-board variant | 112 | Inline head | Same bank, larger board envelope, heavier frame |
Parallel rotary | 76 | Rotary heads | High output with a tight bank; product split may be required |
Fast rotary | 56 | Rotary head, small-chip focus | Small-chip specialist; wide tapes and odd parts belong elsewhere |
A slot budget assumes every position stays loaded for the whole run, and real production breaks that assumption on a schedule. Reels end. When a reel ends mid-lot, the operator either splices a new reel onto the old tape, which introduces a joint that the feeder must index through without a pickup error, or stops the machine to reload, which costs more time than the splice but less risk. The configuration decisions here are unglamorous and effective: choose feeders with reliable low-reel sensing so the warning arrives with enough tape left to plan a splice at a natural break, standardise splice tape and train one method, and keep the most frequently consumed 8 mm parts on the largest reel the feeder accepts. On a high-mix line the reload tax is the dominant hidden loss, and it is best measured rather than argued about — log every stop with cause for two weeks and the ten worst offenders will usually be three or four part numbers on small reels.
If feeders decide what is available, nozzles decide what can actually be lifted. A nozzle is a vacuum cup with a precisely drilled bore, a tip geometry matched to the component face, and a spring or rigid body matched to the head, and the family runs from a fraction of a millimetre to several millimetres. The mistake most often made is to treat nozzles as consumables ordered after the machine, because the set that arrives with a new machine is usually chosen by whoever assembled the quotation, not by the person who owns the component list. The consequence shows up as a mid-run changeover between two component families, a pickup error rate that nobody can explain, or a small chip that will not centre. This section reads the bore ladder as a specification, explains the pickup window that governs selection, counts the holders a product mix really needs, and sets out the reference test that separates a worn nozzle from a wrong one.
The replacement nozzle range published for JUKI machines shows how granular the ladder is, and it is worth reading as a table rather than a list. Part numbers 40001339 to 40001346 map to nozzle codes 501 through 508 with progressively larger bores, from a 0.7 x 0.4 mm tip with a 0.25 mm hole up to a 9.5 mm outer diameter with an 8.0 mm bore, and two further part numbers cover a two-hole 0.3 mm tip and the 500 code. The codes are not decorative: they are how the machine program, the setup sheet and the spare-parts order refer to the same object, and mismatches between the three are a common source of wrong-nozzle errors. Keep the mapping in one controlled document — code, part number, bore, tip geometry, which component families it is approved for, and which holder it fits — and treat any deviation as an engineering change. The full list is on the replacement nozzle part numbers and bore sizes page; print it and annotate it with your own approvals rather than relying on memory.
Part number | Nozzle code | Bore / tip | Where it sits in a typical set |
|---|---|---|---|
E3608-729-OA0 | 500 | 1.0 x 0.5 mm tip, two 0.3 mm holes | Small chips that need two-point support |
40001339 | 501 | 0.7 x 0.4 mm tip, 0.25 mm hole | Smallest chip class; the nozzle that sets the floor |
40001340 | 502 | 0.7 mm outer / 0.35 mm bore | Small chips and small diodes |
40001341 | 503 | 1.0 mm outer / 0.6 mm bore | General purpose chip and small discrete |
40001342 | 504 | 1.5 mm outer / 1.0 mm bore | Larger discretes, SOT and small inductors |
40001343 | 505 | 3.5 mm outer / 1.7 mm bore | SOP and medium ICs |
40001344 | 506 | 5.0 mm outer / 3.2 mm bore | Larger ICs and connectors |
40001345 | 507 | 8.5 mm outer / 5.0 mm bore | Large body parts; check head clearance |
40001346 | 508 | 9.5 mm outer / 8.0 mm bore | Largest class; confirm the head accepts it |
Bore choice is usually described as a rule of thumb, and the rule is worth writing down because it explains why the ladder exists. A bore that is too large for the part lets air leak around the body, so the vacuum at the tip is weaker than the gauge suggests and the part shifts during acceleration. A bore that is too small leaves the part sitting off-centre, so the vision correction has to work harder and the placement offset grows. As a working rule — validate it on your own components rather than treating it as a standard — aim for a bore between roughly one third and two thirds of the shorter side of the component body, then check the two failure modes at the edges of that window. For a 1.0 x 0.5 mm part the window is about 0.17 to 0.33 mm, which points at the 501 or 502 class; for a 3.2 x 1.6 mm part the window is about 0.53 to 1.07 mm, which points at the 503 or 504. Where a family straddles two codes, choose by the failure you can least afford, and confirm the mapping with the supplier's component-to-nozzle matrix.
