The top five productivity killers in the PCB industry

Some solutions

The route to maximizing factory floor productivity lies in a top-down approach that addresses the detailed operational points in the previous sections. Steps need to be taken in the following areas:  Data preparation  Manufacturing process simulation  Manufacturing process preparation  Manufacturing execution systems Technical solutions are needed for all of these challenges. And, crucially, reporting systems must be in place for engineers, line operators and manufacturing managers that identify the specific actions that need to be taken to improve performance. Moreover, the information must be timely enough to enable improvements to be made before the opportunity is lost and fresh problems appear elsewhere. Data preparation Attention to data preparation for both component model input and design data input is a mandatory first step: Component modelling – Manufacturers need to put in place accurate physical modelling of all the parts they plan to use on the line, including pincontacts for solder joints, integrated with the CAD data. This should comprise:  Consistent, CAD-library neutral, modelling of the parts to enable standardised DFM and process preparation functions downstream.  Normalized component off-set, rotations and polarity statements to a standard.The worldwide electronics industry has sales of $750 billion, two thirds of which is accounted for by PCB assembly. PCB manufacturing is characterised by an obsessive drive for increased productivity in the context of three significant industry drivers: Shorter product lifecycles – The pressure is on to develop better products and bring them to market before the competition does, at lower cost, while simultaneously developing the next generation product. Only five years ago, product lifecycles were measured in years; now they are measured in months, putting pressure on designers and manufacturers to accelerate the process of moving from prototype stage to high-volume manufacture. More complexity – Manufacturers are producing more complex, higher density designs with increased miniaturization and more sophisticated boards. A typical bill of materials (BOM) for a PCB assembly can now have thousands of parts in total, made up from hundreds of unique line items. The “bought-in” items – capacitors, resistors, diodes and so on – will each have one or more “alternative parts” to enable minimum BOM cost and maximum parts availability. More complex bills of materials (BOM) put a premium on increased component quality and better traceability. Outsourcing is growing fast – Shorter product lifecycles and increased complexity have forced OEMs to embrace outsourcing, now the fastest growing segment of the PCB industry. Electronics Manufacturing Service (EMS) companies accounted for 21% of the market in 2004, but their share will reach about 30% by 2008. The market overall will grow just 16% in that time. EMS providers offer lower prices, accelerated speed-to-market and better order-fulfilment performance because they leverage massive aggregated purchasing power derived from serving hundreds of different customers, and by consolidating their manufacturing assets and managing them to achieve minimum unit cost. EMS providers focus on their core competency of manufacturing and component procurement; OEMs are free to focus on the design and marketing of new products. These industry trends are well understood and have contributed to making PCB assembly one of the most competitive industries in the world. With pressure to cut costs, while simultaneously improving yields and speed to market, the search is on for those changes to factory floor operations that can improve competitiveness. Typically, 60-70% of invested fixed-asset capital in PCB assembly operations is locked up in the machines in the assembly lines. SMT assembly is especially capital intensive, for example, with single lines costing more than $1 million and the price is increasing. Hard pressed manufacturing plant managers are asking themselves how they can ensure that their invested capital delivers maximum productivity and competitiveness. The answer lies not only at the level of the individual machines, but also at the level of the complete line or factory-floor. PCB assemblers use many measurements of manufacturing performance from the productby-product specifics of cycle-time, line beat-rate and first-pass yield, to higher-level benchmarks such as “BOM conversion cost” and return on capital employed. Whatever Key Performance Indicators (KPIs) are used, the goal is to generate the maximum output of acceptablequality product from the available assembly lines, materials, fixtures and human resources available. This white paper examines the five key factory floor challenges that must be overcome by manufacturers who want to become productivity champions in the SMT assembly business:

  • Parts chaos
  • Inefficient line set up
  • Slower than optimal beat rates
  • Low machine peak performance
  • PCB/process combination is sub-optima
  1. Parts Chaos The first issue affecting productivity is that materials are not in the right place at the right time, ready for use on the assembly lines. Many believe that having complete coverage in the ERP or master stock control system of all BOMs to be assembled is enough. But the critical factor is to have the correct quantities of parts and materials available and installed on the machines at the exact time when needed. Verification of availability of component part numbers en masse does not prevent failure to manufacture due to non-availability of parts on the factory floor because:

i. Parts already committed to other set-ups – Components for assembly onto PCBs are typically handled in bulk – either in reels holding thousands of parts or in stacks of trays holding hundreds. If the same parts are needed simultaneously for two production orders, neither line can be set-up correctly. Multiplying the impact of this problem across the hundreds of component reels or trays present on a typical PCB factory floor, magnifies the risk of being unable to deliver the right quantities of parts to the lines for every production order, despite the fact that, in aggregate, the required total quantities of parts for the production orders matches the total quantities of parts in the master stock control system.

