Uploaded by 1973376779

Probe Card Mark Analysis Case Studies 2010

advertisement
Probe Card Analysis and Probe Mark Analysis
case studies from the sort floor
Rudolph Technologies, Inc.
INTRODUCTION
Probe card analysis (PCA) and probe mark analysis (PMA) allow engineers to optimize the performance of the equipment and
procedures they use to test the electrical functionality and measure the performance of integrated circuits after the devices have
completed the wafer fabrication process but before they have been diced and packaged. PCA tools examine the physical probes
used to make mechanical/electrical contact with test pads on circuits. The probes are carried by a probe card and there may be
anywhere from a few tens to tens of thousands of probes on a single card. PCA can determine the precise location, size, shape,
overtravel deflection and more for each probe in the array.
Probes are designed to “scrub” across the test pad surface as the probe card is moved from the nominal contact z position to
the overtravel z position. The scrubbing action breaks through any surface oxide or contamination that may be present, thereby
ensuring good electrical contact. The scrubbing action also creates a probe mark which carries important information about the
probing process. Probe mark inspection (PMI) has long been used by semiconductor manufacturers to detect defective marks
that may interfere with device operation. Probe mark analysis differs from probe mark inspection in its intent to derive from the
marks statistically valid information that can be used to improve the performance of the probing process. It does this primarily by
detecting patterns in the data that would be difficult or impossible to find manually. PMA might identify a specific probe in the
array that is creating defective marks, or determine from the pattern of position errors that the probe card is tilted with respect to
the wafer, or even that the screws holding the wafer chuck have been loosened by temperature changes and probing forces.
Together PCA and PMA provide an extremely powerful toolset for investigating and optimizing almost every aspect of
probing operations. They can be used to monitor and detect excursions in the process; to validate new processes, such as testing
at temperature, pad shrinks, new pad materials; to qualify new or reconditioned probe cards; to compare the performance
of alternative probe technologies; to qualify new equipment; to make tools-to-tool comparisons, to monitor maintenance
requirements; to evaluate and select new tools; and more.
This paper offers PCA and PMA examples drawn directly from practical experience on the sort floor.
REAL TIME PROCESS MONITORING
Figure 1
PMA provides a wealth of information about the probing/
sort process making it easy to identify and isolate errors from
multiple sources. Figure 1 shows two wafer maps plotting “left
distance”, the minimum distance between the left edge of
the test pad and the probe mark. The right map was acquired
four hours after the left map. The probe card in this case was
smaller than the wafer and required six touch downs to cover
the whole wafer surface. A number of observations can be
made. Left distance differs by approximately 4 µm between
the first (upper left) and third (middle left) touchdowns, a
wafer load error. Left distance differs by approximately 7 µm
between touchdowns on the left and right halves of the wafer,
a prober error. Left distance differs by approximately 5 µm
Page 1 of 7
2010
between the left and right sides of the probe array, a probe
card error. Left distance changes by approximately 7 µm
between the left map and the right map, due to thermal drift
over the 4 hour elapsed time.
Wafer Fabrication Process Monitor
of the reference surface. The parallelism of the chuck to the
reference surface and the response of all components to the
compressive forces generated by overtravel are important
parameters routinely characterized in overtravel optimization.
In this case, the discrepancy in performance was ultimately
traced to differences in the flatness of the card rings in the two
testers.
VALIDATING NEW PROCESSES
Thermal Movement of Probe Card
Figure 2
Figure 2 plots variations in probe mark size (scrub area) over
the wafer surface and (right) displays the detected extent of
one mark within the outline of the probe pad. The process
and error components had been well characterized and an
increase in probe mark area was quickly identified as an outof- tolerance condition, which triggered further review and
evaluation. The increase in probe mark size was traced to a
discoloration of the pad, which was in turn linked to an error
in the fabrication process.
