Title: Methods and systems employing infrared thermography for defect detection and analysis
Abstract: Described are methods and systems for providing improved defect detection and analysis using infrared thermography. Test vectors heat features of a device under test to produce thermal characteristics useful in identifying defects. The test vectors are timed to enhance the thermal contrast between defects and the surrounding features, enabling IR imaging equipment to acquire improved thermographic images. In some embodiments, a combination of AC and DC test vectors maximize power transfer to expedite heating, and therefore testing. Mathematical transformations applied to the improved images further enhance defect detection and analysis. Some defects produce image artifacts, or "defect artifacts," that obscure the defects, rendering difficult the task of defect location. Some embodiments employ defect-location algorithms that analyze defect artifacts to precisely locate corresponding defects.
Patent Number: 6,840,666 Issued on 01/11/2005 to Enachescu, et al.
| Inventors:
|
Enachescu; Marian (Richmond, CA);
Belikov; Sergey (Palo Alto, CA)
|
| Assignee:
|
Marena Systems Corporation (Hayward, CA)
|
| Appl. No.:
|
348940 |
| Filed:
|
January 22, 2003 |
| Current U.S. Class: |
374/5; 250/341.4; 250/341.6; 324/501; 324/765; 324/770; 374/124; 382/149 |
| Intern'l Class: |
G01N 025/72; G01R031/02; G01R031/318.3 |
| Field of Search: |
250/341.1,341.4,341.6,342
374/4,5,57,121,124
324/770,750,752,765,501
382/149,151
345/87-90,92,93,904,207,211-214
|
References Cited [Referenced By]
U.S. Patent Documents
| 3991302 | Nov., 1976 | Danner | 235/151.
|
| 4733175 | Mar., 1988 | Levinson | 374/124.
|
| 4755874 | Jul., 1988 | Esrig et al. | 382/149.
|
| 5032727 | Jul., 1991 | Cox et al. | 374/5.
|
| 5208528 | May., 1993 | Quintard | 250/341.
|
| 5250931 | Oct., 1993 | Misawa et al. | 345/92.
|
| 5294198 | Mar., 1994 | Schlagheck | 374/4.
|
| 5309108 | May., 1994 | Maeda et al. | 324/501.
|
| 5495535 | Feb., 1996 | Smilansky et al. | 382/145.
|
| 5504438 | Apr., 1996 | Henley | 324/770.
|
| 5740272 | Apr., 1998 | Shimada | 382/149.
|
| 5774573 | Jun., 1998 | Caspi et al. | 382/141.
|
| 5775806 | Jul., 1998 | Allred | 374/5.
|
| 5982945 | Nov., 1999 | Neff et al. | 382/271.
|
| 6111424 | Aug., 2000 | Bosacchi | 324/770.
|
| 6269194 | Jul., 2001 | Nichani | 382/270.
|
| 6288782 | Sep., 2001 | Worster et al. | 356/394.
|
| 6289492 | Sep., 2001 | Dutta-Choudhury et al. | 716/5.
|
| 6381355 | Apr., 2002 | Goonetilleke | 382/141.
|
| 6381372 | Apr., 2002 | Loce | 382/261.
|
| 6389163 | May., 2002 | Jodoin et al. | 382/173.
|
| 6411731 | Jun., 2002 | Saito | 382/173.
|
| 6456899 | Sep., 2002 | Gleason et al. | 700/212.
|
| 6714017 | Mar., 2004 | Meier et al. | 324/501.
|
| 2002/0093354 | Jul., 2002 | Enachescu et al. | 324/760.
|
Other References
"Thermography Receives a Warm Welcome from Microelectronics Industry," by
Doug Little. EE-Evaluation Engineering, Aug. 1999. 3 pages.
"Thermography in the Microelectronics Industry," by Chris Alicandro and
Doug Little. EE-Evaluation Engineering, Aug. 1999. 2 pages.
|
Primary Examiner: Gutierrez; Diego
Assistant Examiner: Pruchnic, Jr.; Stanley J.
Attorney, Agent or Firm: Silicon Edge Law Group LLP, Behiel; Arthur J.
Parent Case Text
This application claims benefit of Ser. No. 60/350,074 filed Jan. 23, 2000
and claims benefit of Ser. No. 60/350,055 filed Jan. 23, 2002.
Claims
What is claimed is:
1. A test configuration comprising:
a. a display panel including:
i. a plurality of source lines;
ii. a plurality of control lines;
iii. a plurality of common lines;
iv. a two-dimensional array of display elements, each display element
including a transistor and a capacitor, wherein the transistor has a first
current-handling terminal connected to one of the source lines, a second
current-handling terminal, and a control terminal connected to one of the
control lines, and wherein the capacitor has a first capacitor terminal
connected to the second current-handling terminal and a second capacitor
terminal connected to one of the common lines; and
v. a defect amid surrounding features;
b. an infrared detector positioned to receive infrared radiation from the
display panel, including infrared radiation from the defect and
surrounding features;
c. a signal generator having:
i. a first test-signal output terminal connected to at least one of the
source lines;
ii. a second test-signal output terminal connected to at least one of the
control lines; and
iii. a third test-signal output terminal connected to at least one of the
common lines;
iv. wherein the signal generator simultaneously applies a first test vector
on the first test vector output terminal, a second test actor on the
second test-signal output terminal, and a third test vector on the third
test-signal output terminal;
v. wherein application of the first, second, and third test vectors heats
the defect and surrounding features; and
vi. wherein the detector captures a thermal image of the heated display,
the thermal image including defect data representative of the defect and
defect-artifact data representative of the surrounding features; and
d. an image processor analyzing the defect-artifact data to locate the
defect data within the defect-artifact data.
