The physical convention defines the visibilties as:
V(u,v,w) = int I(l,m) e^(-i2pi w(n-1)) e^(-i2pi (ul+v*m) dldm
And gridding as the adjoint (+ in exponent). The engineering convention switches sign convention. We can compute everything in the physical convention and adjust the inputs and results accordingly via the following relation:
complex conjugation of input and result is equivalent to swapping conventions.
Thus going from image to visibilities, since images are real this is the same as negating w followed by the conjugating the resulted visibilities. In the other direction, this is the same as complex conjugating the visibilities, performing wgridding. The final complex conjugation does nothing because images are real.
Where this matters is in dish simulation, since beams are complex in both image and aperture space.
The physical convention defines the visibilties as:
V(u,v,w) = int I(l,m) e^(-i2pi w(n-1)) e^(-i2pi (ul+v*m) dldm
And gridding as the adjoint (+ in exponent). The engineering convention switches sign convention. We can compute everything in the physical convention and adjust the inputs and results accordingly via the following relation:
complex conjugation of input and result is equivalent to swapping conventions.
Thus going from image to visibilities, since images are real this is the same as negating w followed by the conjugating the resulted visibilities. In the other direction, this is the same as complex conjugating the visibilities, performing wgridding. The final complex conjugation does nothing because images are real.
Where this matters is in dish simulation, since beams are complex in both image and aperture space.