Controlling the Sort Process
Figure 4
Metrology is a key tool when testing devices at operating
temperature (test-at-temperature). Since everything moves as
it heats, it is critical to isolate and measure all contributing
components: needle movement, array movement, pad
movement, etc. What happens to the support material,
flatness and planarity of the array, the reference surface, or the
prober chuck? Do they all move in the same direction or does
overall planarity change? With metrology all of these variables
can be measured and a steady state or optimal process defined.
Figure 4 (left) shows probe array z-movement of up to 110µm
in response to heating. The images on the right show marks
made by individual probes in nine different areas of the wafer.
The marks drift within the pad as a result of different thermal
scaling coefficients between the wafer, the probeneedle/card
and the prober. Individual probe drift varies from 3 to 15 µm.
Thermal Movement of Probe Card
Figure 3
Plots of continuity vs overtravel (Figure 3) demonstrated a
significant overtravel needed to make full electrical contact
with 2 different probe cards, PC10 and PC11. The scrub
signature corresponds to a flatness issue - the reference surface
is not flat. The plots at bottom left and right demonstrate
excessive pad damage. The bottom middle shows the flatness
Figure 5
Page 2 of 7
2010
Figure 5 compares “Y Scale” measurements made on two
wafers at different temperatures 88 °C (left) and 150°C (right).
Y scale measures the accuracy of the die-to-die step size in the
y direction. This test used an 8 X 1 probe card array (outlined
in black). The 88°C measurement shows a Y scale error
range of 7 µm. The range increases to 33 µm at 150°C.
Process modifications will be needed to support test at
different temperatures.
Validating Pad Shrink
Thermal Drift
Figure 8
Figure 6
Probe marks were found to be drifting off the pads (Fig. 6)
during the probing process because of thermal scaling. PMA
identified a thermal drift signature. The standard serpentine
test pattern did not allow for temperature stabilization. A
different pattern provided more stable thermal performance of
the probe card.
Supporting a pad shrink can be a daunting problem. First,
the total error of the process must be broken down into
its components. The left side of Figure 8 shows the total
error (Cpk) of the process. The right side shows the error
associated with just the probe card. In this case the probe
card is performing well, so it can be eliminated from the
investigation and other error sources can be pursued.
Probe Card Verification
Probe Card Verification at Temperature
Figure 9
Figure 7
Figure 7 shows hot (left) and cold (right) data collected while
verifying a probe card at temperature. The Scrub X/Y position
error plots (upper) use a vector display mode that clearly shows
patterns in the magnitude and direction of the error. This card
cannot be qualified to run at multiple temperatures because
the scaling is not linear across the temperature ranges.
The key to a stable process is consistency among tool sets.
Getting all of the tools sets to match some standard is the
key. The standard can be somewhat arbitrary and may not
necessarily be the same standard used by the tool supplier.
Being able to quickly adjust incoming probe cards to an
appropriate standard is key to maintaining the overall process.
Figure 9 illustrates a typical verification process flow, in this
case resulting in a significant improvement in probing process
stability.
Page 3 of 7
2010
PROBE CARD QUALIFICATION
Validating New Probe Technology
Modeling or understanding how the probe card performs
under load is key to maintaining a uniform process.
Using metrology equipment it is possible to model probe
card deflection and optimize the programmed overtravel
appropriately. Incorrect overtravel leads to a variety of sort
related issues, including low Cres, under-pad metal cracking,
punch through, excessive probe wear, excessive cleaning,
excessive debris, etc. Figure 12 is a deflection plot, showing
high probes moving up with the fixture as load increases even
at this low overtravel (1 mil), this reduces the applied Z on
high probes by 5µm. This will affect the scrub and CRES
performance of the high probes.
Automated Deflection Test
Figure 10
Figure 10 shows the probe card signature (vector view of
scrub X-Y position) acquired for a new card using a blank,
unpatterned wafer. The plot shows a strong thermal scaling
effect. This card did not pass incoming inspection.