2. The test configuration of claim 1, wherein at least one of the first and
third test vectors is an AC signal.
3. The test configuration of claim 2, wherein the second test vector is a
DC signal biasing on the transistor.
4. The test configuration of claim 1, wherein the second test vector turns
the transistor on.
5. The test configuration of claim 1, wherein the signal generator applies
a test vector across the source and common lines to develop a test current
through the first and second current-handling terminals of the transistor.
6. A test configuration comprising:
a. a display panel including:
i. a plurality of source lines;
ii. a plurality of control lines;
iii. a plurality of common lines;
iv. a two-dimensional array of display elements, each display element
including a transistor and a capacitor, wherein the transistor has a first
current-handling terminal connected to one of the source lines, a second
current-handling terminal, and a control terminal connected to one of the
control lines, and wherein the capacitor has a first capacitor terminal
connected to the second current-handling terminal and a second capacitor
terminal connected to one of the common lines;
b. an infrared detector positioned to receive infrared radiation from the
display panel; and
c. a signal generator having:
i. a first test-signal output terminal connected to at least one of the
source lines;
ii. a second test-signal output terminal connected to at least one of the
control lines; and
iii. a third test-signal output terminal connected to at least one of the
common lines;
iv. wherein at least one of the first and third test vectors is of a test
frequency and encounters a resistance and a capacitance, and wherein the
test frequency is matched to the capacitance and the resistance to provide
maximum power transfer from the signal generator to the panel.
7. A test configuration comprising:
a. an electrical circuit having at least one defect amid surrounding
features and disposed within a defect region, the defect region having an
initial temperature;
b. an infrared detector positioned to receive infrared radiation from the
defect region, including infrared radiation from the defect and
surrounding features, wherein the detector captures a thermal image of the
defect region, the thermal image including defect data representative of
the defect and defect-artifact data representative of the surrounding
features;
c. a signal generator having at least one test-signal output terminal
connected to the electrical circuit, the signal generator applying a test
vector to the defect via the electrical circuit, wherein the defect
exhibits a thermal response to the applied test vector; and
d. an image processor analyzing the defect-artifact data to isolate the
defect data within the defect-artifact data;
e. wherein the infrared detector captures an image of the defect after
application of the test vector and before the defect reaches 95% of a
difference between the initial temperature and a final equilibrium
temperature.
8. The test configuration of claim 7, wherein the infrared detector
captures the image before the defect reaches about 86.5% of the
difference.
9. The test configuration of claim 7, wherein the infrared detector
captures the image before the defect reaches about 63.5% of the
difference.
10. The teat configuration of claim 7, wherein the signal generator
includes a second test-signal output terminal, the electrical circuit
comprising a transistor having a first current-handling terminal connected
to the first-mentioned test-signal output terminal, a cortrol terminal
connected to the second test-signal output terminal, and a second
current-handling terminal.
11. The teat configuration of claim 10, wherein the test vector is an AC
signal.
12. The test configuration of claim 11, wherein the signal generator
applies a second test vector to the control terminal via the second
test-signal output terminal.
13. The test configuration of claim 12, wherein the second test vector
turns the transistor on.
14. The test configuration of claim 10, wherein the signal generator
includes a third test-signal output terminal, the test configuration
further comprising a capacitance connected between the second
current-handling terminal of the transistor and the third test-signal
output terminal.
15. A method for identifying a defect on an electrical circuit, the method
comprising:
a. applying test vectors, at a first instant, to the defect via the
electrical circuit, wherein the defect exhibits a thermal response to the
test vectors, the thermal response being characterized by a thermal time
constant;
b. capturing a thermal image of the electrical circuit, the thermal image
including representations of the defect and a defect artifact, at a second
instant separated from the first instant by less than three time
constants; and
c. analyzing the representation of the defect artifact.
16. The method of claim 15, wherein the second instant is separated from
the first instant by less a than two time constants.
17. The method of claim 15, wherein the second instant is separated from
the first instant by less a than one time constants.
18. The method of claim 15, further comprising repeating steps (a) and (b)
at least twice to capture a plurality of thermal images.
19. The method of claim 18, further comprising capturing a reference image
between ones of the plurality of thermal images.
Description
BACKGROUND
Electrical circuits, such as printed circuit boards (PCBs), integrated
circuits, and flat-panel displays (FPDs), may be tested for defects using
infrared (IR) thermography. In general, power is applied to a device under
test (DUT) to heat various of the device features. An infrared detector
then captures a test image of the heated DUT. The resulting image, a
collection of pixel-intensity values spatially correlated to the imaged
object, is then compared with a similar collection of reference image
data. Differences between the test and reference data, typically stored as
a "composite image," indicate the presence of defects.
Defect identification algorithms analyze composite images to automatically
identify defects, and consequently improve throughput and quality in
device manufacturing. Examples of such inspection systems include, but are
not limite to, inspection of FPDs, PCBs, MEMS-based devices, semiconductor
devices, and biomedical specimen. One purpose of such systems is to test
for defects potentially present on a device at some critical point during
manufacture of that device. Once identified, the defects can then be
repaired by a repair system, or a choice can be made to reject the device,
leading to manufacturing cost savings in both cases. Other applications
include inspection and identification of artifact-like features in
research specimens, e.g. in biology.
