Bursts and time lapses:

The time lapse between two images must be sufficiently short so that the displacement remains proportional to the local velocity, and correlations are not lost by deformation(3.5). Furthermore a majority of the particles must not move in and out of the laser sheet by their transverse velocity. Shorter time lapse leads to a larger relative error in the measurement of displacement. A typical displacement of 5 to 10 pixels is generally optimum for a first CIV processing, and possibly increased with hierarchical iterations(3.5).

To be able to optimize the time lapses, we record ``bursts'' of 3 or 4 images (expressed in units of the frame period in Civit2.7). We can use any pair of images chosen in this set in the CIV software, so that during data processing, the time lapse can be chosen (with four images) among , ,, or the sums , , .

For instance, for a field 2.5x2.5 mwith typical velocity 2.5 cm/s, one pixel (2.5 mm) is travelled in 0.1 s, so we recommand a frame period of 1/15 s. To allow an optimum choice of time intervals, it is good to use a burst with three or four images, as sketched in fig. 7. This allows to adjust the time interval for image pairs. If the maximum range of intervals is wanted (e.g. if the velocity values strongly depend on position or time), the choice of fig7a is appropriate. The choice of fig. 7b allows less range, but it is optimum to estimate the precision by comparing the velocity field from the middle interval (with 6/15 s) to the field averaged from the two side intervals (with 5/15 s), and the field obtained with the total interval 16/15 s (see ).

Figure 7: example of time intervals in a burst of 4 images, expressed in units of 1/15 s, a) allows a large range of intervals, b) allows to compare two intervals centered at the same time.