Concatenated coding schemes with iterative decoding structures require the exchange of soft information between component decoders to improve the decoding results, which are heavily dependent on the accuracy of statistical properties of the channel, i.e., the signal to noise ratio (SNR). Incorrect estimation of the channel SNR is known as SNR mismatch (SNRM) and can lead to serious degradation in iterative decoding performance. We study the relationship between the number of pilot symbols and the performance of a Maximum Likelihood Estimation (MLE) channel estimator operating over a Binary Input Additive White Gaussian Noise (BIAWGN) channel. The simulation shows that using a larger number of pilot symbols leads to more precise estimations of both the noise variance and symbol energy. Nonetheless, the accuracy comes at the expense of bandwidth efficiency. There exists a substantial tradeoff between bandwidth efficiency and estimation accuracy. Furthermore, We investigate the impact of the SNRM on the performance of the Soft Output Viterbi Algorithm (SOVA) decoded serially concatenated continuous phase modulation (SCCPM) system. Our simulation shows that under-estimation of the channel SNR has a more detrimental effect on the SCCPM system compared to over-estimation.