One particularly important use of IR thermography is the testing of the
active layer, or "active plate," of liquid-crystal display (LCD) panels.
Defect analysis can be used to improve processing and increase
manufacturing yield. Also important, defective panels can be repaired,
provided the number and extent of defects are not too great, again
increasing manufacturing yield.
FIG. 1 (prior art) depicts portions of an active plate 100 for use with an
LCD panel. (FIG. 1 was taken from U.S. Pat. No. 6,111,424 to Bosacchi,
which issued Aug. 29, 2000, and is incorporated herein by reference.)
Active plate 100 includes a first shorting bar 105 connected to each pixel
in an array of pixels 110 via a collection of source lines 115 and a
second shorting bar 120 connected to each pixel 110 via a collection of
gate lines (control lines) 125.
According to Bosacchi, active plate 100 is tested by evaluating the IR
emission of active plate 100 with voltage applied to shorting bars 105 and
120. With power thus applied, portions of plate 100 operate as resistive
circuits, and consequently dissipate heat. The heating response
characteristics of plate 100 are then evaluated, preferably after plate
100 reaches a stable operating temperature (thermal equilibrium).
In the absence of defects, the pixel array should heat up uniformly.
Non-uniform thermal characteristics, identified as aberrant IR intensity
values, therefore indicate the presence of defects. Reference intensity
values can be obtained by averaging the pixel intensity values of a given
image frame, or by means of a reference frame corresponding to an ideal or
defect-free reference plate.
FIG. 2 (prior art) details a portion of a conventional pixel 110, and is
used here to illustrate a number of potential defects. The depicted
features of pixel 110 are associated with the active plate of a
liquid-crystal display, and include a thin-film transistor 200 having a
first current-handling terminal connected to one of source lines 115, a
control terminal connected to one of gate lines 125, and a second
current-handling terminal connected to a capacitor 210. The second
electrode of capacitor 210 connects to a common line 212. Pixel 110 also
includes a second capacitor 211 having a liquid-crystal dielectric.
The defects, which are illustrative and not exhaustive, include both shorts
and opens. The shorts are between: source line 115 and gate line 125
(short 215) or common line 212 (short 216); the two current-handling
terminals of transistor 200 (short 220); the gate and second
current-handling terminal of transistor 200 (short 225); and the two
terminals of capacitor 210 (short 226). The opens segment the source,
gate, and common lines (opens 227, 228, and 229), and are between: source
line 115 and transistor 200 (open 230), gate line 125 and the control
terminal of transistor 200 (open 232), capacitor 210 and common line 212
(open 235), and transistor 200 and capacitor 210 (open 233).
Each defect of FIG. 2, plus a number of others, adversely impacts the
operation of pixel 110. Unfortunately, many of these defects are difficult
to discover using conventional test methods. There is therefore a need for
improved methods and systems for identifying and locating defects.
Some inspection systems include an excitation source that excites the
object under test in a way that highlights defects to an imaging system.
The type of excitation depends upon the imaging system, which may acquire
images based on visible light, infrared, combined spectroscopy, magnetic
fields, etc. Whatever imaging system is employed, test images of the
object under test are contrasted with some reference image to obtain a
composite image: significant differences between the test and reference
images show up in the composite image, and identify potential defects.
Some forms of excitation produce defect artifacts, which are differences
between test and reference images that are caused by defects but that do
not physically correlate to defect areas. A short between two lines
increases current through those lines, and consequently elevates the
temperatures of the lines along with the short. Thus the lines, though not
themselves defective, nevertheless appear with the short in the composite
image. The defect data representing the short is thus imbedded within
defect-artifact data (i.e., a "defect artifact"). Defect artifacts often
obscure the associated defects, rendering them difficult to precisely
locate. Human operators can locate a defect within a defect artifact by
careful study under a microscope, but people are relatively slow and are
quickly fatigued. There is therefore a need for means of automatically
distinguishing defects from their related artifacts.
BRIEF DESCRIPTION OF THE FIGURES
FIG. 1 (prior art) depicts portions of an active plate 100 for use with an
LCD panel.
FIG. 2 (prior art) details an exemplary pixel 110, and is used here to
illustrate a number of potential defects.
FIG. 3 depicts a test configuration 300, including a conventional panel 305
and an inspection system adapted in accordance with an embodiment of the
invention.
FIG. 4 depicts a portion of a panel 400 adapted in accordance with one
embodiment to provide enhanced testability.
FIG. 5 depicts a portion of an LCD panel 500 adapted in accordance with one
embodiment.
FIG. 6A is a diagram 600 illustrating the thermal response of the
above-described sample defect and the surrounding area.
FIG. 6B depicts a diagram 630 illustrating the thermal response of sample
defect 610.
FIG. 6C depicts an experimentally obtained image 680 highlighting a line
685 indicative of an open.
FIG. 7 depicts a composite image 700 showing three representative defects,
a point-type defect 705, a line-type defect 707, and a corner-type defect
710.
FIG. 8 is a composite image 800 exhibiting a point-type defect 805, a
line-type defect 810, and a corner-type defect 815.
FIG. 9 is a flowchart 900 depicting a defect-location algorithm 900 in
accordance with one embodiment.
FIG. 10 depicts an embodiment of MFA 910 of FIG. 9.
FIG. 11 depicts an embodiment of type-specific MFA 935 of FIG. 9 adapted
for use with line-type defect artifacts.
FIG. 12 depicts an embodiment of defect-location algorithm (DLA) 945 of
FIG. 9.
