How to Match JUKI CPH, Feeder Capacity, and PCB Size to Your SMT Line

Line-balance guide · Published by I.C.T · For production engineers, process owners and equipment buyers

A mounter purchase is usually argued with one number, and the number is the published CPH. That figure is real but narrow: it describes what the head can do when everything around it is already perfect, and the JUKI pick and place machine CPH feeder PCB line balance question is about everything else. What the head can do per hour, how many BOM lines the bank can hold, how large a panel the conveyor and clamps accept, and how fast the printer and the oven in front and behind can feed and absorb that output are four separate constraints that meet in one place: the boards per hour your line actually ships. Match them and the machine disappears into the line; mismatch one of them and it becomes the station everyone works around.

This guide treats the purchase as a line-balance problem rather than a specification comparison. It starts with what a published rate actually measures, then builds the three clocks that run on an SMT line, then shows how board size and clamping change the arithmetic, how feeder positions and changeover decide the effective rate, and how component mix discounts the headline figure before any board is run. The last chapters compute a hypothetical line balance, name the cases where the mounter is not the bottleneck, and lay out a configuration ladder by volume step. Every calculated figure is labeled as an assumption for illustration; the published numbers come from the specification table on the product pages in the placement platforms listed in this range.

What a published CPH actually measures

CPH means components per hour, and the number on a specification sheet is measured under conditions that flatter it. The head moves over the shortest possible path, the program uses parts already loaded, the vision system sees a component it recognises immediately, and nothing stops for a reel change, a paste shortage or a board that arrives late. The figure is still useful, because it ranks heads against each other on a like-for-like basis, but it is not the number you can put in a capacity plan. Treating it as one is the most common error in machine selection, and it is the error that JUKI pick and place machine CPH feeder PCB line balance work exists to correct: the published rate is an input to the calculation, never the answer.

The gap between the published rate and the delivered rate is not a trick by the manufacturer. It is the difference between a component being placed and a board being produced. Between the two sit the feeder advance, the vision check, the board transfer, the clamp, the program's own sequencing and the changeovers that happen between products. Each of those consumes time that the optimum measurement does not include, and each is affected by your BOM rather than by the head. The rest of this chapter separates those effects so that later chapters can put them back into a number you can defend.

Optimum rate, IPC-9850 rate and the gap between them

Two kinds of rate appear in supplier documentation. The optimum, or maximum, rate describes the equipment under the most favourable conditions the manufacturer can construct, and it is the figure most often quoted in a first conversation. A rate characterised under IPC-9850 describes the same equipment measured through a standardized procedure with a defined mix and defined counting rules, so different manufacturers' numbers can be compared without renegotiating the conditions. Neither is wrong; they answer different questions. The optimum figure tells you what the head is capable of at its limit, and the characterized figure tells you what to expect when the machine is asked to do normal work.

When a supplier quotes one and you plan with the other, the shortfall appears as a line that never reaches its target, and the conversation that follows is about whose assumption was wrong. Ask which figure a quotation uses, and ask for the conditions behind it: the component mix, the number of feeders in use, the board size and whether transfer time is included. A supplier who can answer those questions is quoting a number you can use. A supplier who cannot is quoting a headline, and the headline will not survive the first month of production. The specification rows on this site distinguish optimum figures from characterized ones for exactly this reason, and where a model has no characterized figure published, that absence is itself information.

Three things a rate figure hides

The first hidden cost is feeder advance. Every placement needs a component presented, and on tape that means an index. Narrow tape indexes quickly; wide tape and trays index more slowly, and tray presentation may require a table movement that costs an order of magnitude more than a tape index. A program with a high proportion of wide-tape and tray parts therefore places fewer components per hour than the same program made entirely of 8 mm tape, even though the head moves identically. The published rate does not distinguish between these cases, and the difference can be twenty percent or more on a BOM weighted towards large parts.

The second hidden cost is vision. A chip part recognised by silhouette takes a fraction of the time a BGA needs for ball recognition, and a part taught with a complex recognition window takes longer again. The third is transfer: the board must index in, clamp, be measured for fiducials, and index out, and on a long panel that transfer is a fixed cost per board that does not shrink when the placement count does. All three are visible in the machine's own cycle data, which is why a demonstration run on your board is worth more than any comparison of headline rates. Ask for the cycle breakdown, not just the total.

