Accelerating Test Socket Design
08/01/2003
By Jon Diller
Today's semiconductor test engineer faces frustrations that 10 years ago hardly could be imagined. The time span between requirement notification and first silicon has been reduced to a matter of weeks, and the task of preparing a workable test interface has spiraled upward in difficulty dramatically.
Typically, the largest component of the delay between the first device drawing and the working test interface is the time required to specify and manufacture the test socket. The other elements of the process all have been compacted — device under test (DUT) boards are produced in two to three weeks, change kits can be had in a week or so and technicians have become capable of setup in a matter of hours.
Socket delivery itself has been compressed greatly; custom vendors have streamlined six-week queues into two or three weeks. However, the design process has remained relatively unchanged, and one to two weeks are still absorbed in this element. This retards socket manufacture and delays the starting date of DUT board and change kit design and manufacture.
Accelerating the Process
If today's test engineer is to be successful, socket vendors must streamline the design process. Taking the socket specification and design process from a matter of weeks to hours requires design.
A key feature of socket design software must be ready availability. The growth in programming languages that operate with a Web interface has allowed socket design automation to emerge that anyone with a browser and access can use, regardless of platform or firewall. Further, the Web's communication capabilities allow instant dissemination of the design and attendant requirements, improving the socket vendor's reaction time.
To understand the impact of readily available socket design automation, the socket design process evolution over the past decade must be examined.
Test socket specification at one time was a relatively straightforward process. The engineer would select a molded burn-in socket loaded with bent metal contacts from a catalog, along with its receptacle. The components were available off-the-shelf, the prices were low enough to be considered incidental expenses not requiring detailed approval and the footprints were sufficiently common to allow DUT board vendors to work from a part number rather than a socket drawing.
Socket Definition
The type of automated test handler determines the socket's general shape. Handlers generally require some form of mechanical interface to the socket, which usually is unique to a given handler type. This can limit or fix the overall envelope into which the contactor must fit, the location and type of handler alignment features such as bushings that receive guide pins from the handler, the depth of the device alignment pocket, and the minimum available compression of the spring contact probe.
Beyond the general requirements placed on socket shapes by specific handler platforms, many socket users have specific feature packages of their own design. Any such customer socket standards must be communicated to and obeyed carefully by the socket vendor. Semiconductor test operations frequently will purchase sockets from multiple suppliers. It is, of course, desirable that design of both socket and interface board be standardized as much as possible. Therefore, the socket's footprint (the number and position of dowel pins that align it to the interface board, and the number, position and type of features that allow it to be affixed to the interface board) may be required to conform to a specific format.
It is common for large semiconductor manufacturers to attempt to capture the key socket feature demands of several handler types in a single socket design template. Such designs become complex, but tight adherence to socket standards lets the semiconductor manufacturer transfer sockets easily from one handler platform to another as test requirements and capacities dictate. The nature of the test also may require that the interface board bear decoupling capacitors or other passive and active electronics close to the device. This would require areas of the socket bottom to be relieved, or that the socket have an unusual, asymmetrical shape.
Test Socket Body Features
Precise Alignment. Precise alignment of the device to the spring contact probes is critical to the performance of the test socket. Poor alignment, even when it does not generate open failures due to miscontact between device leads and spring probes, can induce lateral sliding of the device across the spring contact probes during compression, resulting in rapid contamination and false failures.
Device alignment to the spring contact probes generally is accomplished by means of a chamfered pocket. The pocket's outline matches that of the device substrate, and the chamfers guide the device to center as the handler plunges it towards the spring contact probes. The chamfers transition to a straight wall prior to engagement of the device by the probes, so that the device is not forced to move laterally while in contact with the probe tips.
Pocket Size Calculation. The pocket is sized so that, at its minimum dimensions, it is providing the maximum amount of guidance to the package without causing a device to stick or jam in the pocket. Generally, this pocket size is calculated as nominal outline of the device, plus worst-case tolerance for device size, and a small clearance; this value will have a positive tolerance only, to bias the equation toward free device movement.
The pocket plays a key role in socket performance. As such, it is absolutely vital that the tolerances affecting pocket size are collected and fully understood by the socket vendor. Even better examples of device drawings are frequently missing such tolerance information.
Floating Device Nest. In some cases, the tolerance on the outline dimensions of the device substrate may be so large that the potential offset of the package is greater than the area of the lead. In such a case, if the spring contact probe has a pointed or domed tip, it will miss the lead completely. This can be countered by using a probe with a larger surface area tip, such as a flat or crown; but ultimately, the best solution for devices with projecting leads is a floating device nest.
A floating nest guides the device by its leads, rather than its substrate outline. Taking the example of a ball grid array (BGA), the pocket is removed to a separate plate, which is mounted on a spring so that in the socket's normal condition it is several millimeters above the probe plane. The normal device alignment pocket is widened significantly to provide only rough alignment, and an additional ledge is added to go under the substrate and find the periphery of the outside rows of balls. This ledge fits closely to the maximum ball periphery.
