Kelvin measurement using spring probes for packaged IC testing
Ila Pal and Behrouz Sadrabadi, Antares Advanced Test Technologies, Gilbert AZ USA
Test engineers often look to leverage a Kelvin-measurement configuration when testing high-power devices, particularly QFN packages — even as the finer-pitch terminal in latest-generation packages makes it difficult to accomplish. A spring probe interconnect technology, as demonstrated through validation, enables test of QFN packages using the Kelvin setup.
Consumer electronics are driving half of today’s semiconductor market; smaller mobile phones with more capabilities, MP3 players, camcorders and other gadgets command what can become insatiable demand for next-generation ICs. This means smaller devices with more functionality. The evolution of smaller and faster ICs can tax the socket market heavily, as device manufacturers and test houses need reliable socket solutions for these high-performance devices. Although customers have their own set of requirements for each application, there are basic universal demands: interface with the device, out-of-the-box integration, electrical and mechanical performance, and lowest-possible costs. Smaller and faster devices are typically more sensitive to changes in the electrical configurations used for testing, particularly in the case of resistance. A very low resistance setup is required to measure the performance of modern ICs. More engineers want to leverage a Kelvin-measurement configuration when testing — even though the finer-pitch terminal in latest-generation packages makes it difficult to accomplish.
Since the advent of the quad flat no lead (QFN) package, there has been high demand for Kelvin test sockets. A Kelvin connection helps minimize the effects of the socket interconnection and PCB traces and vias on IC test results. The objective is to measure a low-resistance device using an apparatus, or set up, that has zero resistance, allowing users to “see“ the effects, or readings, of the IC itself.
Spring probe technology
QFN packages have small and fine-pitch features, which traditionally made reliable Kelvin contacting a nearly insurmountable task for IC testing. A test socket using a spring probe interconnect technology, however, presents engineers with a reliable, production-worthy, and high-performance enabler of Kelvin testing, down to 0.5mm pitch, on QFN packages. Using an offset plunger design, the socket provides two electrically independent contact members that interface each terminal on the package. A series of experiments designed to test robustness were conducted, characterizing electrical and mechanical performance.
An initial test examined the relationship between deflection of the spring probe, force, and the contact resistance. Tricor’s Model 921A Displacement Force (DF) test station measured the spring probe deflection and its corresponding force. A total of 32 pins were assembled into a test fixture, which then was mounted on a board connected to a tester for contact-resistance measurements. The return electrical path was connected to the force-gauge plunger. Test was initialized by moving the force-gauge plunger to the tip of the spring probe. Then, the force-gauge plunger was moved down in 0.01mm increments, and the corresponding force and contact resistance recorded. When the force-gauge plunger reached full travel of the spring probes, it retracted in 0.01mm increments. Corresponding force and contact resistance were also recorded during retraction. Figure 1 shows the force vs. deflection vs. resistance curve. The graph depicts force increasing linearly as the displacement increases. Similarly, contact resistance decreases as the displacement increases. Desired displacement is based on the compliance requirement of each application, in this case it is 0.3mm. Force and average contact resistance corresponding to this displacement are 33g and 50mΩ, respectively. The graph also shows standard deviation curves for both force and contact resistance, which represent the spread in the data set. Test engineers use this information to set up failure criteria when performing spring probe device tests.
Figure 1. Force-Deflection-Resistance curves for Kelvin probe.
Figure 2. Mechanical lifecycle data for Kelvin probe.
The second test examines contact resistance over the spring probes’ lifecycle count. A cam-operated lab cycler was used for this experiment. Again, 32 pins were assembled onto a test fixture that was mounted on the test board, which was connected to a tester. A gold-plated shorted device simulator was mounted on the bottom of head, which has a cam on the top. Set up was adjusted such that when the cam rotates, the head moves down 0.3mm, the chosen travel for the spring probe. Initial contact-resistance data was measured via tester. A motor rotated the cam, which cycled the spring probe. A digital counter inserted into the test set up measured cycle count. Contact-resistance data collected at different cycle intervals is shown in Fig. 2. The average contact resistance is less than 60mΩ over 250,000 cycles. Individual data points also provide an understanding of the data spread. Barring a few outliers, data points clustered around the 60mΩ line. Based on the graph, it can be concluded that the spring probe operates over 250,000 cycles with 60mΩ average contact resistance.
Figure 3. Kelvin measurement of known-good sample showing 1mΩ accuracy.
The third test reveals spring probe performance in a Kelvin setup. A device simulator with known resistance is used with a similar test set up to prior tests, with a pneumatic-actuated cylinder instead of a cam for repetitive cycling. This closely emulates the handler environment where actual devices are tested. The device simulator had two pads and a 10mΩ resistor connecting both pads, with two pins per pad. Pins 1 and 2 were connected to pad A and pins 3 and 4 were connected to pad B. One pin was used for forcing current and the other for sensing voltage drop between pads. When applied, current traveled to pad A via pin 1 (force) and returned from pad B via pin 3 (force). A voltmeter was connected between pin 2 (sense) and 4 (sense) to measure the voltage drop precisely. This is a true Kelvin measurement. After the initial data points were collected, the device simulator was cycled to a specific interval count and measurements taken. This was repeated until 250,000 cycles were reached; the data was plotted. The experiment was repeated using five more 10mΩ resistors and plotted in Fig. 3. The measurement was right on at 10mΩ except few data points over 250,000 cycles. This minimal variation could be attributed to measurement accuracy, as well as tolerance of known-good resistor values itself.
IC devices have to be functionally tested by ATE before being sent for field installation. The test socket is the heart of an ATE system. The reliability of test sockets is critical to ship defect-free products. Validating the performance of test sockets gives confidence to engineers testing their end product. Three tests were performed to validate the spring probe’s use in Kelvin applications. Force-deflection-resistance (FDR) testing validated the specification for force and contact resistance at recommended spring probe travel, while cycle testing validated the number cycle up to which the spring probe would perform without degradation. Testing known-good resistors using this spring probe technology in the Kelvin set up provides repeatable measurements with ??1mΩ error over 250,000 cycles, meaning these pins do not have to be replaced. This leads to reduced ATE downtime and increased throughput.
The authors would like to thank Randy Dowell for support in the testing of this contact technology.
Ila Pal, product development manager, holds a M.S. degree in mechanical engineering from Iowa State University, Ames. He holds an M.B.A. degree from the University of St. Thomas, Minneapolis, and is certified on the Six Sigma Black Belt process. He has received six patents, presented many papers, published articles and has spent 15+ years developing new technologies in the packaging and interconnection field. He may be contacted at Antares Advanced Test Technologies, Gilbert, AZ., ph. 480/682-6230; email email@example.com.
Behrouz Sadrabadi, product development engineer, received his M.S. in mechanical engineering from San Jose State University. He has eight years of experience in design, finite element method (FEM), stress and thermal analysis. He may be contacted at Antares Advanced Test Technologies, Gilbert, AZ., ph. 480/682-6215; email: firstname.lastname@example.org.