Reading the four columns that matter on a specification table

A specification table is long, and most of it describes capabilities that rarely decide a purchase. Four columns decide it: the placement rate, the maximum board size, the feeder input count and the largest component the head can handle. Those four map onto the four constraints of a line, which are rate, geometry, bank size and component ceiling, and a machine that fails any one of them cannot be rescued by strength in the others. Reading them together, before reading anything else, saves the time that would otherwise go into comparing heads on price or on a fifth column that does not constrain your product.

The table below reproduces the four decision columns from the specification table published on this site's product pages. Note how the families trade: the RS models carry the largest component ceiling and the biggest bank with the lowest rate, and the RX models invert that trade. Note also that the maximum board size differs between the RX models' single-lane figure and the RS models' single or double figure, which is the subject of a later chapter. Where a figure is described as approximate on the source page, it is shown here as approximate, because a planning number that hides its own uncertainty is worse than no number.

The table is a filter, not a shortlist. A plant whose largest part is a 30 mm connector removes the two RX models at the largest-component row and never needs to compare their rates, while a plant placing 03015 chips at volume removes the RS models on the smallest-component row if its process depends on that capability. Applying the four columns in the order rate, geometry, bank and ceiling, or geometry first if the product dictates it, is the discipline described in the machine-selection steps that come before any rate comparison, and it is what keeps a rate conversation from happening before it can be useful.

The three clocks on an SMT line

A line produces boards at the rate of its slowest station, and each station has its own clock. The printer has a cycle built from stencil separation, squeegee speed, cleaning and inspection. The mounter has a cycle built from placements divided by effective rate, plus transfer and fiducial time. The oven has a fixed transit determined by the profile, and the handling equipment has a magazine cycle that may be faster or slower than either. Those clocks rarely match, and the mismatch is what a line-balance calculation measures. The JUKI pick and place machine CPH feeder PCB line balance question is really a question about which of those three clocks governs, and whether the purchase changes the governing clock or merely moves the queue.

Most buyers compare only the second clock, because the mounter is the most expensive station and therefore feels like the one that should be optimized. In practice the printer is the more frequent constraint on fine-pitch work, and the oven is the more frequent constraint on long or heavy boards. A mounter that is twice as fast as the rest of the line produces inventory, not output, and the inventory accumulates in front of the oven where it becomes a quality risk. Building the three clocks explicitly, with a variance for each, is the only way to know whether the machine you are about to buy will be the one the line waits for.

The printer clock and its variance

A printer cycle typically contains print, separate, clean and inspect, and the variability comes from the cleaning interval and from operator intervention rather than from the print itself. A printer that cleans every five boards has a different average cycle than one that cleans every twenty, even when the print time is identical, and a printer that needs a paste addition every forty boards carries a stop that no mounter can compensate for. The number to record is not the fastest cycle but the median cycle plus the frequency and duration of the stops, because the stops are what the mounter downstream experiences as starvation.

Measure the printer over a full shift, not a demonstration. Log the cycle for every board, mark the stops and their cause, and compute the median and the ninety-fifth percentile. A printer with a median of 22 seconds and a ninety-fifth percentile of 40 seconds will starve a mounter that needs a board every 20 seconds, even though the medians look compatible. That difference is the reason a line can hit its target on a good day and miss it on every other day, and it is invisible in any comparison of machine specifications.

The mounter clock in placements per board

The mounter clock is the placement count divided by the effective rate, plus the fixed cost of transfer, clamp and fiducial recognition. On a board with 400 placements and an effective 26,000 CPH, the placement time is about 55 seconds; add 6 seconds of transfer and fiducial time and the board takes 61 seconds. The same machine on a board with 40 placements takes 5.5 seconds of placement plus the same 6 seconds, so the fixed cost is more than half the cycle. That ratio explains why small boards never reach a machine's rated output and why the transfer and fiducial design of a line matters as much as the head on short-cycle work.

Record both numbers separately when you test a machine: the placement time and the overhead. Suppliers sometimes quote a combined cycle for a demonstration board that is larger than your product, which flatters the overhead ratio. Asking for the split costs nothing and tells you how the machine will behave on your board rather than on theirs. If the split is not available, observe the machine's own cycle display during the demonstration and time ten consecutive boards, discarding the first two.

How to measure all three clocks in one shift

The three clocks can be measured without instrumentation in a single shift, and the record is worth more than any supplier demonstration. Put an observer at the printer for one hour with a stopwatch and a note of every stop and its cause. Put a second observer at the mounter for one hour recording the cycle display or timing board exits, and separately time ten consecutive boards to get the transfer and fiducial share. Ask the oven operator for the profile transit and the number of boards in the tunnel, and time the unloader for a magazine change. That is one shift of observation and it produces the four inputs every later calculation needs.

