Long-Haul Fiber Optical Communication Systems [Optical System]

[Damian Marek]

Question:

I have a design with multiple lengths of SMF and multiple EDFA’s. I want to increase the overall length of my system, but the layout is starting to get very large and cumbersome. Is there a better way to do this?

[Damian Marek]

Answer:

It is often difficult to experiment on long fiber lengths in a lab setting. Usually a lab will perform the experiment on a shorter length fiber, but loop the signal through multiple times. This is performed using a loop control.

In OptiSystem, the component Loop Control performs this function. You can learn more, by following the tutorial at:

Lesson 3: Optical Systems WDM Design

[alistu]

Dear Mr Marek
I have been trying to simulate a CO-OFDM system using Optisystem v12.2 and the example “Coherent Detection Optical OFDM System.osd”. As I increase the number of loops, the simulation time becomes very long. For example, if I run the simulation with 2 loops, the simulation is done rather quickly. But now that I have simulated the system with 5 loops (whose .osd file is attached), it has taken more than four hours (so far) just to finish calculations for the second loop. I would like to know if this is due to some kind of software bug and if there occurs the same problem if newer versions are used. I need the software to work on long-haul CO-OFDM systems (eg. 6000km-long) and with such a problem, I will not be able to get by. Really appreciate your help.
Thank you in advance.

Attachments:

1. Coherent-Detection-Optical-OFDM-System-Loop5.osd

[Damian Marek]

The fact that it works quickly for low number of loops and takes much longer for more loops, leads me to believe that this is a memory issue. In your project file you are using quite large sequence lengths, which can lead to using too much memory. Once the memory is full the computer starts to use the hard drive to store information and slows down the simulation considerably.

[Alessandro Festa]

Thank you Damian, this is very helpful for long repeated systems!

[alistu]

I am using a computer with an 8Gig DDR3 RAM and an Intel 4-core I7 2600k CPU. The image file I have attached shows my CPU and memory usage while the simulation is run and calculations for the second loop are being done. Apparently the calculations do not even take a quarter of the memory capacity.

Attachments:

1. 

   CPU-and-Memory-Usage.jpg
   

[Damian Marek]

I missed something else and I think I’ve found the main problem. I had some help from the technical team working on OptiSystem and they pointed out that the power levels for your amplifiers are quite high. What this does is cause large amounts of non-linearity in the fiber, which in turn makes the simulation much longer. If you set the amplifier gain to around 10 dB the simulation runs fine again. Try that.

Regards

[alistu]

Thank you very much Damian. You also mentioned something in one of your posts about Optisystem problem with OFDM. Does it really make a significant difference if I simulate the system with 5 loops and high power levels of amplifiers with Optisystem 13? I am using Optisystem 12.2 now.

[Damian Marek]

Yes it does, and I tried it with lower amplifier levels. Since each loop both amplifiers produce a gain of 30 dB and the total loss is around 28 dB. That means each loop the signal is getting larger by 30 dB! The split-step fourier method, that calculates the nonlinear and dispersion effects, then needs to use many more sections to split up the fiber, which can cause a slow down for the simulation. You could also try the CUDA option in the Layout Parameters. Using CUDA, fourier transforms can be calculated quicker and in this case might improve the simulation time considerably.

Regards

[alistu]

It was helpful. Many Thanks.

[Author]

Posts