Sines amplitudes when their frequencies do not fall on the grid

Hi, I have a time record of the sum of two pure sinusoidal tones. Naturally, the length of the record is finite, say T. As long as the tone frequencies are multiples of 1/T, a DFT (FFT) calculation will produce pure delta functions at the tone frequencies, from whose height I can determine the amplitude of the two tones. However, if the tone frequencies are NOT multiples of 1/T, each tone will be characterized by several non-zero frequency points. How should I estimate the tone amplitudes (the frequencies are KNOWN). It appears that I can use Parseval theorem and sum the square of the values of the relevant non-zero frequency points, as long the two tones are not too close.

I'll appreciate any comments/recommendations/tips(Hamming/Hannig) to help get as close estimates for the true tones amplitudes as possible.

Thanks, moshe