Non-Uniform Sampling for Efficient Analog-to-Information

Advanced Signal Processing, Reducing Irrelevant Data

Innovation Digital’s architecture and signal processing approach to highly efficient data collection using a sparse sampling clock optimally tuned to signals of interest. This technique significantly reduces the amount of data being captured by concentrating on the signals of interest and ignoring irrelevant information.

This is a very powerful technique in performing the following:
1) Reducing the average sample rate, thereby lowering power of the receiver.
2) Reducing the digital data rate, thereby reducing the back-end digital signal processing requirements (size, power)
3) Efficiently implementing common functions (e.g., mixing, down conversion, interpolation, decimation, etc.
4) Leveraging priori knowledge of the signal for more efficient processing (e.g., more efficient bandwidth allocation).

Classical Nyquist sampling essentially convolves the spectrum of the sample clock with that of the input signal. A uniform sample clock produces copies of the input spectrum uniformly spaced in the frequency domain. Non-uniform sampling also convolves the spectrum of the sample clock with that of the input signal, which produces copies of the input spectrum which are not-uniformly spaced in the frequency domain. The non-uniform sample clock can be constructed to position the desired signals in non-overlapping frequency bins while undesired signals can be overlapped in out-of-band frequency bins and discarded. This provides significant improvements in spectral efficiency by maintaining only the signals of interest, and effectively discarding undesired signals.