How does a time domain OFDM signal look like

Hi,

The ofdm symbols when placed in the freq domain say for QPSK, all the
spectrum components have the same value. 
Hence it results in a flat square frequency spectrum.

Hence the time domain OFDM symbol shoul be a sinc pulse. 

Am I right ??

On Oct 23, 1:10 pm, "walter2k" <walte...@gmail.com> wrote:
> Hi,
>
> The ofdm symbols when placed in the freq domain say for QPSK, all the
> spectrum components have the same value.
> Hence it results in a flat square frequency spectrum.
>
> Hence the time domain OFDM symbol shoul be a sinc pulse.
>
> Am I right ??

No.

Firstly, whilst the PSD (power spectral density) may be flat, each of
the subcarriers will be at one of four different phases, so the
frequency spectrum is not flat.

Secondly, once you've oversampled in the time-domain, the PSD won't
even be flat; it will be a series of impulses.

--
Oli

>On Oct 23, 1:10 pm, "walter2k" <walte...@gmail.com> wrote:
>> Hi,
>>
>> The ofdm symbols when placed in the freq domain say for QPSK, all the
>> spectrum components have the same value.
>> Hence it results in a flat square frequency spectrum.
>>
>> Hence the time domain OFDM symbol shoul be a sinc pulse.
>>
>> Am I right ??
>
>No.
>
>Firstly, whilst the PSD (power spectral density) may be flat, each of
>the subcarriers will be at one of four different phases, so the
>frequency spectrum is not flat.
>
>Secondly, once you've oversampled in the time-domain, the PSD won't
>even be flat; it will be a series of impulses.
>
>--
>Oli
>
>

I have simulated an OFDM freq spectrum using 64QAM in matlab, using 256
fft full channel allocation for wimax. 
Each subcarrier in the spectrum has its own amplitude and phase based on
IEEE specs.. When i take the inverse fft of the spectrum, the time domain
signal is a sinc function. That I presume because we have a square wave
multiplied with the subcarrier spectrum when we take the ifft. Hence the
ifft looks as if , in the time domain, the subcarrier spectrum time signal
has been convolved with the time signal of square wave spectrum ie the sinc
pulse.

I tried searching on the internet but could not find any answers. Am I
going in the right direction. Does the 64 QAM ifft time domain signal look
like a sinc pulse.

I am new to communication theory and need some help to figure it out.


Walter

On Oct 23, 6:03 pm, "walter2k" <walte...@gmail.com> wrote:
> >On Oct 23, 1:10 pm, "walter2k" <walte...@gmail.com> wrote:
> >> Hi,
>
> >> The ofdm symbols when placed in the freq domain say for QPSK, all the
> >> spectrum components have the same value.
> >> Hence it results in a flat square frequency spectrum.
>
> >> Hence the time domain OFDM symbol shoul be a sinc pulse.
>
> >> Am I right ??
>
> >No.
>
> >Firstly, whilst the PSD (power spectral density) may be flat, each of
> >the subcarriers will be at one of four different phases, so the
> >frequency spectrum is not flat.
>
> >Secondly, once you've oversampled in the time-domain, the PSD won't
> >even be flat; it will be a series of impulses.
>
> I have simulated an OFDM freq spectrum using 64QAM in matlab, using 256
> fft full channel allocation for wimax.
> Each subcarrier in the spectrum has its own amplitude and phase based on
> IEEE specs.. When i take the inverse fft of the spectrum, the time domain
> signal is a sinc function. That I presume because we have a square wave
> multiplied with the subcarrier spectrum when we take the ifft. Hence the
> ifft looks as if , in the time domain, the subcarrier spectrum time signal
> has been convolved with the time signal of square wave spectrum ie the sinc
> pulse.
>
> I tried searching on the internet but could not find any answers. Am I
> going in the right direction. Does the 64 QAM ifft time domain signal look
> like a sinc pulse.
>

The only way the time-domain signal can look like a sinc function is
if the frequency-domain representation is the DFT of a sinc function
(more correctly, a truncated discrete-time version of a sinc
function).

For this to be the case, your IFFT would have to be wider than the
number of active subcarriers, and all subcarriers would have to be the
same amplitude and phase (alternatively, there could be a constant
phase offset between neighbouring subcarriers).

So in general (i.e. with arbitrary data), you shouldn't be seeing a
sinc pulse!  I can only assume that you must be doing something wrong,
but without further details, I can't tell what.

--
Oli

On Oct 23, 8:10 am, "walter2k" <walte...@gmail.com> wrote:
> Hi,
>
> The ofdm symbols when placed in the freq domain say for QPSK, all the
> spectrum components have the same value.
> Hence it results in a flat square frequency spectrum.
>
> Hence the time domain OFDM symbol shoul be a sinc pulse.
>
> Am I right ??

OFDM looks like white noise in the time domain.

John

John wrote:
> On Oct 23, 8:10 am, "walter2k" <walte...@gmail.com> wrote:
>> Hi,
>>
>> The ofdm symbols when placed in the freq domain say for QPSK, all the
>> spectrum components have the same value.
>> Hence it results in a flat square frequency spectrum.
>>
>> Hence the time domain OFDM symbol shoul be a sinc pulse.
>>
>> Am I right ??
> 
> OFDM looks like white noise in the time domain.
> 
> John
> 
One man's white noise is another's robust and stealthy signal. :-)

Steve

>On Oct 23, 8:10 am, "walter2k" <walte...@gmail.com> wrote:
>> Hi,
>>
>> The ofdm symbols when placed in the freq domain say for QPSK, all the
>> spectrum components have the same value.
>> Hence it results in a flat square frequency spectrum.
>>
>> Hence the time domain OFDM symbol shoul be a sinc pulse.
>>
>> Am I right ??
>
>OFDM looks like white noise in the time domain.
>
>John
>
>


Thanks for the help.

I got the symbol now and it looks like random data.

There was some understanding problem with OFDM theory and some matlab ifft
issues (I am new to matlab)..

Thanks once again.

Walter