Lock-in thermography is increasingly becoming popular as a non-destructive testing technique for defect detection in composite materials for its low heating excitation. The experimental data is processed with Fourier transformation to produce phase and amplitude images. Phase images, though immune to surface emissivity variation, suffer from blind frequency effect, where a defect becomes invisible at a certain excitation frequency. There exists no analytical model to predict this 3-dimensional heat flow phenomenon. This paper presents a study of blind frequency using electro-thermal model based numerical simulation on a piece of thermally anisotropic carbon fibre composite. The performance of the simulator is optimized for spatial mesh size. Further the effect of paint layer, which is often applied to the sample surface for better thermal imaging, has been incorporated in the simulation. Finally, both experimental and simulation results are presented side-by-side for easy comparison. © 2013 AIP Publishing LLC.