ryujin_ssdt wrote:

> So maybe TDoA estimation of OFDM signals is a trivial problem no one is
> interested in??  Is this topic already covered or maybe applying CDMA
> methods (i.e. Generalized Correlation Methods) is good enough?

OFDM is a sort of modulation. OFDM allows building simple yet efficient 
receivers for the wideband signals; that's the only advantage of it.

The task of estimating TDoA is unrelated to the modulation. The main 
parameter is the autocorrelation function of the signal.

For TDoA estimation, it would be good to have the zonde signal with the 
constant envelope. The OFDM signal is very inconvenient for that purpose.

> I have searched about TDoA of OFDM signals and can hardly find anything. I
> can tell from what I have read that OFDM physical characteristics
> (multicarrier) may allow implementation of more efficient and accurrate
> TDoA estimation algorithms using GCC method and/or Bispectral methods but
> my mathematics knowledge on these topics are slowing me down.

You can compute the autocorrelation in time domain or in frequency 
domain. It is a matter of preference. The width of the peak of ACF is 
determined by the bandwidth of the signal. It does not matter if the 
signal is formed by OFDM or by any other method.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com

On May 9, 8:18 am, "ryujin_ssdt" <hsan...@gmail.com> wrote:

> My guess is that since OFDM is spectral efficient it can transmit a lot of
> data using a small bandwidth. This translates to less accurrate estimation
> ot delays.
>
> Does this mean that OFDM cannot be used for localization purposes??

Let's be clear here:  OFDM is a communication scheme, not a
localization
scheme.  It is simply a signaling format.

Now let's go back to how people do delay estimation in a typical
single-carrier,
narrowband communication system.  When the modulation symbols are
known,
the answer is straightforward:  it's based on finding the maximum of
the
correlation in the time domain.  Of course, an equivalent operation
can be
done in the frequency domain.  And as I said, the quality is better
when the
bandwidth is wider.

When the modulation symbols are not known exactly, either they are
first
decoded and then used to estimate the delay (or iteratively, a la
expectation-
maximization), or they are considered to be a nuisance parameter.
Most of
what I think is the good research in the area try to solve this
problem, or
side step it by finding a good "training sequence of symbols" to use
to get
good estimation performance.  There's a bunch of really good papers by
Georghiades et al.

Suppose now you want to do the same for OFDM.  OK, that is fine.  But
in
order to get quality proportional to your *overall* bandwidth, you
have to consider
all the tones all at once.  And I said above that correlation can be
computed
either in time or frequency domain.  There is no advantage to using
OFDM.

Think of it as if you were *designing* a signal that you wish to send
to get the
best time delay estimate.  It is already known that the best signal is
the one
with the most narrow autocorrelation.  So what does it matter if the
signal were
an OFDM signal or not?

Hope this clarifies things a bit.
Julius

>On May 8, 9:54 pm, "ryujin_ssdt" <hsan...@gmail.com> wrote:
>> So maybe TDoA estimation of OFDM signals is a trivial problem no one
is
>> interested in??  Is this topic already covered or maybe applying CDMA
>> methods (i.e. Generalized Correlation Methods) is good enough?
>>
>> I have searched about TDoA of OFDM signals and can hardly find
anything. I
>> can tell from what I have read that OFDM physical characteristics
>> (multicarrier) may allow implementation of more efficient and
accurrate
>> TDoA estimation algorithms using GCC method and/or Bispectral methods
but
>> my mathematics knowledge on these topics are slowing me down.
>>
>> Any papers, references or books that have related information would be
>> great and appreciated.
>>
>
>The mean-square error of the best linear unbiased estimator
>of signal delay is inversely proportional to the bandwidth
>of the signal.  Can you make a guess as to why using OFDM
>may be an issue here?
>
>Julius
>
>

My guess is that since OFDM is spectral efficient it can transmit a lot of
data using a small bandwidth. This translates to less accurrate estimation
ot delays.

Does this mean that OFDM cannot be used for localization purposes??

_____________________________________
Do you know a company who employs DSP engineers?  
Is it already listed at http://dsprelated.com/employers.php ?

On May 8, 9:54 pm, "ryujin_ssdt" <hsan...@gmail.com> wrote:
> So maybe TDoA estimation of OFDM signals is a trivial problem no one is
> interested in??  Is this topic already covered or maybe applying CDMA
> methods (i.e. Generalized Correlation Methods) is good enough?
>
> I have searched about TDoA of OFDM signals and can hardly find anything. I
> can tell from what I have read that OFDM physical characteristics
> (multicarrier) may allow implementation of more efficient and accurrate
> TDoA estimation algorithms using GCC method and/or Bispectral methods but
> my mathematics knowledge on these topics are slowing me down.
>
> Any papers, references or books that have related information would be
> great and appreciated.
>

The mean-square error of the best linear unbiased estimator
of signal delay is inversely proportional to the bandwidth
of the signal.  Can you make a guess as to why using OFDM
may be an issue here?

Julius

So maybe TDoA estimation of OFDM signals is a trivial problem no one is
interested in??  Is this topic already covered or maybe applying CDMA
methods (i.e. Generalized Correlation Methods) is good enough?

I have searched about TDoA of OFDM signals and can hardly find anything. I
can tell from what I have read that OFDM physical characteristics
(multicarrier) may allow implementation of more efficient and accurrate
TDoA estimation algorithms using GCC method and/or Bispectral methods but
my mathematics knowledge on these topics are slowing me down.

Any papers, references or books that have related information would be
great and appreciated.



_____________________________________
Do you know a company who employs DSP engineers?  
Is it already listed at http://dsprelated.com/employers.php ?