During manufacture, testing adds to the cost of the final product significantly. Sometimes the cost of testing can be even greater than the cost of product manufacturing itself. Test and measurement manufacturers have been quick to respond by coming up with innovative ways to lower the test cost and increase throughput
DECEMBER 2011: Testing is a critical component of product development and production process. It can improve a product’s performance, increase quality and reliability, and lower return rates. It is estimated that the cost of failure decreases ten times when an error is caught during production instead of the field, and decreases ten times again if it is caught in design instead of production. By catching these defects and collecting the data to improve a design or process, test delivers value to an organisation. At the same time, manufacturing tests add to the cost of the final product. So it’s time to drive innovation into this process through technology insertion and best-practice methodologies in order to generate large efficiency gains and cost reductions.
Manufacturing a product involves a lot of in-line equipment. In a surface-mount technology (SMT) line, not one machine is capable of handling the entire process. There are various machines used for various purposes. Roughly, a basic SMT line requires a printer, a mounter and a reflow oven to give an assembly output. To reduce the final cost, certain measurements can be done on the manufacturing line itself.
R.S. Gupta, senior manager-Sales & Marketing, DVS India, informs that manufacturers have started using high-quality, repeatable printers capable of 2.5D and 3D inspection of the paste printing itself, so no separate testing is required at this stage. The printer introduces almost 60-65 per cent of the problems of the process in line. So by taking care of the first stage, your product doesn’t show up the kinds of problems that are incorporated at the first stage. Problems detected at an earlier stage can be corrected at a very low cost. This reduces the final product cost.
Once the product comes out of assembly, the manufacturer does various kinds of testing to check whether the product is manufactured as per the requirements or not. Testing is done for manufacturing defects and functionality.
During the product assembly, some problems may get incorporated due to human or machine mistake, or due to a faulty part put on the assembly from the supplier’s side. Out of a thousand components, one or two parts may be failing. As component manufacturers today are taking a lot of care, this is very rare. So it may not be failure from the supplier’s side but there are devices which are very electrostatic discharge (ESD) sensitive and may fail when handled with bare hands. Also, the process itself may fail due to over-heating, etc.
The basic requirement of testing an assembly is to detect whether the product is free from manufacturing defects or not. It is the cheapest testing carried out in an assembly. It detects whether the component used is correct or placed correctly. Apart from these component problems, there could be soldering-side problems as well. When solder is used in excess it could create a bridge between two nearby pins, while less solder creates a dry point. Manufacturing defect testers detect the analogue component type and its value, but can’t test digital components. These aren’t very costly either. One can use manufacturing defect analysers (MDAs), flying probe testers (FPTs), automated optical inspection (AOI) systems or automated X-ray inspection testers depending on the equipment’s return on investment. Of these, manufacturing defect analysers are the cheapest.
In the same kind of environment, one can even use a high-end tester called in-circuit tester (ICT). An ICT is a full-capacity tester capable of testing analogue, digital and mixed-signal devices. It costs ten times more than an MDA. The rate of fault detection is very high using an ICT. So a manufacturer using an ICT charges more for its products. Besides doing what an MDA does, the ICT performs power-on testing as well. That is, when you apply the power, it checks individual components on the assembly for their functionality (whether they are performing as per their specifications or not). It, however, can’t test the full product for its functionality.
ICTs are very generic. These are user-programmable. You can use an ICT to test different components by using a dedicated fixture and program written for the PCB. The fixture ensures electrical connection between the assembly and the tester. These fixtures and programs are very dedicated. That means if you have a product with ten components, you need ten fixtures and ten programs for component testing. Of course, even a manufacturing defect analyser requires a dedicated fixture and program but these two cost very less. For ICT the cost is higher due to full power-on testing.
Next comes the functional testing of the product. A functional tester (FT) ensures that the product is working as per the requirements. There are various kinds of functional test equipment. Sometimes you may use a million-dollar machine and sometimes not even a million-rupee machine. It goes like this: If your product is very simple, you can develop a smaller fixture and some back-end signalling for your product. For example, electronic energy meter manufacturers need to check whether the LED displays properly or not when electricity is fed, etc. They want to see the sequence of their product in powering condition as well as in functioning condition. For this, a very small FT which is very cheap can be developed. But if the product is critical, the cost of an FT goes very high.
Functional testers are dedicated testers. That means if you develop a fixture for an energy meter, it can be used for a particular energy meter or board only. This functional testing is cheaper. But if you want a universal functional tester, it will be a huge machine that can test a hundred types of boards. This machine is very costly.
MDAs, also called ICTs (but not power-on ICTs), cost $25,000-35,000. The machine’s cost is based on its pin configuration. If the board has 500 points, an ICT with 500 outputs is required.
