Process Preparation for World-class Competition

Estimated reading time: 28 minutes

Process preparation is not an optional part of printed circuit board assembly (PCBA) manufacturing. It is mandatory and critical for high-quality production performed profitably. The tools and methods used in process preparation will determine if the goal for introducing a new product to the production line can be achieved: Right first time production, right away. No exceptions.

Let's start by defining what we mean by a "world class" process preparation tool. The PCBA world is not static by any means. Customers demand larger varieties of more complex products, all of which have to be delivered with perfect quality, on time, and at a low price. All these factors lead potentially to instability of an operation.

Process preparation is there to bring order to the chaos. Process preparation systems should be able to cope with all aspects and requirements of production--all machines, all processes--without mistakes and without confusion. Only in this way can we scale the flexibility and responsiveness of the production operation in line with customer and market expectations.

Process preparation then is the critical link between the completion of the design and the start of physical product realization. The gap between these two processes can be quite wide, geographically and organizationally, especially where external contract manufacturing is used.

One of the biggest challenges in process preparation is to understand the actual product that will be built, to know the processes that will be needed, and to assess and understand the cost of making the product. This requires communication, project skills, process and resource information, and a lot of product information.

Objectives of Process Preparation

1. Minimize Time-to-market

For leading companies in the market, the significance of fast and accurate process preparation is the ability to continuously bring new "hit" products into the market ahead of competitors without the overhead of disruption to other product schedules and the threat to quality and delivery. The short time from when an innovative product hits the market is the key time for a company to make profit. Being late is not a good business option!

2. Right First Time Setup

It is a good target to have a fast process, all very well unless something goes wrong. Unexpected issues can cause the most loss. Mistakes and inconsistencies in the set-up of processes is a key cause of uncontrolled, unplanned delay. There is no tolerance any longer for mistakes. The key is to bring as much of the preparation into an off-line specialist simulation tool in which all aspects of machine and process performance can be managed, checked and confirmed. This can eliminate the need for live on-line trials reducing very significantly the disruption of production as new products are introduced. It also eliminates the need for specialist new product introduction platforms, eliminating the follow-on step from NPI to main production. Perhaps even no special treatment for new products is necessary if the process preparation can be managed perfectly.

3. Accurate Quotation for Product

A strong additional benefit of process preparation for contract manufacturers is in the quotation process. The price of producing a product is based on the work and cost to produce the product. The variable element in pricing from the risks. Too low a quotation with unexpected issues then occurring in production can ruin the profitability. Too high a quotation, trying to cover the cost of any risk, and the competitive edge is lost. The key is simply to eliminate as much of the risk as possible, no surprises.

4. Agility of Line Transfer

Process preparation is not only about the introduction of new products. There are other elements which are a continuous challenge. What happens when a product has to be moved from line to line or site-to-site. The machines on each line may be different models, maybe even different platforms, with different configurations. How much work is it to transfer the product information from one line configuration to another? How many mistakes can be made in the translation of all of the parameters and libraries?

The work required for product transfers and risk it carries represents a high barrier to an agile operation. An agile operation will want to perform product transfers more often, since as customer demand fluctuates, different line configurations may become more efficient and productive. Moving production to match supply and demand ensures that high levels of asset utilization are maintained.Figure 1: The ability for a process preparation solution to facilitate easy line or site transfer is key to utilization and productivity.

5. Manages Changes to the Product

There are also changes in the product. These are engineering changes, sometimes temporary sometimes permanent, both of which must be managed effectively. Even minor changes in the product can lead to a major change in production processes. For example the supply-form of a component may change from reel to tray which causes placements of components to move between machines, throwing off the line balancing. To keep an operation running at maximum potential, close management of the engineering data and the ability to re-engineer the environment without artificial barriers, is essential.6. Support Work Order Creation Per Requirements

The engineering platform created by the process preparation tool is an essential part of the execution process also. Information created by process preparation defines and governs every action needed to make sure that the product is built according to the desired specification. Work-orders are created for execution. Engineering information in the form of machine programs, work-instructions and test parameters can be attached to each work-order at each process to describe the specific work to be done for the specific version and instance of the product. Having this information enables the enforcement of each operation against what is required. This is the origin of compliance and conformance.