FIG. 13 depicts an array of LDF structures 1205 in accordance with one
embodiment.
FIG. 14 depicts an illustrative filtered composite image 1400 similar to
what one might expect from step 907 of FIG. 9.
FIGS. 15A-15D depict binary, type-specific images 1505, 1510, and 1515 like
images 940[1 . . . i] of FIG. 9.
FIG. 16 depicts a tree 1600, a compilation of the information in keys 1525,
1530, and 1535 of FIGS. 15B-D.
FIG. 17 is an illustrative peak profile 1700 showing the relationship
between four defect artifacts 1705, 1710, 1715, and 1720.
DETAILED DESCRIPTION
FIG. 3 depicts a test configuration 300, including a conventional panel 305
and an inspection system 310 adapted in accordance with an embodiment of
the invention. Panel 305 is similar to panel 100 of FIGS. 1 and 2,
like-numbered elements being the same or similar. Panel 305 includes a
shorting bar 312 that is not depicted in FIG. 1, but is nevertheless
conventional. Inspection system 310 includes an IR detector 315 (e.g., an
IR camera) oriented over panel 305 to provide image data to a computer 320
via a frame grabber 325. An excitation source, signal generator 330,
provides electrical test signals, or "test vectors," to panel 310. The
test vectors heat features of panel 310 to produce thermal characteristics
useful in identifying defects.
Computer 325 controls signal generator 330 to apply test vectors to panel
305. These test vectors enhance the thermal contrast between defects and
the surrounding features, and consequently allow IR detector 315 to
acquire improved thermographic images for defect detection and analysis.
Computer 320 additionally instructs IR detector 315 when to acquire image
data, receives and processes captured test-image data from frame grabber
325, and provides a user interface (not shown).
IR detector 315 should have excellent temperature sensitivity. In one
embodiment, detector 315 is an IR Focal-Plane Array Thermal Imaging Camera
employing a 256.times.320 element InSb (Indium Antimonide) detector. The
minimum temperature sensitivity of this camera is less than 0.020 degrees
C. Some embodiments include multiple IR detectors, for example a
relatively low magnification IR camera for defect detection and a higher
magnification IR camera for defect detection and analysis. Additional
cameras might also be used to increase inspection area, and thus
inspection bandwidth.
Signal generator 330 provides a source test vector V.sub.TS to shorting bar
105, a gate test vector V.sub.TG to shorting bar 120, and a common test
vector V.sub.TC to shorting bar 312. Referring back to FIG. 2, testing for
some types of defects (e.g., opens 230, 233, and 235) requires transistor
200 be turned on to create a signal path between respective source and
common lines 115 and 212. Signal generator 330 thus applies a DC test
vector V.sub.TG to gate line 125 (via shorting bar 120), thus turning
transistor 200 on, while applying the source and common test vectors
V.sub.TS and V.sub.TC.
Even with transistor 200 forward biased, a non-defective pixel 110 will not
pass direct current, absent short 226, because capacitor 210 blocks direct
current. The source and common test vectors V.sub.TS and V.sub.TC are
therefore selected to produce an AC signal that passes through capacitor
210. The frequency of the AC signal is matched to the impedance of the
load provided by panel 305 to maximize power transfer to panel 305, which
expedites heating, and therefore testing. Maximizing power transfer also
allows for testing with lower applied voltages, which are less likely to
damage sensitive components. Also important, as detailed below, a
combination of faster heating and specific image-capture timing provides
improved thermal contrast. In one embodiment, source test vector V.sub.TS
oscillates from zero to 30 volts at about 70 KHz and common test vector
V.sub.TC is ground potential.
Some embodiments apply either AC or DC test vectors between source line 115
and common line 212, source line 115 and gate line 125, and between gate
line 125 and common line 212. Still other embodiment employ AC signals to
turn transistor 200 on. The simultaneous application of AC and DC test
vectors, as detailed above, facilitates more comprehensive testing than is
obtained by application of only one type of waveform (e.g., only DC, AC,
or pulsed DC test vectors).
FIG. 4 depicts a portion of a panel 400 adapted in accordance with one
embodiment to provide enhanced testability. Panel 400 conventionally
includes an array of pixels 405, each connected to a source line 410, a
gate line 415, and a common line 420. Four sets of shorting bars (source
bars 425, gate bars 435, and common bars 430) allow inspection systems,
such as that of FIG. 3, to test subsets of pixels 405. Alternatively, four
sets of one type of bar (e.g., four source bars or four gate bars) can be
used to energize selected columns or rows. Energizing some features while
leaving adjacent features de-energized can improve image contrast. In
other embodiments, only one or two types of shorting bars are provided in
sets. There may only be one gate bar 435 and one common bar 430, for
example, in which case pixels 405 can be excited in four sets of columns.
Moreover, one or more sets of shorting bars can include more or fewer than
four shorting bars.
FIG. 5 depicts a portion of an LCD panel 500 adapted in accordance with
another embodiment. Panel 500 conventionally includes an array of pixels
505, each connected to a source line 510, a gate line 515, and a common
line 520. Source bars 525, gate bars 530, and common bars 535 are
segmented to allow an inspection system to power test areas (e.g., area
540) one at a time. Alternatively, fewer types of bars need be segmented.
For example, the common bar need not be segmented to power area 540 if the
source and gate bars are segmented. Area 540 might be coextensive with the
field of view of the IR detector used to capture images. The number of
pixels 505 in a given area is generally much greater than depicted in this
simple example.
