Sensor Domain Signal Processing
ABSTRACT
ADCs have improved but haven't fundamentally changed for decades. Neural Networks and AI teach us new methods of processing data and SiliconIntervention uses these insight to radically change the principles of gathering analog data into a system. The sensor, MEMS or otherwise, connects to an analog circuit that converts the signal into the frequency domain: it performs a continuous FFT operation directly on the sensed data. Each bin of the FFT is digitized by two low power sigma-delta modulators. This combination reduces the power of data conversion by at least 10x, more typically 100x. Since FFT is lossless, the receiving processor can recreate the full bandwidth signal if necessary, but more often than not the FFT data is more useful - to drive an AI for example.
BIOGRAPHY
Martin Mallinson has a Masters degree in Physics from Oxford but always had a passion for electronics. Founder of multiple companies, SiliconIntervention (SI) since 2019 now leverages Martin's many patents, 116 and 22 in SI alone. Recognized as a leading software expert as well as an inspired mixed-signal designer, Martin created the CAD tool that SI uses to create and manage complex and novel analog circuits. Everything from the engine controllers used by Boeing and Airbus, X-ray systems at Varian, Medtronic devices and world renowned Audio devices. Now his attention is focused on reducing cost, power and size of Edge and IoT systems.