The number of nozzles a machine can carry at once is a physical limit set by the head, and it is the reason a shared nozzle set matters. If a product needs four nozzle families and the head carries four, the run proceeds without a change. If it needs five, one swap happens mid-run, and that swap costs the machine stop plus the verification that follows it. The planning step is therefore to list the families each product needs, count the distinct families across the family of products run in the same week, and compare that count to the holder positions available. Two outcomes are common: either the count fits and the only work is to standardise the approved mapping so nobody improvises, or the count exceeds the holders and the fix is a configuration decision — move a component family to a second machine, accept a scheduled swap at a batch boundary, or change a component package on the design side so it falls into a family already carried. Deciding this during quotation is cheap; discovering it in the third week of production is not.
Nozzles wear, and the wear is slow enough to be invisible until a pickup rate drifts. Solder paste dust, flux vapour and component debris accumulate in the bore; the tip deforms after enough contact with hard ceramic bodies; the spring in a spring-loaded body weakens. None of this announces itself. The practical control is a reference test run on a schedule: a fixed board or a fixed set of component positions, picked and placed under fixed conditions, with the pickup error rate and the placement offset recorded. When the reference numbers drift beyond a set band, the nozzle is inspected, cleaned and if necessary replaced — before the drift reaches a production lot. Pair the test with a cleaning interval driven by placement count rather than calendar time, and keep a small stock of the two or three codes that wear fastest. A line that replaces nozzles only when a defect appears is using the customer's boards as the inspection instrument.
Options are where configuration planning most often collapses into a wish list, because each item is presented with a benefit and none is presented with a boundary. The useful discipline is to ask three questions of every option: which specific failure does it remove, what does it consume in slots, head time or floor space, and how would we prove on our own boards that it worked. Options that survive those three questions earn a place; options that do not are usually bought because a competitor has them. This section works through the option families that matter most in feeder and nozzle planning — tray and stick supply, feeder identification, component verification, and automatic nozzle change — with the boundary of each made explicit. Naming differs between platforms and generations, so treat the descriptions as functions to look for and confirm the exact scope with the supplier.
A tape feeder requires the component to be in a pocket on a carrier tape, and a growing share of modern parts is not: fine-pitch BGAs and QFNs arrive in trays, some connectors and odd-form parts arrive in sticks or tubes, and bare die and specialised sensors arrive in waffle packs. Without a tray option, those parts leave the machine and become a manual placement operation, which drags a hand station into the middle of an automated line and adds a defect source the process cannot control. A tray changer addresses this by presenting stacked trays to a pickup position, and a stick feeder does the same for tubes. The planning cost is real: tray supply consumes head travel and cycle time, needs floor space beside the machine, and usually needs a specific nozzle family that the tray parts already require. The decision test is simple — count the tray and stick part numbers per product and the number of boards per shift; above a threshold that you set, the option pays, below it, manual placement with a documented control may be cheaper.
A feeder identification system — typically a reader at the slot plus a tag on each feeder — lets the machine know which physical feeder has been loaded where, instead of relying on the operator to key it in. The value is not the data entry saved; it is the class of error removed. With identification, a feeder loaded in the wrong slot, a feeder still carrying the previous product's reel, and a feeder whose tag does not match the program can all be flagged before the first board is placed. On a high-mix line where carts are staged ahead of the changeover, that check is what makes offline setup safe: the cart can be prepared away from the machine and verified at the moment it docks. The boundary is that identification proves identity, not quantity or orientation — a correctly identified reel can still be nearly empty or loaded upside down. Budget for the readers on the slots you actually stage, and treat the tag population as a controlled asset with its own replacement routine.
