Translinear circuit for 2-D analog error correction

Analog error correction allows discrete-time continuous-alphabet signals to be protected against channel impairment without the need for a digital representation or digital circuit. This is a powerful way to distribute the representation of an analog value across multiple wires or multiple devices in a manner similar to how digital representations distribute bits across multiple wires and devices to create digital words. In this paper we propose a translinear circuit implementation of a class of analog error correction schemes proposed in [1]. We use subthreshold analog CMOS circuits as our medium of choice due to their strong potential for enabling ultra-low power processing and powerful native universal computational ability. We focus on the simplest case of a two-dimensional analog codeword (one analog value represented across two wires) and contrast this formulation with simple two-dimensional digital quantization. We demonstrate through SPICE simulations that our analog codec attains a strong performance advantage over digital quantization methods, and we give an example of analog source-channel coding applied to images.