Record the tail, not only the average. A printer with a 58-second median and a 74-second ninety-fifth percentile behaves very differently from one with a 60-second median and a 62-second tail, even though both average about the same. The tail is what starves the mounter, and the tail is what a buffer has to cover. Note also when the stops occur: a stop every forty boards is a pattern that can be planned around, while a random stop with the same average is a source of variability that has to be buffered. The two look identical in a monthly report and behave completely differently on the floor.

The oven and handling clock: the accumulator that hides imbalance

The reflow oven runs at a fixed transit time determined by the profile, and it cannot be sped up without changing the metallurgy. Its capacity is the number of boards in the tunnel divided by the transit, which for a typical profile and a typical board is often higher than the mounter's capacity, which is why the oven is rarely the constraint on a simple line. Where it does bind is on long boards, on heavy assemblies that need a slower profile, and on lines running two products with different profiles through one tunnel, where a changeover includes waiting for the tunnel to stabilise.

The handling equipment is the accumulator that hides imbalance, and that is both its value and its danger. Loaders, buffers and unloaders let a fast mounter run ahead of a slow printer without stopping, and they let a slow oven be fed continuously. The danger is that the buffer absorbs the mismatch so effectively that nobody notices the line is running at the oven's rate rather than the mounter's, and the purchase that was justified by mounter throughput never delivers the expected output. Count buffer capacity in boards and in minutes, and compare those minutes with the variability you measured at the printer. If the buffer holds less time than the printer's tail variance, the line will still starve.

PCB size, lanes and clamping: the geometry that sets the clock

Board size affects output in two ways and buyers usually count only one. The obvious effect is capability: a panel larger than the published maximum simply cannot be run, and no configuration fixes it. The less obvious effect is rate, because transfer time scales with board length while placement count scales with board area. A board that is twice as long takes roughly twice as long to index and clamp, but it may carry only the same number of placements if the extra length is mechanical space, which means the fixed overhead per board doubles while the productive work stays flat. That is why two plants with identical machines and identical placement counts can report different boards per hour, and why board geometry belongs in the rate calculation rather than in a separate envelope check.

The lane architecture multiplies the same effect. A machine that clamps one board at a time cannot overlap transfer with placement; a machine with two clamping positions can load the next board while the current one is being populated, and a machine with buffers can stage several boards to absorb upstream variability. The benefit is real but conditional on the upstream and downstream stations being able to feed and absorb at the rate the machine consumes, which is the balance question from the previous chapter. This chapter quantifies the geometry effect, then the clamping effect, then the buffer trade on a specific family where the trade is explicit in the published data.

How board length changes boards per hour

Take a hypothetical 300 × 250 mm control board with 900 placements and a 500 × 250 mm power board with 1,100 placements, both run on a machine with an effective 26,000 CPH and a transfer-and-fiducial overhead that scales from 6 to 10 seconds with length. The control board takes about 125 seconds of placement plus 6 of overhead, or 131 seconds, which is 27.5 boards per hour. The power board takes about 152 seconds plus 10, or 162 seconds, which is 22.2 boards per hour. Output per board falls by nineteen percent even though placements per board rose by only twenty-two percent, because the overhead grew with the board and the placement count did not grow proportionally to area.

Now convert to placements per hour, which is the number the machine is actually good at: 27.5 × 900 is 24,750, and 22.2 × 1,100 is 24,420. The machine delivers almost the same placement rate on both boards, and the difference in board output comes entirely from the geometry of the product. That is the correct way to read a capacity shortfall: if placements per hour are close to the plan and boards per hour are not, the constraint is board geometry or overhead, not head speed. Buying a faster head to fix it is expensive and ineffective, because the head is not what changed.

The last row is the panelisation trade in numbers. Running two small boards as one panel halves the number of transfers per pair, which looks like a gain, but it doubles the placement time per cycle and the boards per hour counted in individual boards is 28.2, only marginally better than 27.5. Panelisation wins where transfer and fiducial time is large relative to placement, which is the case on short-cycle boards and on machines with slow indexing. It loses where the panel forces a longer conveyor stroke, more support pins or a stiffer clamp, all of which add their own time. Measure the panel option rather than assuming it.

