Wafer probe is one of the last frontiers for yield improvement and process development in the IC industry. To test the devices, contact to the bonding pads of each chip on the wafer must be established using a probe card. As the chips are tested, a map is created of good and bad chip locations. This map is used to determine which chips—hopefully the good ones—should be assembled in packages.
For the most part, each chip on the wafer is contacted at wafer probe using technology developed more than 20 years ago. The cantilever-beam epoxy-ring probe card is produced today with virtually the same techniques used since its inception (Figure 1). The difference primarily is in the skill of the probe-card manufacturers and the sophistication of the product they produce. Although several other probe-card technologies are available, the cantilever-beam technology still is the most pervasive.
Since the cost of producing the wafers is spent at this point, any yield loss is expensive. Because the package and assembly processes also are expensive, it is important to package only good chips. Until recently, IC manufacturers spent little effort or money to improve this rather crude process.
History
In the past few years, IC manufacturers have realized there is significant gold in the wafer-probe mine. Processes, probe cards, and equipment have improved significantly. Some of the most dramatic improvements have been in the area of testing the probe card.
To build higher-performance and more sophisticated probe cards, it is critical to perform accurate, repeatable tests to evaluate the results. The newer IC devices require increasingly smaller pads, tighter pitch, and higher pin-counts.
In 1995, a high-end production probe card contained 500 probes with 60-micron dia tips contacting 100-micron pads on 125-micron pitch. Today’s high-end production probe cards could have 1,500 tips with 20-micron diameters. They are built with three, four, or more level technologies to contact 55-micron pads on 70-micron pitch. This requires very tight tolerances and control on probe-tip alignment, diameter, and coplanarity. Testing and characterizing these cards demand even more improvements in metrology.
For several years, probe cards were tested and aligned by very basic methods—a simple light box was used to test coplanarity. Alignment consisted of contacting a wafer and adjusting probes until the scrub marks were properly positioned in the pads. Contact resistance, leakage, probe-tip diameter, and gram force rarely were tested, but merely were sampled.
As probe cards became increasingly more complex, manufacturers and users needed methods to test these parameters to guarantee compliance to specifications and characterize the building process. To provide this capability, probe-card analyzers were developed to test the probe cards. These analyzers have evolved to combine video, electrical, and mechanical techniques to test the various parameters.
Probe-Card Specifications
First, let’s define the kinds of tests an analyzer must perform (Table 1).
Using the alignment test as an example, we will show how probe-card specifications are determined, how the tests are performed, and what measurement precision and repeatability are required. Integrated Technology’s Probilt Probe-Card Analyzer performs the alignment test using a video technique to locate the X, Y positions of the probe tips (Figure 2).
After using an initialization routine to find and identify at least one probe in the array, the system uses a file of the relative X, Y positions of each probe tip to move the stage and video-capture system around the probe array. The image of each probe-tip is captured by a video frame grabber. Its location at zero and the specified Z overdrive are determined by a combination of stage position and probe-tip position within the video image.
Then, the set of probe-tip positions is compared to the file of correct positions using a best-fit algorithm to correct for offset and theta. To make a pass/fail determination for each probe, its actual position is compared to the specified position, and an offset is calculated. Since the probe tip cannot touch the edge of the pad, the path it makes as it scrubs across the surface of the metalization must be compared to the allowed contact region within the pad.
Figure 3 represents a bonding pad with a probe scrubbing on its surface. The safe area is the part of the pad a probe can contact (Xs, Ys). The area is inside the passivation opening of the pad by a sufficient margin to prevent the probe from cracking the glass or lifting the polyimide layer.
Positions P1 and P2 are the start and end of the scrub mark made by the probe as it is overdriven on the pad. The typical probe-card specification requires that the center of the scrub (CS) must be within a certain tolerance of the center of the pad (CP), and the entire scrub mark must remain within the safe area.
The specifications required for a specific card may be determined by considering the safe area, the maximum tip diameter, and the maximum scrub length. The worst case is when the Y dimension is in the general direction of the scrub.
The same tolerance also can be used for X with square pads. If a rectangular pad with Ys longer than Xs is used, a tighter Xs tolerance can be used. The alignment tolerance in Y with Ys = 50 micron (safe area), 25-micron tip dia, and 10-micron scrub length is calculated as:
CS-CP = ±½(Ys – tip diameter – ½ scrub length)
= ±(50 microns – 25 microns – 5 microns)/2
= ±10 microns
The parameters affecting the alignment tolerance of the probe card other than simple X, Y positioning are the tip diameter and the scrub length. The tip diameter must be controlled by the probe-card manufacturer.
Unfortunately, with use, the tip is worn away and increases in diameter. The scrub-mark length is directly related to the shape and angle of the probes with respect to the probe card and the overdrive. This means that the tolerance allowed for misalignment and increase in tip diameter is ±10 microns. Obviously, less overdrive results in shorter scrub marks.
