English Views: 0 Author: Elsa Long Publish Time: 2026-11-10 Origin: Site
Choosing a JUKI RX-7R vs RX-8 high speed chip shooter is usually framed as a throughput comparison, and the framing is the first thing that needs fixing. Both machines belong to the high-speed rotary class, both are aimed at small chip components, and both will place faster than the printing and reflow stations around them can keep up with on most boards. The difference that decides a purchase is rarely the headline rate. It is where each machine's rate collapses — on which component range, at which board size, with which feeder load, and against which neighbouring station — and that collapse point is a configuration question rather than a specification question. This guide works through the comparison in that order, treating the published figures as a starting point rather than an answer, and finishes with the documents and test boards that turn a claim into evidence.
A note on the numbers used here. Published specification tables for high-speed mounters are frequently written for a machine family and copied across models, so a table can carry a model name at the top and another model's row values below it. Every figure in this guide that is not attributed to a verified document is labelled as an assumption, and the calculation methods are written so they can be re-run with the buyer's own verified values. The platform classes referred to throughout are the ones in the placement platforms this comparison is drawn from.
A high-speed chip shooter buys one thing: a lower cost per placement on boards that contain a lot of small components. It does not buy a faster line, and confusing the two is the most expensive mistake in this category. The placement station is one of three clocks on a line, and doubling its rate changes the line's output by whatever fraction of the cycle it represented before. On a board with 600 placements where placement consumed three quarters of the cycle, a much faster head can nearly double line output. On a board with 180 placements and a fine-pitch stencil, placement may already have been the shorter clock, and the faster head buys almost nothing while adding changeover complexity. The machine is purchased for a board population, not for a specification sheet, and the population decides whether the purchase is real or decorative.
There is a second thing high speed buys that buyers rarely price: a harder set of constraints downstream. A machine placing several times faster consumes feeder tape several times faster, which converts a comfortable reload cycle into a bottleneck of its own. It also delivers boards to the reflow oven in tighter succession, which turns conveyor buffering and oven loading into scheduling problems. Those constraints are manageable, but they cost money and floor space, and they belong in the same budget as the machine. A comparison that stops at the placement rate has compared the cheapest part of the decision.
A placement rate quoted as components per hour is measured under conditions that the buyer will not reproduce: a specific board pattern, a specific component mix, a specific feeder arrangement, and no stops of any kind. It is a valid number in the sense that the machine can achieve it, and an invalid number in the sense that it predicts nothing about a real shift. The gap between the quoted rate and the achieved rate is usually 30 to 50 percent on a well-run line, and larger on a mixed one, because the quoted figure excludes pick errors, nozzle changes, feeder reloads, vision rejects and every stop that is not a placement.
The practical way to use the number is as a ceiling rather than a forecast. Divide it by a realistic operational factor, then compare the result with the other clocks on the line. If the adjusted figure is already above what the printer or the oven can support, the extra headline rate has no economic value and the comparison should move to the second-order questions: feeder capacity, component range, board handling, and what the machine does when it is not placing at full speed. That reframing is where an RX-7R versus RX-8 discussion becomes useful, because the two machines differ more in their behaviour at the edges than in their behaviour at the optimum.
Both machines sit in the rotary high-speed family, and that family shares a defining trait: a rotating head that carries multiple nozzles and picks several components per movement, which is what makes small-chip placement fast. Within the family the differences that matter to a buyer are not the ones printed largest. They are the number of nozzles the head carries and how the machine trades between picking and placing, the range of component sizes the head can handle without a nozzle change, the feeder interface and how many positions are reachable, and the board handling arrangement that sets the maximum panel the machine can accept.
Those four differences produce very different behaviour on real boards. A head with more nozzles wins on a board dominated by 0402 and 0603 chips and loses some flexibility on a board with a wider component spread, because a wider spread forces more nozzle exchanges or a second pass. A machine with a larger reachable feeder count wins on boards with many unique part numbers and loses nothing on boards with few. Board handling decides whether a panel format that is efficient for the printer can pass through the mounter at all. None of these is a specification-sheet headline, and all of them decide which machine fits a given product list. The shortlisting method that produces the candidate list in the first place is set out in the platform shortlisting method that precedes any model comparison.
Unused rate is not free. A machine bought for a capability the line cannot use still consumes floor space, still needs power and air, still requires maintenance and still appears on the depreciation line. On a line where the printer is the constraint, the additional placement capability contributes nothing to output and something to overhead, and the correct decision may be a smaller machine, a used machine, or no machine at all. That conclusion is unwelcome in a capital request, which is why the three-clock comparison is often completed after the purchase rather than before it.
The way to make the argument honestly is to express the unused capability as a percentage of the machine's working time. If the placement station is blocked or starved for a third of the shift because of the printer or the oven, then roughly a third of the capital sits idle, and that fraction is the real cost of overspecifying. A plant that presents this figure alongside the quotation usually finds the discussion moves quickly to the neighbouring stations, which is where the money should have gone. It is a more useful conversation than a debate about which model is faster, and it is grounded in the plant's own data rather than in a brochure.