Single or double clamping: when the second board pays

Double clamping lets the machine hold two boards and populate them in one program, which removes one transfer per two boards and lets the head work across the pair. On the RS-1R the published maximum board size is 650 × 370 mm described as single or double, which means the same envelope supports two boards side by side when each is small enough. The saving is real on boards whose transfer time is a large share of the cycle, which means short-cycle boards and small panels. On a board that already takes two minutes of placement, removing three seconds of transfer changes the cycle by two and a half percent, which will not justify the programming complexity of a dual program.

The condition to check before choosing double clamping is support and access. Two boards side by side need support pins under both, need fiducials visible on both, and need the clamp to hold both flat without tilting either. A thin board supported over a long span will deflect, and a deflected board places badly no matter how good the head is. Ask how many support pins the configuration provides, whether the pin layout can be changed per product, and how long that change takes during a changeover. Double clamping that adds fifteen minutes to every changeover has a poor return on short runs, and a good one on long runs of a single product.

Dual-lane and buffer configurations: the RS-2 trade

Buffers and lanes trade board length for staging capacity, and the trade is written explicitly into the published envelope of at least one model. The RS-2 publishes 800 × 370 mm with one buffer and 410 × 370 mm with three buffers, which means the same machine holds a long panel or stages more boards but not both. That is not an arbitrary limitation; it is the physical consequence of putting buffer positions inside the same conveyor length. A plant that runs 780 mm LED strips needs the single-buffer configuration and cannot also have three staging positions, while a plant running 400 mm boards at high mix may prefer three buffers to absorb changeover interruptions.

The choice therefore follows the product rather than the preference, and it should be made with the largest panel in hand. The length-versus-buffer breakdown across the three RS models sets out how the RS-1R, RS-1XL and RS-2 differ on both axes, and it is worth reading before specifying a buffer count, because the buffer decision can quietly decide which model you need. A plant that wants three buffers on the RS-2 has accepted a 410 mm maximum board, and if any product in the family exceeds that, the buffer count is the wrong variable to be optimizing.

Feeder positions and the changeover clock

Feeder positions are usually quoted as a maximum, and the maximum is only reachable with the narrowest tape. A bank of 112 positions loaded with 8 mm double-lane feeders holds more lines than 112; loaded with 24 mm single-lane feeders it holds far fewer. The consequence for line balance is not about placement rate at all, it is about how often the line stops to change product. Every changeover consumes time that no head speed can recover, and the frequency of changeovers is set by how many lines the bank can hold relative to how many lines the product needs. That is why feeder capacity belongs in a line-balance calculation and not only in a procurement checklist.

The arithmetic is simple enough to run before the purchase. If a product needs 96 lines and the bank holds 112 positions in a convenient mix, the product can stay loaded and changeover is a program selection. If the same product needs 96 lines and the bank practically holds 70 in your tape mix, the line changes over twice per production run, and each changeover carries a reload, a verification and a first-article check. Multiply the changeover time by the frequency and subtract it from the available hours, and the effective rate falls by an amount that has nothing to do with CPH. This chapter quantifies that loss and then describes the two ways to reduce it without buying a larger machine.

Positions are capacity, not rate

A feeder position does not add placements per hour; it adds the ability to keep a product loaded. That distinction matters because buyers sometimes trade rate for positions without noticing, choosing a machine with a larger bank and a lower head speed, then discovering that the bank solved their changeover problem and the rate did not hurt them, or the reverse. The right comparison is not positions against CPH in the abstract but the annual hours of changeover saved against the annual hours of placement lost, computed on your own product mix and your own batch sizes.

A practical way to express it is in hours per month. Suppose a product family needs 96 lines, the bank holds 70 in your mix, and a changeover takes 90 minutes including verification. At eight changeovers a month that is 12 hours, or about three percent of a single-shift month. Choosing a machine with a larger bank might recover all 12 hours; choosing a machine with a ten percent higher rate might recover a different number of hours that depends on the placement count. Compute both in hours and compare them in the same unit, because the unit is the only thing that makes the two options comparable.

Changeover frequency sets the effective rate

Effective rate is published rate multiplied by availability and by the fraction of time spent producing rather than changing over. A machine with 47,000 CPH, 90 % availability and 85 % productive time delivers about 35,955 CPH, while the same machine with 70 % productive time delivers 29,610. The difference between those two numbers is larger than the difference between most competing models, which is why a plant that ignores changeover ends up buying on a spread that its own schedule destroys. The productive-time fraction is a management outcome, not a machine property, but the bank size and the cart design set the ceiling on how good it can get.

