thanks for your patient and detailed answer, I really appreciate it. I
knew the default parameters worked very well, but if I didn't know how
to set these parameters, I kinda was stuck with this fft_length: if I
wanted to use another fft_length, I had to change the sync words
simultaneously, because it had to be as long as the fft_length.
So I tried to change the sync words just like the default ones. When I
extended the fft_length to 256-- four times as the original one, I
just copied the default sync_word1 four times, the signal was
successfully recovered on the receiver side, but on the transmitter
side, the sent signal kept flashing four-time repetition of the
original frequency spectrum every now and then. Then I changed the
sequence a little bit, to make the sequence more random, then I got a
4-time wider spectrum flash every once in awhile. So I guess the
sync_word setting can not be changed randomly. But can you tell me
why? In case you don't understand my description I attached several
screenshots.
Before solving this I think I'd better keep trying with the default
fft_length and sync_word.
Thanks for your time!
Yang
On Tue, 26 Sep 2017 11:34:48 -0400
Michael Dickens <michael.dickens@ettus.com> wrote:
> [off list]
>
> Hi Yang - Just out of curiously, why are you playing around with the
> OFDM settings? The default GR OFDM tx and rx set these parameters to
>(if
> I recall correctly) those used with 802.11a, and they do work. You
>can
> change the packet length or fft length or some others, but some of
>the
> parameters are touchy & might result in system failure.
>
> The below are my best understanding of the variables for OFDM. I
>might
> be wrong in some places; I'm still learning, like you!
>
> Although you are correct that, at least in theory, one could set the
> sync words and pilot symbols to any random values, you really don't
>want
> to do that. The sync words are used for, well, sync. They should
>have
> high autocorrelation and low cross correlation (with other sync
>words or
> random data), so that an OFDM burst starting point is easily found.
> Barker codes are a good example of possible (real only) sync words;
>I'm
> sure there are other types of sequences that could be used. Note
>that
> sync words tend to be real / BPSK modulated, though they don't have
>to
> be.
>
> You can choose to use 1 or 2 sync words, each of which is fft length
> long in symbols. Using 1 will, in general, result in lower
> autocorrelation and higher cross correlation, but reduces the total
>OFDM
> burst by 1 word, which reduces transmission latency slightly. Adding
>in
> a 2nd sync word improves correlation with only slight added latency.
>
> The pilot symbols are used for equalization. The values really
>should
> not matter since they will be normalized somewhere in the OFDM Rx
> flowgraph, so just start with 1 for simplicity. I think the GR OFDM
>uses
> [1, 1, 1, -1] and 4 pilots. I doubt the pilots are as critical as
>the
> sync words.
>
> As I said, I'm still learning about these variables / topics. The
>above
> is my current interpretation of them. Hope this is useful! Please
> correct me anywhere you see an obvious mistake. - MLD
>
> On Mon, Sep 25, 2017, at 09:54 AM, Shi Yang wrote:
>> I created a simple flow graph for ofdm transmission. My problem is,
>> how to set the value of pilot symbols and sync words? Since they are
>> basically just symbols that are already known for some certain
>> function, any random values, as long as the form are correct, should
>> work, right? But what is sync_word2 for? After setting it my flow
>> graph could not work (without that it worked just fine). And why it
>> does not need to fill with zeros on alternating carriers, since the
>> preamble is repeated twice? The grc file is attached.
No comments:
Post a Comment