Validating Probe Card Design
Figure 13
Figure 13 plots (Probe Card from Figure 12) the expected
planarity versus the actual planarity at load with probes sorted
from lowest to highest. High needles translate up as increasing
load causes card deflection.
Figure 11
PROBE TECHNOLOGY
Comparing Probe Card Technologies
In Figure 11, the probe card showed inconsistent results at
temperature. Analysis determined that the change occurred
after the prober stopped to check probe marks. This allowed
the probe card to cool down causing a shift in scrub
performance. The root cause for the rapid cool down was
found to be the probe card stiffener, which was acting as a
heat sink. The left image is the original stiffener; the right
image is the same card with a new stiffener design. Changing
the design improved the overall performance.
Automated Deflection Test
Figure 14
Gauge studies provide a means to quantify measurement
capability when comparing different technologies (Fig. 14)
Figure 12
Page 4 of 7
2010
Comparing Probe Card Technologies
Translation Effects
Figure 15
Figure 15 compares the scrub lengths measured for two
different probe card technologies, cantilever Dual DUT (left)
and MEMS Quad DUT (right).
Comparing Probe Card Technologies
Figure 18
Some marks near the edges of the die were failing probe mark
inspection. In particular, the area, shape and direction of
marks left by the same probe were changing from touchdown
to touchdown. In Fig. 18, the lower images show probe marks
from the same needle in a touchdown on the left and right
side of the wafer. These changes were found to be effects of
probe card translation and deflection.
Figure 16
MONITORING MAINTENANCE
Figure 16 compares probe mark area measured for two
different tip types of cantilever probe cards using the same
prober, prober settings, time and temperature. The probe tip
on the right clearly leaves significantly smaller marks. Tip
shape has a large impact on scrub performance for different
types of applications.
Equipment Problems
EQUIPMENT COMPARISON
Prober to Prober
Figure 19
Figure 17
Prober to prober comparisons can help to standardize tools
across the sort floor. Figure 17 shows the results of a test run
to determine prober to prober variance. The test used the
same card, prober settings, time and temperature. Errors in
chuck tilt and head stage to chuck parallelism were found and
fixed in the left tool.
The wafer map in Figure 19 plots “left distance”, the
minimum distance from the scrub mark edge to the left pad
edge. The observed pattern is typical of wafer scaling errors,
which result from a mismatch between prober step size and
die spacing on the wafer, in this case in the x direction. The
error was immediately apparent from the prober report chart.
Page 5 of 7
2010
Validating Maintenance Problems
The “swirl” pattern apparent in Figure 22 (left) indicates
errors in prober setup and alignment. Adjustments to the
prober setup resulted in a 50% improvement (right).
Probe Card Setup Errors
Figure 20
In Figure 20, the left wafer map of scrub X/Y position shows
a pattern typically caused by loosening of the screws used
to adjust the chuck planarity the wafer chuck to the prober.
Retesting after tightening the screws (right map) confirmed the
source of the problem and the efficacy of the corrective action.
Stage Stepping Accuracy
Figure 23
Incorrect setup by the operator caused an X-axis offset error.
Probe Card Setup Errors
Figure 21
The pattern shown in Figure 21 indicates scaling and offsets
not applied correctly by the prober.
Figure 24
Prober stage error resulted in theta alignment issues
Setup Validation
Figure 22
Page 6 of 7
2010
TOOL EVALUATION
Prober vs Probe Card
Figure 22
The wafer plot on the left shows inconsistent pad damage
which will ultimately result in higher failure rates. The data
were acquired using a standard vertical probe card on which
all needles scrub in the same direction. The resulting force
caused stage translation. A new vertical head was designed on
which half of the needles scrub in the opposite direction, thus
balancing out the translational force applied to the stage. The
wafer map on the left shows the improved performance (reduced
and more consistent damage) achieved with the new design.
www.rudolphtech.com
1 952 820 0080
info@rudolphtech.com
Page 7 of 7
2010
Download