Component verification options check that the part presented at the pickup point is the part the program expects, usually by measuring the body with a sensor or by reading a mark. This addresses a failure that identification cannot: the right feeder, correctly loaded, carrying the wrong reel. The cost of that failure is asymmetric — one wrong part can fail an entire lot at functional test, and the defect is often not visible at AOI. When evaluating the option, ask what it measures, at what point in the cycle, what it does on a mismatch, and how the mismatch is cleared and recorded. A verification system that stops the machine but leaves no record is a nuisance; one that logs the event against the feeder, the slot and the program is a control. On products where a single wrong passive can scrap a board, the option is usually justified by one avoided lot. On products where every part is visually inspected downstream, it may duplicate existing control, and that duplication should be declared rather than assumed.
An automatic nozzle changer holds a magazine of nozzle codes and swaps them under program control, which converts a manual changeover into a short machine stop. Its value scales with how often the product mix crosses nozzle families: a line that runs one product for weeks gains almost nothing, while a line that changes product twice a shift gains a great deal. The boundary is that the changer only removes the handling time, not the reason for the swap. If the mix needs five families and the head carries four, the changer makes the fifth swap cheap but the swap still happens, and the program still has to be written to schedule it at a sensible point. It also adds a magazine to keep stocked and calibrated, and a new failure mode if a nozzle is missing or misfitted. Before ordering, count the swaps per week that the changer would absorb and multiply by the observed minutes per manual swap; if that figure is small, the money is better spent on a second holder position or a wider nozzle set.
Configuration planning becomes manageable once it is done against a mix rather than a single product. A single product can usually be configured on instinct; a family of products run on the same line cannot, because each one pulls the slot budget, the nozzle set and the option list in a different direction and the configuration has to satisfy all of them with one set of hardware. The method here is deliberately plain: build the slot budget from the widest product, build the nozzle budget from the union of component families, price the options against the failures they remove, then check the whole assembly against a three-product example that is small enough to do on paper. The numbers in this section are assumptions chosen to make the arithmetic visible; they are not recommendations and they are not taken from any customer installation.
Start from the bill of materials of every product the line will run, and for each line item record the package, the tape width and the reference designator count. Collapse the list to part numbers: ten positions of the same 0603 capacitor share one feeder, not ten. Then convert each part number's tape width into 8 mm positions using the ladder from the previous section, and sum. Do this per product, write the totals in a column, and take the maximum — that figure is the base load. Add a staging reserve for reels that must be mounted during a changeover, and add the tray and stick positions if those options are being considered, because they consume access space even when they do not consume slots. The output of this exercise is a single number with a traceable derivation, which is far more useful in a supplier conversation than a statement that the line runs many part numbers. Keep the worksheet with the product name and revision on it, because the budget will be re-checked every time the mix changes, and a number without its derivation cannot be re-checked.
The nozzle budget is built the same way but groups by component family rather than by part number. Cluster the BOM into families that share a bore window — the smallest chip class, small discretes, larger discretes and SOT parts, SOP and medium ICs, large ICs and connectors — and assign each cluster the narrowest bore that holds the part reliably, using the one-third to two-thirds window as a starting point and validating on real components. The result is the number of distinct nozzle codes the mix needs, which is usually smaller than the number of component sizes and larger than the number of codes a machine arrives with. Then compare that count to the holder positions on the head and to the magazine size if an automatic changer is being considered. Two adjustments commonly follow: consolidating two component packages onto one nozzle family by changing a package on the design side, or splitting the mix across two machines so that the small-chip specialist never has to carry the large-body codes. Both are configuration decisions with a direct production consequence, and both are cheaper than living with a mid-run swap that nobody scheduled.
The option ledger is a table with four columns, and filling it honestly removes most of the emotion from the decision. The first column names the option; the second states the specific failure it removes, in the language of a production event rather than a feature; the third states what it consumes — slots, head time, floor space, operator attention, maintenance; the fourth states how the claim will be proven on your boards. An option with an empty second column is a habit. An option whose fourth column cannot be filled is a promise. This does not mean options should be refused; it means the ones that survive get ordered with a test attached, which is exactly what an acceptance conversation needs. On most feeder and nozzle projects, three or four options pass this filter and the rest are deferred to a second phase once the line has real production data to argue with.
