Making Simulation Time Fast

[Aadil Raza]

Dear All,

Suppose I have a system where i need it to run for 20 Iterations and it takes more than 60 Minutes. Is there some way by which we can reduce the simulation time as maximum as possible? please do let me know…

[Fady]

This depends mainly on your hardware (CPU; RAM etc)

[Aabid Baba]

hello aadil,
It has been already discussed. I believe simulation time , apart from power level used for transmission , is directly proportional to the number of sweeps we are using and the system complexity ( number of components and the type of components we are using in our system).
Please refer to the link below it may give you an idea about different factors.

SIMULATION TIME

[Aabid Baba]

Also as mentioned by Alistu in one of the comments that It really depends on how much the sequence length is reduced. An the bigger the network, the more time is saved when using a shorter sequence, since the signal is processed as a discrete vectors in any of these components and the shorter the signal, the more reduction in the time it is processed.

Regards

[Aadil Raza]

Thanks for reply. Don’t you think by reducing the sequence length, the obtained results will be effected? I mean degraded…

[FAYIQA NAQSHBANDI]

Hi aadilqau

As far as i have observed the simulation time mostly depends on the power level you are using in your system design.
Try simulating a system with 40 iterations at low power level and compare it with a system with less number of system sweeps but use a high value of power in this case.The simulation time of the first case will be small compared to the second one even when the first case is using large number of sweeps.
Reason being the fiber takes a very long time to allow a high power signal to transmit.

But ofcourse it is not always possible to work at very low power levels just to have a faster simulation.So even if sometimes a system takes long to simulate it is worth to wait to achieve a the required performance.

Thanks and regards

[Naazira Badar]

Hi Aadil..

That sure shall take quite a long time to get simulated. Well I’m sure you won’t be able to bring changes in the design to reduce the time. But I think you should divide the no. of sweeps, like you could have 10 iterations in one simulation and the rest of the 10 in another. This would at least help you in analyzing half of your results without having to wait for more than an hour.

In addition to this , you could refer to the link posted by Aabid in reply #32484.

Regards.

[Aadil Raza]

Thanks Faiqa.
@ Nazira..
As you said divide it in two parts both having half of total iteration in each simulation. Then can we run simultaneously both simulation parrallel if yes then how?

[Naazira Badar]

Hi Aadil..

That’s not what I exactly meant. I meant that you could simulate for 10 iterations first. And then after that for another 10.
For parallel simulations, I am not aware if optisystem could have that inherent feature, but you could practically do that in case you can use two systems at the same time. You could use one system for 10 iterations, and another system for another 10 and simulate your projects in parallel.

Also, I believe if we have two operating systems in our PC, and Optisystem installed in both, we could divide our task on the two . I hope I have made myself clear.

Regards.

[ZULKARNAIN]

HI AADILQUA….
As discussed already the simulation time depends on input power level.High power levels will indeed make the simulation time very long.
As already told by Damain in one of his comments that you can try modifying the Numerical parameters in the optical fiber parameters to tailor the engine. You can also try to enable CUDA simulation with your graphics card to speed up the simulation as it also depends on your hardware specifications.
with regards

[Ranjeet Kumar]

Hi aadilqau,
The simulation time depends on system complexity and also on RAM of our Personal Computer. E.g if we use 3R regenerator in place of PBRS and NRZ/RZ , system will be less complex and hence takes less time, It also depends upon no. of sweep we are using.

[Aadil Raza]

Thanks all for reply.

I have benefited from reducing sequence length to an optimum level. This gave me same results at very less simulation time.
Also by considering other suggestions like making power less, reducing number of iterations, having good RAM and CPU further reduce the simulation time. Thanks alot

[Aabid Baba]

Hello aadil,
The sequence does play a role in reducing the simulation time but there is also a certain effect on the performance ( it has been already discussed in a topic ).
If your results are optimum then its fine. You could also vary other parameters like power if it certainly does not affect the performance of your system.
Regards

[Aadil Raza]

Thanks aabid. Can we have some exact relationship of setting exact sequence length for getting good results for a typical setup or it is just hit and trial thing…

One thing which I know is that by increasing sequence length we have a greater number of samples which ultimately always gives good results….

[Aabid Baba]

hello aadil,
As i said we have already discussed the pros and cons of increasing the sequence length and you are right the results get better by increasing the sequence length because the probability for each bit reduces by increasing the sequence length. And as far as values are concerned, it was just hit and trail thing we had presumed but there must be some relation.
I would suggest you to refer to this link. it has already been discussed in detail. I hope it helps.

SIMULATION TIME

Regards

[FAYIQA NAQSHBANDI]

HI AADIL
you are right performance increases with increasing the sequence length… Normally it is 128 and you can increase it as far as i know. ..you can use a 256 long sequence. As aabid mentioned error rate decreases by increasing the sequence length and i have observed it by doing same with my major project work.
Thanks and regards

[Author]

Posts