The first question is what fraction of your placement volume is chip components below 0603. If the answer is high and stable, the high-speed rotary class is the correct tool and the comparison is genuinely about rate and feeder capacity. If the answer is moderate and mixed with connectors, shields and odd shapes, a rotary head will spend part of every cycle waiting for a nozzle change that a flexible machine would not need, and the comparison should include a second machine rather than a bigger first one. This single ratio explains most of the cases where a buyer is disappointed by a machine that performed exactly as specified.
The second question is what the second-fastest clock on the line is. If printing is close behind placement, the placement upgrade has a hard ceiling and the money is better spent on the printer or on buffering. If reflow is the limit, the placement upgrade is wasted entirely on that line. The third question is how often the product changes. A dedicated high-speed machine running one product for weeks extracts the full value of its rate; the same machine in a mixed schedule spends its advantage on changeovers it was not designed to make quickly. Answering those three before opening a quotation is what makes the RX-7R versus RX-8 decision a comparison rather than a preference.
Capacity planning for a high-speed mounter is arithmetic with three inputs: placements per board, an adjusted placement rate, and the availability the line actually delivers. None of the three is on a specification sheet. Placements per board comes from the bill of materials and the panelization; the adjusted rate comes from the quoted rate divided by an operational factor measured on a comparable line; availability comes from the plant's own downtime and changeover record. With those three, the machine's output becomes a number that can be compared with the printer and the oven, which is the only comparison that matters commercially.
The reason to do this before choosing between two models is that the models usually differ by a margin smaller than the uncertainty in the three inputs. If the adjusted difference between an RX-7R and an RX-8 on your boards is ten percent and your availability estimate is uncertain by twenty, the models are equivalent for planning purposes and the decision should be made on other grounds: feeder capacity, service, spare parts, or the board sizes you expect in three years. Being honest about that is more useful than manufacturing a precision the data does not support.
Worked example, explicitly illustrative: a board carries 620 placements, the line runs panels of four, so each panel carries 2,480 placements. Assume an adjusted placement rate of 42,000 placements per hour for the machine under evaluation, a figure chosen only to make the arithmetic visible. One panel takes 2,480 divided by 42,000 hours, or 0.059 hours, which is about 3.5 minutes. At 85 percent availability, the practical panel rate is about 14.5 panels per hour, or 58 boards per hour. Multiply by the shift length and subtract changeover time and the daily output appears.
Now run the same board at an adjusted rate of 52,000 placements per hour, a 24 percent improvement. The panel time becomes 2.86 minutes, and the practical rate rises to about 17.8 panels per hour, or 71 boards. That looks like a large gain until it is compared with the printer: if the printer takes 4 minutes per panel including its cleaning cycle, it is now the constraint and the extra boards never materialise. The arithmetic is trivial and that is precisely why skipping it is indefensible. The numbers above are placeholders; substitute verified rates and the plant's own availability before using any of them in a justification.
Capacity conversion: the same board at three adjusted placement rates (illustrative assumptions) | |||
Input | Case A | Case B | Case C |
|---|---|---|---|
Placements per board | 620 | 620 | 620 |
Panels per board count | 4 | 4 | 4 |
Adjusted rate (placements/hour, assumed) | 42,000 | 52,000 | 62,000 |
Panel time (minutes) | 3.54 | 2.86 | 2.40 |
Availability (assumed) | 85% | 85% | 85% |
Practical panels per hour | 14.4 | 17.8 | 21.3 |
Boards per hour | 58 | 71 | 85 |
Printer ceiling (panels/hour, assumed) | 15 | 15 | 15 |
Useful gain | none | capped | blocked by printer |
Five losses account for most of the distance between a quoted rate and a shift average. Pick errors and vision rejects are the first, and they rise with component density and with nozzle condition. Nozzle and feeder exchanges are the second, and they are larger on mixed boards. Feeder reloads are the third, and they are the loss that scales most directly with placement speed, because a faster machine empties tape faster. Board transfer and clamping are the fourth, and they become visible when panels are small, because the fixed transfer time is amortised over fewer placements. Stops for paste, stencil cleaning and material replenishment are the fifth, and they are line-level rather than machine-level, which is why a machine specification never includes them.
A plant that wants a defensible number should measure those five on an existing line, express each as a percentage of available time, and apply the combined factor to the quoted rate. The result is usually sobering and always more useful than the brochure. The measurement takes a week of production data and can be done by the line leader without new equipment. Its value is not only the number: the breakdown shows which loss to attack first, and in a high-speed installation the first three losses usually dominate, which is exactly where feeder planning and nozzle discipline pay.