Option family | Failure it removes | What it consumes | How to prove it |
|---|---|---|---|
Tray changer | Tray parts hand-placed off line, uncontrolled | Head travel, cycle time, floor space, a nozzle family | Run a tray-heavy product and compare cycle time and placement defects with manual placement |
Stick / tube feeder | Tube parts becoming a manual operation | Slots and head travel; orientation risk | Run the tube part through a full lot and record jam frequency |
Feeder identification | Feeder in wrong slot; stale reel from previous product | Readers on staged slots, tag population, tag maintenance | Deliberately misload a feeder and confirm the machine refuses to start |
Component verification | Right feeder carrying the wrong reel | Cycle time per verified part, false-reject rate | Present a wrong part and confirm the stop is logged against slot and program |
Automatic nozzle change | Manual swaps inside a run | Magazine, calibration, a new failure mode if a code is missing | Time ten swaps manually, then ten with the changer, on the real mix |
Extended feeder bank or second cart | Staging forces the line to stop between products | Capital, floor space, tag and cart management | Change over two products back to back and measure the gap |
Take three assumed products on one line. Product A is a dense controller board with 96 part numbers, mostly 8 mm tape, converted to 104 positions, and four nozzle families. Product B is a power board with 41 part numbers but a heavy tail of 16, 24 and 32 mm tape, converted to 71 positions and three families, plus two tray parts. Product C is an LED driver board with 28 part numbers, all small chip, converted to 30 positions and two families — the kind of board that the four-head LED-driver machine built for software placement is aimed at, and a good argument for running it on a second, cheaper platform rather than loading it onto the fast machine. The maximum slot load is 104, the union of nozzle families is four plus the tray family, and the option list that survives the ledger is a tray changer, feeder identification on the staged slots and a second cart. That configuration fits the flexible class comfortably and does not fit the 56-input rotary class at all — which is the whole point of doing the arithmetic before the shortlist.
Configuration earns its money twice: once by making the correct setup possible and once by making it fast. The second half is where most projects leave value on the table, because the hardware is bought and the working method is not. A line with pre-set carts, identified feeders, a controlled nozzle mapping and a program that references all three can change product in a fraction of the time a line with the same hardware and no method needs, and the gap widens as the mix gets more varied. The changeover budget that governs this is the same one that decides whether high-mix production is viable at all — the high-mix changeover budget that feeder planning feeds into is the right frame for it. This section covers the three practices that convert configuration into minutes: offline staging, a single revision across feeder, nozzle and program, and standardisation across a fleet.
Offline staging means the next product's feeders are mounted, checked and identified on a cart while the current product is still running, so that the changeover at the machine is a dock-and-confirm operation rather than a build. The prerequisites are physical and administrative: enough carts for the products in rotation, a staging area with power and light, a documented cart layout per product, and an identification system that lets the machine confirm the cart matches the program. Without identification, offline staging moves work but adds a new error class, because the cart can be built correctly and docked into the wrong product's program. The practical rule is that no cart leaves the staging area without a completed checklist signed against the product revision, and that the check at docking is a machine check rather than a second human check. Lines that adopt this routinely cut changeover by more than half; lines that adopt the carts without the identification usually see the gain eaten by wrong-part events.
A configuration only works if the three descriptions of it agree: the physical cart, the nozzle set on the head, and the program that tells the machine what to pick and with what. In practice these drift, because they are maintained by different people at different times, and the drift is silent until a run fails. The control that fixes it is a single configuration record per product revision, holding the part number list with slot assignments, the nozzle code per part, the feeder tag identifiers, the tray positions and the verification settings, and a change process that requires all three to move together. When a component is substituted, the record changes, the cart layout changes and the program changes as one action. This sounds bureaucratic and is in fact the cheapest error-proofing available, because it costs documentation time rather than capital, and it converts a class of failures that are hard to diagnose into a class that is impossible to reach.
Once more than one machine is involved, standardisation decisions start to dominate. Nozzle codes are the usual first candidate: a fleet that uses six codes in total is far easier to control than one that uses fourteen, even if each machine could justify its own optimum. Feeder width sets follow, because a shared pool of 8, 12 and 16 mm feeders that can move between machines keeps utilisation high and simplifies spare holding. Cart layouts come third: a common layout convention across products lets an operator move between lines without relearning, and lets a cart be built by anyone. The counter-pressure is genuine — standardising on a nozzle code that is slightly wide for one component family trades a little pickup margin for a large reduction in control complexity. The way to decide is to price both: the pickup margin shows up as a defect rate, the complexity shows up as the number of distinct codes to stock, calibrate, audit and train, and in most factories the second number is the larger one.
