GitLab

Scaling of signal traces is useful if a linear combination of a set of traces
shall be computed rather than the correction of a signal. This kind of
application has to be carried out when predicting disturbances caused by
magnetic field in seismic recordings, for examples.

For large time series operation in the Fourier domain is computationally more
efficient than discrete time domain convolution or cross-correlation.

program operates correctly in convolution mode; though results are not yet as
expected in cross-correlation mode

Merge branch 'timeseries' into cross

Taking a reference to a public base class object in a copy constructor allows
implicit type conversion even if the constructor is declared explicit. This
causes undesired type conversion in instatiation of timeseries containers. In
such cases the header data commonly is lost silently. I get rid of all such
constructors, forcing users to explicitly express the desire of type
conversion.

code compiles, program still does not produce desired output

Merge branch 'timeseries' into cross

Provide recent updates of interfaces to time series containers.

Merge branch 'odatstream' into timeseries

Conflicts:
	src/ts/wf/autocorr.cc

Provide proper input/output facilities on branch timeseries.

- place error handling in separate compilation unit
- place type definitions in separate header file
- provide input/output operators
- use debug code of libtsxx

Assignment and copy operations were ambiguous, because all class templates use
the same series container type as public base class.

FourierProcessor is a support tool to provide FFT of a padded input series. In
order to supply Fourier transforms for different padded versions of the same
series efficiently, a persistent workspace is used.