Spring Contact Probe Selection
Once the fundamental features of the socket body have been established, the most critical element of socket specification may be approached: spring contact probe selection. The impact of this choice on performance of the socket — both in the DC and RF spectra, and in the sense of durability and ease of maintenance — cannot be underestimated.
The first limiter of probe selection is device pitch. In purely DC and mechanical terms, the largest probe that will fit on a given pitch is preferable. Probes that are larger in diameter tend to last longer, be less expensive and provide a better match between crown and ball diameter. RF considerations may drive the selection of a smaller probe in some cases. While decreasing diameter relative to pitch drives inductance up, it also reduces capacitance and provides bandwidth gains that have application to some high-speed digital devices.
 Figure 1. This RFQ of a socket illustrates the design from several angles. A preliminary design of the socket allows engineers to inspect its specifications. |
null
Required RF performance usually is the second limiting factor. A high-frequency test may place significant demands on the allowable inductance or minimum bandwidth of the spring contact probe. Primarily, inductance is a function of length and pitch-to-diameter ratio. Spring contact probes now are available that produce self-inductance values as low as 0.3 nH, by virtue of being as short as 1.6 mm long when compressed.
High bandwidths typically are accomplished by increasing the probe's DC performance. Better contact of the probe's disparate members reduces capacitance. Probe designs now are available that offer as little as -0.5 dB at 10 GHz.
A careful description of the RF requirements of the probe — expressed as bandwidth, maximum self inductance or test frequency — permit the socket vendor to select the best probe for the customer's application. Other electrical characteristics may be critical: Sockets may have to handle significant amounts of current or voltage; a low ground inductance, accomplished by placing several probes on the ground pad of a leadless package, may be required; or contact resistance may be defined as particularly critical, in some cases necessitating that two probes be populated per device lead to allow four-wire measurement of resistance.
Tip style is the final critical consideration in probe selection. Quad flat packages (QFPs) and other leaded devices require four-point crowns large enough to prevent lead from slipping off. Leadless packages may require penetration of oxides on lead-free lead coatings, requiring a single point; and in many cases, it is considered desirable for probes used in a BGA socket to have tip diameters that are coincidental with the ball's diameter. This will ensure that the ball is struck near its equator, preventing ball damage at its soldering point that may give rise to air pockets during the soldering process.
Lid Design
Even a socket that is to be used purely for automated test will require some sort of manual actuator so that the tester can be set up before the handler is put online. Manual actuators, or lids, are critical to socket performance in engineering characterization and failure analysis applications, and lid design often is as critical as any other element of the socket's makeup.
Lids may be simple — and may sacrifice ease of use for low cost — in the case of simple setup actuators. For characterization applications, ease of use is critical, and if heat sinks or other special features are added, the lid may cost as much as the socket body.
Design Automation Reduces Time
Definition of these factors historically has required careful interviewing of customers by socket design engineers. Naturally, if design engineers spend significant amounts of time defining requirements, it detracts from their ability to devote their attention to the design process itself. By automating the collection of this information, the customer can be assured that his request will be complete to allow the socket design engineer to proceed directly to the critical task of design.
Socket design automation software currently fielded breaks the collection of this information into four simple steps:
- Device definition
- Operating (test) parameters for the socket, device and environment
- Lid selection and component clearance cutout definition
- Generating socket footprint and DUT board recommendations.
The device definition process gathers those elements of the device that are most critical to socket design — the device substrate dimensions, array pitch and pattern, and lead characteristics and dimensions. This stage also is where users can select their units of measurement, define their handler, if applicable, and choose to have the socket drilled out with a full array so that it can be re-populated to accommodate minor changes in the device personality. Speeding the collection of this information are easy-to-use drag selection tools and an intuitive inquiry process.
The handler definition done in this step determines the template for socket features and outline used. The design automation software includes several templates for common handlers and customer-specific templates defined for frequent users who have their own socket standards.
Operating parameters, such as test frequency, operating temperature ranges and maximum current values are gathered next. These chiefly affect the selection by the design software of the appropriate spring contact probe, based on the core dimensional information gathered during the description of the device's characteristics.
Critical to both these steps are extensive failsafe mechanisms within the software, which disallow entry of information that produces invalid results. Extensive pop-up help prevents confusion and helps inexperienced users.
The information gathered in each of these stages also produces alerts and helps if tolerance stacking indicates a possible lead miss. Remedies such as floating nests and tightened pockets will be suggested, and the software will work to produce a socket that hits its targets.
Lid definition must be considered before generating the DUT board footprint because the lid will impose certain keep-out areas on the DUT board to avoid interference when the lid opens. The topography of the device's topside also is considered; the lid's plunger will render automatically with projections that avoid contact with any exposed circuitry on the top of the device.
Once the above information is collected, a suggested DUT board footprint is generated and displayed to the user. At this point, the user can change socket mounting features to conform to a design or standards, producing an exact configuration.
JON DILLER, director of business development, may be contacted at Synergetix, 310 South 51st Street, Kansas City, KS 66106; e-mail: [email protected].