Walker et al. (2010) specify that their decimated envelopes are demeaned prior to the application of AGC. When I do this in my implementation of their algorithm, the resulting AGC'd envelopes go negative and look strange. The code currently uses the line st_out.detrend('demean'), but perhaps it’s better to deviate from Walker here and remove it. Doesn’t look like Richard uses it in his implementation of the Walker AGC algorithm.
Original report by me.
Walker et al. (2010) specify that their decimated envelopes are demeaned prior to the application of AGC. When I do this in my implementation of their algorithm, the resulting AGC'd envelopes go negative and look strange. The code currently uses the line
st_out.detrend('demean')
, but perhaps it’s better to deviate from Walker here and remove it. Doesn’t look like Richard uses it in his implementation of the Walker AGC algorithm.