# Oversampling in OFDM

Why do we need an oversampling in OFDM signal? and how to do it? If around half of the Sub-carriers are only used to send data, then do we need to perform oversampling to the signal? Is it necessary to have LPF in OFDM transceiver?

• Could you provide some more detail about what you're asking specifically? Sampled OFDM signals require the same Nyquist criterion that any other discrete-time signal would in order to avoid aliasing. I would say that in general, though, using oversampling in OFDM systems wouldn't be typical. Feb 18 '14 at 13:02
• I have seen some implementation of OFDM where they insert an interpolation and windowing block. I don't understand the need for this block? Is it a must to have it or I can eliminate it if I am using less number of data sub-carriers compared to IFFT size? I am new to communication field so can you please clarify to me what to do with OFDM signal after IFFT block in transmitter side if I am going to send it at IF or RF frequency.
– Sara
Feb 18 '14 at 13:13
• If possible you should provide a reference to what you're asking about then so it will be easier to provide an answer. Feb 18 '14 at 13:16
• In the book of "OFDM for Wireless Communications Systems" by Ramjee Prasad 2004. There is a block diagram where they include windowing block.
– Sara
Feb 18 '14 at 13:19
• Also, in this reference they are using raised cosine function cresis.ku.edu/~rvc/documents/862/862_ofdmreport.pdf
– Sara
Feb 18 '14 at 13:20

Oversampling in OFDM is usually implemented by modulating some subcarriers at the spectrum margins with zero. This zero-padding in frequency domain corresponds to oversampling in time domain. If only 50% percent of the subcarriers are used, then this corresponds to an oversampling factor of 2. Specifically, if the total number of subcarriers is $N$ and if $M$ of these subcarriers are set to zero, then the oversampling factor is $N/(N-M)$. Note, that the $M$ zero subcarriers have to be arranged in a row, cyclic shifting allowed. Normally, they're arranged symmetrically at the lowest and highest frequencies.

An OFDM transmitter might require a low pass filter in order to keep out-of-band radiation low. The cut-off frequency is usually half of the sampling frequency ($f_\mathrm{s}$).

An OFDM receiver requires a low pass filter before A/D conversion (sampling) with cut-off frequency $f_\mathrm{s}/2$. Otherwise aliasing can occur.

• So, the LPF is optional in transmitter to remove out of band radiation while its a must in the receiver to avoid aliasing? In a transmitter the LPF is placed before DAC and in receiver before ADC?
– Sara
Feb 19 '14 at 11:32
• I would say the transmitter LPF is optional, yes. The filter is analogue and has to be placed after the DAC and before the ADC.
– Deve
Feb 19 '14 at 13:58
• I am confused why in a transmitter the LPF is placed after the DAC? can you please clarify. In the book I referred in my previous comment in section 5.2.1, figure 5.1 they are putting windowing block before the DAC! many thanks
– Sara
Feb 19 '14 at 15:26
• The LPF that I'm talking about is an analogue device that suppresses alias spectra caused by the digital-to-analog conversion. It therefore must be placed after the DAC. The windowing block in the figure you're referring to is for spectral shaping.
– Deve
Feb 19 '14 at 16:13
• Thanks. Just to clarify if I got it right. In this case the windowing block is optional in the transmitter while the LPF after the DAC is a must to avoide aliasing? or the analog LPF after the DAC is also optional since we are using LPF in receiver before the ADC?
– Sara
Feb 19 '14 at 17:00