Successfully reported this slideshow.

Enhanced Surface Metrology1

520 views

Published on

Accurate measurement using breakthrough scientific approach

  • Be the first to comment

  • Be the first to like this

Enhanced Surface Metrology1

  1. 1. Enhanced Surface Metrology<br />Russell M. KurtzRAN Science & Technology, LLC<br />Ryder NesbittHexagon Metrology, Inc.<br />
  2. 2. There is a constant search for more accurate measurement<br />Traditionally 3 Parameters<br />COST<br />MEASURMENT SPEED<br />ACCURACY<br />Traditional Tradeoffs<br />Cheap, Fast, Inaccurate<br />Expensive, Slow, Accurate<br />Enhanced Surface Metrology<br />
  3. 3. We Present a Method of Adding One More Tradeoff Parameter<br />MEASUREMENT PRECISION<br />Measurement Precision<br />(perpendicular to the surface)<br />vs.<br />Sample Correlation<br />(along the surface)<br />Enhanced Surface Metrology<br />
  4. 4. Through application of statistical sampling and curve fitting...<br />... we can improve precision by approximately the same amount that we increase correlation<br />Enhanced Surface Metrology<br />
  5. 5. Theory & Methods<br />POINT CLOUDS<br />The number of points in a 3-D space, each corresponding to a point on the surface of the item<br />Enhanced Surface Metrology<br />
  6. 6. Described as Deviation from CAD<br />Three possible deviation causes:<br />The real surface deviation from nominal<br />Consistent measurement inaccuracies<br />Random measurement inaccuracies orother random noise<br />δr<br />δm<br />δn<br />Enhanced Surface Metrology<br />
  7. 7. Slide 1<br />Enhanced Surface Metrology<br />
  8. 8. Slide 2<br />Enhanced Surface Metrology<br />
  9. 9. Slide 3<br />Enhanced Surface Metrology<br />
  10. 10. Slide 4<br />Enhanced Surface Metrology<br />
  11. 11. Our Focus: Noise (δn) Reduction<br />visible with low noise<br />hidden by moderate noise<br />defect<br />10 x 10 mm portion of a surface, nominally spherical, defect height = ~1.5 µm<br />Enhanced Surface Metrology<br />
  12. 12. Methods of Noise Reduction<br />Arithmetic Mean over a Square Sampling Area<br />Arithmetic Mean over a Circular Region<br /><ul><li>wmn = 1 over the range m2 + n2 ≤ (M/2)2, again with M = N
  13. 13. This statistical process uses 78.5% as many samples as the square, and is expected to reduce noise by a factor of 0.886M.</li></ul>Gaussian Sampling Process over Square Region<br />Enhanced Surface Metrology<br />
  14. 14. Sample Correlation, 5 pixel wide<br />Square<br />Circular<br />Gaussian<br />Enhanced Surface Metrology<br />
  15. 15. Targeting Task Performance Model<br />Probability of Identifying a 3-µm Feature with No Noise Reduction<br />Enhanced Surface Metrology<br />
  16. 16. Experiments<br />Flat Sample: WLS400<br />block made of ceramic with a polished white surface finish<br />Enhanced Surface Metrology<br />
  17. 17. Experiments<br />Flat Sample: WLS400<br />Measurement Noise and Actual Surface Deviation<br />Before Statistical Noise Reduction<br />After Statistical Noise Reduction<br />Enhanced Surface Metrology<br />
  18. 18. Experiments<br />Sphere Sample: WLS400<br />Test on a known sphere, whose diameter was specified to be 1.5000±0.0001 in. but whose surface was specified smooth only to ±0.001 in. We scanned a section of the upper half of the sphere, a circular area whose diameter was 27.20 mm.<br />Enhanced Surface Metrology<br />
  19. 19. Experiments<br />Sphere Sample: WLS400<br />Measurement Noise and Actual Surface Deviation<br />Before Statistical Noise Reduction<br />After Statistical Noise Reduction<br />Enhanced Surface Metrology<br />
  20. 20. Experiments<br />Sphere Sample: WLS400<br />Surface of the sphere with locations measured from its center. Surface color shows deviation from nominal (after noise reduction)<br />Enhanced Surface Metrology<br />
  21. 21. Experiments<br />Interferometric Measurement<br />Test on a flat, metallic surface (aluminum), mechanically polished to RA ≈ 60. We used a Mach-Zender interferometer to obtain an interferogram covering a square 1 cm on a side.The illumination wavelength was 532 nm.<br />Enhanced Surface Metrology<br />
  22. 22. Experiments<br />Interferometric Measurement<br />Original Interferogram<br />Original fit to a sinusoid with intensity varying from 0 to 1<br />A<br />B<br />The difference is the noise pattern<br />After noise reduction<br />C<br />D<br />Enhanced Surface Metrology<br />
  23. 23. Summary & Conclusions<br />We have developed a statistical method for noise reduction of three-dimensional metrology systems. <br />The method involves combination of measurement values over a range of points, and permits weighting the points (enabling the user to, for example, assign confidence values to the various measurements). <br />Enhanced Surface Metrology<br />
  24. 24. Summary & Conclusions<br />On the flat surface, we reduced the deviation from nominal by more than a factor of two.<br />The sphere was specified to an accuracy of ±25.4 µm, and we measured that its accuracy was ±15.8 µm.<br />With the interferometric measurement we determined that the surface deviated 16.0% from perfect flatness.<br />Enhanced Surface Metrology<br />
  25. 25. Summary & Conclusions<br />These improvements indicate that the statistical surface metrology enhancement method can be useful in detecting small features.<br />Enhanced Surface Metrology<br />
  26. 26. Enhanced Surface Metrology<br />Russell M. KurtzRAN Science & Technology, LLC<br />Ryder NesbittHexagon Metrology, Inc.<br />

×