Hello, I have a hit a wall in my work, where I have all the DFT values except for the DC component. And what I want is to reconstruct the time-domain signal. I know this probably makes no sense, but I wanted to ask nevertheless because I seem to have exhausted all the possible ways. Is there any method I can use to reconstruct the signal when the DC component is missing, or any method to find the DC component? Thank you very much.
When the DC component is missing in DFT
Started by ●March 5, 2009
Reply by ●March 5, 20092009-03-05
On 5 Mar, 11:42, "m26k9" <maduranga.liyan...@gmail.com> wrote:> Hello, > > I have a hit a wall in my work, where I have all the DFT values except for > the DC component.How the heck did you manage that?> And what I want is to reconstruct the time-domain > signal. > > I know this probably makes no sense, but I wanted to ask nevertheless > because I seem to have exhausted all the possible ways. > > Is there any method I can use to reconstruct the signal when the DC > component is missing, or any method to find the DC component?Well, you could compute the DC component as the mean of the original signal (make sure to use the proper scale factors!), but I can't see why you would need to reconstruct the signal if you already have the original... Apart from that, are you sure you need the DC component? Some times it's OK just to get 'close enough', like to within a scale factor or to within a constant additive term. Rune
Reply by ●March 5, 20092009-03-05
Thank you very much Rune. Actually what I have is a (circularly) convolved signal. I'm trying to extract/reconstruct the original signal. I will try if I can try to calculate the average before hand and use it. Thank you.
Reply by ●March 5, 20092009-03-05
On 5 Mar, 13:36, "m26k9" <maduranga.liyan...@gmail.com> wrote:> Thank you very much Rune. > > Actually what I have is a (circularly) convolved signal. I'm trying to > extract/reconstruct the original signal.Why circular convolution? Why don't you do like everybody else and zero-pad to avoid the wrap-around? Rune
Reply by ●March 5, 20092009-03-05
This is regarding OFDM and is a transmission through air. The circular convolution is intentional because then the FFT cant be treated as a multiplication of the channel and the data. Thank you.
Reply by ●March 5, 20092009-03-05
On 5 Mar, 16:20, "m26k9" <maduranga.liyan...@gmail.com> wrote:> This is regarding OFDM and is a transmission through air. The circular > convolution is intentional because then the FFT cant be treated as a > multiplication of the channel and the data.What are you talking about? If you deal with physical EM wave propagation, the channel response is convolved with the signal. No circular convolutions or multiplications. Rune
Reply by ●March 5, 20092009-03-05
Rune Allnor wrote:> On 5 Mar, 16:20, "m26k9" <maduranga.liyan...@gmail.com> wrote: >> This is regarding OFDM and is a transmission through air. The circular >> convolution is intentional because then the FFT cant be treated as a >> multiplication of the channel and the data. > > What are you talking about? If you deal with physical > EM wave propagation, the channel response is convolved > with the signal. No circular convolutions or multiplications.Isn't is very difficult to propagate DC as an EM wave through air? Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●March 5, 20092009-03-05
On 5 Mar, 17:58, Jerry Avins <j...@ieee.org> wrote:> Rune Allnor wrote: > > On 5 Mar, 16:20, "m26k9" <maduranga.liyan...@gmail.com> wrote: > >> This is regarding OFDM and is a transmission through air. The circular > >> convolution is intentional because then the FFT cant be treated as a > >> multiplication of the channel and the data. > > > What are you talking about? If you deal with physical > > EM wave propagation, the channel response is convolved > > with the signal. No circular convolutions or multiplications. > > Isn't is very difficult to propagate DC as an EM wave through air?Static potentials? But yes, if this question, however weird it might seem, is a EM propagation problem, I can't see any reasons why one can't set the DC coefficient to 0. Rune
Reply by ●March 5, 20092009-03-05
On Mar 5, 5:42�am, "m26k9" <maduranga.liyan...@gmail.com> wrote:> Hello, > > I have a hit a wall in my work, where I have all the DFT values except for > the DC component. And what I want is to reconstruct the time-domain > signal. > > I know this probably makes no sense, but I wanted to ask nevertheless > because I seem to have exhausted all the possible ways. > > Is there any method I can use to reconstruct the signal when the DC > component is missing, or any method to find the DC component?it's kinda hard to squeeze the toothpaste back into the tube, once it's out. one you (or something else) multiplies a quantity by zero, you can't divide by zero to get it back. but you *can* reconstruct the signal less the DC component. it's like you would have passed it through a DC blocking filter (which, for some of us, is a *good* thing). r b-j
Reply by ●March 5, 20092009-03-05
On Thu, 5 Mar 2009 08:31:14 -0800 (PST), Rune Allnor <allnor@tele.ntnu.no> wrote:>On 5 Mar, 16:20, "m26k9" <maduranga.liyan...@gmail.com> wrote: >> This is regarding OFDM and is a transmission through air. The circular >> convolution is intentional because then the FFT cant be treated as a >> multiplication of the channel and the data. > >What are you talking about? If you deal with physical >EM wave propagation, the channel response is convolved >with the signal. No circular convolutions or multiplications. > >RuneIt's probably an OFDM system with a cyclic prefix. That makes the convolution of the FFT signal window a cyclic convolution with the channel. That's pretty much the reason the cyclic prefix is there. Eric Jacobsen Minister of Algorithms Abineau Communications http://www.ericjacobsen.org Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php