Everything above assumes the configuration can be described precisely enough to be ordered, tested and accepted. That assumption fails more often than it should, because quotations describe machines and installations describe intentions, and the layer in between — feeder list, nozzle list, option list, tag assignments, cart layouts — is frequently left to the installation week. The corrective is to make the configuration a document set with named owners before the order, and to attach tests that can only pass if the configuration is what it claims to be. This section lists the documents, the tests to run with the real component mix, and the acceptance criteria that turn a demonstration into an obligation. It follows the same logic as the model shortlisting sequence that comes before configuration: decide what you are buying, then make the supplier prove it, in that order.
Five documents do most of the work. First, a feeder schedule listing every part number, its tape width, its slot assignment, its feeder type and its reel size, signed against a product revision. Second, a nozzle mapping listing every part number against an approved nozzle code, with the holder position and the fallback code where a family straddles two. Third, an option schedule naming each option with the failure it removes, the settings it requires and the operator procedure it introduces. Fourth, a cart layout drawing for each product in rotation, including where the staged reserve reels sit. Fifth, the program configuration file itself, exported and stored with the other four so that the physical and digital descriptions can be compared. The test for whether the set is complete is simple: could a competent technician, who has never seen this product, build the setup from these documents alone and be checked against them? If the answer depends on asking someone, the set is not finished.
Demonstrations are usually run on a convenient board. Acceptance should be run on the boundary product — the densest BOM, the widest tape mix, the smallest component, the tray part — because that is where configuration defects live. Four tests cover most of the risk. Load the full feeder schedule and confirm every position picks first time, with pick errors recorded rather than retried silently. Run the nozzle set through a full component cycle and confirm no mid-run swap is required that the plan did not anticipate. Load one feeder deliberately in the wrong slot and one reel with the wrong part, and confirm the identification and verification systems stop the machine and log the event. Finally, perform a timed changeover from the current product to the next, with the cart staged offline, and compare the measured minutes to the plan. Where the acceptance protocol needs a formal vocabulary for rates and accuracy, the same measurement conditions used for the high-speed capacity conversion for the RX class apply here: state the board, the mix and the definition, or the number cannot be compared with anything.
Acceptance criteria should be written as conditions on the configuration, not only on the machine. Each criterion needs a measurement, a threshold and a consequence: every scheduled feeder position picks within a stated error rate; no unplanned nozzle change occurs in a stated lot; a deliberately misloaded feeder is rejected; changeover completes within a stated time; every tray and stick part is placed by the machine, not by hand. Reference a measurement standard where one exists — IPC-9850 is the usual vocabulary for placement equipment performance, and SMEMA covers the line interface — so that accuracy, repeatability and transfer behaviour are defined identically on both sides. Then attach the punch list mechanism: open items with owners and dates, a re-test for each item, and a documented close-out that both parties sign. The structure is the same as the acceptance test sequence run on your own boards, with the feeder schedule, the nozzle mapping and the option settings added as extra gates. Platform documentation is worth reading in advance for this exercise, because it shows what the supplier already publishes; the RX-8 rotary-head platform on the manufacturer page is a useful example of what a published configuration table looks like, and of why the model-specific datasheet rather than the family page is the document to accept against.
Send the component list before the quotation, not after
A JUKI feeder nozzle option configuration is only as good as the component list it was built from, and the component list is the one input the buyer always has before the supplier does. Send the bill of materials with tape widths, the panel drawing, the tray and stick part numbers, the products in rotation and the changeover time the line needs to hit. With those five inputs the feeder schedule, the nozzle mapping and the option ledger can be built and checked before a price is discussed, which is the point at which changes are still free.
We will return the slot budget converted to 8 mm positions against the machine class you are considering, the nozzle codes the mix actually needs and how many holder positions that implies, and the short list of options that survive the three-question test. Where the arithmetic says a single machine cannot carry the bank, the answer will say so and show the split rather than move the problem to the installation week.
