For product manufacturers, first-pass yield is perhaps the single most important indicator of production profi tability. The high cost of field recalls, the possible loss of reputation, and the growing need to meet legal and safety regulations make product quality and production reliability top priority. The cost of reject boards—fully populated with expensive components—is high enough but a low first-pass yield will also demand a whole department to test, trace faults, and repair boards. That is becoming more diffi cult with each generation of new, smaller components. The latest 01005 components, for example, are almost too small to see. With low first-pass yields, a large proportionof the valuable output will have to be scrapped.The most important factor for determining first-pass yield is the defects per million level of the pick-and-place machine. Machines from different manufacturers have widely different figures, with between 50 and 75 defects per million being common in the industry. The figure largely depends on the variation allowed by the placement cycle, from component pickup to placement on the solder pad. To ensure proper placement, each component needs to be inspected at the key points during placement. Machines have to correct for factors like offset angles during pickup, offset of solder pads, and board warpage. This is becoming increasingly difficult, since miniaturization is driving down the availability of substrate real estate for fiducials, which demands artwork recognition for board alignment, and vision systems (figure 1) with higher resolution. The trend towards finer-pitch components and conductive adhesives similarly requires higher placement accuracy and wellcontrolled placement forces.  Repeatable placement means that, if the pick-and-place variation is well within customer specifi cation limits, any drift in performance can be detected early, before it starts producing rejects. That gives the highest possible first-pass yield for day-to-day manufacture. And in the longer term, product manufacturers need to work on reducing variation in the production process as a whole to continually reducethe cost of placement (CoP). Reducing cost of placement Optimizing the placement cycle is fundamental to minimizing CoP. The time taken to actually place a component is the only productive time in the placement cycle. Time taken for board run-in/run-out and positioning, and for component pick and alignment, adds no value to the product. There are five major aspects to reducing cost of placement: eliminating wasted time, reducing operator intervention and maintenance effort, increasing output, and eliminating wasted boards. The first three are mainly down to the pick-and-place machine manufacturer. Waste time can be reduced by improving the efficiency of SMD placers. That calls for fast PCB transport, fast changeover modules like feeder trolleys, on-the-beam and on-the-fly vision systems, on-the-fly toolbit exchange, and smart placement optimizers. Operator intervention can be reduced by proactive placement quality control systems, and reducing the feeder replenishment actions by increasing the number of feeders per machine. Bulk feeders can, for example, bring cost savings of up to 15 percent by reducing mispick parts per million, operator intervention for feeder replenishment, storage space requirement, and tape waste, while improving feeder uptime. Reduced maintenance largely demands good initial mechanical design, and modular construction. Product manufacturers can have more influence over the last two aspects. Steadily increasing gross output comes from increasing speeds, increasing the number of parallel placement heads, and using modular (scalable) machines and servo-controlled placement heads. And the major influence product manufacturers can have on eliminating waste products is zero-defect quality. Alongside the pick-and-place machine design, reducing waste means improving the capability of the process as a whole. The objective of maximizing line uptime and placement efficiency is to optimize CoP on the assembly line, which is the key to calculating potential operational profi tability. Downtime is only one of many components that combineto determine CoP, defined as:  where the top line is the sum of all downtime, changeover, defects, and mispick costs together with ownership, space,labor and capital costs. Figure 2 gives a measure of actual cost of production of the major different placement concepts (multiple pickand- place, turret, collect-and-place and gantry systems). The graph highlights the benefi ts of a modular expansion path, which eliminates the ‘sawtooth’ characteristic of less flexible concepts requiring substantial incrementalinvestment for volume expansion.  Of course, no single placement machine can deliver the ideal combination of output, accuracy and flexibility for every production environment. There are some general guidelines, though. The PCB should not move during placement for better placement accuracy and no speed limitations. Modular systems are scalable and easy to upgrade In future. Off-linerepair and maintenance helps minimize uptime.Placement quality is intrinsically better with multiple placement robots working in parallel and indexed substrate transport. Multiple pick-and-place machines give the best fit between required and installed output with lowest CoP, relatively independent of output. Reducing placement variation Parallel placement leads to defects per million figures that are routinely below 10 – an industry benchmark. First-pass yields of above 98.8 percent can save more than 50,000 phones a year from scrap or rework. These results can cut rework costs by two thirds, which streamlines the whole product manufacturing process. This is even true with the newest miniature components like 01005 and 0201 types. These have incredibly narrow alignment windows, and it is becoming harder for these and bare dies or flip-chips to self align. A-Series machines can even place 01005 components in between QFP (Quad Flat Pack) leads that only allow for 15 microns free space, and at full speed. Such small components also crack more easily, and must be placed very delicately (figure 3). Components like large snap-fi t connectors need more force, which therefore has to be determined per individual component.  