Are you sure that you need a discrete PDF? You have not specified that your time series takes on values only in a discrete set, e.g. {0,1}.
Just because you have a discrete-time process, it can very well be that you should model the random variables as continuous.
For example:
x = randn(100,1);
gives a realization of a discrete-time white Gaussian noise process, but it is not a discrete random variable.
Do you have a specific pdf that you would like to assess for your data?
You can always use ksdensity to estimate the pdf
x = randn(100,1);
[f,xi] = ksdensity(x);
plot(xi,f);
If you want to see if your data follow a certain distribution, see histfit, or dfittool
