The formula I am familiar with for sample size for the confidence interval of a proportion is
n = (p hat)(q hat)((z sub c/E)^2
p hat = .75 q hat = 1- .75 = .25
E = .02 confidence level = 18/20 = .9
z sub c at .9 = 1.645
so putting it all together you get
n = (.75)(.25)(1.645/.02)^2 = 1268.4 or round up to 1269
Don't know if you use the same symbols or if this is the type of problems you are doing.