The article considers the issues of stability of cascade integrator-comb (CIC) filters used in digital signal processing, including decimation and interpolation. A brief review of modern publications on the architectural optimization of CIC filters is given. The main attention is paid to increasing the stability of filters to the overflow of the bit grid, analyzing their stability and the method of synthesis of recursive FIR filters (filters with a finite impulse response). For a better understanding of the nature of the stability of CIC filters, the paper presents mathematical calculations illustrating the features of the accumulation of the constant component for various block configurations. A change in the structure of the CIC filter is proposed, consisting in the permutation of the integrator and comb filter blocks. It is proved that such a change prevents the accumulation of the constant component of the signal in the integrators and, therefore, eliminates the overflow of the bit grid due to the accumulation of the constant component in the integrator. This approach is based on the property of linear filters, according to which changing the order of inclusion does not affect the transfer function. amplitude-frequency characteristic, but in the case of digital implementations it allows to significantly reduce the probability of overflow. The possibilities of hardware and software implementation of such structures are considered from the point of view of minimizing the loss of accuracy and increasing the reliability of digital signal processing systems. It is proposed to use integers or numbers with a fixed point to eliminate the accumulation of quantization errors. In addition, a program in Python was developed that implements a CIC filter taking into account the stability of the constant component in the input signal and the accurate execution of operations. The obtained results are compared with modern approaches presented in scientific research in recent years. The proposed solutions can be useful in developing digital filters for systems with limited computing resources and increased stability requirements.