ICTs cost minimum $100,000-150,000. For a very good brand, the cost can go up to $300,000. Universal FTs cost million dollars or more.
Automatic optical inspection (AOI) systems (used to detect manufacturing defects) do optical testing only. Sometimes designers in the manufacturing line try to inspect the assembly for component defect with naked eye. AOI system helps to ensure that faults aren’t skipped. It can be desktop or inline type. Desktop AOI system costs $40,000-50,000, while inline AOI system costs $70,000-90,000.
Another machine which the manufacturer may want to use is the flying probe tester (FPT). Its speed of testing is very slow but no fixture is needed—fixture adds a lot of cost. But it needs a program, which is very easy to develop. FPT costs about $150,000.
X-ray testers are very dedicated for ball-grid array (BGA) components. Because an AOI system can’t inspect this particular type of device, a few manufacturers who want 100 per cent testing—as required in critical applications like avionics, automotive electronics and space electronics—use automated X-ray inspection (AXI) testers also. Automated X-ray testers cost $100,000-200,000.
What adds to the cost of manufacturing test?
Testing, once confined exclusively to the end of the line, now occurs at several points to provide greater failure isolation and feedback to the manufacturing process. Process tests such as ICT, AXI and AOI sit at various points along the line to immediately catch process flaws. Functional test, which tests the operation of the system of components, typically resides at one or more points along the line and at the end of the line as the final test of the fully assembled product.
With decreasing component size and correspondingly increasing circuit board density and complexity, many products no longer have access points for the bed-of-nails fixtures used for in-circuit testing. Although AOI and AXI systems continue to become more common and more comprehensive, functional test is used more extensively to find and even isolate faults.
With the need for higher throughput and more efficient production and testing, contract electronics manufacturers (CEMs) and OEMs demand sophisticated test platforms for functional testers that are built on open industry standards.
New approaches to testing
There have been efforts to make manufacturing test cost-effective, simpler and more efficient.
Naresh Kumar, general manager-South Asia Pacific Application Engineering Organization, Agilent Technologies, explains that the cost of a material to be tested is amortised over a certain period of time. To reduce the cost of test, pay-per-use option has come up. Here the payment depends on how you test. For testing a product like Apple iPhone, more features are required and therefore you will have to shell out more. Whereas for a product like USB device, you need to pay much lower than for iPhone. How much you pay depends on two-three parameters: For how much time you test, how many boards you test and how complex are the boards to be tested. The more complex the boards, the more the amount to be shelled out for testing the same number of boards.
Today, most of the manufacturing is done by CEMs. That is, if you design a product today, the boards are manufactured by some CEM which has the full facility to service a hundred customers. CEMs survive on multiple orders. Sometimes there could be big orders, sometimes there could be very low volume of manufacturing. The pay-per-use test system works very well for them. It dramatically reduces their cost of test—while maximising flexibility to test boards in a wide range of sizes and complexities. Let’s say, a CEM has a huge order which requires three months of complete activity and then two months of low activity. In these two months, they hardly spend anything on testing. So they purchase a full-capability test system at a much lower price than a conventional system and pay only for the capabilities they need, only when they use them.
In order to increase the efficiency of testing, one of the things that manufacturers have been doing over the years is trying to see whether they should test every component on the board or do selective testing.
Let’s say, you have a cellphone which has 15-20 components—one big ASIC, amplifier, receiver, antenna and so on. Ideally, you would like to test every component on the board to ensure that everything is working perfectly.
With functional testing, you need not test every component. You just test the functions that the cellphone is expected to do. So, typically, functional testing is end-to-end testing where an input is given and output checked. If the whole device is working, it is assumed that all its components are working.
There is also a possibility that critical limits aren’t met. So manufacturers do sample testing. In sample testing, some samples are tested for complete test. The samples are tested for every aspect—voltage level, current level, functionality, etc.
If the manufacturing volume is high and you have a very standard parts supplier, you have a certain level of confidence over the supplier that the parts being supplied are certified and wouldn’t like to do the testing exercise over all the components. However, you still do sample testing to ensure parts’ quality. For a new product launch, obviously every component is tested because you haven’t built the confidence yet over the vendor that the components supplied are good.
So for mass manufacturing, there could be a mix of functional testing and complete parametric measurements for sample devices.
To make testing simpler, boundary scan has become an essential tool. It enables much of a board to be tested with minimal access. Over a period of time, JTAG port has become a standard connector on many devices by accessing which you can do many things on the device—you can test the device, program the device and so on. It was introduced primarily to reduce the cost of test. The maximum cost of test in manufacturing is involved in ICT. ICT means every pin of the device has to be accessed for testing. Traditionally, for a 144-pin device, you need to probe every pin, apply signal and check the response. With JTAG, even if it’s a 144- or 200-pin device, through only five pins you can check the functionality of the device.