ii. Available parts cannot be found – Often, in large factories, the ERP systems do not track materials very accurately once they are released to the manufacturing floor. Key data – concerning the line the parts are allocated to, whether the set-ups they are committed to are still in production, and the exact quantity of parts that have been tied up in those set-ups – is frequently missing. While the available data shows the parts are available to start manufacturing, they frequently cannot be located. Unnecessary delays result at the start of a production run while “expediters” are frantically searching for missing material. Equally, and due to the same lack of visibility of which parts are where, often parts can be delivered unnecessarily to a line, to support set-up, when actually a sufficient supply of those parts is already loaded on the line, leftover from a previous production order. These tracking inaccuracies result in unnecessary increases in factory floor inventory cost. due to miss-picks or nozzle failure before placement. By and large, this forces PCB manufacturers towards overestimating

iii. Parts in quarantine – Another factor affecting parts availability at the pick-and-place machine is the sensitivity of some categories of components to exposure to normal atmospheric conditions on the factory floor. Sometimes components are affected by atmospheric humidity after unpacking from sealed containers and, after just a few hours, they must be baked in an oven to remove moisture from the component bodies. Discontinuities brought about by oven-baking cycles mean that certain parts go through cycles of being “available” and “not-available” for assembly, even though they are “in stock” all the time.

iv. Inaccurate stock control – Parts stock availability held in the ERP system is frequently inaccurate as a result of unrecorded wastage. When components are returned to the warehouse following use on the factory floor, gathering an accurate picture of how many parts remain on the reel is problematic. Should production managers simply take the starting quantity and deduct the number of placements defined on the BOM? Probably not, because it ignores the parts lost by the machine stock levels, leading to unexpected stock-outs on the factory floor in subsequent production orders. Expensive (and unplanned) production shutdowns like these create the need to purge the materials from the lines for subsequent orders and urgent parts purchasing to correct shortfalls. Inaccurate stock control also forces the time consuming and costly practise of sitewide inventory audits, when manufacturing is essentially put on hold while records in the ERP system are manually synchronized with the reality of the manufacturing floor.

v. stock levels, leading to unexpected stock-outs on the factory floor in subsequent production orders. Expensive (and unplanned) production shutdowns like these create the need to purge the materials from the lines for subsequent orders and urgent parts purchasing to correct shortfalls. Inaccurate stock control also forces the time consuming and costly practise of sitewide inventory audits, when manufacturing is essentially put on hold while records in the ERP system are manually synchronized with the reality of the manufacturing floor.

  1. Inefficient line set up An efficient SMT assembly line depends on the ability to coordinate hundreds of set-up variables simultaneously. If any aspect of the line set up is incorrect, poor quality output is the result. There are several common reasons for slow line set-up and debug:

i. Set-up instructions do not match machine programs – In many cases, the engineering data arriving on the lines comes from multiple, disconnected data flows. The kitting list for each machine is driven from the BOM in the ERP system, yet it does not take into account the BOM-splitting and balancing decisions taken by the machine programmers. CAM systems used for generating machine programs are often working from a different database than the CAM system used to generate the factoryfloor traveller. And CAM systems used to program AOI machines are different to systems used to program the pick and place machines. The fragmentation of data flows can be extensive; each point of disconnect between engineering databases offers another opportunity to generate unsynchronized data or instructions for different parts of the assembly lines. All set-up errors have to be either eliminated at source, by design, or discovered at the “first-off” stage and eliminated by editing set-up instructions while the line is down and unproductive

ii. Parts-data on the machines is missing or incorrect – Every SMT pick and place machine, AOI machine and in-circuit tester needs a library of data to describe key characteristics of every component to be assembled, inspected or tested. Only when the component library of the machine is filled with data describing the components for the production order can the machine do its job. Every new part loaded onto the factory floor means that the library data for that part must be entered into the machines and verified. Once created, the data must also be managed properly as any changes that are made can potentially result in unnecessary down time if not performed by a qualified operator. Without a controlled and centralised solution to manage the machine-level component data, the data must be painstakingly entered into multiple machines, causing unnecessary downtime and a high risk of data nconsistencies between multiple machines.

iii. Full off-line set up is not achieved – Many manufacturers are incapable of offline component loading and set up verification. This forces line strip-down and set up to be undertaken before manufacturing can begin, leading to wasteful downtime. No doubt total feeder inventory cost can be minimized by performing set-up on-line, but a high price is paid in terms of lost line output and machine utilization.

