Filtering out 60Hz + harmonics

On Saturday, April 27, 2013 8:56:25 AM UTC-5, Vladimir Vassilevsky wrote:

> You mean, do the opposite of what conventional ALE does: estimate the 
> stationary part and subtract it from signal ?

Yes; exactly.

> There should be optimum adaptation rate vs rejection vs artifacts. 

I seem to recall in my adaptive systems class in graduate school that the two cases were presented as "equivalent" -- it was simply a matter of which you called the "desired signal" and which you called the "interfering signal".

Greg

I'd approach this as follows, in real time. It assumes that the frequency
is, in average, accurate.

- Construct a 60 Hz sine- and cosine wave
- Multiply signal with either and sum up in leaky integrator (/filter)
- multiply the output of the integrater again with its sine (cosine) wave
and subtract from the input signal

The 'leaking factor' of the integrator, or filter bandwidth in general,
determines the bandwidth (tracking a jumpy guitar player needs wider BW).

BTW, using an ideal integrator this is mathematically identical to FFT-ing
the whole signal with zero padding, then zeroing out one bin, then IFFT.