|
Complex programmable logic devices (CPLD) such as Flash memories and embedded microcontrollers are rapidly finding their way into virtually every electronics product and for good reason. Whether found in the smallest handheld smart phones or on a massive network router board, whether used in an application that is simple or complex, programmable parts deliver significant benefits: • Elimination of hardware logic circuits, reducing parts count and circuit size. • Shrinking product design cycle times and cost. The designer can correct many design problems or add in new features for a product “re-spin” by altering the device’s program code rather than the parts list and board layout. • OEMs can supply multiple versions or models of the product from just one or two board types by using variations in the program code. • Suppliers can create reasonably priced “one-off” and “custom versions” with fast turnaround whencustomers actually order the product. Parts such as microcontrollers and Flash memories impact the electronics manufacturing process because they must be programmed at some point before, during or after board assembly. Until recently, the most popular programming approach has been to pre-program individual parts offline before they’re attached to the board during the assembly process. However, pre-programmed parts complicate inventory management, especially where multiple program versions for different product configurations exist or pre-programmed code needs to be updated. Moreover, part- or board-specific data (such as serial number, measurement values, date codes, etc.) cannot be programmed into the device ahead of time, forcing an additional programming step after the part is attached to the board. In an alternative process called “on-board” or “insystem programming” (ISP), manufacturers first attach “blank” parts to the board and then program them at incircuit test (ICT) or at final functional test. Tester based on-board programming is attractive because the part to be programmed is usually already connected electrically to the tester based electronics needed to program the part via a bed-of-nails fixture (at the in-circuit tester) or a connector (at the functional tester). Advantages of on-board programming Pre-programming parts before placing them on the board is straightforward for stable designs and/or those that have only one or two small memory parts. However, with the proliferation of multiple board types and revisions together with a growing variety of part types, on-board programming eliminates the requirement to track multiple versions of pre-programmed parts, simplifying inventory management and eliminating rework (reprogramming) costs. Equally important, on-board programming makes it easy to insert unique part- and/or board-specific data into the main programming code “on-the-fly,” eliminating an additional downstream programming step. On-board programming facilitates “one-off” and custom product versions, as well. No free lunch: on-board programming tradeoffs Nevertheless, despite these benefits, on-board parts programming involves costs and compromises. Specialized hardware modules (often called “dongles”) that provide signal conditioning and the proper bus and device algorithms to deliver programming code to the part are required, usually one dongle for each programmable part on the board. When on-board programming is performed on the ICT, the dongles must be incorporated into the same bed-of-nails test fixture used for normal in-circuit testing, adding complexity and cost to an already complex and costly bed-of-nails fixture. The issues are magnified when software debugging and fixture maintenance are required later. In addition, each test fixture for each board type requires its own set of dongles along with the custom engineering to integrate the dongles into the test fixture, usually creating documentation, rework difficulties and still higher costs. Then there is the question of resources needed to write and debug the part programming code, driving up cost and implementation time. Nor are these specialized skills always readily available in a test engineering department. Further, on-board programming time—especially for devices with large memories— is typically much slower than the time required to conduct an in-circuit or functional test of the entire board. (Today’s smaller boards simply exacerbate this programming time/ test time difference.) In high-production volume environments, lengthy parts programming times can idle expensive tester resources, reducing overall equipment productivity. As the popularity of portable products grows, the physical size of “typical” circuit boards continues to shrink. Production engineers increasingly are maximizing assembly productivity via multi-board panels, which (depending on size of the individual board) may consist of up to twenty to forty boards. Panelization multiplies production throughput, adapts well to SMT placement and soldering equipment physical specifications, simplifies board handling and reduces cost. However, panels can also create a major production roadblock: if each individual board has one or two programmable parts, panels can rapidly multiply the number of parts needed to be programmed at one tester pass (figure 1). If a given circuit design uses, say, two programmable parts, a 6 x 4 24-board panel means 48 parts must be programmed. 