Two machines with identical rates can differ by a third in shift output because of availability. Availability here means the fraction of scheduled time in which the machine is able to place, and it is driven by reload stops, nozzle exchanges, vision rejects, material shortages and unplanned maintenance. A high-speed machine has a structural disadvantage on this measure, because its consumption rate generates more of every event that interrupts it. That disadvantage is manageable and it has to be planned for rather than assumed away.
The measurement is a stop log kept for two weeks: every event, its duration and its cause. Plants that do this typically find that the three largest causes account for most of the lost time, and that at least one of them is not a machine issue at all — material availability and feeder preparation are frequent offenders. Once the causes are named, availability becomes an improvable number rather than a fixed property, and the improvement usually costs less than the difference between two machine models. A JUKI RX-7R vs RX-8 high speed chip shooter decision that ignores availability is comparing the machines on the one metric they share and the one that decides least.
Placement stops being the constraint when its panel time falls below the printer's panel time including cleaning, or below the oven's panel time at the profile being run. Both are common once a fast machine is installed, and both are missed because the comparison is made per component rather than per panel. A printer that takes 45 seconds of actual printing plus a 20-second clean every ten panels has an effective panel time of 47 seconds, not 45, and that two-second difference is enough to make it the constraint on a fast line. An oven running a six-zone profile with a long soak can hold a panel for longer than the mounter needs to fill it.
The planning consequence is that a high-speed purchase should always be paired with a decision about the neighbouring stations. Either the printer is upgraded to match, or a buffer is installed to decouple the two, or the machine is specified below its maximum rate on purpose so that the line stays balanced. The third option is unpopular and often correct: running a fast machine at a fraction of its capability, with a stable cycle, produces more good boards than running it at maximum against a printer that cannot keep up. A buyer who understands this stops asking which machine is faster and starts asking which machine is faster at the rate the line can actually absorb. Where the answer points to a machine already in service, the checks that apply are in what a pre-owned RX-class machine must prove before it ships.
A high-speed machine has a supply problem that a slower machine does not, and it appears within the first month of operation. Placement consumes tape in direct proportion to rate, so a machine placing two or three times more components per hour empties reels two or three times faster. The operator who could reload a reel without stopping a slower line now stops the fastest machine on the floor, and the loss is not the reload time itself but the queue that forms behind it. This is the single most under-planned consequence of buying speed, and it is a feeder logistics problem rather than a machine problem, which is why it is absent from the specification and present in every commissioning report.
The planning question is therefore not only how many feeder positions the machine offers, but how the plant intends to keep them fed. That answer has three parts: the reel inventory that must be staged, the reload procedure that allows a splice or a cart exchange without a full stop, and the width mix that determines how many of the nominal positions are actually usable. Each of the three can be specified and costed before the order, and each is cheap at that stage and expensive afterwards.
Take a board with 620 placements and a component mix where 70 percent of placements come from twelve high-usage 8 mm reels. At a practical rate of 58 boards an hour, that is roughly 25,000 placements an hour from those twelve reels, or about 2,100 picks per reel per hour. A reel carrying a typical quantity of small chip components will not survive a full shift at that consumption. The arithmetic is simple and it is routinely missed, because the purchasing calculation is done in placements per hour and the operational reality is in reels per shift.
The countermeasure is to compute reel consumption per shift for the ten highest-usage part numbers before the machine is ordered, then decide how those reels will be replenished. Options include larger reels where the feeder accepts them, a splice procedure that allows a mid-run join, a second cart staged with the next reels, or accepting a scheduled stop at a chosen moment rather than an unplanned one. Each option has a cost and each is preferable to discovering the problem during the first week. A machine that is fast enough to empty a reel in ninety minutes needs a reload plan more than it needs another ten percent of rate.
Nominal feeder counts are quoted in the narrowest tape width, and every wider feeder consumes more than one position. A board with 90 part numbers can easily need 120 to 140 equivalent positions once 12, 16 and 24 mm tape is counted, which means a machine advertised at a comfortable position count may be nearly full on a single product. In a mixed plant the union across a family makes this worse, because the family kit must accommodate the widest product rather than the average one. The effective bank, not the nominal one, is what the product list has to fit into.
The practical measurement is to convert every line of the bill of materials to narrow-width equivalents, take the maximum across the products the machine will run, and compare that with the bank. If the result is above 80 percent of nominal capacity, the machine will be effectively full on day one and the plant needs a second cart, a second bank, or a plan to swap banks mid-product. That is a configuration decision made at order time. Which tape widths consume the most bank positions, and how the conversion works in practice, is set out in which tape widths consume the most bank positions.
Every feeder system has a human throughput ceiling, and on a high-speed machine that ceiling becomes the binding constraint. An operator can splice a tape or exchange a cart in a finite time, and the machine's consumption rate sets how often that has to happen. If the required reload frequency exceeds what one operator can sustain across a shift, the line needs either a second person at the material side, a larger staged inventory, or a deliberate reduction in rate. None of those appears in a machine comparison, and all of them determine whether the machine delivers its rate on a Tuesday afternoon.
