In this paper, the anti-logarithmic quantization approach, recently proposed by the authors to be used in multi-channel implantable neural recording systems is studied in depth. It is shown that not only quantization noise does not degrade the quality of the neural signals being digitized, it helps reject significant portion of their noise content. As a result of using this method of quantization, bitrate and power saving is achieved in the system level. To quantify the properties associated with the proposed quantization method, a four-channel proof-of-concept prototype is developed and tested using prerecorded neural data.