These are the questions that come up in almost every configuration review, usually in the last meeting before the order. They are answered here in the same order the decisions are usually made — feeder count first, then nozzle coverage, then the option list — because the answers depend on each other and taking them out of order produces confident answers to the wrong question. Where an answer depends on a number that only the buyer knows, the answer states the calculation rather than inventing the figure. Anything that could be read as a performance promise has been left out deliberately; the specification you should accept against is the model-specific datasheet, tested on your boards, not a sentence in a guide.
Count part numbers, not placements: identical components share one feeder. For each part number, take the tape width and convert it into 8 mm positions — 12 mm uses two, 16 mm uses two, 24 mm uses three, 32 mm uses four, 44 mm uses six, 56 mm uses seven, 72 mm uses nine and 88 mm uses eleven — then sum. Add ten to fifteen percent for reels that must be staged during a changeover.
The result is a slot budget, and it is the number to compare against the published feeder inputs for the machine class. A common surprise is that a board with fewer part numbers can need more positions than a denser board, because wide tapes cost several positions each. Do the conversion for every product in rotation and take the maximum, not the average.
Usually not with one code, and often yes with a small set. Group the bill of materials into component families that share a bore window, assign each family the narrowest bore that holds the part reliably, and count the distinct codes. Most production mixes land between four and six codes, which is why the number of holder positions on the head matters as much as the head's speed.
If the count exceeds the holders, the choices are to accept a scheduled swap at a batch boundary, split the mix across two machines, or change a package on the design side so it falls into a family already carried. What you should not do is let the machine's delivered nozzle set define the answer, because that set was not chosen against your component list.
It depends on boards per shift rather than part count. A tray changer earns its place when the tray part appears on enough boards that manual placement becomes a throughput and quality problem, and it is hard to justify when the part appears occasionally and a documented manual operation with a second check is acceptable.
The costs are not only capital: tray supply consumes head travel, so cycle time rises on every board carrying that part, and it needs floor space and a nozzle family the tray parts already imply. A fair test is to run one tray-heavy product with manual placement and record the time and the defects, then compare that against the option price over the same period.
Hand loading is the case where identification earns most, because hand loading is where slot mistakes and stale reels come from. Without identification, the only check is a second human check against a printed schedule, which works until the line is under time pressure. With identification, the machine refuses to start on a mismatch, which is a stronger control than any checklist.
The boundary is that identification proves which feeder is in which slot, not what is on the reel or how much is left. Pair it with a reel-level check if wrong-part events matter to you, and keep the tag population controlled — a tag that falls off or is copied defeats the system and does so silently.
Verification at the placement head catches the wrong part before it is placed; AOI catches it after reflow, when removing and replacing the component is far more expensive and sometimes impossible without scrapping the assembly. They are not substitutes: AOI also catches solder and alignment defects that head-level verification cannot see, and verification catches wrong-value and wrong-body errors that AOI may classify as acceptable if the package is similar.
The decision is usually economic rather than technical. On boards where one wrong passive fails functional test and scraps the assembly, verification is cheap insurance. On boards where every part is inspected and reworked downstream anyway, it may duplicate control — and that duplication is worth declaring explicitly rather than assuming.
A wide tape consumes several slot positions, so a bank with a long tail of 16, 24 and 32 mm parts has less room to stage the next product's reels. The changeover then has to happen in two waves — load what fits, run, then reload — which is far more expensive than a single dock-and-confirm step. The placement cycle time barely notices the tape width.
The practical fixes are a second cart, a tighter staging reserve, or moving the widest parts to tray or stick supply where that makes sense. Whichever is chosen, the effect should show in the slot budget as a number rather than in a complaint about changeovers being slow.
Most feeder, nozzle and identification options can be added later, but the cost and the disruption are higher, and some options require preparation that is easier at build time — wiring for readers, mounting points for a tray changer, magazine space for automatic nozzle change. The honest way to plan is to decide now which options are certain, which are likely, and which are speculative, and to make sure the machine is prepared for the likely ones.
Preparation is usually a small fraction of the option price and it preserves the decision. What should not be deferred is the JUKI feeder nozzle option configuration analysis itself: the slot budget and the nozzle mapping are cheap to do before the order and expensive to discover afterwards, and they are what tell you which options were worth preparing for.