iv. Set-up is incorrect at first-off stage – If overall line set up is not verified in parallel with inventory checking at the outset, errors must be detected at the time of producing the first-off. This is the most expensive way to find and eliminate a set up error, since the elapsed time between creating the error and detecting is maximized. Multiply the error/detect/fix opportunities according to the number of feeders, machines, programs, and the opportunity for escalating set-up debug time becomes clear, as compared to verifying every aspect of the set up as it is carried out. Once the first off stage is complete and the line is in full production, it is also vital that errors are avoided when new parts are put on a machine to replenish an exhausted feeder. Worst-case, incorrectly placed parts will be detected after assembly of the full batch, at the inspection or test stage. Such repairs have maximum cost and impact on the overall productivity of the plant.

v. Failure to exploit existing machine setups – The best way to minimize set up downtime is to eliminate the need to strip lines down and set them up again between production orders. Because of the complexity of managing the huge variety of components, feeders, feeder positions, component quantities, and the factors which affect an optimized set-up for minimum cycle time, most manufacturers strip all the feeders and components from the lines between production orders. This maintains control, but dramatically reduces productivity. By analysing production orders in advance and identifying product groups that can share the same set-up (or majority of the set up) on an assembly line without sacrificing beat rate to an unacceptable degree, massive savings in downtime can be achieved. Using product-grouping techniques delivers significant productivity improvements in High Mix/Low-to-Medium Volume operations where changeovers are one of the major contributors to line downtime.

vi. Failure to anticipate parts replenishment requirements – In high-volume, low-mix manufacturing environments, lack of advance for an accurate simulation-based approach to programming the line as a whole. ii. Machine programming is not based on full kinematic simulation – If the line-level simulation and programming (balancing) is separated from the machine programming, there will be conflict between the two; the balancing depends on accurate information about individual cycle times, and the machine programming may generate a different machine cycle time to that assumed by the line balancing function. The key is very accurate simulation of every machine’s configuration (feeders, nozzles, …) and its motion kinematics. Without accuracy in machine cycle time simulation, not only will individual machine performance suffer, but also the line overall will not be balanced for optimum overall output. visibility of the need to replenish parts on the line is the single most important cause of downtime. The worst case occurs when all of the components in a feeder are exhausted but it comes as a surprise to the line operator (who has to supervise hundreds of feeders simultaneously). This forces the line down while the feeder is removed, a new reel loaded (assuming it is at hand), and the feeder reloaded onto the machine.

  1. Slower than optimal beat rates Once the lines have been set up, production settles into its repeatable rhythm, with assembled PCBs coming off the line at a fixed frequency determined by the line balance, machine capabilities, and the level of optimization embedded in the product-specific machine programs themselves. At this point, productivity is affected in an expensive yet invisible way, if lines are not programmed to run at the maximum possible beat-rate. This can happen for several reasons: i. Simulation, BOM splitting/balancing, and machine programming are not performed at the full line level – Individual machines can be programmed to an optimal level, but if a full-line approach is not taken to the programming task, based on a complete kinematic simulation of all the machines that make up the line, overall performance suffers, primarily caused by machine workload imbalances. The cycle time, or beat rate, of the line is determined by the slowest machine in the line, emphasizing the needfor an accurate simulation-based approach to programming the line as a whole.

ii. Machine programming is not based on full kinematic simulation – If the line-level simulation and programming (balancing) is separated from the machine programming, there will be conflict between the two; the balancing depends on accurate information about individual cycle times, and the machine programming may generate a different machine cycle time to that assumed by the line balancing function. The key is very accurate simulation of every machine’s configuration (feeders, nozzles, …) and its motion kinematics. Without accuracy in machine cycle time simulation, not only will individual machine performance suffer, but also the line overall will not be balanced for optimum overall output.

iii. Machine-level parts data is not programmed for optimum handling performance – The parts-data used by each machine defines how to handle the components: at what speed, with which nozzle, how long should the various dwell times be, what offsets should apply to the pickup point and so on. Completing the first-off is enough to verify that the product is assembled correctly, but this does not expose any low assembly speed effects due to sub-optimal handling instructions embedded into the parts data library of the machine. An operator will sometimes choose to reduce the placement speed of a component to ensure assembly, often masking maintenance issues that should be addressed while greatly reducing the overall productivity of the line. As with the optimization of the machine programs themselves, without access to detailed performance data it is virtually impossible for humans to identify these effects; and without detection they cannot be corrected.

