This paper concerns the computation of derivatives from particle image velocimetry (PIV) velocity fields with the goal of obtaining the vorticity component normal to the plane. A variety of derivative schemes are characterized by their transfer function, taking into account the truncation and noise amplification. The PIV measurement noise is supposed to be a white one in the Fourier space. A spectral approach is used in order to choose the best filter for turbulent flows. The derivative spectra are discussed. An application is presented on a real turbulent flow with two interrogation window sizes and different derivative schemes. The most significant schemes are also applied to a velocity field containing a single vortex. A comparison of the maximum of vorticity obtained with each scheme and through a least-square fit with an Oseen vortex, allows us to quantify the effect of the band pass filter and to select the best scheme.

47.32.C- Vortex dynamics

47.11.-j Computational methods in fluid dynamics

02.70.Rr General statistical methods

06.30.Gv Velocity, acceleration, and rotation

47.27.-i Turbulent flows, convection, and heat transfer

06.20.Dk Measurement and error theory

Fluid dynamics

Computational physics

Instrumentation and measurement

Issue 7 (July 2002)

Received 11 December 2001 , accepted for publication 27 May 2002 in final form 27 May 2002

Published 20 June 2002