strange results

[feras]

hello everybody.
i tried to draw a 2D-graph to see the relationship between the BER and the fiber length using 5 sweeps in fiber length , but i got a strange result cause i know that the BER is increase with fiber length increasing but in my result the BER first decreasing then start to increasing ( the graph is attached with the post ) , can anyone give me the reason of this please .
thank you all

[alistu]

Hi Feras,

Can you please also attach the design file if it is possible? If the design is the one used in the below topic, please make sure about symbol rotation before using sweep iterations. Otherwise it is really expected to get plots like the one you have attached:

optiwave.com/forums/topic/sweep-proplem/

Regrads

[feras]

hi alistu,
yes it is the same design in the topic (optiwave.com/forums/topic/sweep-proplem/) , but really in my simulation i did not see any rotation , i will attach you the constellation diagram for the first(25KM) and the third sweep (62.5KM)
and thank you very much

[alistu]

You’re right, I also see no rotation, but I also see no error in the constellations while your plot shows about 0.35 BER for each of these steps. Of course, BER analyzer will show non-zero error for these cases but much less than what is shown above. Which parameter have you chose for BER from project browser?

[feras]

hi alistu,
actually i used min.BER result in the graph with fiber length , but there is a strange issue in the simulation cause when i re-run the simulation i got different result for the same parameter ( each time give me different result ) , i attached the new graph result here and it show a good result as it expect.

[alistu]

The reason why I asked you to do so was the fact that sometimes short bit sequences cause simulation results to change each time the simulation is done. If you can’t get the same results each time you do the simulation, I suggest you use a longer bit sequence and not use sweep iterations to avoid problems due to lack of memory.

[Ravil]

Hi Feras,

There are few reason that can be the reason of obtaining the graph that you have. Your fiber length is not the main reason of your measured BER decrease (since it attenuates noise as well as a signal). The main parameters that decrease your BER are the number of your amplifiers and nonlinearities. Therefore, I agree with Alistu that it would be helpful to look at your design file first.

[feras]

hi Ravil ,
thank you very much for your interested , actually the design file was attached in a previous topic and i will attach it again here ( it is run for 5 sweeps to fiber length )
Regards

[alistu]

I noticed the attached design and realized bit sequence seems to be long enough. Even so, I suggest you only conclude this after a few times of repeating sweep iterations to make sure you get the same results (BTW, I also used BER test set and got BER of about 0.3 for your first iteration even though constellation looked rather good).

[feras]

i changed to BER test set too , but the different results still appears , and it still changed every time i run the simulation (actually this is very strange ) , but i noticed something that when i run the simulation for N iteration , the N’th run has the better results from the previous and the next runs , and i don’t know why .
there is very important thing i noticed also that even i used one iteration ( one length for the fiber without using sweep mode ) it is also different results on each time i run the simulation and that is very strange
Regard.

[alistu]

How about doing the calculations one by one with extended word length without using the sweep iteration. When the sequence length is extended, fluctuations in Q-factor and BER are reduced (using BER analyzer). You cannot use long bit sequences with sweep iterations due to the memory problem you mentioned in the other topic and maybe the reason for fluctuations in your results comes from this.

[feras]

hi alistu
i will try to do that with long sequence length.
thank you for your interested .
Regards.

[Author]

Posts