  1. Low machine peak performance With investments in lines running to millions of dollars, clearly machines should be maintained to perform at maximum productivity for the maximum time. However, there are many aspects of machine condition that have an insidious effect on pulling down overall performance.
  1. Nozzle vacuum pressure – If this is out of spec., it causes components to be dropped in transit between the pick-up point and their position on the PCB.
  2. Sticky nozzle vacuum switching – If the vacuum switch is sticky it leads to nozzleskips. To pick the components from the feeder without error requires positive and fast switching of the vacuum supply to the nozzles. The same applies to the placement; slow or imprecise switching of the vacuum causes imprecise pick-up or placement.
  3. Worn feeders – This leads to high misspick rates. Component feeders are mechanical indexing devices which wear over time. As the mechanism wears with normal use, the accuracy of presenting the component for pickup declines, leading to failure to pick correctly, which wastes components and cycle time. iv. Poor maintenance instructions – SMT lines place components at rates of tens or hundreds of thousands of parts per hour. This lightning machine speed makes it difficult to observe declining performance. Miss-picks happen too quickly to be seen, but a delay of a few milliseconds on a repeating function leads to detuned performance. Without accurate and timely notification of where the performance drop-offs are, line operators and maintenance personnel have little chance of taking the right action to raise performance.
  4. Poor maintenance instructions – SMT lines place components at rates of tens or hundreds of thousands of parts per hour. This lightning machine speed makes it difficult to observe declining performance. Miss-picks happen too quickly to be seen, but a delay of a few milliseconds on a repeating function leads to detuned performance. Without accurate and timely notification of where the performance drop-offs are, line operators and maintenance personnel have little chance of taking the right action to raise performance.
  5. lightning machine speed makes it difficult to observe declining performance. Miss-picks happen too quickly to be seen, but a delay of a few milliseconds on a repeating function leads to detuned performance. Without accurate and timely notification of where the performance drop-offs are, line operators and maintenance personnel have little chance of taking the right action to raise performance.
  1. PCB/process combination is sub-optimal PCBs can be designed to be assemblyprocess friendly or process-hostile. Most PCBs can ultimately be assembled, but higher costs than necessary due to sub-optimal design, rework levels and line efficiencies vary as a result of design features such as:

i. The PCB is not machine- or line-friendly – The fiducials are hidden, components conflict with conveyors, assembly-panel design is not optimization-friendly. Design constraints such as component distribution on the board, or variety on the BOM, is such that one type of machine cannot achieve a low placement cost and this does not become visible until running the product on the line.

ii. Solder-stencil design leads to sub-optimal solder joints – This results in high rework. The primary objective of assembly is to create reliable solder-joints. Apart from good control on the soldering process, the combination of component pin, pad-pattern and solder-stencil aperture must be optimized to give the process the best chance of achieving joints that are within acceptable tolerances (typically measured in tens of poor joints, per million manufactured).

iii. PCB design layout encourages bow and twist – The panels of PCBs loaded onto the line for assembly should be perfectly flat, so as to avoid conveyor “hang-ups” and processing errors in the machines. By designing the PCB with an even distribution of copper in all axes, the tendency of the PCB to bow and twist during processing will be minimized.

iv. Pad/track patterns encourage tombstoning during reflow – With the trend towards smaller passive chip-components, such as the 0201 packages now being handled in volume, the design of pad and track patterns to allow equal heat sinking effects on either side of the component is of increasing importance. As the components get lighter, the effect of surface tension effects during reflow become more important; if one side reflows before the other, surface tension can cause the dry side of the joint to lift, causing the “tombstone” effect.

Some solutions

The route to maximizing factory floor productivity lies in a top-down approach that addresses the detailed operational points in the previous sections. Steps need to be taken in the following areas:  Data preparation  Manufacturing process simulation  Manufacturing process preparation  Manufacturing execution systems Technical solutions are needed for all of these challenges. And, crucially, reporting systems must be in place for engineers, line operators and manufacturing managers that identify the specific actions that need to be taken to improve performance. Moreover, the information must be timely enough to enable improvements to be made before the opportunity is lost and fresh problems appear elsewhere. Data preparation Attention to data preparation for both component model input and design data input is a mandatory first step: Component modelling – Manufacturers need to put in place accurate physical modelling of all the parts they plan to use on the line, including pincontacts for solder joints, integrated with the CAD data. This should comprise:  Consistent, CAD-library neutral, modelling of the parts to enable standardised DFM and process preparation functions downstream.  Normalized component off-set, rotations and polarity statements to a standard.

Tags:
,
No Comments

Post A Comment

X