FIG. 6A is a diagram 600 illustrating the thermal response of an
illustrative sample defect and the surrounding area. The sample defect is
assumed to be a short having a resistance R of about twenty-five thousand
ohms, a volume V of the combined defect and associated electrodes of about
10.sup.-12 m.sup.3, and an exposed suface area A of about 10.sup.-5
m.sup.2. The specific heat Cp of the electrodes is assumed to be about
2.44.times.10.sup.6 J/m.sup.3 K, and the convection heat transfer
coefficient of the surronding air h.sub.air is about 10 W/m.sup.2 K. For
an applied power of about 6 milliwatts the equilibrium temperature at the
defect location is about 6.5 degree C. above the initial temperature. The
following heat transfer model illustrates the thermal response of the
sample defect in response to applied power:
##EQU1##
where:
1. V is the volume of the combined defect and associated electrodes;
2. Cp is the average specific heat of the defect and associated electrodes;
3. T(t) is temperature, in Kelvin, of the defect over time, in seconds;
4. P.sub.applied (t) is the applied power excitation over time, in seconds;
5. h.sub.air is the convection heat transfer coefficient of the surrounding
air;
6. A is the exposed surface area of the combined defect and associated
electrodes; and
7. T.sub.air is the temperature of the surrounding air or initial
temperature (e.g., about 300K).
Equation (1) means, in essence, that the power applied to the defect area
is, at a given instant, equal to the sum of the power absorbed by the
defect area and the power dissipated into the surrounding environment.
Initially, when the temperature difference between the defect area and
surrounding environment is nominal, the first added dominates the
equation. The second addend gains influence as the temperature of the
defect area rises.
A diagram 605 illustrates the sample defect area 610 and surrounding area
615 as a collection of boxes, each box representing the image intensity
recorded by an image pixel. For ease of illustration, diagram 605
illustrates defect area 610 as a single pixel. Diagram 600 assumes about
six milliwatts is applied between the source and gate lines for a time
sufficient for defect area 610--a short between the source and gate
lines--to rise from an initial thermal equilibrium temperature T.sub.I of
about 300K to a final equilibrium temperature TE.sub.F of about 306.5K.
Under the specified conditions, traversing this temperature window takes
about 0.1-0.2 seconds for a typical active plate.
A first response curve 620 illustrates the thermal response of defect area
610. The vertical axis of diagram 600 represents temperature as a
percentage of the temperature span between initial temperature T.sub.I and
final equilibrium temperature TE.sub.F of response curve 620. The
horizontal axis represents time, in thermal time constants .tau.. The
thermal time constant .tau. is the time required for the temperature of
defect area 610 to rise 63.2% of the way from a given temperature to final
equilibrium temperature TE.sub.F. For practical purposes, the defect
temperature is at the final equilibrium temperature TE.sub.F after four or
five time constants .tau..
The thermal response of area 615 surrounding defect area 610 differs from
the thermal response of defect area 610. If the defect is a short, heat
from area 610 diffuses into area 615, causing the temperature of area 615
to rise with defect 610. The rising temperature of area 615 lags that of
defect 610, however, and rises to a lower final thermal equilibrium
temperature than defect 610. Test vectors applied in accordance with some
embodiments provide increased temperature contrast between areas 610 and
615 to allow IR imaging systems to more easily resolve defect features. IR
inspection systems then employ unique image-acquisition timing to capture
test image data well before defect area 610 reaches final equilibrium
temperature TE.sub.F. (If defect area 610 is an open, the temperature of
area 610 lags surrounding area 615, but nevertheless eventually reaches a
final equilibrium temperature.)
FIG. 6B includes a diagram 630 illustrating test vectors and
image-acquisition timing that enhance thermal contrast between defects and
the surrounding areas. Diagram 630 includes a thermal response curve 640
representing the repetitive heating and cooling of defect area 610 in
response to test vectors. FIG. 6B additionally includes a pair of
waveforms IMAGE and EXCITE that share a common time scale with diagram
630. The high portions of waveform IMAGE represent windows of time during
which IR images of areas 610 and 615 are captured. One or more reference
images are captured during each reference window 645, while one or more
test images are captured during each test window 650. The high portions
655 of waveform EXCITE represent times during which test vectors are
applied to defect 610 to introduce thermal contrast between defect 610 and
surrounding area 615.
To capture test images, an inspection system (e.g. inspection system 310 of
FIG. 3) applies test vectors to a device under test during times 655. The
inspection system then captures one or more IR images of the defect area
well before the features of interest reach the final thermal equilibrium
temperature TE.sub.F.
It is desirable to keep the maximum temperature (i.e., the peaks of
response curve 640) below 95% of the difference between the initial
temperature T.sub.I and the final equilibrium temperature TE.sub.F. In
some cases, the maximum temperature may even endanger DUT functionality.
The optimal upper limit for temperature thus varies for different DUTs,
test procedures, etc., but is preferably less than 86.5% in many cases.
Our experimental data suggest that excellent results are obtained when the
maximum temperature does not exceed about 63.5% of the difference between
the initial temperature T.sub.I and the final equilibrium temperature
TE.sub.F (i.e., before the passage of one time constant). The
heating/imaging steps are repeated a number of times, and the results
averaged or otherwise combined, to reduce the effects of noise. The
maximum and minimum peaks of response curve 640 should be sufficiently
spaced for the selected detector to resolve the temperature difference.