The way to test this before buying is to write the reload schedule for a realistic shift: the ten highest-usage reels, their consumption at the planned rate, and the moment each will need attention. If two reels need attention within the same ten minutes, the plan already fails and needs a staged cart. If the schedule shows a comfortable gap, the feeder plan is adequate. This exercise takes an hour with a spreadsheet and prevents the most common form of high-speed disappointment: a machine that meets its specification for the first ninety minutes of every shift and then settles into a slower rhythm that nobody can explain in the weekly report.
The component range of a high-speed rotary machine is usually described as a span — smallest to largest — and the span is misleading in both directions. At the small end, the ability to place 0201 components is a capability, not a routine, and it depends on nozzle condition, feeder accuracy and board support as much as on head resolution. At the large end, a machine that can technically place a connector may do so slowly enough that the placement is better assigned elsewhere. The useful question is not the span but the distribution: what fraction of your placements falls in the band where this machine is genuinely efficient, and what happens to the rest.
That distribution decides whether one machine can carry the whole board or whether the line needs a second, more flexible mounter. It also decides the nozzle set, the feeder mix and the cycle time, because the parts outside the efficient band are the ones that consume the extra seconds. A comparison between two high-speed models that ignores the distribution is comparing two machines on the part of the board where they are most similar.
The small-chip floor is the size below which placement becomes unreliable at speed, and it is set by the interaction of three things: the nozzle orifice and its condition, the feeder's tape pocket accuracy, and the board's support during placement. A machine rated for a small component size can place it, and may not place it consistently once the nozzle has a few thousand cycles on it. That is why the floor should be treated as a maintenance-dependent figure rather than a fixed capability, and why a plant intending to run small chips at high rate needs a nozzle condition regime rather than a larger nozzle inventory.
In practice, the floor question is answered by a test rather than a table. Take the smallest component in the actual product mix, run it at the intended rate for a defined number of placements, and record pick errors and placement offsets. If the error rate is acceptable and stable, the floor is below that component. If it degrades over the run, the floor is above it at that rate, and the choice is to slow down, to add nozzle maintenance, or to assign that component to a different machine. The test costs a shift and produces a number the plant can rely on, which no specification sheet can provide.
Large parts are where the economics of a chip shooter change most sharply. A rotary head placing small chips at high rate is efficient because it moves many components per cycle; a large connector forces a different motion, often a different nozzle, and sometimes a vision routine that takes longer than the placement itself. The result is that a machine which places 600 chips in three minutes can spend another minute on eight connectors, which is a disproportionate share of the cycle. On boards where large parts are a small minority of placements but a large share of cycle time, the correct configuration is often two machines: the high-speed one for chips, a flexible one for the rest.
Odd-form parts — shields, tall electrolytics, parts that must not be marked, parts supplied in trays or sticks — push the same conclusion further. They require feeder types the high-speed bank may not accommodate, nozzle types outside the shared set, and often a slower placement mode. A buyer who counts them as placements and expects the high-speed rate to apply will be disappointed by the first board that contains them. Counting them separately, and pricing the second machine or the second pass, produces a realistic cycle time and a realistic comparison between the two RX models.
0201 metric components are the standard proof point in high-speed placement discussions, and they deserve a specific check because they expose the whole system rather than just the head. Placing them reliably requires a nozzle in good condition, a feeder that presents the tape consistently, a board that is supported directly beneath the placement area, and a vision routine that can distinguish the part from the pocket. Any one of those failing produces the same symptom — a pick error — so a machine that fails on 0201 is not necessarily a machine with a head problem.
The check is to run the smallest component in the mix at the planned rate, on the actual board, with the actual feeder, and to record where errors occur. If they cluster on one feeder position, the problem is presentation. If they cluster on one region of the board, the problem is support. If they are uniform, the problem is nozzle condition or vision. That diagnosis takes less than a day and it tells the buyer whether the machine's small-chip capability is real for this product or only real on a demonstration board. The nozzle range that sets this floor, and how it maps to a real component list, is described in the nozzle range that sets the small-chip floor.
A rotary high-speed machine is built around tape, and everything that does not arrive on tape has to be accommodated deliberately. Trays occupy positions that are wider than their tape equivalent and often need a dedicated feeder; sticks need a magazine and a different presentation height; bulk parts need a feeding method the machine may not offer at all. On a board where those parts are a small minority of placements but a regular occurrence, the accommodation is a fixed cost in bank space that has to be planned before the machine is configured.