“In today’s production process, automation is the key for optimising the cost of the final product. With interfacing of test equipment with PCs and PLCs, and features like pass/fail, these are being integrated within the process chain. These instruments not only measure the parameters but also do data logging for documentation, statistical quality controls, etc. When a component fails to comply with specifications, a solenoid control signal is generated and the faulty part moved out. An example is a high-precision multimeter with pass/fail output used in a resistor, capacitor or inductor component manufacturing unit,” adds P. Prabhu, general manager (technical), Scientific Mes-Technik.
According to Jayaram Pillai, managing director, National Instruments IndRA (India, Russia and Arabia), key technology trends that increase test efficiency and reduce test costs include organisational test integration, system software stack, heterogeneous computing and IP to the pin.
Organisational test integration. Throughout the electronics design and manufacturing industry, test teams are improving integration across the organisation to gain a competitive edge. In the past, validation (the process of testing a product during design to guarantee that it meets feature specifications) and manufacturing test teams had few opportunities to work together. However, test managers seeking to decrease time to market and reduce test costs see that improving. Increased use of automation across both the groups has shown that they can share a common software and instrumentation. Hence one of the top goals for test engineering organisations is to increase reuse between validation and production.
System software stack. Software has been a critical component of automated test systems since it was first used to control standalone instruments more than 40 years ago. In fact, software development costs are often two to ten times more than capital costs of most test systems today. In response to rising software development costs and accelerated product development cycles, today’s industry-leading companies emphasise on designing a robust system software stack to ensure maximum longevity and reuse of their software investments.
From a system software perspective, most companies are moving away from monolithic software stacks that often contain fixed-constant code and direct driver access calls to the instruments. Alternatively, they are seeking modular software stacks in the form of separate yet tightly integrated elements for test management software, application software and driver software. This type of system software stack helps engineers apply optimal tools for each area and choose between standardised commercial off-the-shelf (COTS) and in-house tools at each level. A key trend is the extension of modularity into each layer of the software stack, including the increasing use of process models, code module libraries and hardware abstraction layers.
Heterogeneous computing. Automated test systems have always comprised multiple types of instruments, each best suited for different measurement tasks. An oscilloscope, for example, can make a single DC voltage-level measurement, but a digital multimeter provides better accuracy and resolution. It is this mix of different instrumentation that enables tests to be conducted in the most efficient and cost-effective manner possible. The same trend is now affecting how engineers implement computation in test systems. Applications such as RF spectrum monitoring, for example, require inline, custom signal processing and analysis not possible using a standard PC CPU.
Furthermore, test systems are generating an unprecedented amount of data that can no longer be analysed using a single processing unit. To address these needs, engineers have to turn to heterogeneous computing architectures to distribute processing and analysis.
IP to the pin. For decades, the electronics industry has pursued its version of the holy grail—concurrent design and test. Many have believed this to be an unattainable goal, considering how far apart the two worlds appear. In the design world, most engineers design at a system level using the latest electronic design automation (EDA) software, which has seen tremendous innovation over the last decade. The test industry has, however, not innovated as quickly, and many companies have chosen to invest more in their design tools than their test engineering tools. The consequence is test engineers are typically outmatched when testing the latest software-centric electronic devices.
The next phase in integrating design and test is the ability for engineers to deploy design building blocks, known as intellectual property (IP) cores, to both the device under test (DUT) and the reconfigurable instrument. This capability is called ‘IP to the pin’ because it drives user-defined software IP as close to the input/output (I/O) pins of next-generation reconfigurable instruments as possible. The software IP includes functions/algorithms such as control logic, data acquisition, generation, digital protocols, encryption, math, RF and signal processing.
Strategies to lower the test cost
A functional test platform can lower the cost of manufacturing test in three key ways: By increasing the test system throughput, increasing the engineering productivity and lowering the overall cost of ownership.
Increasing the test system’s throughput. Increasing the throughput is a goal of every high-volume manufacturer. In many manufacturing lines, functional test cannot keep up with the beat rate of devices coming out of production. Because it is usually highly undesirable to slow down the production process to accommodate test, the most common solution is to have multiple functional testers at the end of each line to keep up with the production throughput. This condition is commonly called test fan-out. As this proliferation of test equipment can be avoided by reduced test times, lower capital equipment expenditures can be achieved.
