This paper investigates the channel estimation and reconstruction in fluid antenna systems (FAS) using Nyquist sampling and maximum likelihood estimation (MLE) methods. FAS, a flexible antenna technology, allows the radiating element to switch positions within a predefined space, offering additional diversity and multiplexing gains. However, obtaining accurate channel state information (CSI) over this space is crucial for leveraging FAS benefits. The paper addresses the challenges of channel estimation and reconstruction in FAS, revealing a fundamental tradeoff between the accuracy of the reconstructed channel and the number of estimated channels. It is shown that half-wavelength sampling is insufficient for perfect reconstruction, and oversampling is essential to enhance accuracy. A suboptimal sampling distance is proposed to facilitate efficient channel reconstruction, and MLE is used to bound the channel estimation error within a specific confidence interval (CI). The paper also demonstrates that FAS with imperfect CSI can outperform traditional antenna systems (TAS) with perfect CSI in point-to-point scenarios. The study introduces an electromagnetic-compliant channel model for FAS, covering both 1D and 2D fluid antenna surfaces. It shows that the minimum number of estimated channels and the suboptimal sampling distance required for efficient reconstruction depend on the channel's spectral properties. The results highlight the importance of oversampling in FAS and the tradeoff between channel estimation accuracy and the number of estimated channels. The paper also evaluates the rate performance of FAS and TAS, showing that FAS with imperfect CSI can achieve higher rates than TAS with perfect CSI. The findings contribute to the understanding of channel estimation and reconstruction in FAS and provide insights into the design of efficient channel estimation schemes for FAS.This paper investigates the channel estimation and reconstruction in fluid antenna systems (FAS) using Nyquist sampling and maximum likelihood estimation (MLE) methods. FAS, a flexible antenna technology, allows the radiating element to switch positions within a predefined space, offering additional diversity and multiplexing gains. However, obtaining accurate channel state information (CSI) over this space is crucial for leveraging FAS benefits. The paper addresses the challenges of channel estimation and reconstruction in FAS, revealing a fundamental tradeoff between the accuracy of the reconstructed channel and the number of estimated channels. It is shown that half-wavelength sampling is insufficient for perfect reconstruction, and oversampling is essential to enhance accuracy. A suboptimal sampling distance is proposed to facilitate efficient channel reconstruction, and MLE is used to bound the channel estimation error within a specific confidence interval (CI). The paper also demonstrates that FAS with imperfect CSI can outperform traditional antenna systems (TAS) with perfect CSI in point-to-point scenarios. The study introduces an electromagnetic-compliant channel model for FAS, covering both 1D and 2D fluid antenna surfaces. It shows that the minimum number of estimated channels and the suboptimal sampling distance required for efficient reconstruction depend on the channel's spectral properties. The results highlight the importance of oversampling in FAS and the tradeoff between channel estimation accuracy and the number of estimated channels. The paper also evaluates the rate performance of FAS and TAS, showing that FAS with imperfect CSI can achieve higher rates than TAS with perfect CSI. The findings contribute to the understanding of channel estimation and reconstruction in FAS and provide insights into the design of efficient channel estimation schemes for FAS.