Defect 610 can be heated to higher temperatures, for more than three or
four time constants, for example; in such cases, the images can still be
taken well before defect 610 reaches the final equilibrium temperature
TE.sub.F. Because heating takes time, selecting relatively low maximum
temperatures speeds testing. Moreover, the test voltages selected to
maximize image contrast may, if applied too long, raise the temperatures
of areas 610 and 615 high enough to damage sensitive components. In such
cases, the test vectors are applied long enough to achieve a desired level
of thermal contrast without raising the temperatures of areas 610 and 615
above some maximum temperature.
The applied test vectors differ depending in part on the DUT. In one
embodiment suitable for testing an active panel from a liquid-crystal
display (LCD) having SXGA resolution, the low portions of the EXITE
waveform represent periods of about 20 ms when no test vectors are
applied, and the high portions 655 represent periods of about 20 ms during
which time AC source and common test vectors oscillate from zero to 30
volts at about 70 KHz.
Conventional infrared imaging systems represent images using arrays of
numbers, each number representing a pixel intensity value. Each pixel
intensity value, in turn, represents the image intensity of a
corresponding area of the DUT. In inspection system 310 of FIG. 3, for
example, frame grabber 325 delivers to computer 320 an array of pixel
intensity values for each captured image.
Some conventional inspection systems average sequences of images to reduce
the effects of noise. Averaged test images are then contrasted with a
reference image to identify differences, which indicate the presence of
defects. The above-described methods and systems can be used to produce
enhanced test and reference images.
One embodiment employs an image transformation in place of conventional
averaging. The image transformation, defined below in equation 2, is
applied to both a sequence of test images and a sequence of reference
images. The resulting transformed test and reference images I.sub.T and
I.sub.R are then contrasted to identify differences.
##EQU2##
where:
1. D.sup.-1 represents an image transformation (casting) from sixteen-bit
numbers provided by frame grabber 325, each number representing the
intensity of one pixel, to their floating-point inverses (quasi-inverse to
D, below);
2. I.sup.i is the i-th image of the sequence;
3. n is the number of images in the sequence;
4. F represents image filtering, e.g. low-pass filtering to reduce noise
and eliminate data provided by defective pixels in IR detector 315;
5. L is the application of a look-up table to the image to translate the
range of intensity values to a different scale, e.g., a different range or
from linear to quadratic; and
6. D is an image transformation (casting) from floating-point values to
sixteen-bit numbers.
In carrying out the image transformation of equation 2 on a sequence of
images (test or reference), each pixel intensity value in each of the
sequence of images is converted to a floating-point number. The resulting
image arrays are then averaged, on a per-pixel basis, to combine the
images into a single image array. Next, the resulting image array is
filtered to reduce the effects of noise and to remove data associated with
defective pixels in the imaging device. Each datum associated with a
defective pixel, identified by an extreme intensity value, is replaced
with a new intensity value interpolated from data representative of
neighboring areas.
The intensity values of the combined, filtered image array are applied to a
look-up table that translates the range of intensity values to a different
scale, e.g., a different range or from linear to quadratic. Finally, the
values in the resulting translated image array are converted back from
floating-point numbers to digital numbers to produce the transformed image
I.
The image transformation of equation 2 is applied to a series of test
images and a series of reference image to produce respective combined test
and reference images I.sub.T and I.sub.R. The test and reference images
I.sub.T and I.sub.R are then contrasted, using well-known image processing
techniques, to produce a composite image. The composite image highlights
temperature differences between the test and reference images;
unexpectedly warm or cool areas are indicative of defects.
In general, short circuits produce relatively high currents, and
consequently grow relatively hot. Open circuits reduce current, remain
relatively cool, and are therefore more difficult to image using IR
thermography. The improved thermal contrast provided by the foregoing
embodiments allows sensitive IR detectors to capture images that highlight
many types of defects previously difficult or impossible to view with
conventional IR thermography. Such defects include open circuits of the
type depicted in FIG. 2. The image transformation of equation 2 further
enhances the results.
FIG. 6C depicts an experimentally obtained image 680 illustrating how the
above-described embodiments produce sufficient thermal contrast to
highlight opens. A line 685 represents a relatively cool defect artifact
resulting from the presence of an open.
Defect-location Algorithms
A short between two lines increases current through those lines, and
consequently elevates the temperatures of the lines. Thus the lines,
though not themselves defective, may nevertheless appear in the composite
image. The portion of the composite image highlighting the lines is termed
the "defect artifact" of the short. Opens block current, and consequently
reduce the temperature of associated features. These features then appear
in the composite image as a "defect artifact" of the open. (FIG. 6C
depicts a line-type defect artifact associated with an open). Test images
thus include defect data spatially correlated to the defect region and
defect-artifact data spatially correlated to defect-free regions of the
imaged object. Unfortunately, defect-artifact data can obscure defect data
during image analysis, rendering it difficult to precisely locate defects.
Image-processing algorithms in accordance with some embodiments analyze
the defect data and defect-artifact data to address this problem. (In the
context of images, a related collection of defect data may be referred to
as a "defect," for brevity; likewise, defect-artifact data may be termed
"defect artifacts." Whether the term "defect" or "defect artifact" refers
to a physical feature of an imaged object or image-data representative of
a physical feature will be clear from the context.)
