My experience and that of a few papers for speckle size is more like 8 to 10 pixels. The results were more reliable. Easy math for speckle size = dimension as seen in camera divided by the number of pixels of the camera in that direction (2000) times the 8 to 10. For example, the Specimen size visible in the camera is 10 inches. Again using 2000 pixels and 10 pixels per speckle finds: speckle size = 10 inch / 2000 * 10 = 0.05 inch. Double-check by sampling speckles and counting pixels across them. And about half that spacing between them if they are circular. BTW: Correlated Solutions speckle stampers are great.
Thank you for your comment. While larger speckles will generally yield improved confidence in the subset correlation uncertainty, they are not necessarily better. This is because larger speckles require larger subsets, and larger subsets reduce spatial resolution. With an ideal setup (well-positioned cameras, diffuse illumination, good stereo calibration, etc.), 3 to 5 pixel diameter speckles with appropriate density will yield the best correlation confidence at an ideal subset size, which will produce the best spatial resolution and overall results. However, when the ideal pattern / setup is not achievable, larger speckles and subsets may be used to reduce the risk of noisy and/or aliased data.
And double check the complete setup by taking a dozen frames of an undisturbed specimen. The strains should of course be zero, but they will not be. The magnitude of the reported strains will be the noise floor of results. If the values are too high for your satisfaction, move the camera closer for larger apparent speckles, refocus carefully, or increase the f-stop of the camera. DIC has an inherent noise floor of about 50 microstrain compared to strain gages with a 25 microstrain error. Of course, DIC has the enormous advantage of measuring all 3 surface strains over the entire image in each measurement. And generally at a cost of only 20% of strain gages for only a few measurements.
Thanks for your comment. While strain gauges do have a lower noise floor, they are time consuming to install, and they produce directional data at only one point. DIC has a huge advantage due to the non-contact and full-field nature of the data producing millions of data points at each time step in less time. Also, the ability to visualize 3D data helps users interpret and share results with ease.
You should be getting a much better noise floor than 25µStrain with a strain gauge - I regularly resolve down to 0.1µε and usually quote a noise floor of about 1µε. DIC noise floor seems about right though
My experience and that of a few papers for speckle size is more like 8 to 10 pixels. The results were more reliable. Easy math for speckle size = dimension as seen in camera divided by the number of pixels of the camera in that direction (2000) times the 8 to 10. For example, the Specimen size visible in the camera is 10 inches. Again using 2000 pixels and 10 pixels per speckle finds: speckle size = 10 inch / 2000 * 10 = 0.05 inch. Double-check by sampling speckles and counting pixels across them. And about half that spacing between them if they are circular.
BTW: Correlated Solutions speckle stampers are great.
Thank you for your comment. While larger speckles will generally yield improved confidence in the subset correlation uncertainty, they are not necessarily better. This is because larger speckles require larger subsets, and larger subsets reduce spatial resolution. With an ideal setup (well-positioned cameras, diffuse illumination, good stereo calibration, etc.), 3 to 5 pixel diameter speckles with appropriate density will yield the best correlation confidence at an ideal subset size, which will produce the best spatial resolution and overall results. However, when the ideal pattern / setup is not achievable, larger speckles and subsets may be used to reduce the risk of noisy and/or aliased data.
And double check the complete setup by taking a dozen frames of an undisturbed specimen. The strains should of course be zero, but they will not be. The magnitude of the reported strains will be the noise floor of results. If the values are too high for your satisfaction, move the camera closer for larger apparent speckles, refocus carefully, or increase the f-stop of the camera. DIC has an inherent noise floor of about 50 microstrain compared to strain gages with a 25 microstrain error.
Of course, DIC has the enormous advantage of measuring all 3 surface strains over the entire image in each measurement. And generally at a cost of only 20% of strain gages for only a few measurements.
Thanks for your comment. While strain gauges do have a lower noise floor, they are time consuming to install, and they produce directional data at only one point. DIC has a huge advantage due to the non-contact and full-field nature of the data producing millions of data points at each time step in less time. Also, the ability to visualize 3D data helps users interpret and share results with ease.
You should be getting a much better noise floor than 25µStrain with a strain gauge - I regularly resolve down to 0.1µε and usually quote a noise floor of about 1µε.
DIC noise floor seems about right though