The planning consequence is that the tray and stick requirement belongs in the feeder calculation, not in a footnote. Count the tray positions needed by the densest product in the list, check whether the tray feeder can stay mounted across products, and confirm that its position does not block access to the tape feeders behind it. A machine that loses four tape positions to a tray feeder it uses once a month is worse configured than one that keeps the tray off the machine and runs that product on a second pass. Those are the trade-offs that separate a working configuration from a specification-compliant one, and they matter more on a high-speed machine than on a flexible one, because the bank is the resource that buys the rate.
Installing a faster mounter does not raise line output by the same proportion, and the reason is that the constraint relocates. Before the upgrade, placement is usually the longest clock; after it, printing or reflow takes that role, and the improvement that was justified by a placement calculation fails to appear in the weekly output. This is not a machine fault and it is entirely predictable, which makes it the most avoidable disappointment in the category. The countermeasure is to model the three clocks before the order and to decide, in advance, what will be done about the new constraint.
The modelling is straightforward once the clocks are expressed in the same unit. Printing, placement and reflow all act on a panel or a board, so converting each to seconds per panel reveals which one is longest and by how much. The interesting output of that comparison is not the winner but the margin: a two-second margin is fragile and will flip as soon as a cleaning cycle or a paste change lands, while a twenty-second margin is structural and can be planned around. A line whose clocks are within a few seconds of each other behaves erratically and blames its machines, when in fact it has no constraint and therefore no stable rhythm.
Printing is the first clock, and its panel time includes more than the squeegee stroke. A printer's effective time per panel is the print cycle plus the separation and transfer, plus a share of the cleaning cycle allocated across the panels between cleans. A printer that cleans every ten panels for twenty seconds adds two seconds to every panel, which on a 45-second print cycle is a five percent change — enough to matter when comparing against a fast mounter. Stencil quality and paste behaviour also stretch the cycle indirectly, through the frequency of wipes and touch-ups.
Placement is the second clock and the one being upgraded, and its panel time should be computed from placements per panel at an adjusted rate rather than from the quoted rate. Reflow is the third, and its clock is set by the profile rather than by the oven's mechanical capability: a profile with a long soak for a heavy board can hold a panel for longer than a fast mounter needs to fill it. The three clocks should be written down in seconds per panel, with the assumptions behind each number stated, and the largest of the three named as the current constraint. Doing this once, properly, prevents years of arguing about which machine is slow.
Three clocks in seconds per panel, before and after a placement upgrade (illustrative assumptions) | |||
Clock | Before | After upgrade | What sets it |
|---|---|---|---|
Printing (incl. allocated cleaning) | 47 s | 47 s | Print cycle, separation, clean frequency |
Placement (adjusted rate) | 212 s | 86 s | Placements per panel divided by adjusted rate |
Reflow (profile duration / pitch) | 62 s | 62 s | Profile length, board pitch, zone loading |
Named constraint | Placement | Reflow | Longest clock, with its margin to second place |
Panel output per hour | 17.0 | 44.7 theoretical | Capped by the constraint, then by transfer and stops |
The choice between a fully linked line and a buffered one decides how the new constraint behaves. A linked line transmits every stoppage upstream and downstream: when the printer cleans, the mounter waits, and when the mounter stops for a reload, the printer has nowhere to deliver. A buffered line absorbs those events up to the buffer's capacity, and the buffer's size determines how long a stoppage can be tolerated before the whole line feels it. In practice one to three board positions between stations converts most short events into non-events, and that is usually the highest return per square metre in the whole installation.
The cost of buffering is floor space and a little extra handling, both of which are cheaper at order time than as a retrofit. The cost of not buffering is a line that runs at the speed of its slowest moment rather than its slowest station, which is a meaningfully lower number. For a high-speed placement machine the buffer question is not optional, because the machine's own stops — reloads, nozzle exchanges, vision rejects — are frequent and short, which is exactly the pattern a buffer handles well. A plant that installs the fast machine into a fully linked line will conclude after a month that the machine is not as fast as advertised, when the machine is performing exactly as specified and the line is simply transmitting every pause.
Once placement is fast, the question that follows is whether the plant needs a second, slower and more flexible mounter for the parts the fast machine handles badly. This is the configuration pattern most high-volume small-chip plants converge on: a rotary high-speed machine for the chip population and a flexible machine for connectors, odd forms and short runs. The pattern is efficient and it has a cost that is rarely counted — the second machine adds a placement pass, which adds handling, another first article and another program to maintain.
Whether the second machine pays depends on the share of placements it would take and the cycle time it would remove from the fast machine. If it removes a minute of chip-shooter time per panel while adding thirty seconds of its own, the net gain is thirty seconds only if the chip shooter was the constraint. If reflow is the constraint, the second machine adds handling and removes nothing. The decision therefore belongs after the three-clock analysis, not before it. For boards where the non-chip population is dominated by LED drivers and similar parts, the alternative to a general flexible machine is a dedicated multi-head class, and the trade between that approach and a general flexible mounter is described in the four-head LED-driver class on the other side of this trade.
