Large, gray-scale CNN chips employ analog signals to achieve high-density in the internal distribution of the template parameters. Despite the design strategies adopted at the circuitry employed to implement the weights, accuracy is ultimately limited by the controlling signals. This paper presents a buffering strategy intended to achieve 8-bit equivalent accuracy in the distribution of the internal analog signals, as employed in the chips ACE4k, ACE16k, and CACE1k.