The effects of spatial sampling on random noise for gyrokinetic PIC simulations in real space

We study the effects of cloud-in-cell sampling and gyroaveraging on random noise in real space (as opposed to the common Fourier space presentation), and show that together, these can reduce the noise by a factor of 3 compared to nearest grid point sampling without gyroaveraging. Hence an order of magnitude less test particles are needed for the given noise level. We derive equations for noise level as a function of Larmor radius and also investigate the effect of gyroaveraging on noise in local gradients. The effect of number of gyropoints on noise is also discussed.