fft ifft
Let me try to give you a more detailed example.
First, on FFT/IFFT. Let's say you have a signal, and you sample it N times at a certain sampling frequency fs. So you'll have the discretized signal x
for n=0 to N-1.
If you do a FFT on this signal, you will get a discrete spectrum f(k) for k=0 to N-1. Now, f(k) is related to the real frequency by f=(k/N)*fs, so the interval between each frequency bin is fs, and there are N frequency bins in total.
Now, we move on to OFDM. You have to decide what subcarrier frequency interval you want. Let's say you decide on M*fs. So the subcarriers f(k) where k=M*c for c=0 to (N/M)-1. Now, there are N/M subcarriers in total.
For practical reasons, we space them out evenly over the discrete spectrum. Also, you have to understand that the sampling frequency fs must be equal or higher than the frequency of the input stream, otherwise you would have to buffer the input stream to prevent overflow.
Now, you take the N/M samples of the input stream and set it as the magnitude of each subcarrier. You then do the IFFT on this spectrum and you get your transmitted signal. Since this spectrum has N frequency bins (of which only N/M are bins occupied by subcarriers), you will get an N-point signal.
Added after 10 minutes:
Lastly, just a note on terminology. We call each discrete time signal a sample, and each discrete frequncy spectrum a frequency bin. We call carriers carriers, or in the case of OFDM, subcarriers.
Baskets and eggs are better used for picnics and breakfast.