Improving Socket Design and Validation
A Fast, Low-cost Solution
BY TERENCE Q. COLLIER
The life of a socket begins when a designer sits at a workstation and models a finite element analysis of the force required to make stable electrical contact between the test pin and device. Typically, other than pad pitch, pad composition, and I/O count, few requirements are specified. Historical data acts as a barometer to suggest target loading that yields acceptable test sockets. In some cases, the design takes a hit-or-miss approach until the desired results are obtained.
The work becomes more challenging when lead-free is taken into consideration. Variation in probe-tip geometry, DUT metallization, contamination, and spring parameters such as spring constant, history, age, and temperature, contribute to design complexity and are potential sources of variability. Because a socket is a set of individual pins and springs that make up a complete system, how those individual components behave greatly influences the complete socket.
To validate the design of the finished socket, a known-good-device (KGD) - or golden unit - is placed in the socket and verified both electrically and mechanically. Mechanical validation is critical. If the mechanical load crushes solder balls, damages the component, socket, or equipment in any form, the design is not acceptable. The ability to specify requirements to the card builder and provide in-process validation and final QA on assembled sockets would benefit the end-user.
Sockets in this study were tested using a modified micro-hardness tester (MHT) with a probe tip assembly to simulate a single socket contact. Coupled with an XYZ positioning stage and control of vertical load and displacement, the MHT permits testing at any predefined coordinate with an accuracy of less than 1 µm.
The MHT can be used to make load- or depth-controlled experiments for evaluating additional test requirements. A friction setting helps determines the exact scrub length and force for electrical first contact for sockets designed to scrub across the pad to make electrical contact. To measure electrical performance, the system incorporates two exterior-user channels to record resistance vs. load in real-time with the test data.
Figure 1. Simulated load (blue) vs. resistance (red) demonstrates the force required for both initial and stable electrical contact.
The set of experiments illustrated in Figure 1 show the results of 1, 50, 500, 1,000, and 5,000 cycles. Under repetitive loading/unloading cycles, the fatigue properties of a probe/spring pair can be assessed to evaluate and model the number of cycles-to-failure and preventative maintenance (PM) schedule. The MHT software includes a mode which allows the user to view the sample surface and select areas of interest. Pin locations can be selected optically or loaded from a database and stored, so samples can be tested in real-time or off-line.
How to Build a Better Socket
To build a better socket, the designer and test engineer need to be able to generate certain requirements quickly and economically. Optimum spring loading for a given probe tip geometry determines the sweet spot for stable electrical contact. It must be possible to evaluate contact resistance (CRES), the impact of intermetallic formation at the probe tip, and the impact of contamination on the DUT pads. Life-testing - including PM schedules - of individual components is necessary to assure correct choices are made for the spring and probe. Incoming inspection capability on assembled sockets is required; each probe pin can be verified at receiving inspection to documented requirements and specifications. Lastly, the designer and test engineer must generate the impact of temperature on performance and components.
The MHT setup used in this experiment provides the tool to remove “art” in favor of science. In this experiment, a preexisting socket was used, but the MHT setup can be used as a template to generate design requirements the socket-user can provide to the socket builder. By simply placing a single probe point into the test head, the force to make electrical contact can be modeled. That single pin data is then extrapolated into a multiple pin socket. The data provides the template for successful first-time delivery and consistency.
The MHT was set up using a flat-tipped probe to simulate an individual socket pin in contact with an LGA pad. In the first pass, individual pins were selected randomly and the probe measured the amount of deflection a pin moves under an applied load. The target deflection was set at 50 µm and the force similarly measured. Finally, the MHT was used to run cycles of 50, 150, 500, 1,000, and 5,000.*
Tthe pin does not move until a primary spring constant is exceeded, then moves relatively easily with an actual reduction in force; typically up to 30% less than the initial deflection force. Both initial and secondary forces were found to vary with repeat cycling. When tested to 5,000 cycles, the force to move the pin increases by up to 5 g. Between individual cycles, the pins were consistent until debris caused additional friction, restricting pin movement.
Table 1: Sample loading of 5 random pins with deflection and load deltas.
Five pins were randomly selected and summarized in Table 1 and Figures 2a, 2b, and 3. Thirty grams of force show that 2 of 5 pins would not deflect enough to make electrical contact with the DUT. The likelihood of false failures is real with deltas ranging from 2.6 to 78 µm. Load values ranging from 12 to over 34 g were also measured.
A modified MHT can be used to help design and verify probe card parameters and model PM schedules. In this specific design, the spring has a primary and secondary aspect to loading. Initially, the spring loads to a primary k value with little deflection before settling into position at a secondary k value. This secondary load is required for stable electrical contact and determines the depth of the probe deflection. The primary load accounts for a principle deflection of the probe.
Figure 2a. Peak load of 12.7 g and secondary load at 50 µm deflection point.
Figure 2b. Demonstrates the variance that can be found within a probe card.
The primary load can lead to equipment or hardware mechanical failures if not taken into consideration. If all the pins were equal, the socket would show an initial load delta equivalent to 37 g with the secondary load at 18 g if all the pins were equal (Figure 2b). A 1,000 I/O device would result in almost a 20-kg load delta (37-kg primary vs. 18-kg secondary). Figure 2a shows that the initial load would be around 12 kg. If the test hardware were designed to support 12 kg, the second card would likely show false failures or result in equipment damage. This is too much variation.
Figure 3. 5,000 cycles show that the performance changes over time.
The spring constant changes over time. The amount of variation present in a small sample of pins shows that sockets are harbingers of variation. The modified MHT demonstrates it can be used successfully in both the design phase and potentially as a QA tool. For example, the card in this experiment demonstrated that 2 pins failed to move at least 20 µm under 40 g of loading, with two moving less than 3 µm at 30 g. Whether this is good or bad is unknown since the design requirements were not established prior to building the socket. However, such variation in force could lead to premature life failure in some pins.
*This article shows 2 pins with one cycle each and a third chart at 5000 cycles. Contact the author for additional information.
For a complete list of references, contact the author.
TERENCE Q. COLLIER may be contacted at CVInc., 1155 E. Collins Blvd., Suite 200, Richardson, TX 75081; 214/557-1568; E-mail: firstname.lastname@example.org.