Two habits move the fraction without capital spending. The first is sequencing: grouping products that share feeder lines so that a changeover touches fewer reels. The second is pre-staging: building the next cart while the current one runs. Both require that the machine has somewhere to put the staged cart, which is a floor-space question that should be answered before the machine is ordered rather than after it arrives. A bank of 112 positions with no room for a second cart is a bank that will always be reloaded in place, and in-place reloading is the slowest possible changeover.

What a changeover actually contains

A changeover is not one action, it is four, and only the first is visible on a stopwatch. The first is the physical reload: swapping reels, changing carts, moving the bank into position. The second is verification: confirming that every reel matches the program, which on a ninety-line product means ninety checks and is the step most often skipped and most often responsible for a wrong-part defect. The third is the first-article check, which on a fine-pitch product may need inspection before the run can continue. The fourth is the ramp, where the first few boards run slower while the operator watches for problems.

Timing only the reload understates a changeover by a large margin, and understating it is how a plant convinces itself that a small bank is adequate. Measured properly, a ninety-minute changeover is often forty minutes of reload, twenty-five of verification, ten of first article and fifteen of ramp, and only the reload can be reduced by buying a bigger bank. Verification falls with barcode scanning and cart discipline, the first article falls with better process control, and the ramp falls with experience. Those three are management improvements rather than capital purchases, and they are usually cheaper per hour saved than an extra feeder position.

Carts, staging and the feeder units themselves

The physical feeder units matter more than their count once the mix includes wide tape. A 44 mm feeder consumes five and a half 8 mm slots, and a bank that looks half empty with narrow feeders can be full with wide ones. The feeder units sized to your tape mix range from 8 mm through 88 mm on the listing, and the width you need is set by the component pocket rather than by preference. Buying the narrow set only, to save cost, means the wide parts go somewhere else: a tray feeder, a second machine or a manual station, each with its own cost per board.

Spare feeder stock is part of the same decision. A feeder that fails mid-run stops the line until it is swapped, and the swap is only fast if a replacement of the same width is on the shelf. Most plants hold one spare per width in regular use and accept a wait for the rare widths. That policy is worth writing down next to the bank size, because a machine specified with 112 positions and no spare feeders will lose more hours to feeder faults than to changeovers, and the loss will be attributed to the machine rather than to the spares policy.

Component mix: the fourth clock inside the machine

The fourth clock is the one inside the machine and it is the one buyers can least easily see. A head that places small chips at 47,000 CPH does not place large connectors at the same rate, because each placement costs more: the feeder index is longer, the vision recognition window is larger, the nozzle may be different, and the head has to decelerate to place a heavy part without displacing it. A BOM that is ninety percent chips and ten percent large bodies will run close to the published rate; a BOM that is fifty-fifty will run well below it, and no amount of programming skill closes the gap. This is the effect that most often explains why a new machine measures slower than the demonstration.

The effect compounds with part count. A board with 900 placements takes 900 decisions about which nozzle to use, and every nozzle change inside a program costs time in the head's tool-change sequence if the machine carries multiple nozzles, or costs a stop if it does not. A BOM with three nozzle families will run close to continuous; a BOM with seven will stutter. The same logic applies to vision: a program that mixes chip recognition, ball recognition and body-edge recognition spends a larger fraction of its cycle in the camera, and the camera time is not visible in the CPH figure. Quantifying the mix factor before the purchase is the difference between a plan that holds and a plan that has to be renegotiated in month two.

Mix factor: why 47,000 becomes 26,000

The mix factor is the ratio between the rate the head achieves on your board and the optimum rate on the demonstration board, and it can be estimated from the placement breakdown. Assign each part class a relative cost: a chip at 1.0, a small IC at 1.4, a large IC at 2.2, a connector or odd-form part at 3.5, and a tray-presented part at 4.0. Multiply each class by its share of placements, sum, and divide by 1.0. A BOM that is seventy percent chips, twenty percent small ICs, eight percent large ICs and two percent connectors has a weighted cost of 0.70 + 0.28 + 0.176 + 0.07, or 1.226, which implies a mix factor of about 0.82. Applied to 47,000 CPH that is roughly 38,500 CPH before any other loss.

These weights are hypothetical assumptions for illustration, not published machine constants, and they should be replaced with the ratios you observe on your own equipment. The method matters more than the numbers: it forces the mix into the calculation in a way that a single rate figure never does, and it makes the difference between two candidate machines visible on your product rather than on a generic one. A plant that applies its own weights consistently will find that the ranking between models occasionally changes, which is exactly the information a headline comparison hides.