Components of Process Preparation

Process preparation as a whole is actually made up of numerous processes. This is the nature of electronics PCBA. Each of these processes is quite different in nature, and within each process there can be many variations depending for example on which vendor of SMT machines is used. In most companies today, the process preparation for each of these types of processes is done separately.

In many cases introducing a new product to production is like the process preparation Olympics. No one wants to be the bottleneck. Unlike the Olympics, it is not the first one to the post that is the winner, it is the last one to the post that defines the preparation time. It is more like the weakest link in the chain, both in terms of time and quality. This actually is quite a problem and a significant source of waste in m any PCBA operations.Figure 2: Data preparation tasks are often duplicated by each engineering team creating waste.

Though the processes are all different, the baseline engineering data about the product is the same. Surely it would be better for all teams to work together on a single product data model? However, having different groups of people using different tools can severely limit the ability to collaborate. The end result is many people doing the same thing in slightly different ways in a parallel fashion, with the likelihood of getting slightly different results. To have one tool to perform all of these individual process preparation tasks based on a single common data-set would mean that instead of everyone taking in raw data about the product and processing it separately, the data can be processed one time correctly and then used across the various processes. This reduces the engineering effort, since the common part of process preparation, the data preparation is done just one time instead of many. This can also cut lead time, since the team can work together. The product data created as a result is then a firm base bringing consistency across the operation. A common tool can turn duplicate parallel tasks into a common task!Components of the Engineering Process

What is the definition of the common engineering task as part of process preparation? The answer is in the steps that can be taken to create the single data model, the Data Preparation. As we define these steps, we will identify world-class practices to reduce time and effort as well as eliminating potential mistakes.

1. Transfer Design Data into Production

The first step is to get the available product data from design. What are the challenges? There are many variations of native CAD data formats. To be able to read-in any data format that may be provided is quite a challenge.

Many process preparation tools in use today are unable to utilize the native ECAD formats, which has led to alternative data formats: so-called CAM data exports, such as Gerber files. These formats however are not designed to hold anything other than very simple information about the product design. Using these formats with limited information then requires an extensive reverse engineering of a sample product in order to gather enough useful data to be able to complete the process preparation stages. This is a very significant bottleneck to the process and inevitably results in frequent mistakes being made.

2. ODB++ Data

The better alternative is to utilize the standard industry format ODB++ which can be easily exported from almost any CAD system and contains all of the required information for assembly process preparation. It is a quick and easy transfer completed in a few minutes as compared to days of need-less engineering work recreating what was already known to the designer. This is world-class.Figure 3: Data Preparation using industry standard ODB++ transfer from ECAD creates a more streamline and error process.3. Reverse Engineering

However, a complete process preparation system will also need to cope with the "dumb" formats of information such as Gerber files, should these be the only ones that are made available to the manufacturing operation. Tools that can assist with the reverse engineering process can make a significant difference in the time needed with the minimum number of mistakes. Using the reverse engineering method should only ever be the last resort. Though this method continues to be the norm for many companies, it is such a waste of time and resource, not at all world-class.

4. Bill of Material

Once the product model information has been derived from the design tools, it needs to be qualified against the manufacturing bill of materials. One PCB design can service many products and variations, so it is important to be able to differentiate between these variants to ensure that the product is built correctly. It is necessary to therefore use the bill of materials to define which placements and other components should exist for any particular variant, and what the actual materials used will be.

The bill of materials is based on the PCB design data, with each part number linked to the location on the board by use of a reference designator. However, there are often discrepancies in the naming conventions and content of the BOM. This can happen due to the fact that the bill of materials is managed in the ERP system. ERP Systems are driven by materials represented by part numbers and quantities. The individual component references are often represented only as comments, all references sometimes listed squeezed together into a single field. This leads to significant corruption potential of the reference designators. This is a critical issue since the reference designator is the key piece of information to identify which component positions need to be placed and which precise component part number is location on each and every reference designator.