FIG. 7 depicts a composite image 700 showing three representative defects,
a point-type defect 705, a line-type defect 707, and a corner-type defect
710. These defects are assumed to be of similar size, but the defect
images nevertheless differ due to their respective defect artifacts 715,
720, and 725. Unfortunately, actual composite IR images do not so clearly
distinguish defects from defect artifacts, rendering difficult the task of
precisely locating defects. FIG. 8, though not based on measured data,
more accurately depicts a composite image 800 illustrating how a
point-type defect 805, a line-type defect 810, and corner-type defect 815
might appear in a composite image. The defect artifacts obscure the
location of the defects. One embodiment addresses this problem,
distinguishing defects from defect artifacts to facilitate defect
detection, location, and analysis.
Image processing to distinguish a defect from the associate artifact is
based, in part, on classifying defect-artifact data. Processing differs,
for example, for point-type, line-type, and corner-type defects.
Embodiments of the invention therefore include image-processing techniques
that sort defect artifacts by type. Image-processing techniques that
classify artifacts by type include pattern recognition and morphological
analysis. The "Handbook of Image and Video Processing," edited by Al Bovik
(2000) and "Nonlinear Filters for Image Processing," edited by E.
Dougherty and J. Astola (1999) describe image processing and mathematical
morphology known to those of skill in the art and suitable for use in some
embodiments: these texts are incorporated herein by reference.
FIG. 9 is a flowchart depicting a defect-location algorithm 900 in
accordance with one embodiment. Algorithm 900 receives a composite image
905, optionally filtered using a fast-Fourier-transform (FFT) low-pass
filter (step 907), in which defect data is bounded by defect-artifact
data. Subsequent processing automatically locates the defect data within
the defect artifacts to produce a list of defect coordinates. In one
embodiment, composite image 905 is an IR composite image of the type
discussed above; however, many other types of images depict subjects of
interest surrounded by artifacts of those subjects. Algorithms in
accordance with the invention can be used to localize such subjects. For
example, images obtained from visual optics and/or nuclear-type
experiments depict subjects and subject-artifacts.
The above-referenced Bovik reference describes a connection between
gray-level and binary morphology that is used in some embodiments to
distinguish defects from their associated artifacts. This connection is
based on the observation that an image I(x),x.di-elect cons.D can be
reconstructed from the set of thresholds. Such reconstruction may be
expressed mathematically as follows:
.THETA..sub.v (I)={x.di-elect
cons.D:I(x).gtoreq.v},-.infin.<v<.infin. (3)
where .THETA..sub.v (I) is the threshold of a grayscale composite image I
with a threshold level v, and D.OR right.{character pullout}.sup.2 is the
domain of the image:
I(x)=sup.sub.v.di-elect cons.R {x.di-elect cons..THETA..sub.v (I)} (4)
Algorithm 900 employs a similar type of image reconstruction to define and
optimize threshold levels to yield improved defect localization.
A morphological filtering algorithm (MFA) 910 is applied to the filtered
composite image for each of a number of threshold levels 915. The
repetition of MFA 910 for each threshold level, illustrated by bounding
MFA 910 within a for-loop 920A and 920B, produces a set of i thresholded
composite images 925[1 . . . i]. In each image 925, all pixel values of
the filtered composite image greater than or equal to the applied
threshold level are expressed using one logic level (e.g., logic one) and
all pixel values less than threshold level 915 are expressed using a
second logic level (e.g., logic zero). In one embodiment, threshold levels
915 are in units of standard deviation of pixel-intensity values from a
filtered composite image produced by filter 907. Histogram analysis is
provided to calculate mean .mu. and standard deviation g; the actual level
of threshold is equal to .mu.+.sigma..multidot.T, where T is the input
threshold level in units of .sigma.. MFA 910 is detailed below in
connection with FIG. 10.
The next for-loop (930A and 930B) treats each image 925 to a second MFA
935, this one type-specific. For line-type artifacts, for example, MFA 935
removes other types of defect artifacts (e.g., corner- and point-type
artifacts), leaving only lines. For-loop 930 produces a set of i images,
each having j line-type artifacts. In the illustration, the top image
940[1] includes two line-type defects: the point- and corner-type defects
of image 925[1] are removed. The remaining images 940[2 . . . i] may have
more or fewer artifacts. MFA 935 is detailed below in connection with FIG.
11.
In a final sequence of operations, a defect-location algorithm DLA 945
analyzes the defect artifacts in images 940 to precisely identify the
defect coordinates 950 from among the defect artifacts. A version of DLA
945 specific to line-type defects is detailed below in connection with
FIG. 12.
FIG. 10 depicts an embodiment of MFA 910 of FIG. 9, which is repeatedly
applied to composite image 905 to produce a sequence of i filtered images
925. First, one of threshold levels 915 is applied to the FFT-filtered
composite image 905 (step 1000). A morphological closing step 1005 (a
dilation followed by an erosion) applied to the product of step 1000
smoothes defect artifacts, and a subsequent filter removes small image
effects induced by noise or defective detector pixels (step 1010). MFA 910
thus produces one of the i images 925[1 . . . i]. MFA 910 is repeated, as
shown in FIG. 9, for each one of threshold levels 915.
FIG. 11 depicts an embodiment of type-specific MFA 935 of FIG. 9 adapted
for use with line-type defect artifacts. MFA 935 receives filtered images
925 from MFA 910 and employs type-specific filtering operations that
remove any artifacts other than line-type artifacts. Step 1110 provides
type-specific morphological operations based on an artifact's position on
the image rather than its geometrical properties. For example, an artifact
located in areas where that type of artifact is not expected may be
eliminated. Any holes in the surviving artifacts are then filled (step
1115) using any of a number of conventional hole-filling techniques.
