Variable cut-off frequency LPF with high oversampling

Started by January 23, 2014
Hello all,

I'm using a Xilinx Kintex-7 FPGA to do some low-pass filtering using System Generator/Simulink, and had a question about the architecture of the filter.

My requirements include having to use a fixed sampling rate of 80MSPS. This can't really change. I appreciate the difference b/w sample rate of the filter and the clock rate of the DSP primitives in the FPGA.

The customer wants to be able to change the cut-off frequency during runtime, going from 1kHz all the way to 8 MHz. I'd normally accommodate variable cut-off with loadable coefficients adn just change the sampling rate, but here I don't know what to do... At low cut-off rates building a filter with THAT much oversampling is not really practical, and for various reasons I can't really lower the sampling rate. What do I do?

Are there techniques, similar to superhet architectures for analog radio, to mix and move the signal to a higher bandwidth, and then filter? Such that the filter always sees the same IF band, while the VCO (e.g. its digital equivalent) tunes the mixing appropriately? What kind of Simulink blocks would I use for modeling the mixing and upconversion?

Thanks for the help!