Dynamic Notch Filtering

A new technique, dynamic notch filtering (DNF), is made possible by the absence of a conventional flat passband in the Prism frequency response. The figure below shows two Prisms having identical values of m but with h = 1 and 2 respectively. In each case only the Gch output is generated.

These outputs have matched phase at all frequencies, but different gains. Accordingly, a selected frequency component can be notched out by forming a weighted sum of the two Gchoutputs, where the weighting is the ratio of the respective gains at the notch frequency. This dynamic notching can be done sample by sample at low cost, and where the notch frequency can be selected in real time, so that components with varying frequencies may be successfully separated and tracked. Different weightings, each notching a separate frequency, can be generated from the same Prism outputs, so that it is possible to split a multi-component signal into a corresponding set of single component signals to be tracked individually.

An example of signal splitting using the DNF structure shown above is given in the following figures. The input signal s(t) is shown below; it contains two frequency components at 91 Hz and 75 Hz.


The two corresponding output signals show a successful separation of the two frequency components.