WaveDeform is a linear algebra-based algorithm for deconvolving the DOM response to photons. Using a non-negative linear least squares algorithm (Lawson and Hanson, 1973), it deterministically deconvolves the SPE response function from the waveforms recorded by the DOM’s FADC and ATWD digitizers. Unlike other algorithms, it deconvolves the response from both digitizers simultaneously, improving the accuracy of the fit and preventing edge effects at the boundary of the ATWD.
Name of the input waveform series map, containing waveforms from all digitizer channels (including SLC hits).
Name of output pulse series map.
This controls the number of basis functions to place in the range of each waveform. For example, if SPEsPerBin is set to 3, the ATWD will be unfolded with pulses spaced 1.1 ns apart, comparable to the PMT transit-time spread. The greater the number of SPEs per bin, the greater the time resolution, until the fundamental limit is reached around SPEsPerBin of 5.
The time required by the NNLS algorithm is (very roughly) proportional to SPEs^3. Thus, the speed of the algorithm can be increased substantially at the price of accuracy by reducing the number of unfolded SPEs, and vice versa.
This controls the stopping tolerance of the algorithm, in the amplitude of the gradient of the least-squares residual. When the improvement per PE of allowing one more non-zero pulse drops below this number, the algorithm will terminate. Larger numbers will produce a faster, coarser fit using fewer pulses, and smaller numbers will provide a slower, better fit with more.
For use only running on simulation produced by DOMsimulator, which uses incorrect pulse templates on all channels. They are particularly bad for the low-gain ATWD channels (ATWD1 and 2). Use of the this option will cause the pulse templates used by DOMsimulator to be used in unfolding instead of the ones used for unfolding data in order to provide a self-consistent simulation. Once a replacement for DOMsimulator is available, this option will be removed.
