86 P. L. Reu and E. M. C. Jones Table 11.1 Stereo-rig setup parameters Cameras FLIR 2.3 mega-pixel Lens Edmund optics 35-mm Software Vic3D 7.2.6 Image scale ≈20 pixel/mm Stereo-angle 22.7◦ Correlation ZNSSD Subset 35 ×35 pixels Step 15 ×15 pixels Strain window 15 ×15 data points Shape function Affine Normalization ZNSSD Pre-processing Low pass filter Fig. 11.1 Setup photo of cameras inside the large vacuum chamber. A hotplate was placed immediately under the target to create heat waves 11.3 Results The stereo-images were analyzed using DIC software with the settings in Table 11.1. Approximately 900 images were analyzed (5 fps for 3 min) to look at both the spatial distribution of noise and the temporal distribution of noise. We noted that after removal of the air from the chamber, other error sources started to become important. For example, sample or camera motion, on the order of 200 nm and small temperature changes of 2 ◦C were measurable, particularly for the setup with no hotplate. We attributed the improved measurement results, even at atmospheric temperature, to the fact that the lights and other heat sources, along with disturbances from building HVAC systems were isolated outside the measurement chamber. This gave the system, even at atmospheric pressure, a better noise floor than is typically observed in a standard laboratory setting. At this point it becomes difficult to separate the different error sources. For simplicity, only the horizontal displacement (U) from the experiment with the heat source are shown. Temporal errors are measured by taking the standard deviation of the center subset through the 3-min record time, approximately 900 image pairs. Spatial errors are the standard deviation of the U-displacement at frame 125 at 25 s after the start of the experiment. Figure 11.2 shows the results at vacuum on the left and atmospheric pressure on the right. An order of magnitude improvement is seen in the error. Figure 11.3 shows the steady improvement of the displacement errors, both spatial and temporal, as the air is removed. Plots are in engineering units, but the corresponding errors in terms of pixels are 0.046 pixels at atmosphere and 0.0032 pixels in vacuum. The atmospheric noise floor is a little larger than the typical rule of thumb of 0.01 pixels, mainly due to the positioning of the hotplate immediately below the sample. However, in vacuum, the value is now approaching the theoretical noise floor for DIC—even in a very poor experimental setup with a heat source immediately below the sample. Results for the hotplate behind the cameras and without the hotplate are similar in trends, but typically with lower errors. However, once below 200 Torr, vibrational motion of either the camera or the sample corrupted the results. To improve on these results, a floating vacuum chamber will likely be required as we are approaching the resolution of interferometry.
RkJQdWJsaXNoZXIy MTMzNzEzMQ==