Forums

Guard Interval removal methods

Started by irsha November 25, 2009
hi all! i'm currently working on my thesis about FFT windowing.. but i'm a
little short of ideas on how to remove the cyclic prefix components in the
receiver.. i can't seem to find a literature that points specifically on
this issues.. and one more thing, does timing synchronization is relevant
on this problem?

thanks before for the help..




On Nov 25, 9:03=A0am, "irsha" <irsha.prima...@gmail.com> wrote:
> hi all! i'm currently working on my thesis about FFT windowing.. but i'm =
a
> little short of ideas on how to remove the cyclic prefix components in th=
e
> receiver.. i can't seem to find a literature that points specifically on > this issues.. and one more thing, does timing synchronization is relevant > on this problem? > > thanks before for the help..
In fact obtaining good sync in the presense of noise is at the heart of a good performing receiver design. Once you get the sync working well, the rest follows. Often you will find the sync and error correction components intertwined together. You know that by running your data through a buffer and once you have good sync you can actually backup to the start of the valid data. Good Luck, Clay S. Turner
>On Nov 25, 9:03=A0am, "irsha" <irsha.prima...@gmail.com> wrote: >> hi all! i'm currently working on my thesis about FFT windowing.. but
i'm =
>a >> little short of ideas on how to remove the cyclic prefix components in
th=
>e >> receiver.. i can't seem to find a literature that points specifically
on
>> this issues.. and one more thing, does timing synchronization is
relevant
>> on this problem? >> >> thanks before for the help.. > >In fact obtaining good sync in the presense of noise is at the heart >of a good performing receiver design. Once you get the sync working >well, the rest follows. Often you will find the sync and error >correction components intertwined together. You know that by running >your data through a buffer and once you have good sync you can >actually backup to the start of the valid data. > >Good Luck, >Clay S. Turner >
Thanks for the information about the sync.. but can anyone please explain to me the general method on removing the cyclic prefix components? i found one way to do it is by correlating the CP with last few symbol on the received ofdm symbol.. for example, if I have 64 ofdm symbol transmitted with 1/4 CP, the method is to sample 16 first symbol and 16 last symbol (because the cyclic extension is copied from the original data) then correlate both data. If I get a high correlation rate, (for example > 0.9) then the data between the sampled symbol must be the FFT window. If the correlation rate is low, then the sampling moved one symbol before or after depending on the program. The iteration continue until a high correlation rate is achieved. Is the method efficient? or is there a method that used generally.. please help me on this issue.. Thanks in advance! Irsha Primanda
On 11/25/2009 7:54 PM, irsha wrote:
>> On Nov 25, 9:03=A0am, "irsha"<irsha.prima...@gmail.com> wrote: >>> hi all! i'm currently working on my thesis about FFT windowing.. but > i'm = >> a >>> little short of ideas on how to remove the cyclic prefix components in > th= >> e >>> receiver.. i can't seem to find a literature that points specifically > on >>> this issues.. and one more thing, does timing synchronization is > relevant >>> on this problem? >>> >>> thanks before for the help.. >> In fact obtaining good sync in the presense of noise is at the heart >> of a good performing receiver design. Once you get the sync working >> well, the rest follows. Often you will find the sync and error >> correction components intertwined together. You know that by running >> your data through a buffer and once you have good sync you can >> actually backup to the start of the valid data. >> >> Good Luck, >> Clay S. Turner >> > > Thanks for the information about the sync.. but can anyone please explain > to me the general method on removing the cyclic prefix components? i found > one way to do it is by correlating the CP with last few symbol on the > received ofdm symbol.. for example, if I have 64 ofdm symbol transmitted > with 1/4 CP, the method is to sample 16 first symbol and 16 last symbol > (because the cyclic extension is copied from the original data) then > correlate both data. If I get a high correlation rate, (for example> 0.9) > then the data between the sampled symbol must be the FFT window. If the > correlation rate is low, then the sampling moved one symbol before or after > depending on the program. The iteration continue until a high correlation > rate is achieved. > > Is the method efficient? or is there a method that used generally.. please > help me on this issue.. > > Thanks in advance! > Irsha Primanda
I was a bit confused as to what you were asking, as removing the CP is pretty trivial once symbol synchronization is achieved. Now it appears you're asking how to achieve symbol synchronization. What you've described is a very common method for determining symbol boundaries in an OFDM signal. The CP will be corrupted by ISI from the previous symbol in a channel with delay spread, so the cross correlation won't be ideal. It is usually good enough, though, at least to achieve initial coarse time synchronization. Fine tuning can be done by other means if needed, but CP correlation is a great place to start. -- Eric Jacobsen Minister of Algorithms Abineau Communications http://www.abineau.com