how to calculate Suffcient Sample amount for one distribution I am trying to answer the following question.
I want to know if the distribution of my samples follows some well-known distributions. How I can compute the minimum required samples in order to be able to compare if this distribution look like another on
for example:
if we count the total number of each character in the dictionary and we then plot the distribution we will have the unigram distribution for such dictionary. now if I have a text file how I can know or estimate the minimum number of characters should be in that text file in order to sufficiently decide if that text have the same characters distribution or not.
I am not expert in this domain but I tried so far to calculate based on the maximum probability. if we say that "e" repeated 50 times more than "j" then I need 50 "e"s to before I expect one "j". 
what I want to estimate how many character in the file I should have before I see all characters.
regards,
 A: [At the moment this is a request for clarification; it's too long for a comment, and hopefully will grow to an answer.]
You seem to be asking two different things there. 
In respect of the first thing: 

if I have a text file how I can know or estimate the minimum number of characters should be in that text file in order to sufficiently decide if that text have the same characters distribution or not



*

*you can't actually conclude they're the same, but you can (maybe) do one of two things: 


(i) define an amount of difference you want to be able to pick up with some probability (standard sample size / power calculation for a goodness of fit test), or possibly
(ii) see if it's "close enough" by some criterion (such as via some kind of equivalence test, if we can construct one)  
I'd assume that you'd just want to treat this as a multinomial and do standard goodness-of-fit type testing -- such as a chi-square type test, but possibly using the exact (discrete) null distribution of the test statistic.

but I tried so far to calculate based on the maximum probability. if we say that "e" repeated 50 times more than "j" then I need 50 "e"s to before I expect one "j". what I want to estimate how many character in the file I should have before I see all probabilities.

I assume you mean "before I see all of the characters" - and if so, please edit your question to reflect that. 
This is a quite different question; it's a generalization of the coupon collector problem. While it would be possible to define a goodness of fit test based on the time to see all characters, I don't think it would be an especially powerful test. However, some variation on that basic idea might do reasonably well.
