ft - Fourier transform operations on stack 1
ft see:; pl ch: hst: Fourier transform stack 1, and show the
amplitude in stack 1 and phase in stack 2.
ft ift: Inverse of the above.
ft see:; pl ch:; el 18; ft ift: Filter out the 18th Fourier component
ft rm:3,5 Zero out the 3 brightest Fourier components,
ignoring the 5 lowest frequency components.
ft does a Fourier transform on stack 1. The "ft see:" mode transforms
the channels shown in the current plot of stack 1 and places the
amplitude in stack 1 and the phase (measured in turns) in stack 2.
The inverse operation "ft ift:" takes amplitude in stack 1 and phase in
stack 2 and converts back to spectrum. Because the Fourier transforms
are done using fast Fourier transform techniques, the spectrum is padded
with zeros on the right to the next biggest power of 2 (e.g. 1024 or
2048 channels). The re-transformed spectrum will show these added channels.
The added channels have zero weight. If the Fourier spectrum was modified,
for example by averaging or zeroing out components corresponding to baseline
ripple, then these added channels will no longer have a value of zero.
Since "ft see:" operates only on the channels shown in the current plot,
the restored stack using "ft ift:" will contain only those channels and
may be smaller than the original full spectrum. The "observation type"
identification number in the stack header has 100 added to it for the
amplitude plot and 200 added to it for the phase plot. These values are
subtracted in the inverse operation.
"ft rm:" does a Fourier transform of the plotted channels of stack 1, then
zeros out selected high amplitude channels, then re-transforms the
selected channels and replaces them in stack 1. This is a fast method
of removing well-defined baseline ripple components. The parameter "nc:"
gives the number of Fourier components to zero out. The parameter "ex:" is
the number of low-frequency components to ignore, since these may contain
significant data. The data shown in the plot of stack 1 is transformed,
and the "nc:" brightest channels in the transform with channel number
higher than "ex:" are set to zero. The data is then re-transformed,
and inserted back into stack 1. Since the single-precision stack
data is converted to double precision and subsequent operations are
all double precision, there is no loss of accuracy. The number of
components which have been zeroed out appears in the header as header
variable ".ftrm" A value of "nc:" in the range 1 to 5 can effect an
80% reduction in baseline ripple.
This document was last updated on
Tue Oct 17 16:24:05 EDT 2000
by Marc W. Pound.