As novice to OFDM, I have a question, related to modulator. Here, I present modulator block diagram, I found somewhere. OFDM modulator AFAIK, final output signal has to have frequency spectrum consisted of $f_0, 2f_0, \ldots$ components positioned around $f_c$. Because inverse FFT produces complex numbers at it's output, my opinion is that signal at point A consists of pulses with different amplitudes having some frequency. If I am right so far, than the question is - what's the magic behind DAC block in order for him to have $f_0, 2f_0, \ldots$ components at his output B ?
The samples at the input to the IFFT block $X_0$, $X_1$, ... represent the magnitude and phase for each output frequency represented by the index. The inverse FFT translates this block to the equivalent time domain signal which has real and imaginary components (complex time domain representing your baseband signal). The dual DAC combined with a quadrature mixer converts this to a real signal centered at $f_c$ whose spectrum is identical to the baseband signal.
By declaring our data at the input to represent frequency with magnitude and phase represented by the constellation is central to the OFDM concept. You could imagine (properly) that each individual input $X_n$ to be a QPSK modulated waveform (such as your picture suggests), which is then up-converted onto its own sub-carrier, and banded together with the other sub-carriers with their own modulation, side by side in frequency, which is exactly what is occurring. The use of the IFFT just does this efficiently by putting each QPSK modulation onto it's appropriate sub-carrier by assigning it to an input of the IFFT (which must have frequency inputs), and allowing the IFFT to compute what the composite time domain signal of all of those sub-carriers would represent.