Using Gold Deconvolution on time series data

Hi All,

I’m a Geophysicist looking at applying Gold Deconvolution to time-series data from the TSpectrum library.

I note that this code is applied on histograms in this instance. Is it possible for the TH1 classes to work with time series style data, i.e. positive and negative sample amplitudes? I considered that if each bin was a sample it could work but negative sample values would be a problem. I noted from this documentation a image of the applied deconvolution gave bins with negative counts, so in theory it should allow waves to be stored and manipulated?


The TSpectrum::Deconvolution method (from the TSpectrum class) uses ordinary “double source[ssize]” and “double response[ssize]” arrays (and an ordinary “int ssize”, of course) so, try it with your own data.

Dear James,

I hope this is not too late to answer your question.

I’ve personally used the Gold Deconvolution method in my time-series data and it works. Just be careful with negative values and make sure that your reference signal is ‘periodic’.

If you want, here is a bit more details:
I used the method in signals in Silicon-Photomultipliers where I deconvolved the signal of 1 photon from the pulses response of the device. This allowed to get back the time profile of liquid argon scintillation quite accurately.
The problem was when we had undershoot on the signal. The negative values and the ‘non-periodicity’ of the signals made it impossible to deconvolve the signals.