Optimal weights/Tobin separation

Question

I'm having troubles understanding the following: I understand given, let's say 5 stocks and their historic data, how to compute the efficient frontier of optimal assets. However, when I introduce Tobin separation the confusion starts. Given a constant required return, how do I compute the optimal portfolio when risk-free assets are introduced (e.g. RF=2%), that is the portfolio with lowest standard deviation? For example if I want an expected return of 20%, how do I compute the proportion I should invest in risk-free asset, and the proportion I should invest in my optimal portfolio? From what I understand the optimal portfolio is the tangent point from (0,RF) to the efficient frontier. But I don't understand how to compute the proportion of the optimal portfolio, and the proportion risk-free assets.