variance-stabilizing transformation for simple linear regression

studying for a test.. need some help.

Consider a simple linear regression analysis with β0 and β1 both positive. Suppose that the mean of the response variable with µy and that its variance is proportional to 1/µ^2y.

a) How would this be seen in the plot of residuals vs. independent variable?
=> need help

b) What is the appropriate variance-stabilizing transformation?
=> E(Y) = µy. Var(Y) is proportional to 1/µ^2y.
Need to find a function f such that [f'(µy)]^2 prop. to µ^2y.
f'(µy) prop. to µ^y.
f(µy) prop. to .. I don't know. Am I on a right track here? Any solutions?