Re: [Discuss-gnuradio] high-level OFDM question

[Tom Rondeau]

On Fri, Feb 12, 2010 at 10:34 AM, Achilleas Anastasopoulos
<address@hidden> wrote:
> Hi there,
>
> before delving into the details of the OFDM implementation in GNURAdio I
> wanted to ask a high-level question:
>
> What is the method used for initial timing/frequency acquisition?

We use, by default, the standard Schmidl and Cox. There are two other
methods that we wrote which you can use to varying degrees of success.
My presentation from the address@hidden symposium, 2007, has details of
all of these methods and why we use the one we do.

> Is there one or two training OFDM symbols that precede the transmission of
> data? and how many OFDM data symbols are following? Is this number
> determined by the first training symbols or is it fixed?

We use a method that only requires 1 training symbol. The much of the
code is actually set up to allow you to put in however many you want,
but you'll need to redesign some of the the receiver blocks to make
use of them.

The number of symbols following depends on the frame length you
specify. In the benchmark_ofdm* examples, we default to 400 bytes with
200 occupied tones per symbol, so for BPSK this is (400*8/200) = 16
symbols. So there are 16 symbols between preambles.

> Also, my understanding is that each data OFDM symbol has pilot tones for
> continuous frequency/timing tracking, right?
>
> Thanks
> Achilleas


While I wrote the transmitter to allow for pilot tones, we don't
actually make use of them in the receiver. Instead, we use a decision
feedback equalizer to keep on track. Making use of the pilot tones in
the receiver is definitely something we need done, though.

Tom