RSEIS (version 4.2-4)

AUGMENTbutfilt: Butterworth filter with Augmentation

Description

Design and apply butterworth low/high/band pass filters with augmentation of the signal on either end to suppress edge effects.

Usage

AUGMENTbutfilt(a, fl = 0, fh = 0.5, deltat = 1, type = "BP",
proto = "BU", npoles = 5, chebstop = 30, trbndw = 0.3,
RM = FALSE, zp = TRUE, pct = 0.1)

Value

Filtered time series with the augmentation removed after filter.

Arguments

a

vector signal

fl

low frequency cut-off, default=0

fh

high frequency cut-off, DEFAULT= (1/2dt)

deltat

sample rate, s, deFAULT=1

type

type of filter, one of c("LP", "HP","BP" ,"BR" ), DEFAULT="BP"

proto

prototype, c("BU", "BE" , "C1" ,"C2"), DEFAULT="BU"

npoles

number of poles or order, DEFAULT=5

chebstop

Chebyshev stop band attenuation, DEFAULT=30.0

trbndw

Chebyshev transition bandwidth, DEFAULT=0.3

RM

Remove mean value from trace, default=FALSE

zp

zero phase filter, default=TRUE

pct

Percent augmentation applied to each side, default=0.1

Author

Jonathan M. Lees<jonathan.lees.edu>

Details

Creation of butfilt is a described by the following arguments:

LP

low pass

HP

high pass

BP

band pass

BR

band reject

BU

Butterworth

BE

Bessel

C1

Chebyshev type 1

C2

Chebyshev type 2

Arguments chebstop , trbndw are ignored for non-chebyshev filters. LP and HP filters are seet by specifying fl for HP filters and fh for LP filters, the other argumentin each case is ignored.

Mean values should be removed prior to calling this function, and then set RM=FALSE. This is true especially if tapering is applied prior to filtering.

Zero phase filter is achived by running filter back and forth. Otherwise a single pass is returned. This should be equivalent to package signal filtfilt (from MATLAB).

Augmentation involves copying the first and last percent of the signal, reversiing the time and adding to the signal on each end. This is then filtered, and removed after filter is complete. It is assumed that the important part of the signal is in the center of the time series and the edges are less critical. Then the augmented part has the same statistical content as the edges of the signal (presumably noise) and will not affect the filtered signal considerably. This is then thrown away prior to return.

See Also

butfilt

Examples

data(CE1)

ts1  <-  CE1$y
zz  <-  AUGMENTbutfilt(ts1, fl=1, fh=15,  deltat=CE1$dt, type="LP" ,  proto="BU",
npoles=5 )


##############    second example with plotting


data(KH, package ='RSEIS' )
w = KH$JSTR[[1]]
dt = KH$dt[1]

x = seq(from=0, by=dt, length=length(w));
plot(x,w, type='l')

par(mfrow=c(2,1) )
        
i=1
       fl = 1/50
fh= 1/2
       ftype = 'BP'
       ##########  normal band pass filter
        
 zz = butfilt(w, fl, fh,  dt, ftype ,  "BU")
     f.stamp =   filterstamp(fl=fl, fh=fh, type=ftype)

plot(x, zz, type='l', xlab='s', ylab='amp', main= f.stamp)
title(sub='butfilt')
       
       ####  
   zz1 = AUGMENTbutfilt(w, fl, fh,  dt,    type=ftype ,  proto="BU", zp=TRUE, pct=0.2 )
     f.stamp =   filterstamp(fl=fl, fh=fh, type=ftype)
plot(x, zz1, type='l', xlab='s', ylab='amp', main= f.stamp)
title(sub='AUGMENTbutfilt')