Board size is usually treated as a pass or fail check — does the board fit — when it is really a decision about panelization and about how much of the transfer time is amortised across placements. A machine accepting a large maximum panel can run a panel format that keeps the printer and the oven efficient, and it can also run small boards in a frame that reduces the relative cost of transfer. A machine with a smaller maximum forces a different panelization, which changes stencil cost, handling and the effective cycle time of every station. The maximum board size therefore propagates through the whole line, and comparing two machines on that number alone misses why it matters.
The second half of the question is the transfer boundary: how the board enters and leaves, whether the conveyor can be adjusted without stopping the line, and whether the machine can accept the panel thickness and warpage the plant actually produces. Those are the details that decide whether a technically acceptable machine is operationally smooth, and they are cheap to check before the order.
A dual-lane machine carries two boards or panels side by side, which reduces the relative cost of transfer and can improve throughput on small boards by allowing one lane to load while the other places. The benefit is real and it depends entirely on the board population: on large panels the second lane may not be usable at all, and on very small boards the gain can be substantial because transfer time is amortised over twice the placements. The cost is floor width, a more complex conveyor setup and a changeover that has to be done twice.
The decision rule is arithmetic. If transfer time is a significant fraction of panel time — which it is for panels with few placements or for small boards run singly — a dual lane earns its space. If panels are large and dense, transfer time is already amortised and the second lane adds complexity without output. A mixed board list makes the decision harder, because the machine has to be configured for the common case and adjusted for the rest. In that situation a single lane with a good buffer is often the better compromise, and the money is better spent on feeder preparation than on a lane the plant will use half the time.
Panelization decides how much of the machine's capability reaches the board. A panel with four boards and a frame consumes machine time on the frame's transfer and clamping but not on its placements, so the effective rate per good board is lower than the rate per placement. A panel with more boards amortises transfer better and increases the cost of a single defect, because a reject can take the whole panel. The frame itself has a cost in material and in the space it occupies on the machine, and on a large-panel machine that cost is small while on a small-panel machine it can be dominant.
The planning method is to compute, for two or three candidate panel formats, the placements per panel, the panel time at the adjusted rate, and the good boards per hour after a realistic panel reject rate. The format that wins is rarely the one with the most boards, because the reject rate rises with panel complexity and because a denser panel needs better stencil and better paste control. Running that comparison before choosing the machine matters, because the machine's maximum board size sets which formats are available. The same board-size decision behaves differently once the product list becomes mixed, and how a mixed schedule changes it is described in how a mixed product list changes the same board-size decision.
Conveyor width adjustment is a small feature with a large operational footprint in a plant that runs several board widths. A manual adjustment takes minutes and requires the line to stop; a motorised one takes seconds and can be part of a program. Across a mixed schedule the difference accumulates: ten width changes a week at four minutes each is forty minutes, which is small, but the same changes at four minutes each with a line restart are considerably more. The feature is inexpensive relative to its operational value and is routinely omitted because it does not appear in a throughput comparison.
Board thickness and warpage belong to the same discussion. A machine that handles the nominal thickness but struggles with a warped panel will produce placement errors that look like accuracy problems, and the fix is support and clamping rather than a different head. Checking the actual thickness range and the actual warpage of the plant's boards against the machine's support arrangement is a pre-order check that costs an hour with a micrometer and a straight edge. It is also the check most likely to reveal that a machine which passed every specification comparison needs an additional support option to work on real boards.
Board support is the least discussed and most consequential detail in high-speed placement, because at high rate there is no time for the machine to recover from a board that flexes. A panel that bows slightly under the placement force will place accurately at low speed and drift at high speed, and the drift appears as an accuracy problem that no amount of calibration fixes. The support arrangement — pins, a vacuum plate, a magnetic table, a dedicated carrier — has to match the panel, and the panel's behaviour under load has to be known before the machine is ordered.
The check is mechanical rather than electrical: place the actual panel on a flat surface, apply light pressure at the placement area, and measure the deflection. Then compare that with the support the machine offers. A panel that deflects noticeably needs either a denser pin layout or a vacuum support, and both are cheaper to specify with the machine than to add later. Thickness range matters for the same reason, because a conveyor set for a nominal thickness will grip a thin panel unevenly and let a thick one ride high, and either condition produces errors that look like a head fault. A plant running several thicknesses should confirm the adjustment range and decide whether motorised width and height adjustment is worth the option price.
The comparison so far has assumed that verified figures exist. They often do not. Specification tables for high-speed machines are frequently written once for a family and reused across models, which means a document can carry the right model name at the top and a different model's numbers underneath. This is not necessarily deceptive — family tables are a normal engineering shortcut — but it makes the buyer responsible for confirming which row applies to which machine. A JUKI RX-7R vs RX-8 high speed chip shooter comparison built on an unverified table compares nothing, because both sides of the comparison may be describing the same machine.