Next, step 1120 filters the images from step 1115 using a type-dependent
list of constraints 1125 derived from artifact type and area information.
Constraints 1125 are a type-specific list of measurable artifact
parameters and ranges. For example, line-type artifacts in simple cases
may be filtered out by their circular factor ("lines" are not circular),
orientation (e.g., line-type artifacts must be close to vertical or
horizontal), and their edge intersections (e.g., line-type defects,
distinct from corner-type defects, do not interest two adjacent image
boundaries). The resulting filtered binary image 940 contains only the
desired type of artifact. In one embodiment, image 940 has two pixel
values: background (0) and artifact (1). The artifacts appear as areas of
touching pixels, all set to 1; the surrounding areas appear as pixels set
to 0. The details of the implementation of each block of MFA 900 are well
known to those of skill in the art of image processing. MFA 935 is
repeated, as shown in FIG. 9, for each one of threshold levels 915.
The set of morphological operations applied in MFA 935 is specific to
line-type defects in the example, but these operations can be modified as
desired to accommodate e.g. point- and corner-type defects. If only
point-type artifacts were of interest, MFA 935 can be adapted to remove
the artifacts that touch the border of the image, artifacts greater than a
specified maximum area, etc., which would correspond to line-, corner-,
and other type artifacts. If only corner-type artifacts were of interest,
MFA 935 can be adapted to remove artifacts that do not represent
intersecting perpendicular lines of a specified minimum area, etc. MFA 935
can be modified to select these and other types of artifacts using
well-known image-processing techniques.
FIG. 12 depicts an embodiment of defect-location algorithm (DLA) 945 of
FIG. 9, which generates a list of physical coordinates for defects on an
object under test using the array 940 of type-specific images. DLA 945 is
specific to line-type defect artifacts, but can be adapted for use with
other types of defects, as will be evident to those of skill in the art.
In the first step, a line-defect filter LDF algorithm 1200 produces an
array of i LDF structures 1205[1 . . . i], one LDF structure for each
image 940. In addition to type-specific images 940, LDF algorithm 1200
receive as inputs 1210 the original filtered composite image (from step
907 of FIG. 9), the value of the initial threshold level used to create
images 940, a threshold step value indicative of the difference between
threshold values used to acquire successive images, and the number i of
threshold levels (the number of threshold levels i is the same as the
number of images 940 because 940 are thresholded images).
FIG. 13 depicts a linked-list array of LDF array structures 1205 (FIG. 12)
in accordance with one embodiment. In this example, four threshold levels
and associated processing are used to produce four LDF structures 1205[1]
through 1205[4], though the actual number of arrays can be more or fewer.
These arrays are adapted for use with line-type defect artifacts, but can
be modified for use with other types. Each LDF structure 1205[i] includes
the below-listed features.
1. A threshold field 1300 that stores the threshold level applied to the
composite image to produce the respective binary image 940[i].
2. A number field 1305 that stores the number n of identified defects, two
in the example of FIG. 12.
3. An image field 1310 that stores the respective binary image 940[i].
4. An array 1315 of defect structures 1320[1-j], where j is the number of
defect artifacts in the respective image 940[i]. Each defect structure
1320 stores the X and Y coordinates of the pixel at the tip, or "peak," of
the respective line-type defect artifact (defects are assumed to be near
the tip of line-type defect artifacts). Each defect structure additionally
includes a pixel-value field 1325 that stores the pixel intensity value
corresponding to the X and Y coordinates of the peak pixel in composite
image 905.
5. An array 1330 of line structures 1335[1-j], each associated with a
defect artifact in the respective filtered image 940[i]. Each line
structure 1335 includes an area field 1340 that stores the area of the
defect artifact (in pixels); a rectangle field 1345 that stores the left,
top, right, and bottom coordinates of a rectangle encompassing the defect
artifact; a belongs-to field 1350 that stores a line index relating a
line-type artifact to features of an image taken at a lower threshold
level (-1 of no such line exists); and a contains field 1355 that stores
an array of line indices relating a line-type defect in the filtered image
with one or more related lines in another filtered image taken at a higher
threshold level. The purposes of Belongs-to field 1350 and contains field
1355 are detailed below in connection with FIGS. 14 and 15A-D.
Returning to FIG. 12, a LDF structures 1205[1 . . . i] are analyzed (step
1210) to develop j LDF gradient peak profiles 1215[1 . . . j], one for
each type-specific defect artifact in each image 940. The following
discussion of FIGS. 14, 15A-15D, and 16 illustrates the process of peak
profiling 1210.
FIG. 14 depicts an illustrative filtered composite image 1400 similar to
what one might expect from step 907 of FIG. 9. Image 1400 includes a pair
of vertical line-type defect artifacts A1 and A2 and a point-type artifact
A3. The boundaries of the defect artifacts are blurred to depict the lack
of an emphatic image boundary between defect- and defect-artifact data. It
is assumed, however, that defects associated with line-type defect
artifacts are near artifact tips, or "peaks," and defects associated with
point-type defects are centered within the related artifact.
FIGS. 15A-15D depict binary, type-specific images 1505, 1510, and 1515 like
images 940[1 . . . i] of FIG. 9. Each image represents composite image
1400 with a distinct applied threshold value, per MFA 910; point-type
artifact A3 is removed from each binary image by the subsequent
application of type-specific MFA 935. Each image is stored in field 1310
of an LDF structure 1205 (FIG. 13) along with the other image- and
artifact-specific information described above in connec