Modular machine design is essential. For the A-Series, it allows the three machines to share common user interface, software, feeder range, trolleys, trays and placement heads. A modular approach also allows a better balance between front and back end processes and minimizes machine overor under-capacity. Capacity can be redistributed between lines without even rearranging the footprint for easyhandling of changing application mixes (capacities). Modular design also eases the trend from ultra highvolume to high-mix manufacturing. The focus is now on changeover times, and this impacts board transport, nozzle and feeder exchange, and set-up programming. Offline trolley preparation and auto calibration can reduce product changeovers to typically less than six minutes. Production can continue while tapes are being replenished. A production run can even be transferred ‘on the fly’ to another machine without loss of accuracy or quality because machine-specificdetails are kept out of the production recipe. Product mixes and batch sizes can change quickly and, without fast product changeovers and high first-pass yield, this puts pressure on product margins. Prototypes may only have a single set of components, and a single faulty component can waste the whole set. All this requires perfectly functioning hardware, but also the software to tiethe system together, and tie it into the factory. Integrated manufacturing The various hardware and software modules have to work seamlessly. For the A-Series, they are tied together by the Assembléon Manufacturing Suite (AMS). AMS optimizes production processes at machine and line level, and supports all the key SMT manufacturing processes. The front end has an open architecture to integrate machines into existing factory and enterprise environments. AMS links to MES (manufacturing execution systems) and ERP (enterprise resource planning) systems, which helps cutdesign-to-production and ramp-up times. Offline board teaching, vision preparation and program creation and optimization together allow ‘single-click’product changeover. Offline feeder and trolley setup and component verifi cation ensure the correct parts are loadedat the correct positions. The line ramps up immediately to full speed and starts collecting performance related data like machine status, active program name, board count, component consumption, PPM data, and error and effi ciency data. The data can be used directly for performance monitoring, orviewed on a manufacturing information system (MIS). Advance warning of empty reels ensures that machines are kept running. There is also a record of exactly where and when components were placed, which simplifies fault finding and complies with customer and legal traceability requirements. All this, with remote operation and monitoring,brings the ‘lights off’ factory another step closer. Improving performance through amachine’s lifetime Since pick-and-place is the heart of the assembly process, Assembléon has broadened placement improvement to include the whole manufacturing process. That also addresses the problem of keeping machine parts operating competitively. It does this with its installed base solutions(IBS). Assembléon first performs a fundamental review of factory requirements and performance, and agrees on the key performance indicators (KPIs). It then addresses the whole manufacturing process to optimize both quality andcost of operation. The simple and modular construction and lowspeed robot movements (figure 4) translates to lower maintenance demands for the A-Series. IBS builds from this base, and brings continual process improvement to the manufacturing equipment itself, even several years after it has been installed. EM  ---------------------------------------------------------------------------------------------------------------------------------------------- Sequential placement vs.parallel placement Conventional placement is sequential – overhead gantries use up to four robots, each with around 20 heads. This approach has lowest initial costs but has fundamental disadvantages. All process steps are executed sequentially, leading to limited process times and related high speeds and accelerations. With machines using revolver heads or multiple nozzle heads, continuous individual component monitoring is impossible or very expensive. And with typically 40 or more pipettes/nozzles per machine, there is a relatively high risk of nozzle contamination and condition variation, and a higher risk of component loss. These types of errors tend to come in bursts, until the ‘special cause’ of variation is corrected. Component monitoring is particularly important because of the high acceleration and deceleration forces acting on each individual component that increase the risk of component shift or loss. That means more missed components, and so more variation in the placement process. For sequential machines, though, there is often no component position monitoring between component alignment and placement position (‘blind placement’). More recently, parallel placement systems like Assembléon’s A-Series have reduced placement variation. Parallel placement— with multiple heads placing com-ponents in parallel—reduces the accelerations components are subjected to, and gives extra time to accurately align each component. That makes for a very stable placement process. Keeping the speed of each head relatively low gives time to monitor and dynamically compensate the important process parameters for each component over the complete pick-toplace cycle. About the author Sjef van Gastel is the Manager of Advanced Development at AssembléonNetherlands BV. | 