Verification has three parts and each is cheap if done before the order and awkward afterwards: a document set that names the exact model and configuration, a test board that exposes the differences that matter, and an acceptance run that measures the result on the buyer's own product. None of them requires a laboratory. All of them require the buyer to ask for something specific rather than accepting a brochure, and the asking is easier when the request is framed as a standard purchasing step rather than as distrust.
The first document is a model-specific specification sheet that carries the full model designation, the head configuration, the feeder interface and the board handling arrangement as ordered — not as a family default. If the sheet shows a model name that differs from the quotation, or a parameter table whose column header names another model, that is the signal to ask for a corrected sheet before proceeding. The second document is the configuration list: exactly which heads, which feeder banks, which nozzle set, which conveyor and which options are included in the price. Family specification sheets describe the platform; the configuration list describes what will be delivered.
The third document is the verification protocol: how the machine's accuracy, repeatability and rate will be measured, with what board, at what sample size and by whom. Placement equipment performance standards such as IPC-9850 provide the vocabulary and the measurement approach for accuracy and repeatability, and referencing that standard in the protocol removes most of the ambiguity about how a claim will be tested. A supplier who can point to a specific clause is easier to hold to a number than one who describes general capability. The same three documents apply to the placement machines at either end of a line, and the sequence for testing them on site is set out in the site test sequence that proves a published rate on your boards.
The right test board is not the supplier's demonstration board and not the plant's easiest product. It is the product that sits closest to the boundary between the two machines — the one with the highest chip density, or the widest component spread, or the largest panel, whichever is the reason the plant is considering the upgrade. Running that board on both machines, or on the same machine in both configurations, produces a comparison with meaning. Running a comfortable board produces a demonstration.
The test should record four things: the panel time achieved, the pick error rate, the number of nozzle or feeder interventions required, and the first-article result. Those four separate the machine's rate from its operational cost, and the second is usually where the differences between two models of the same family actually appear. A machine that is marginally slower but needs no intervention across a shift may deliver more good boards than a faster one that stops twice. Recording interventions is the part most often skipped and the part most likely to change the decision.
Verification checklist for a high-speed placement comparison (what to ask for, and what settles each item) | ||
Item | Document or test | What settles it |
|---|---|---|
Model identity | Model-specific spec sheet matching the quotation | Model designation on the sheet equals the model in the contract |
Head and nozzle configuration | Configuration list | Nozzle count and types listed, not described as "standard" |
Effective feeder capacity | Bank layout plus width conversion | Your densest product fits at 80% or less of nominal |
Component range at rate | Test run on the smallest and largest parts | Error rate stable across a defined number of placements |
Board handling | Panel drawing against machine envelope | Your panel passes, including frame and thickness range |
Accuracy and repeatability | Protocol referencing IPC-9850 | Sample size, board and method agreed before the run |
Line interface | SMEMA handshake and transfer time | Automatic transfer with the neighbouring stations |
Operational cost | Intervention count during the test | Number of reloads and nozzle stops per shift |
An acceptance run should prove three things in this order: that the machine does what the configuration list says, that it does it on your board at the rate your capacity plan assumed, and that it can be handed over in a state the plant can maintain. The first is documentation plus a physical check of heads, feeders and options. The second is the timed run on the boundary product, with interventions recorded. The third is the program set, the parameter baseline and the training, which are the items most often reduced to a signature at the end of a long day.
Where the machine is an RX-8 and the reference for the comparison is the platform page rather than a contract, it is worth confirming that the page and the quotation describe the same thing before the timed run. The reference page used in this comparison, the RX-8 platform page that this comparison references, describes the platform's role and options; the contract describes the machine you will receive, and the two should be reconciled line by line. That reconciliation is not a formality: it is the step at which family-level claims become a specific obligation.
Every comparison needs one person who owns the figures. In practice the number that decides a purchase is often assembled by several people — process engineering supplies the placements per board, production supplies the availability, purchasing supplies the rates — and each contribution arrives with its own assumptions, none of which are written down. When the machine underperforms, the assumptions are reconstructed from memory and the discussion becomes personal rather than technical.
The fix is a one-page record that states each input, its source, the date it was taken and the person who supplied it. That page travels with the capital request and with the acceptance protocol, so that the same numbers are used for justification and for verification. It is a small discipline and it changes the character of the post-installation conversation, because the question becomes whether the measured rate matches the recorded assumption rather than whose memory is better. A JUKI RX-7R vs RX-8 high speed chip shooter decision supported by such a page can be audited; one supported by a brochure cannot, and the difference shows up precisely when the plant needs it least.
Want the comparison run on your own board rather than a demonstration panel?
Send the bill of materials, the panel drawing, the current cycle time for each station on the line, and the component sizes that sit closest to the boundary between the two machines. We will return a three-clock model of your line before and after the upgrade, an effective feeder capacity calculation for your densest product, and the test plan that would settle the RX-7R versus RX-8 question on your boards rather than on a specification table.
