OFDM synchronisation

Hi

I'm trying to implement an OFDM system.  I decided to use the lowest 
frequency carrier as a constant tone with a cyclic prefix so that I 
could lock the receiver to this by adjusting RX frequency to achieve 
zero real signal on this channel, IOW a single sine cycle with a cyclic 
prefix before it starts.

It locks and tests very well on the bench wired straight through, and 
when I connect up through a transmission line it locks well too. 
However, the cyclic prefix discontinuity has now moved along the carrier 
by about 90 degrees due to the frequency dependant phase shift of the 
transmission line.

Well, of course it would.  Oops, should have thought of that.

But what to do?  How can I lock so that the cyclic prefix discontinuity 
is always in the right place, ie just after the FFT sampling has ended?

I don't much care about the phases of the various carriers, as I can 
correct for that on TX or RX as needed, but the cyclic prefix 
discontinuity needs to be in the right place.

Any ideas?  Preferably practical ones, I'm a bit resource limited.  Is 
there some way of locking onto this discontinuity?  It disappears into 
the mush as far as I can tell on the 'scope when I start using the data 
carriers.

Cheers
-- 
Syd

Hi,

some material for reading:
http://en.wikipedia.org/wiki/Zadoff%E2%80%93Chu_sequence

in LTE, the radios remain synchronized continuously.
A different concept is used in WLAN, as transmissions are bursty and each
gets synchronized anew (picked a random reference)

http://140.117.160.140/course/pdfdownload/93_2/%E6%96%B0%E4%B8%96%E4%BB%A3%E7%84%A1%E7%B7%9A%E9%80%9A%E8%A8%8A%E7%B3%BB%E7%B5%B1%E5%9F%BA%E9%A0%BB%E8%A8%8A%E8%99%9F%E8%99%95%E7%90%86/BBIC-WLAN-Algorithm-Synchronization.pdf

On Tue, 29 May 2012 16:53:12 +0100, Syd Rumpo wrote:

> Hi
> 
> I'm trying to implement an OFDM system.  I decided to use the lowest
> frequency carrier as a constant tone with a cyclic prefix so that I
> could lock the receiver to this by adjusting RX frequency to achieve
> zero real signal on this channel, IOW a single sine cycle with a cyclic
> prefix before it starts.
> 
> It locks and tests very well on the bench wired straight through, and
> when I connect up through a transmission line it locks well too.
> However, the cyclic prefix discontinuity has now moved along the carrier
> by about 90 degrees due to the frequency dependant phase shift of the
> transmission line.
> 
> Well, of course it would.  Oops, should have thought of that.
> 
> But what to do?  How can I lock so that the cyclic prefix discontinuity
> is always in the right place, ie just after the FFT sampling has ended?
> 
> I don't much care about the phases of the various carriers, as I can
> correct for that on TX or RX as needed, but the cyclic prefix
> discontinuity needs to be in the right place.
> 
> Any ideas?  Preferably practical ones, I'm a bit resource limited.  Is
> there some way of locking onto this discontinuity?  It disappears into
> the mush as far as I can tell on the 'scope when I start using the data
> carriers.

It sure seems to me as if you are synchronizing your transmitter and 
receiver through some channel other than _just_ the OFDM signal.  Unless 
it's crappy coax or you're sending some hugely wideband signal, there's 
not enough dispersion in the coax to cause this.  It sounds more like 
your "frequency dependent phase shift" is just plain old lag that's 
constant across the board.  If that's true, and your receiver even 
_notices_ the delay, then your receiver is depending on information 
that's just not going to be there when you try this over the air.

-- 
Tim Wescott
Control system and signal processing consulting
www.wescottdesign.com

On Tue, 29 May 2012 16:53:12 +0100, Syd Rumpo <usenet@neonica.co.uk>
wrote:

>It locks and tests very well on the bench wired straight through, and 
>when I connect up through a transmission line it locks well too. 
>However, the cyclic prefix discontinuity has now moved along the carrier 
>by about 90 degrees due to the frequency dependant phase shift of the 
>transmission line.

I do not understand why a matched coaxial cable would have frequency
dependent phase shift, at least with moderate relative bandwidths (say
10 %).

However, in a mismatched coaxial line, reflections will occur and at
some frequencies, the forward and reflective wave will cancel, causing
a deep notch in the spectrum, perhaps taking out your synchronization
carrier. Around this notc(es), the phase will vary violently. For
instance an open 1/4 wave stub could do this easily. 

Of course, if your system will not survive the simple reflections in a
mismatched coax, I very much doubt that it would survive in any real
on the air environment with complex reflections.

On 30/05/2012 02:41, Tim Wescott wrote:

<snip>

> It sure seems to me as if you are synchronizing your transmitter and
> receiver through some channel other than _just_ the OFDM signal.  Unless
> it's crappy coax or you're sending some hugely wideband signal, there's
> not enough dispersion in the coax to cause this.  It sounds more like
> your "frequency dependent phase shift" is just plain old lag that's
> constant across the board.  If that's true, and your receiver even
> _notices_ the delay, then your receiver is depending on information
> that's just not going to be there when you try this over the air.

Yes, it's very crappy coax, which I can't change.

Cheers
-- 
Syd