Small parts, big parts and the head travel cost

Head travel is a physical cost that scales with distance, and distance scales with how far apart the parts are on the board. A board with all its chips clustered in one region places faster than a board of the same placement count spread across its full area, because the head spends less time in transit and can maintain a higher average speed. This is why two identical boards with identical BOMs can differ in cycle time, and why a program's feeder layout matters: putting the most-used parts near the region where they are consumed shortens the average travel and can recover a meaningful share of the mix loss.

Big parts add a second effect. A heavy connector requires the head to decelerate, place with controlled force and accelerate away, and the acceleration limit is set by the part rather than by the head. On a board where connectors are placed early, the head travels over them for the rest of the program, which may force a higher travel height and a slower traverse. Sequencing the tall and heavy parts last, where the geometry allows it, keeps the rest of the program at full speed. That sequencing decision is free and is usually worth more than a small rate advantage between two models.

Nozzle and vision overhead inside a run

Every distinct nozzle family in a program costs something. If the head carries multiple nozzles, a change costs a few tenths of a second; if it carries one, the change costs a stop and an operator. A BOM with three families avoids most of the cost; a BOM with seven pays it repeatedly. This is the practical reason to standardize the nozzle set rather than to match every part exactly, and the nozzle set that covers your BOM's orifice range should be chosen with the family count in mind rather than with a part-by-part match.

Vision overhead behaves the same way. A part recognised by silhouette costs less than one needing ball or lead recognition, and a part taught with a wide tolerance window costs less than one requiring a tight window with multiple cameras. Products with formed LED leads or unusual body shapes sit at the expensive end, and a configuration built around vision-guided form placement handles them without falling back to manual work, as the vision-guided build for formed and irregular LED parts illustrates. Where a product family has a high share of these parts, the vision cost belongs in the mix factor rather than in a separate discussion about quality.

Balancing the line: where the bottleneck really is

Line balancing is arithmetic once the clocks are known, and the arithmetic is unforgiving in a useful way. For each station, compute the time per board and convert it to boards per hour; the lowest number is the line's capacity, and every other station is idle for the difference. Running that calculation before a purchase shows whether the new machine will be the constraint or will simply add capacity that the line cannot use. It also shows how much of the current shortfall is caused by the mounter and how much by the printer, the oven or the handling equipment, which is the JUKI pick and place machine CPH feeder PCB line balance question in its most concrete form.

The calculation needs four inputs per station: the cycle time per board, the availability, the changeover time per run and the batch size. With those four, each station's effective capacity in boards per hour is computable, and the line's capacity is the minimum. Two refinements matter. First, availability is not the same as uptime: a station that runs but produces scrap is not available in the planning sense. Second, changeover time should be amortised over the batch, because a station with a fast cycle and a long changeover can be the constraint on small batches and irrelevant on large ones. Both refinements make the number less pretty and more true.

Compute boards per hour at each station

The table below is a hypothetical line balance for a single-shift plant running 900 boards a month of the control board from the earlier example, with a mounter whose effective rate after mix factor is 26,000 CPH. Each row shows the cycle, the availability assumption, the resulting capacity and the utilisation against a demand of 900 boards a month, which is about 45 boards per shift day. The mounter is not the constraint in this example; the printer is, and the difference between them is the capacity the plant would gain by improving the printer rather than by buying a faster head.

The mounter row is deliberately extreme to make the point: at 24 boards per hour against a demand of 45 per day, it has capacity for roughly two shifts of work in one, which means the line is not mounter-limited and a faster mounter would add nothing. In a real plant the numbers are usually closer, and the mounter is often the constraint by a modest margin. The value of the table is that it identifies which margin, and it prevents the purchase of capacity at one station to fix a shortage at another. Run it with your own cycles before you decide, and re-run it after every process change, because the constraint moves.

When the mounter is not the constraint

Three conditions reliably make the mounter not the constraint. The first is a fine-pitch product where the printer's variability and cleaning interval dominate, which is common in industrial and automotive boards. The second is a long or heavy board where the oven profile sets a transit the mounter cannot beat, which is common in LED and power products. The third is a high-mix plant where the changeover time at the mounter is large but the batch size is small, so the mounter's effective capacity is limited by setup rather than by placement. In all three cases, the first investment should be in the constraining station, and the mounter should be specified to match it rather than to exceed it.

Recognising those conditions changes the purchase in a specific way: it moves money from head speed to feeder capacity, cart design and programming discipline, which are the items that actually relieve a setup-limited line. It also changes the acceptance test, because a machine bought to relieve changeover should be tested on changeover rather than on rate. A supplier who understands the distinction will accept a test that measures a timed changeover between two products, and that test is worth more than any number of demonstration cycles on a single board.

