Abstract: Deep-detection logging is a significant logging method in oil-gas exploration and development. However, its data amount has sharply increased, bringing greater challenge for the cable transmission to upload all data to the ground system in real-time. Compressing the downhole data before uploading can improve the cable transmission and logging speed. The characteristics of the full waveform data in this logging was analyzed and a downhole compression method was proposed. This method is based on a non-uniform quantization and corresponding adaptive coding. The hardware, a downhole compression algorithm and its corresponding ground decompression software were designed to realize the method. Its performance and scalability were tested using water experimental data. The results show that the compression rate of the algorithm is about 50%, the relative distortions at the peak of the full waveform, the first peak of the direct wave and the reflected wave are all within 3%, the execution time and memory usage can meet the requirements of fast acquisition and real-time transmission. In addition, this algorithm has good adaptability and stability in the conditions of different hardware, software and data characteristics. This work is helpful to achieve downhole data compression in deep-detection acoustic logging and improve logging efficiency.