Therefore the merging of the ERP based bill of materials with the original design data can be quite a challenge requiring a sophisticated tool, able to make decisions about matching and exposing any discrepancies, which often happen. For example, reference designators in the ERP BOM which do not appear in the design, or quantities of components used of a specific part number in the ERP BOM may not match the count of references listed in the design data. These are all things that must be resolved with 100% certainly before moving forward. This process is still done manually in many cases, perhaps with the aid of a spread sheet, and consumes significant time and resources, and carries with it a high risk of errors. Not world-class at all.

Having the design and automated BOM import tools together makes an order of magnitude reduction at minimum of this cost. This is world-class. This is also one of the main strengths of the Valor MSS Process Preparation solution.Figure 4: Example of BOM Merge.5. Simulation

Once the CAD and BOM are imported and merged, is the result something that looks like the actual product? We now know what components are on the product, where they all are and what part numbers they should be. What does it look like though, physically? The simulation of the PCB-A used in Valor MSS Process Preparation is itself a great tool to let the key engineers recognize issues visually that raw data cannot describe. This raw data is what will be sent to the machines for visual recognition during execution, and for all other processes. It needs to be right. Visual tools are very useful here to be able to make sense of the product, before any sample of the product is physically available.

To visualize each part, we need to understand what the part actually is, in terms of type, size, shape, leads and other key physical attributes. These are defined as shapes. Using shapes, we can build up from all components what the finished PCB will look like.

Creating and maintaining shape libraries has two major challenges. Any new part numbers that are added to the design will need to be identified. The shape of a new part number could be the same as an existing part number, and so it can be mapped, or, it may be a completely new and unknown shape.

The limitation of not knowing the actual shape is critical for new products using materials which are new and may not yet be physically on site. These shapes are required for the engineering process, for SMT, test and inspection as well as confirmation of things in the basic data such as rotations. Not being able to know the shape data can delay many of the process preparation stages.

The second challenge, going one level deeper, is the availability of different suppliers for a specific part number, each of which may have a slightly different shape or size. Identifying the representative or "worst case" shape is a key decision.

This is a significant amount of work! No way anyone would want to duplicate this over many different tools, and yet this is the reality for many companies today. We find that many companies are creating and maintaining this difficult data in duplicate ways in many different pieces of software. This represents a huge waste of time and often results in inconsistent treatment of placements on different machines for the same parts.Figure 5: The VPL is a key element to the accuracy of the product data model.

These issues lead to the need for a sophisticated central shape management library tool, and, access to a significant and reliable database of shape information. The Valor Parts Library (VPL) can be an important asset in the race to reduce the lead-time of process preparation. Of course, the accurate verified information is available for all areas of process preparation, and is available at any time.The issues we have described are resolved in the central process preparation tool together with a subscription to the VPL. This unique integration can be a simple cost-effective way to un-block the process for new product introduction.

6. Component Rotation Management

After merging the accurate shapes of the components into the data model, still one more crucial issue plagues nearly every company doing SMT placement today; i.e. component rotation, needs to be addressed.

Component rotation is a pervasive and tricky issue because it is a relative measurement, relative to the original design macro rotation of 0 degrees. This is often a problem, since a standard definition of rotation in design is rarely successfully achieved, and even more rarely communicated. Even within the same design team and even within the same shape of components, there can be cases where the rotation of zero degrees of one part number is different to that of another. Just looking at the rotation, it is often impossible to know the actual correct orientation of the component. Correct orientation is achieved on the board when pin 1 of the component is actually located on the designed pin 1 of the pad stack or component foot print.At the time of importing the design data, automated adjustments can be made. For example, as a first step, a rule can be applied based on the quadrant that pin one or the first electrode of a device appears in. This rule will set consistent normalised rotations.

Once normalized in this way, the rotations can then be adjusted to a defined standard for that shape using a simple look-up table. Very quickly then all the correct rotations can be neutralized in the product data model.

