OFDM preamble symbol

Hi,

I understand that for OFDM, each symbol is 4us. I read some WLAN books
and it says that there are 12 symbols for the preamble period (short +
long). However, the duration for the preamble is only 16us (10 short,
8us + 2 long, 8us). Should the preamble is 4 symbols instead?

Thanks for your insight!

feilip@gmail.com wrote:

> Hi,
> 
> I understand that for OFDM, each symbol is 4us. I read some WLAN books
> and it says that there are 12 symbols for the preamble period (short +
> long). However, the duration for the preamble is only 16us (10 short,
> 8us + 2 long, 8us). Should the preamble is 4 symbols instead?
> 
> Thanks for your insight!

*ERROR ERROR*
You make a obvious error of *PRE* sumption ;]
I know nothing of OFDM except what I read in first few paragraphs of 
http://en.wikipedia.org/wiki/OFDM and
http://www.wave-report.com/tutorials/OFDM.htm

There is some thing missing in your problem definition.

I mean for a 11g WLAN signal. Please see slide 4 from
http://eesof.tm.agilent.com/pdf/DesignConWLAN.pdf

On Wed, 2 Jan 2008 11:02:19 -0800 (PST), feilip@gmail.com wrote:

>Hi,
>
>I understand that for OFDM, each symbol is 4us. I read some WLAN books
>and it says that there are 12 symbols for the preamble period (short +
>long). However, the duration for the preamble is only 16us (10 short,
>8us + 2 long, 8us). Should the preamble is 4 symbols instead?
>
>Thanks for your insight!

Figure 2 shows the preamble configuration, which is divided into 12
sections.   t1 through t10 are ten repetitions of the "short" symbol
sections, followed by a Guard Interval section for two, identical,
long symbols which are both the same length as the usual 3.2us data
symbols.

The second portion of the preamble only needs a single guard interval
for the two training symbols, t11 and t12, since those symbols are
identical.   The end of t11 is cyclic with t12, or in other words, the
tail end of t12 that would be used for the Guard Interval is identical
to the end of t11, so there's no need to formally place a Guard
Interval section for t12.

It can also be observed that t1-t10 can also be viewed as two,
identical 3.2us symbols, t3-t6, and t7-t10, with t1-t2 being the
cyclic guard interval for identical symbols t3-t6 and t7-t10.   So the
full preamble is made up of four symbols, and since they're two pair
of identical symbols only two guard intervals are needed.

So, yes, you're right, the preamble can be viewed as four OFDM symbols
with only two guard interval sections.

FWIW, generally t1-t10 are used for preamble detection, coarse timing,
coarse frequency, and AGC.   t11-t12  are used for fine timing and
frequency and channel estimation.

Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.ericjacobsen.org

Thanks for your tip, Eric! It helps a lot.

I don't understand why some books refers to the preamble (10 short + 2
long) as 12 symbols. 4 OFDM symbols seem to make more sense to me.

Please take a look: in section "5. OFDM PLCP sublayer of the Wireless
802.11a" at http://www.vocal.com/data_sheets/wireless_802.11a5.html

On Thu, 3 Jan 2008 06:07:51 -0800 (PST), feilip@gmail.com wrote:

>Thanks for your tip, Eric! It helps a lot.
>
>I don't understand why some books refers to the preamble (10 short + 2
>long) as 12 symbols. 4 OFDM symbols seem to make more sense to me.

I think it's just an interpretational convenience that makes sense to
some people.   As you saw in Figure 2 of the previous link, in the
time domain the short sequence is repeated 10 times, and then the long
sequence is repeated twice.   So in the time domain this can be
thought of as 12 different "symbols", even though the first ten have
different length than the last two.

In the frequency domain, or even in the time domain for that matter,
it can still be interpreted as four OFDM symbols with two guard
intervals.   All of those descriptions are consistent, they're just
looking at different attributes of the same sequence.

Another way to demonstrate this is that the sequence can be generated
either in the time domain by repeating the short symbols ten times,
then the long symbols twice with a Guard Interval, or by generating
four OFDM symbols (two identical pairs) as described in Section 5 of
the other cite  you linked.    Either way generates the same sequence.

Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.ericjacobsen.org