The PCI eXtension for instrumentation (PXI) platform provides tremendous performance advantages over other hardware platforms. Built on the industry-standard PCI bus, it offers data throughput of up to 132 MB—more than a hundred times faster than the throughput of a general-purpose interface bus (GPIB) device. This high data throughput means that measurement data can be quickly transferred from the measurement device to the computer where a decision can be made or data can be logged and thus the test time for the device be significantly reduced.
To further increase the throughput, PXI provides star-triggering capability. With the star trigger, several PXI devices can be tightly synchronised on common trigger or clock signals. In many applications, the use of the star trigger has resulted in throughput gains of ten times or more.
The Interchangeable Virtual Instrument (IVI) standard for instrument drivers provides a state-caching driver model to eliminate redundant calls to instrumentation and significantly increase performance. IVI instrument drivers can deliver performance gains of ten times or more over traditional drivers without state-caching technology.
To deal with fan-out at functional test, many manufacturers place several independent but identical testers in parallel. However, this method uses capital equipment very inefficiently, because even under-used instruments are duplicated for each tester. A far more efficient method is to test several devices in parallel on a single, parallel tester with multiple test sockets. A parallel test strategy offers the most efficient use of equipment because more expensive, under-used instruments can be shared among sockets while more common instrumentation can be duplicated per socket for maximum throughput.
Increasing the engineering productivity. Getting a product into volume production and on the market as soon as possible can make the difference between a successful product and a failure. In highly competitive industries, for example, a six-month delay in the shipment of a product can result in more than a 30 per cent decrease in cumulative profits over the entire life-cycle of the product. For such applications, developing a manufacturing test system rapidly is essential, as is quickly adapting a tester after product revisions.
Technologies based on widely available commercial technologies have inherent productivity advantages. PXI, for example, takes advantage of standard Intel and Microsoft platforms to deliver plug-and-play device installation and easy, consistent device configuration. PXI measurement devices also offer increased flexibility over proprietary platforms, because the PXI device makes only the raw measurement and uses software for analysis and decision making. Thus when new products are brought to production, systems using PXI can be quickly changed to add new measurement routines in software, often without requiring hardware changes.
Another way to maximise productivity is to use commercially available software whenever possible instead of creating custom software packages in-house. Many organisations, for example, write their own test codes instead of using an off-the-shelf product. Spending valuable engineering time recreating software that already exists, reduces the overall productivity and ties up even more resources over time for software maintenance. Today’s successful manufacturing organisations concentrate on core competencies such as the design of products and test strategies to achieve efficient use of available resources.
Lowering the overall cost of ownership. The cost of a test system is measured by two variables:
1. Initial cost of the equipment, plus development and/or integration services
2. Cost over time for maintenance and support of the test system
PXI measurement devices use commercially available technologies such as standard integrated circuits for analogue-to-digital conversion and PCI bus interfacing. Because these devices are widely used in other high-volume technologies, such as telecommunications and consumer electronics products, their unit cost is substantially lower than highly customised components. This feature results in measurement functionality at a cost much lower than that of devices in proprietary platforms. In addition, the PXI platform uses components already present in your computer—the processor, memory and monitor—to perform the measurement analysis and display instead of duplicating this functionality in a customised box instrument.
PXI also lowers costs by delivering a platform capable of diverse measurement and control capability. In many systems, the platform for analogue measurements is completely separate from functionality, such as in optical inspection, motion control or continuous data acquisition. PXI, however, integrates this diverse set of measurements into a single hardware platform to lower overall equipment costs and to permit a greater degree of integration between these systems.
Much of the total cost of a test system comes after the initial capital expenditure, because software training and maintenance costs can add significantly to the cost of test. Widely used commercial software have standard training, technical support and a large user base. More the commercial technologies you use in your test system, better the advantage of these cost savings.
Maintenance is often a hidden cost in software development. All software require some amount of maintenance over time. Keeping up with the latest operating systems, for example, is a requirement for compatibility with hardware and other systems in the enterprise. Porting customised software to a new operating system can use up valuable resources and time. Some companies spend several million dollars, for example, to port in-house test software from disk operating system to Windows NT. Using commercial software tools offloads much of the maintenance burden to the software vendor and lowers your overall cost.
To sum up
A manufacturing test system using widely used hardware and software platforms built on commercial technologies and open industry standards can dramatically lower the cost of test. Such a platform increases the test throughput, with fast measurement hardware and software capable of managing multiple test routines in parallel. Engineering productivity will increase, because engineers would take full advantage of powerful software programs to develop tests rapidly. The time required to redesign test systems for new products will decrease with the use of flexible, modular software and hardware platforms. Finally, the cost of the test system, including both the initial capital expenditure and the overall cost of ownership, will be drastically lower, thanks to the leverage of commercial technologies and widely used standards.
The author is an executive editor at EFY