Closed Hazboun6 closed 6 years ago
I know what the issue is, but I only know how to fix it in simulate
mode. The boxcar window is a smoothing window, so any of the stochastically sampled pulses will necessarily be smoothed if we use the boxcar afterwards, which someone might do in explore
mode. In other words it is completely reasonable that a student would generate randomly sampled pulses and then want to disperse them, and invoke the DM broadening to make it as realistic as possible. The DM broadening will necessarily smooth the pulses. One way around this would be to completely regenerate the signal, using broadened templates which is what we do in simulate
mode.
This was fixed in filterbank
mode by convolving the per frequency profiles with a boxcar. DM Broadening should happen naturally when the voltage signal is put through a filter bank. I am therefore closing this issue.
There is a new script in the
disperse()
method ofism.py
that convolves the signal with a top hat window to DM broaden the signal after it has been time shifted for dispersion delays. For some reason the convolution obliterates the stochasticity of the signal, i.e. there are no wiggles anymore. I am guessing that it is a similar issue to the dtype issues that were happening in issue #3 where there is clipping happening at some level.