In our example, the planarity of the probe tips would be specified within a 20-micron window. The overdrive typically would be specified at 75 microns to ensure that the highest probe would have at least 55 microns of overdrive. This is to ensure good contact to the aluminum pad. If the card can be built to tighter tolerance or the wafer prober controlled more closely, the overdrive and, consequently, the scrub length in the equation could be reduced to give more margin.
Measurement Repeatability
To grade the metrology requirement and performance, we must use statistical means. These involve standard gage calibration studies and process capability measurement studies called P/T analysis.
The P/T analysis determines how much of the specification tolerance of a given measurement is uncertain because of the repeatability of the tool. For example, if P/T= 0.3, then the repeatability of the measurement tool is equal to 30% of the specification window, assuming that the standard used in gathering the data does not contribute to the error.
The term P is the statistical variation (3 or 6 sigma) of the measurement or the repeatability of the measurement. The term T is the specification window or limits placed on the parameter being measured. The alignment tolerance of ±10 microns would result in a 20-micron specification window. The 6 sigma variation would be used for P since this is a double-ended limit as opposed to single-ended.
The repeatability of the measurement tool required to achieve a P/T = 0.3 for this case is shown as:
= ±(T × 0.3)
= ±(20 microns × 0.3)
= ±6.0 microns
A P/T ratio of 0.3 is about the worst case acceptable for measurements on modern probe cards. A tool with repeatability sufficient for P/T = 0.1 is considerably more useful.
For T = 20 microns, the measurement system must perform an alignment test with a 6 sigma repeatability of 2.0 microns or less. This is not an unrealistic expectation.
The data in Table 2 shows how this can be used to detect degradation of probe- card performance with T = 20 microns. The mask standard is a photomask with dots in an array to simulate probe tips. The standard is used to characterize the measurement-system P/T performance.
The good card is a relatively new probe card taken from the production line. The bad card is one taken from the production line after much use, maybe nearing the end of its life. Both of these cards easily passed the test criteria for a good card established for this process.
Analysis of Results
The P/T data shows the advantage of having a very repeatable measurement system when analyzing probe-card alignment data. The goal is to determine how repeatable the probe positions of a good probe card are with respect to a bad probe card.
The capability to see these variations and determine their magnitude is much better if the uncertainty due to the measurement system is low; that is, the system is very repeatable.
The bad card is statistically much worse than the good card, even though both pass their production test specifications.
If X is the repeatability of the measurement system and Y is the distribution of a particular measured parameter, the combined distribution (P) will be approximatelySince the repeatability of the card alignment must be added to any built-in misalignment of the probes, the ±4.6 microns use up most of the ±10-micron alignment specification. If the P/T ratio of the measurement were 0.3 instead of 0.1 or less, this data might be masked in the measurement.
A very low P/T ratio for the measurement system also allows you to reduce the sample size and still get useful data. This can be invaluable in predicting end of life for the probe card and preventing possible expensive yield loss at wafer probe.
In summary, probe-card metrology has evolved to a high degree of sophistication. The seemingly simple probe card requires very precise and repeatable metrology to accommodate today’s sophisticated devices. In this field, a measurement system that can provide repeatable, dependable data is a necessity.
About the Author
Rod Schwartz is vice president and technical director of Integrated Technology, a company that he and his partner, Gary Orman, started in 1975. Previously, he was employed by Texas Instruments, Integrated Circuit Engineering as a consulting engineer, and Bowmar Semiconductor on the startup team. Mr. Schwartz graduated from Iowa State University with a B.S.E.E. and Southern Methodist University with an M.S.E.E. Integrated Technology, 1228 N. Stadem Dr., Tempe, AZ 85281, (602) 968-3459.Test
Measurement
Technology
Description of Test
Alignment
Video
Determine relative X, Y positions of probe tips
Tip Diameter
Video
Measure the diameter of the probe tip
Planarity
Electrical/Mechanical
Determine relative Z positions of all probe tips
Contact Resistance
Electrical
Measure path resistance of probe-card trace, probe needle, and contact to the measurement chuck
Leakage
Electrical
Measure electrical leakage from each probe to all other probes
Gram Force
Electrical/Mechanical
Measure the spring constant and force exerted by probe at a given overdrive or deflection
Wire Check
Electrical
Check the electrical connection path for each probe to determine if the card is wired correctly
Components
Electrical
Measure the actual values of discrete capacitors and resistors on the probe card
TABLE 1
Sample
Description
P/T Ratio
P (6 sigma)
1
Mask Standard
0.078
1.56 microns
2
Good Card
0.124
2.48 microns
3
Bad Card
0.243
4.86 microns
TABLE 2
Copyright 1998 Nelson Publishing Inc.
September 1998