Self-Calibrating GPS Accelerator Implemented Using Analog Calculation in 65 nm LP CMOS

Many signal processing applications involve manipulating a large amount of data to generate a relatively narrow numerical result, such as, “does a particular object X exist in a picture?”, “which of several possible words was spoken in a sentence?”, or, “how strongly do these two patterns correlate?”. These applications take in noisy inputs with relatively low quantization, such as 8bits/color for video, or 8-bit audio samples, or 1b-to-2b digitized baseband signals. However, a sufficiently large amount of data allows noise rejection, and an accurate answer can be computed. This large-data-to-narrow-result property makes these applications a good fit for analog calculation, which also has limited resolution and inherent noise, but can offer higher energy efficiency than digital implementations.