Neural Radiance Fields (NeRF) have achieved impressive results in view synthesis for various capture settings, including 360° and forward-facing captures. This technical report analyzes the potential ambiguities in NeRF, particularly the *shape-radiance ambiguity*, and discusses how NeRF avoids these issues. It also addresses a parametrization issue in applying NeRF to 360° captures of large-scale, unbounded scenes. The report introduces an *inverted sphere parameterization* method to improve view synthesis fidelity in such challenging scenarios. The method separately models foreground and background, enhancing the ability to capture both elements accurately. Experimental results on real-world datasets demonstrate the effectiveness of the proposed method, showing improved quantitative and qualitative performance compared to NeRF. The code for the method is available at <https://github.com/Kai-46/nerfplusplus>.Neural Radiance Fields (NeRF) have achieved impressive results in view synthesis for various capture settings, including 360° and forward-facing captures. This technical report analyzes the potential ambiguities in NeRF, particularly the *shape-radiance ambiguity*, and discusses how NeRF avoids these issues. It also addresses a parametrization issue in applying NeRF to 360° captures of large-scale, unbounded scenes. The report introduces an *inverted sphere parameterization* method to improve view synthesis fidelity in such challenging scenarios. The method separately models foreground and background, enhancing the ability to capture both elements accurately. Experimental results on real-world datasets demonstrate the effectiveness of the proposed method, showing improved quantitative and qualitative performance compared to NeRF. The code for the method is available at <https://github.com/Kai-46/nerfplusplus>.