In addition, each test fixture for each board type requires its own set of dongles along with the custom engineering to integrate the dongles into the test fixture, usually creating documentation, rework difficulties and still higher costs. Then there is the question of resources needed to write and debug the part programming code, driving up cost and implementation time. Nor are these specialized skills always readily available in a test engineering department. Further, on-board programming time—especially for devices with large memories— is typically much slower than the time required to conduct an in-circuit or functional test of the entire board. (Today’s smaller boards simply exacerbate this programming time/ test time difference.) In high-production volume environments, lengthy parts programming times can idle expensive tester resources, reducing overall equipment productivity. As the popularity of portable products grows, the physical size of “typical” circuit boards continues to shrink. Production engineers increasingly are maximizing assembly productivity via multi-board panels, which (depending on size of the individual board) may consist of up to twenty to forty boards. Panelization multiplies production throughput, adapts well to SMT placement and soldering equipment physical specifications, simplifies board handling and reduces cost. However, panels can also create a major production roadblock: if each individual board has one or two programmable parts, panels can rapidly multiply the number of parts needed to be programmed at one tester pass (figure 1). If a given circuit design uses, say, two programmable parts, a 6 x 4 24-board panel means 48 partsmust be programmed. Dongle-based and traditional ICT-based programmers can program only one part at a time. Multiple part programming times add up quickly in the case of panelized boards, tying up expensive test assets. This is usuallyunacceptable, even in a medium-volume production environment. For the 24-up panel that means a 48 times“hit” to throughput—clearly an unacceptable productivitycompromise. Solving the large memory, large panel problem On-board parts programming at in-circuit or functional test is an excellent strategy when only one or two devices need to be programmed in a single tester “pass.” But once more than three or four parts need to be programmed, the custom engineering task of integrating multiple dongles in a fixture with relatively more complex software becomes unwieldy and expensive. Even if that hurdle is overcome, the productivity degradation effectively rules out on-board programming. Recall that offline parts programming systems can program multiple parts at a single pass. Clearly, a similar “gang programmer” technology is required for on-board parts programming. The other critical requirement for an on-board gang programmer is to separate the on-board programmer from the host tester in order to address productivity considerations such as line balancing. The design objective of a standalone on-board gang programming system is to meld the throughput and cost advantages of offline simultaneous part programming to the process simplification and ability to insert part- and board-unique data into the main programming sequence of the on-board programmer. The on-board gang programmer must have the flexibility to program multiple chips of multiple disparate family types simultaneously, since many circuit designs use different part types on the same board. It is vital that the gang programmer provide all the signal conditioning and bus algorithm flexibility needed to eliminate the requirement for dongles and reduce or even eliminate custom engineering costs. The gang programmer brings standardization to what up to now has been a custom task. CheckSum LLC has recently added the MultiWriter pps (production programming system) to its MultiWriter product family (figure 2). Depending on the number of parts to be programmed in a single pass, the user may now choose between tester-based parts programming and a standalone gang programming system. The system uses the same patent pending simultaneous programming technology originally developed by CheckSum for its Analyst in-circuit testers. With its simple architecture, the MultiWriter pps PC based system provides concurrentprogramming of up to 24 parts of up to 16 different typesor families—up to 384 parts at once. 
The future of on-board production programming Ever shorter product lives and the ability to customize products on a lot-size-one basis, and logistics will make programmable parts increasingly popular. In the manufacturing and test environment, the combination of process simplification and data customization offered by onboard part programming makes it a more attractive option for electronics manufacturers than pre-programming. But the growing popularity of parts with ever larger memories, together with the proliferation of multi-board panels creates a throughput challenge in the production environment. One-device-at-a-time on-board programming approaches such as dongles are better suited to the lab bench than the production floor. For single boards with one or two parts on them, ICT based on-board programming such as MultiWriter-equipped Analyst testers remains an effective solution. But for the throughput demands created by multiple parts to be programmed on a single panel the best production programming tool is the standalone onboardparts programmer. | 