The second platform question

Once the line balance is known, the question of a second platform becomes arithmetic rather than ambition. If the mounter is the constraint and its placements are dominated by small chips, a rotary-head machine placed alongside or in series can take the chip volume and leave the flexible head for the parts that need it. If the constraint is changeover, a second bank or a second small platform loaded for the other family may be cheaper than a larger machine. The position of the rotary and flexible routes in a balanced line is worth understanding before that decision, because the two routes solve different halves of the problem and neither solves both.

A second platform also changes the failure mode of the line. Two machines running in parallel let the plant continue at reduced rate when one is down; two machines in series stop the line entirely. That redundancy is a real benefit for a plant with delivery commitments and a real cost in floor space and programming. The cheaper option on paper, a single larger machine, has no redundancy at all, and the value of redundancy depends on how expensive a stopped line is for your customers. That is a commercial judgment rather than an engineering one, and it belongs in the same comparison as the capital cost. Where a modest capital step buys a second smaller platform, the budget chip platform for the volume family is the kind of option that makes that step affordable.

A configuration ladder by volume step

Most plants do not choose between four machines in one decision; they step up as volume and product range grow, and each step has a configuration that makes sense at that point and stops making sense at the next. Describing the ladder explicitly keeps a buyer from paying for a step three configuration at step one volume, which is the most common form of overcapitalisation in SMT, and from buying a step one machine for step two work, which is the most common form of premature replacement. The ladder below uses the same four constraints as the rest of this guide, applied to three volume bands, and it treats JUKI pick and place machine CPH feeder PCB line balance as a question that is answered differently at each band rather than once and for all.

The bands are defined by demand rather than by machine size: a low band where the line is setup-limited and flexibility matters most, a middle band where the line is placement-limited on one product family and a second head pays, and a high band where two families run concurrently and the line should be split. The boundaries are not universal and should be computed from your own cycle times, but the pattern of what to buy at each band is stable across plants. What follows describes each step, the configuration that fits it, and the signal that tells you it is time to move up.

Step one: one flexible platform with a full bank

At low volume the constraint is usually changeover, not placement, so the configuration should buy feeder capacity and flexibility rather than rate. That means a platform with the largest bank the budget allows, a component ceiling that clears the tallest and largest parts in the family, and enough nozzle families to run the whole BOM without a stop. The RS models fit this shape: 112 feeder inputs, a 25 mm height ceiling, 74 mm square component limit and a flexible inline head, with the RS-1R covering panels to 650 × 370 mm and the RS-1XL covering 650 × 560 mm.

The signal to leave step one is not a rate shortfall; it is a changeover shortfall. When the plant is spending more than about ten percent of its available hours in changeover, or when the same product family is being reloaded more than twice per run, the bank or the staging has run out. The cheapest response is a second cart and a staging discipline; the next is a larger bank; buying a faster head at this point adds cost without relieving the constraint, because the line is not placement-limited.

Step two: add the volume family on a second head

The middle band appears when one product family has enough volume that its placement count dominates the schedule. At that point the flexible platform spends most of its time on chip placements it could do faster on a rotary head, and the parts that need flexibility are a minority of the placements. Adding a second machine for the volume family, with the flexible platform resuming responsibility for the mixed work, is the configuration that fits. The two machines can run in series for one product or in parallel for two, and the choice depends on how the family's BOM divides.

The volume family is often LED work, where placement counts are high, line counts are low and the board may be long. A six-head high-speed configuration for LED boards is the kind of build that takes that family without asking the flexible platform to do it, and it keeps the bank requirement small because the line count is small. The signal that step two was right is that the flexible platform's changeover count falls while total output rises, which is the arithmetic of separating a repetitive family from a mixed one.

Step three: split by family and re-balance

At high volume the line should be split by family rather than balanced across one line, because two families with different constraints cannot both be optimized by the same sequence of stations. A common split is a high-rate line for the chip-dominant family and a flexible line for the mixed family, each with its own printer, oven and inspection, and each balanced internally. That structure costs floor space and duplicates the support equipment, and it buys the ability to change over one line while the other runs, which is the only reliable way to serve a high-mix schedule at high volume.

At this point the balance calculation is done per line rather than per plant, and the question becomes which line has spare capacity for the next product. That is a different planning discipline and it needs the station numbers from the earlier table maintained per line rather than once. Plants that reach step three usually discover that their bottleneck has moved to inspection or to handling, which is why the calculation has to be re-run after every structural change rather than fixed at the moment of purchase.