The seriousness of an error in rotation or offset is such however that positions and rotations must always be 100% checked for new products, no matter what. The standard approach in the industry for a new product introduction to SMT is to run a special board covered in a film of sticky tape in lieu of solder paste, through the SMT placement machine. People will then study this board through a magnifying glass to see whether in fact all of the rotations and positions of all of the components are correct. Corrections can be made to the program and the process repeated until perfect.

This can go on for a while, consuming a lot of line time and materials which cannot then be re-used. The result of this even cannot be fully trusted since having gone through the machine programming software to make program "tweaks" on the shop floor, compensating errors may have been introduced, potentially causing serious quality issues later on. Definitely not world class.

Instead of this, the world-class data preparation tool, Valor MSS Data Preparation, has a built in simulator known as "Virtual Sticky Tape" which shows exactly how each component will look when placed on the surface of the PCB. This is possible only due to the fact that the actual design data and actual shape data are available. When overlaying these two data layers for each component position the result is a significant reduction of time and cost. Quality can also further be improved. Often 180 degree errors on symmetrical polarized parts are missed in visual checks of the physical board, as polarity marks are sometimes obscured.7. DFM/DFA

By this step we have a complete and accurate data model of the product for production. Before we start into the individual process preparation elements for the different kinds of processes on the shop-floor, we first have the opportunity to do some automated tests to find any areas of interest in the design itself. There can be many reasons why we might find issues in manufacturing as a result of a design which may not have a good understanding the needs and limitations for the manufacturing processes. It can be as simple as the actual materials being larger than expected, the leads being different shapes than expected, or the test points being in inaccessible places, etc. There are literally hundreds of categories where compromises may have been made.

The key point is that knowing these issues in advanced gives us the opportunity to plan countermeasures into the operation. If we didn't do this, we run the risk of getting some nasty surprises when running the operation for the first time. Some of these may result in corrective actions on every product produced and the risk of quality issues.

There is the left-shift rule that states that knowing about a problem in advance will reduce the cost of correction by at least 10 times as compared to finding it at the next process. I would say that in manufacturing this can actually be 100 times, 1,000 times, or much more in some cases. An undetected quality issue that makes its way into the market can have fatal consequences as far as the product life and business returns are concerned.Figure 6: Pin-to-pad fit problems can create marginal solder joints which can reduce the product quality.For the contract manufacturer, there is even more significance to this process, since here is the opportunity to use the best information about the product and production processes and to be able to more accurately make the quotation for producing the product to the customer. More accurate quotations based on lower risk reduce business volatility and are more competitive.

For the OEM manufacture, the same principle of DFA is an opportunity to feed issues back to the design team with a view of improvement for the next generation of design, based on actual manufacturing capabilities and limitations.Figure 7: DFM analysis is crucial to protect product quality and the cost of manufacturing.

8. SMT Process Definition

The big news going forward in the process preparation work flow is that now each of these processes is founded on our central consistent known good product data model. One common data model can now be used in many different ways.

The first key challenge of SMT programming is to understand the capability of the machines. There are so many different machine platforms, types, models and variants from many different suppliers. Each are designed to bring slightly different advantages. These machines operate using many different software platforms also. It can be quite a challenge to effectively distribute the placement workload over the different machines in a line. There are often compromises and assumptions that are made to make the job manageable, for example a pre-set decision of which line is to be used for each product.

With a world-class process preparation system however, the capabilities of all machines are modelled in software. This enables the software to automatically and accurately distribute the placement workload over any line configuration made up of any machines from any vendor. Whether a product is run on line A or line B should be based on using the best machine for the job, not limited to which machines include required library data or not.

Doing this so easily and quickly also opens up the flexibility of choice. No more assumptions or compromises. With Valor MSS Process Preparation we have the freedom to model several line configurations and find the best machines for the job. The engineering time required to perform mindless repetitive duplicated manual operations can now be re-focussed using the tools to make real improvements in shop-floor productivity.

