what does the output of ifft in ofdm look like?

i have a problem in ofdm modulation .... there is a stream of bits that is converted from serial to parallel and is applied to an IFFT block. based on the definition of IFFT the output should be an complex valued number. how does it represent the real and imaginary parts in the digital domain( bits 0,1) ?

You are correct that the output of the IFFT will normally have both a real and imaginary component. That is fine and normal. When the signal is modulated from baseband to the carrier frequency the real part is typically modulated by a sine wave and the imaginary part by a cosine (or vice versa). This allows both parts to coexist without interfering with each other, since sines and cosines have zero cross-correlation.

Regarding how the numbers are represented digitally, they are represented the exact same way that real numbers are represented. The only difference is that there is an extra set of numbers for the imaginary part of the signal.

Your question is somewhat ambiguous but i'll try to answer it according to my understanding of it.
You have an OFDM system, a serial input say s(n). Your block diagram will look like this.
The signal is converted from serial to parallel. The number of parallel channels depend on N point IFFT block. ie. an 8 pt IFFT will have 8 parallel channels & hence the input will be represented in 8 PSK constellation. As seen from the diagram, IFFT separates real & imaginary parts.
If your question is how these 2 signals are represented in digital domain then the answer is since you used 8 PSK constellation to represent them before IFFT, they are represented as 8 PSK only. Both Real & Imaginary parts.
Real & Imaginary parts are just numbers when they are separated. It is when they are combined that they have magnitude & phase information.
These are then converted to Analog signals(continuous time) & modulated using cos & sin(orthogonal carriers).
Then they are added to form s(t). This signal s(t) is transmitted.
If your question is how this signal is represented in digital then the answer is it can either be PCM-ed or PSK-ed depending on transmitting media. It is represented as a bunch of 1 & 0 but order depends on coding

• The constellation has nothing to do with the IFFT length. It can even be different for each channel. Your getting to the right point (Real & Imaginary parts are just numbers) but I think your answer can be improved by pointing out the relevant information.
– Deve
Jul 12 '13 at 6:40

Maybe 8 years late on this one. But it appears that topics related to OFDM involving 'IFFT' is not even really physical OFDM as such. The issue seems to be with the complete lack of information about the topic of 'OFDM' involving IFFT.

At first, we expect (or get the impression) that taking an IFFT of complex number sequences is going to help physically generate time domain complex number sequenes such that the time domain sequence 'physically' has a 'frequency spectrum' associated with the 'sub-carriers'. If that is assumed, then it is incorrect, because periodicity is meant to be involved. That is - the IFFT (time-domain sequence needs to be made 'periodic' in order to get some 'physical' meaning between the IFFT complex sequence and the original complex sequence. So if the IFFT sequence is not purposely made to become periodic (repeating) - then we can get the idea that outputting the IFFT sequence just 'once' - isn't going to do much in terms of getting some meaningful frequency spectum.

This OFDM IFFT method is not anything like 'classical' OFDM.

Instead, it is really the sending and receiving of an ad-hoc (on paper) 'discrete-frequency' complex number sequence. And the IFFT is used because if any of the complex numbers in the 'original' sequence happen to be purpose set to zero, then this won't go down well if we use Quadrature Modulation (quadrature carrier) methods to transmit each 'complex number' value (ie. real part and imaginary part independently modulates quadrature sinusoidal carriers). But - it looks like the IFFT of a sequence (where the original sequence has 1 or more 'zero' values in the sequence) is unlikely to have zero values. Comments can always be made on this topic - just to help with understanding, or for correcting certain information that needs to be corrected.

Also - because the method is really just about sending and receiving complex number sequences, then that is really all it is. Also noting that there are details about adding/using cyclic prefixes, which benefit the recovery of the complex sequences at the receiver-side (for cases where the communications channel is not ideal ----- as in multipath channels etc). But using cyclic prefix has nothing to do with OFDM itself. It is only for setting up a condition where mathematical relationships between the input sequence, output sequence, and the channel's impulse response --- to be exploited for purposes of recovery of data from the channel's output sequence (distorted by the channel).

This answer is just given to address some apparent failure of sources to explain the topic in detail that everybody can at least have a chance of understanding.