If the panel format or the component distribution is still in flux, say so and the model will be run for two scenarios with the assumptions written down, so that the decision can be re-checked once the product settles.
The output is deliberately not a recommendation of one model over the other. It is a statement of which machine is faster on your board, by how much, at which point the line stops benefiting, and what the neighbouring stations would need in order to release the gain. Where the honest answer is that neither machine will raise output on the current line, that is what the model will say, and the money is better spent elsewhere.
These are the questions that come up most often when a plant compares the two machines, answered in the same order the guide uses. Any figure below is illustrative and should be replaced with a verified value before it enters a business case.
No machine is faster in every case, because placement rate is a property of a board and a configuration as much as of a head. The two models sit in the same high-speed rotary family, and the situations where one pulls ahead are specific: boards dominated by very small chip components, panels that fit one machine's handling envelope but not the other's, and product lists that fill one machine's effective feeder bank but not the other's.
The practical approach is to stop asking which is faster in general and start asking which is faster on your boundary product, at your rate, with your feeder load. That question has an answer that can be measured in a day, and it is the only answer with commercial meaning. A JUKI RX-7R vs RX-8 high speed chip shooter comparison that produces a single winner for all boards has probably been built on specification tables rather than on a test.
It can when the board is dominated by chip components and its non-chip population is small and well behaved. It struggles when connectors, shields, odd forms or tall parts are a meaningful share of the cycle, because those parts force nozzle changes, slower placement modes or feeder types the high-speed bank does not carry. On such boards a single machine will place the chips quickly and then lose the advantage on a handful of parts.
The test is to compute the share of cycle time taken by the non-chip population. If it exceeds roughly a fifth of the total, a two-machine arrangement usually wins, with the high-speed machine carrying chips and a flexible machine carrying the rest. If it is below that, a single machine with a well-chosen nozzle set is simpler, cheaper to maintain and easier to schedule.
Only as much as the placement clock was limiting the line. If placement took three quarters of the line cycle, a large placement improvement can raise output substantially; if printing or reflow was already the longest clock, the improvement raises output by close to nothing and simply moves the queue. This is why a three-clock comparison should precede the purchase rather than follow it.
A useful discipline is to compute the panel time for printing, placement and reflow in seconds, name the longest, and state the margin to second place. If the margin is small, the line has no stable constraint and will behave erratically regardless of which machine is installed. In that case the highest-return investment is usually buffering rather than placement speed.
Plan for the largest panel your product list is likely to need over the machine's service life, not for the largest board you make today. Board sizes tend to grow as products consolidate, and a machine that fits today's panel can become a constraint when a customer merges two boards onto one panel. The cost of headroom at order time is small compared with the cost of a machine that cannot accept the next panel format.
At the same time, check the handling details rather than only the maximum dimension: thickness range, warpage tolerance, conveyor width adjustment, and whether the panel frame fits within the clamp area. A machine that accepts the dimension but not the warpage will produce placement errors that look like an accuracy problem and are actually a support problem.
It depends on whether the second machine removes cycle time from the constraint or merely adds a pass. If the high-speed machine is the constraint, moving the non-chip population to a second machine shortens the line cycle and the second machine pays. If reflow or printing is the constraint, the second machine adds handling, another first article and another program to maintain, and removes nothing from the line's rhythm.
The decision also has a maintenance dimension that is easy to underestimate: two machines means two program sets, two nozzle inventories and two sets of parameters to keep current. That overhead is manageable and it is real. Plants that run two mounters successfully usually have a single owner for both machines' programs, which is a staffing decision as much as an equipment one.
Ask for the measurement conditions first: the board pattern, the component mix, the feeder arrangement and the definition of the rate. A rate quoted without conditions is a ceiling, not a forecast. Then ask for a timed run on your own boundary product, with pick errors, nozzle interventions and feeder reloads recorded alongside the rate.
Reference a measurement standard in the protocol so that accuracy and repeatability are defined the same way on both sides; IPC-9850 is the usual vocabulary for placement equipment. Finally, compare the measured rate with the capacity plan's assumed rate. If the measured figure is below the assumption, the plan changes before the order rather than after the installation, which is the entire point of measuring.
Three things usually change and only one of them is welcome. The welcome change is that placement stops being the constraint, which allows the plant to see its real bottleneck for the first time. The unwelcome changes are a feeder reload rhythm that is faster than the material handling was designed for, and a nozzle condition regime that becomes visible because small-chip placement at speed is less forgiving than at low speed.
All three are manageable if they were anticipated. The plant that planned the reload schedule, staged a second cart and defined a nozzle inspection interval spends the first month tuning; the plant that did not spends it reacting. Recording interventions during that month — reloads, nozzle stops, first-article repeats — produces the baseline that every later improvement will be measured against.