There is always a large overlap in SMT machine capabilities, meaning that most parts can be placed on a choice of machines in the line. This is where program optimization and line balancing come in. Of course, within a single machine vendor's platform, there will be software provided with the machine to do this. In these cases it can be just a matter of sending the fully prepared data to the machine vendor's software, including of course the details of the shape library.

One key element to do this is the automated rule based generation of what would normally be hand-madesettings for the components to run on the machines. These settings are the personal parts settings for example, lighting, camera, nozzle, things which are related to the treatment of specific types and shapes of material on the specific machines. These can be worked out automatically by our unique auto-generation function which generates these parameters based on defined engineering rules. This leads to consistency and accuracy, even taking care of new parts automatically as they are introduced.Figure 8: Valor MSS Production Planning works in concert with Process Preparation for setup and change-over optimization. When there are mixed platforms in the line, including the case of different platforms from the same vendor, there is some very significant engineering work required for line balancing. First, the data has to be prepared into multiple formats. Then, when each piece of vendor's software has done its work, the estimated working times of each machine may be quite different. Parts assignment will need to be moved manually from machine to machine and the optimisation repeated, often several times, until a good balance is achieved.

In our world-class process preparation system however, the machine optimization and line balancing is done centrally and automatically. Just press a button and a week's worth of work is performed as you drink a cup of coffee. In an increasing number of cases, the Valor MSS program optimization utilises the machine vendors own software to provide the most accurate and consistent performance. Others are optimized using Valor MSS advanced optimization algorithms. Many iterative line balance cycles are performed automatically until the line is as perfect as it can be.

9. SMT Setup/Change-over Optimization

Much of the time lost in SMT production is changing-over the machines from one model to another. This is a much more an everyday occurrence on every line now as the market goes towards higher product mix environments.

The major contributor to this changeover loss time is the time that it takes to set up the materials. Clearly, reducing the number of material changes will reduce the changeover time. Making some or all of the feeder positions common between products can reduce the change-over time to almost zero. The trade off however is that for each individual product, it will not be possible to optimise the feeders as well as if any choice was available. The best optimization of several products together then is the goal. The Valor MSS process preparation tool can consider the optimization of many products run in any combination to find the best possible result in even the highest product mix scenarios.

This particular tool is also put to use in a wider scope integrated with the Valor MSS Production Planning tool, which uniquely allows schedule optimization at the same time as creating common feeder set-ups.

10. Test and Inspection Engineering

In parallel to the SMT process preparation, the test processes can also be prepared from our central product data. For electrical testing, the essential issue is finding the best ways possible to access all circuit nodes using test points or other exposed contacts on the PCB. This heavily draws from the design data to understand the position and nature of every feature of the circuit. Once the best accessibility has been found, the process preparation system chooses the most appropriate probe or pin to be used with respect to the particular tester and fixture parameters.

The logic for flying probe is similar to in circuit test except that accessibility has to be calculated in three dimensions as the probes go around the product. In each case, the process preparation system creates the necessary output data formats to completely describe the test process to the respective machines. Again, many days of work has been replaced with a few minutes.

A similar process is provided for optical inspection machines. Depending on where a machine is placed in the line, the system knows which components are placed on the PCB at that point, or none in the case of solder paste inspection. Again, utilising the design data, being able to simulate the optical views of the PCB based on the each machine's inspection parameters, and knowing the physical shape attributes of the parts themselves, process preparation can calculate the required inspection program parameters.

In the process of creating all of the various forms of test programs, we can get a good measure of the testability of the PCB. We can check whether every element of the PCB can be tested, or whether there are untestable risks.Figure 9: Test Engineering and Testability Analysis are always based the common product data model.If any of the risks are deemed critical, for example an important areas or feature of the PCB cannot be tested, then processes in production can be made or adjusted to compensate. Using our left-shift rule, it is better to put these countermeasures into place and plan in advance rather than face the consequences of quality issues later in the production process or worse still, in the market with the customer.