The bands are illustrative and the boundaries will move with your cycle times, batch sizes and shift pattern. What does not move is the order of the steps: flexibility first, then rate for the family that needs it, then structural separation. Buying the steps out of order produces the two classic failures, a plant that owns a fast machine it cannot keep loaded and a plant that owns two flexible machines competing for the same mixed work. Neither is a machine problem; both are configuration problems, and both are visible in a line balance before any money is spent.

Send the cycles you measure, not the rate you were quoted

If you send us your board sizes with rails, your placement count and BOM line count by product family, your batch sizes and the cycle times you actually measure at the printer and the mounter, we will build the balance with you rather than hand you a specification. We will show which station is the constraint at your demand, what the mounter has to deliver to stop being the queue, how many feeder positions your tape mix consumes and whether a second cart or a second platform is the cheaper answer. We will mark which figures come from the product pages on this site, which come from JUKI's published specification and which are our assumptions, and we will put the availability and mix factors next to the capacity number so you can defend it internally. If the honest answer is that your line is printer-limited and a faster mounter would add nothing, the review will say that, because a machine sold into a line that cannot use it is a machine that will be blamed for a bottleneck it did not create.

Frequently asked questions

Is the published CPH figure usable for capacity planning?

As an input, yes; as the answer, no. The published optimum rate describes the head under favourable conditions with a demonstration board, and it excludes feeder index, vision recognition, transfer, fiducial reading and changeover. Convert it into an effective rate by applying a mix factor for your placement breakdown, an availability factor for stops and maintenance, and a productive-time factor for changeover. A machine published at 47,000 CPH commonly lands between 26,000 and 32,000 CPH on a mixed BOM, and the exact number depends on your product rather than on the machine.

How do I know whether the mounter or the printer is my bottleneck?

Time both stations over a shift and convert to boards per hour. The lower number is the constraint. A printer with a median of 58 seconds and a 95th percentile of 74 seconds has a capacity near 52 boards per hour after availability, and a mounter placing 900 components at an effective 26,000 CPH needs about 131 seconds per board, which is 24 boards per hour. In that case the mounter governs, but only by a small margin, and improving the printer's tail variance would still raise output. Record the tail, not just the median, because starvation is caused by the tail.

Does a bigger feeder bank increase output?

It increases output only when changeover is the constraint. A larger bank keeps more lines loaded, which reduces the number of changeovers per run and the hours they consume. On a plant already running long batches with few changeovers, a larger bank adds little. The test is the share of available hours spent in changeover: above roughly ten percent, bank capacity or staging is worth money; below that, the same money is better spent on the constraining station.

Should I panelise small boards to raise throughput?

Only when transfer and fiducial time is large relative to placement time. Panelising halves the number of transfers per pair of boards, but doubles the placement time per cycle, so the gain in boards per hour is often small. In the worked example, two 300 × 250 mm boards panelised gave 28.2 boards per hour against 27.5 run separately, about two and a half percent, and that gain can disappear if the panel needs a longer conveyor stroke, extra support pins or a stiffer clamp. Measure both options on your machine.

How much does double clamping actually save?

It saves one transfer per two boards, so the benefit scales with how large the transfer is in the cycle. On a board that takes two minutes of placement, removing three seconds of transfer changes the cycle by about two and a half percent. On a short-cycle board where transfer is a third of the cycle, the saving is substantial. Check support and fiducial access for both boards before choosing it, and check how long the pin layout change takes during a changeover, because double clamping that adds fifteen minutes to setup is a poor trade on short runs.

Why does the same machine run slower on one board than another?

Three reasons, and they are measurable. The component mix changes the placement cost: large bodies, tray parts and specialised nozzles each add time. The board geometry changes the travel distance, so a BOM spread across a large panel places slower than the same count clustered in one area. And the transfer and fiducial overhead is fixed per board, so short-cycle boards carry a larger share of it. Comparing placements per hour rather than boards per hour separates the mix and geometry effects from the machine's own performance.

When is a second machine better than a faster one?

When the constraint is changeover rather than placement, or when two product families have genuinely different requirements. A second platform loaded for the other family removes the changeover entirely, and a second machine also provides redundancy that a single larger machine does not. A faster machine wins when one family dominates and its placements are chip-heavy enough that a rotary head pays. Compute both options in hours per month, including floor space, a second program structure and the training load, and treat redundancy as part of the return rather than as a bonus.