sweep proplem

[feras]

hello everybody,
i want to see the relationship between BER and the fiber length in my simulation using sweep mode ,i did that for 4 iteration but when i want to increase the iteration to 5 or higher than 5 the program give me massage said ” out of memory ” , how can i solve this problem if anyone can help me please .

[alistu]

Hi Feras,

Is it possible for you to upload the simulation file here in this topic page? I remember facing this error when doing a simulation using Matlab component and Matlab codes. I think the problem was over when I changed the Matlab codes. It may have occurred in your case due to large sequence length and/or samples per bit.

Regards

[feras]

hi alistu
this is my simulation i tried to sweep 10 iteration in fiber length but the simulation run only first 3 sweeps and the it give me error massage .
thank you for your helping

Attachments:

1. 64QAM-OFDM.osd

[alistu]

You’re welcome Feras. I did not understand if your problem is solved by what I suggested in my previous comment (which is reducing the sequence length and/or samples per bit) or not. By the way, I found no attachment to your comment (if actually you have intended to attach any files).

P.S. I got the attachment. Thank you.

[alistu]

Hi Feras,

I managed to run the simulation successfully. About 7 Gb RAM was used to do the simulation and RAM usage was increasing as the iterations were progressing. So if you don’t have enough RAM, you should consider shortening vectors by decreasing samples per bit and sequence length parameters.

Regards

[feras]

hi alistu
i have 4 Gb ram so i will do reduce the samples per bit maybe it will work better
thank you very much for helping

[feras]

i tried to reduce the sequence length and it is ok for sweeps but the consolation diagram is not ok , anyway the file is available now

[alistu]

Reducing the sequence length should not have that much effect on the constellation diagram (should not worsen it, even though the constellation points become less). On the other hand, reducing samples per bit in your DDO-OFDM system might have an overall negative effect on the system performance.

[feras]

so if i reducing the sequence length it is better for the design from reducing the samples per bit if i understood you correctly .right?

[alistu]

Yes, and if you reduce the sequence length more than a specific amount, the error calculations lose accuracy. I remember Damian recommending 1024 as the least sequence length to give reliable results in terms of BER. Of course, if any more value than this is possible, you’d better choose that value.

[feras]

thank you alistu very much for your helping , i will try to find the best value for both

[alistu]

You’re welcome. There is one issue with using iterations in your design. Since you are using OS12 OFDM components and you are constantly changing the length of the fiber, the symbols in constellation diagram are rotating with each iteration. You have to set receiver laser initial phase value for each value of fiber length manually. So it seems iterations cannot be used in your case unless you use new OFDM components.

Regards

[feras]

actually i can not see any rotating in the design , can u please attach me the file with the changes u suggest pleas

[alistu]

When you get a clear constellation where all symbols are separate, you can notice the rotation. As an example, you can refer to OptiSystem OFDM samples which make use of OS12 OFDM components. I ran you simulation for two different launch power values but did not manage to get a clear constellation.

[Author]

Posts