This information is also one example of data that can be fed back to design so that with the next iteration of this PCB, or the next product, the testability risks can be designed out. This is a key feature of the Mentor complete design through manufacturing flow.

11. Process Documentation

Once again, another lane of functionality and value coming from our prepared single engineering product data is shop floor documentation.

Documentation, whether delivered electronically through a web-browser or on paper is an essential part of any operation where a manual "touch" will be required. Some documentation is static, i.e. a fixed image provided as a reference for any particular operation. More interesting is the dynamic documentation that follows the working process, that can zoom in to features of interest providing a much larger and contextual amount of value.

One key issue with documentation is that it is said that it is not needed beyond the third or fourth usage. This can be true--once an operation is learned, it may not be necessary to follow the document any more, though standards tell us that documents should always be visible.The key issue is what happens when something changes? The learned operation must be un-learned, even if there is only a small and subtle change. This is the major opportunity for human error. Having documentation which dynamically highlights recent changes in the operation is a very critical trigger to avoid such mistakes, and can greatly improve the level of quality of manual operations.

The major usage of documentation is work instructions. For assembly work-stations, the preparation of work instructions involves the assignment and sequencing of hand operations, to ensure that all placements are visible and accessible for the human hand at the time of execution. The order in which components can be placed is driven from this logic, and work-instructions created. The work instructions need to show precisely yet simply the method for correct assembly. These can be used for PCBA assembly as well as final product assembly.

For repair stations, the documentation is triggered dynamically by a test failure report. in this case, the appropriate area of the product needs to be highlighted quickly and automatically so that the area of suspected failure can quickly be found and compared to what it should be. Digging deeper dynamically in to the product information about the PCBA at the repair station is essential for accurate and timely repair.

Real-life Benefits of World-class Process Preparation

For some people, what we have discussed so far represents a huge paradigm change. For others, it is an aspiration that is waiting to be fulfilled. The key decision maker is return on investment.

What are some of the key benefits that we can get from our world-class process preparation tool? Have we met the objectives that we set out to achieve?

Time to market is a major winner. A single automated data preparation stage creates a single product data model for all process preparation stages within minutes, instead of days. Features dedicated to each task then work together on the product data model to complete the preparation. All in a fraction of the time compared with using separate software pieces, even those provided by the machine vendors, in isolation.

Built into all of these tools is the highest degree of error checking, ensuring that nothing is left out, that every element of the production process can be simulated in the software tools eliminating errors avoiding the waste of expensive line time. New products are introduced right the first time without significant cost and resource overhead. For the contract manufacturers, the surprises have been removed, risk is lowered, and pricing can be made more accurate and competitive.

For existing products, the software reduces the time and cost to move products between line configurations to the point that it is insignificant, even if the target site is on the other side of the planet. This reduces the product movement barrier enabling agility throughout the operation, allowing it to scale much more effectively and efficiently in response to changing demand patterns.

A lot of the benefits that we have just mentioned are difficult to measure. How much more profit can be made with a few weeks taken off the introduction lead-time? What is the savings in terms of the cost of poor quality as we eliminate internal quality issues and market quality issues? These are key factors which can be most significant to any business.

More tangible however are the benefits of increased asset utilisation as a result of more efficient line configurations, balancing, and machine programs. The cost of work to perform process preparation has also been significantly reduced, with process preparation engineers now free to focus on actual engineering issues and improved planning technologies rather than mindless manual and duplicate data processing.The benefits of process preparation do not stop there however. Engineering product data is used as part of the integrated factory, therefore:

Scheduling is far more effective when the accurate execution times of each process are known. Planning can be much more flexible when given several line configuration options to choose from, together with the ability to dynamically create common feeder layouts. It is now practical and very cost effective to prepare each product for multi different lines so as to bring real value to schedule optimisation.

For assembly and quality, each process now has complete and accurate information, highlights of changes, and qualified documentation. This leads to the assurance of conformance. Knowing that the product is being built correctly, to the correct revision, with each process knowing exactly what is expected, providing the benchmark against which to measure the actual execution of each piece of product at each process.