Methods and systems for detecting defects in devices using X-rays
In one embodiment, an automated high-speed X-ray inspection system may generate a first X-ray image of an inspected sample at a first direction substantially orthogonal to a plane of the inspected sample. The first X-ray image may be a high-resolution grayscale image. The system may identify one or more elements of interest of the inspected sample based on the first X-ray image. The first X-ray image may include interfering elements that interfere with the one or more elements of interest in the first X-ray image. The system may determine one or more first features associated with respective elements of interest based on variations of grayscale values in the first X-ray images. The system may determine whether one or more defects are associated with the respective elements of interest based on the one or more first features associated with the element of interest.
1. A method comprising, by an automated high-speed X-ray inspection system:
generating a first X-ray image of an inspected sample at a first direction, wherein the first X-ray image is a high-resolution grayscale image, wherein the first direction is substantially orthogonal to a plane of the inspected sample;
identifying, from a plurality of elements of the inspected sample, one or more elements of interest of the inspected sample based on the first X-ray image, wherein the first X-ray image comprises a plurality of interfering elements from the plurality of elements that interfere with the one or more elements of interest in the first X-ray image;
determining, for each of the one or more elements of interest, one or more first features associated with the respective element of interest, wherein the one or more first features are determined based on variations of grayscale values in the first X-ray images;
determining, for each of the one or more elements of interest, whether one or more defects are associated with the respective element of interest based on the one or more first features associated with the element of interest;
wherein the automated high-speed X-ray inspection system comprises an X-ray source emitting an X-ray beam with a white spectrum and one or more X-ray filters filtering out respective spectrum ranges of the X-ray beam, and wherein the one or more X-ray filters comprise a first X-ray filter corresponding to a first spectrum range being responsive to a first material and a second X-ray filter corresponding to a second spectrum range being responsive to a second material, and generating a first material selective X-ray image with the first X-ray filter being applied to the X-ray source, wherein the first material selective X-ray image excludes elements made of the first material, and generating a second material selective X-ray image with the second X-ray filter being applied to the X-ray source, wherein the second material selective X-ray image excludes elements made of the second material; and
generating an enhanced X-ray image by combining the first material selective X-ray image and the second material selective image, wherein the enhanced X-ray image have a higher signal-to-noise ratio than the first X-ray image.
2. The method of claim 1 , wherein the high-resolution grayscale image has a number of pixels greater than or equal to 12 megapixels.
3. The method of claim 2 , wherein the high-resolution grayscale image has a number of pixels greater than or equal to 29 megapixels.
4. The method of claim 1 , wherein the first X-ray image is generated by an X-ray detector having a dynamic range greater than 10 K, and wherein the first X-ray image has greater than 10 K grayscale levels.
5. The method of claim 1 , wherein the first X-ray image is a superposition of multiple low dynamic range X-ray images taken with different X-ray exposure settings using an X-ray detector having a dynamic range lower than 10 K.
6. The method of claim 1 , wherein the one or more first features are determined based on a material thickness along the first direction using a machine-learning model.
7. The method of claim 6 , wherein the one or more first features comprise one or more of:
a material thickness value;
a material thickness variation;
a material thickness profile;
an element shape;
an element size;
a distance between the element of interest to another element;
a position of the element of interest in the first X-ray image;
a relative position of the element of interest to another element; or
a geometric signature associated with the element of interest.
8. The method of claim 1 , further comprising:
generating a second X-ray image of the inspected sample from a second direction different from the first direction; and
determining one or more second features associated with the element of interest by comparing the first X-ray image and the second X-ray image.
9. The method of claim 8 , wherein the one or more second features comprise one or more of:
a stacking order of two or more layers;
a position change of the element of interest;
a shape change of the element of interest;
a position change of an interfering element;
a shape change of an interfering element;
one or more grayscale value changes associated with the element of interest; or
one or more grayscale value changes associated with an interfering element.
10. The method of claim 8 , further comprising:
determining that the one or more defects associated with the element of interest are false positive results based on a determination that the one or more second features of the element of interest invalidate that the one or more defects are associated with the element of interest based on the one or more first features.
11. The method of claim 8 , further comprising:
confirming the one or more defects being associated with the element of interest based on a determination that the one or more second features confirm that the one or more defects are associated with the element of interest based on the one or more first features.
12. The method of claim 1 , wherein the one or more defects are identified by a machine-learning model, wherein the machine-learning model is trained by a sample population comprising at least 10 K samples, and wherein the machine-learning model, after being trained, isolates the element of interest from the plurality of interfering elements.
13. The method of claim 1 , further comprising:
generating two or more reference X-ray images for respective adjacent inspected samples of the inspected sample; and
generating a reference model for the inspected sample based on the two or more reference X-ray images.
14. The method of claim 13 , further comprising:
comparing, by a computer vision algorithm, the first X-ray image of the inspected sample to the reference model, wherein the one or more defects associated with the element of interest are identified based on the comparison between the first X-ray image and the reference model.
15. The method of claim 1 , wherein the element of interest comprises a solder connection, and wherein the one or more defects comprise one or more of:
a void defect;
a head-in-pillow defect;
a non-wet defect;
a non-contact open defect;
a misalignment; or
a bridging/short defect.
16. The method of claim 1 , wherein the one or more defects associated with the element of interest are identified based on the first material selective X-ray image, the second material selective X-ray image, or the enhanced X-ray image.
17. The method of claim 1 , wherein the one or more X-ray filters comprise a low energy X-ray filter filtering out low energy X-ray components from the X-ray beam.