SIMULATION TIME

[FAYIQA NAQSHBANDI]

How can we reduce the simulation time of our system apart from reducing the power level ?

[Aabid Baba]

Hello Fayiqa,
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).

You can reduce it by reducing the sweep iterations if at all there is a window you can can reduce sweeps for your system.

Regards

[Naazira Badar]

Hi all..
Yes, the no.of sweeps has a major role there in. Also, if you are introducing delays in your system, the simulation time would further get increased. Moreover, for some reason I believe if you simulate three individual configurations, the total time taken is less than that of three sweeps set .

Regards.

[FAYIQA NAQSHBANDI]

hello aabid

Even if your system has large number of sweeps for other parameters except power.It doesn’t take long for the system to simulate.
But i am working at high values of power.So reducing or increasing the number of sweeps for other parameters doesn’t really help.

Thanks

[FAYIQA NAQSHBANDI]

Yes naazira.

Use of delays also increases the simulation time.Main problem is that i cannot decrease the power level of my system.
Even if I use a single sweep or I use around 10 sweeps in my design.It takes around the same time to simulate.
Actually the fiber takes a long time to let the high power signal propagate through it.So main issue is with the power.

Thanks

[Naazira Badar]

Okay Fayiqa. So that must be with power only. What’s the value of power per channel in your system?

Regards

[aasif bashir dar]

hi all,

you need sweeps to take results at different values of parameter, hence reducing no. of sweeps can not be logical way to reduce simulation time.
there can be complex systems which can not be further reduced to simpler systems.thus reducing system complexity may not work at times.

with regards

[Aabid Baba]

Hello Naazira,
I find both your responses logical. Naazira you are right delays do have an effect on simulation time in our systems but at the same time you cannot avoid using them because they are important for simulation purpose even though there is no physical significance using them in our designs as it is mentioned in optiwave itself.

Regards

[Naazira Badar]

Yes Aabid, that’s right. (y)

Regards.

[alistu]

Hi Fayiqa,

It is not feasible to change the complexity of number of iterations of a system in many cases, as the design may change somewhat. However, the reduction in the sequence length can be possible in many cases. Even though reduction from a certain point on may result in less accurate answers.

Regards

[aasif bashir dar]

hi all
well mentoned by alistu
we gennerally prefer that sequence is well enough to get accurate results.but reduction in sequence length does not reduce the simulation by great extent. i have this while simulating my .osd file.

however i must mention that my designs were not soo completed. so i dont have idea how great impact of reduction in the sequence length in complicated networks , large sweep iteration,high data rates, will be on the reduction of simulation time.
.

with regards

[alistu]

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

[Aabid Baba]

Hi alistu,
Thank you for making a valuable point here. I have never tried to reduce the sequence length in any of my implementations. How do i exactly do it??
I guess instead of PsuedoRandom sequence generated randomly by the generator i have to use the user-defined sequence..Right??

regards

[alistu]

You’re welcome Aabid. You can change the value of sequence length by accessing it in layout parameters. Then the bit sequence created by PRBS generator changes, since it reads the value in layout parameters via script mode. The default value for sequence length is 128 in layout parameters.

Regards

[Aabid Baba]

Hi Alistu,
yeah i can see it right now. But don’t you think it can have an impact on the performance of the overall system because if we reduce the sequence length don’t you think it can limit the bit rate. Because here i presume that higher the sequence length we keep , the probability of Bit error rate decreases. Please correct me if i am wrong.
Regards

[alistu]

It does not affect the bit error rate in that way. Either two bits are sent or a long sequence is sent, the time interval between the two consecutive bits being sent remains the same and is equal to 1/bit rate ; where the bit rate is determined in the layout parameters of the design.

Regards

[Aabid Baba]

Well. Thank you so much Alistu. it was rather very helpful point you made.
But here i wish to ask doesn’t changing sequence length has any impact on the performance because whenever we design a simple system the sequence length is set to 128 by default and if we reduce it i guess there must be some fluctuations in our system response.

Regards

[alistu]

you’re welcome Aabid. Changing the bit rate definitely has an effect on the accuracy of the calculation of the bit error rate, especially when BER test set is used in the implementation, since it calculates BER on an empirical basis unlike the BER analyzer. The longer the bit sequence, the more accurate BER results up to a point.

Regards

[Aabid Baba]

Hi alistu,
If you read my comment #31080 , i had mentioned the same there that that higher the sequence length we keep , the probability of Bit error rate decrease which means our system becomes more accurate because the error rate in our bits has decreased though i completely agree with you that it doesn’t affect the bit rate.
Thanks for the valuable inputs.

Regards

[aasif bashir dar]

hi aabid,
thanks all for reply
changing sequence length will provide more accurate results(because probability of error decreseses with incresing sequence length),there is no such imapct of the sequence length on the system performance.

with regards

[alistu]

Hi Aabid,

I’d rather put it this way: When the sequence length is increased (of course, up to a certain point), the measurements become more accurate rather than the probability of error is reduced. When the sequence length is too short, each time the BER is calculated with a different results as a consequence.

Regards

[Aabid Baba]

Hi asif,
If you had gone through my comment #31080 , i have said the same thing there that bit error rate would decrease by increasing sequence length.
And you seem to be contradicting with your comment. On one hand you say system becomes accurate and on the other hand you say it does not have an impact on system performance. If your system becomes more accurate , it simply means your system is performing better.

Regards

[aasif bashir dar]

hi aabid,
thanks all for having good discussion

after alistu’s explanation, there could not any better and lucid explanation to under the significance of sequence length.
i apprciate his respone and his efforts.

with regards

[Aabid Baba]

you are welcome asif. It was indeed a helpful discussion. It is a good platform where we can learn about the parameters of the components we use in our system designs. It is good to know more about the characteristics and set parameters of the component than the component itself.
Thank you all.
Cheers!

[alistu]

You’re welcome and thank you all for drawing some interesting question.

Regards

[FAYIQA NAQSHBANDI]

Hi Evryone

So there has already been a prety long discussion on the topic.Since the point was to find ways of reducing the simulation time.What i understand is that thr would always be a trade off between the system performance and time if any of the parameters are altered.So one will have to compromise on the system performance if a fast simulation time is required and vice versa.
Thank you all for responding.

Regards

[Aabid Baba]

Hi fayiqa,
Absolutely i completely agree with you. Indeed, there is a definitely a trade-off between changing various parameters in our system designs. You just have to be sure enough in choosing optimum values for every parameter. But here i would like say that what actually is more important is the system performance rather than simulation time.
regards

[FAYIQA NAQSHBANDI]

Hi aabid.

Yes of course,system performance is definitely more important.I posted this question just to know if there was a way to reduce the simulation time not affecting the performance of the system.But i see thr is no choice than to compromise on either of the two.

Thanks and regards

[Aabid Baba]

hi fayiqa. You are welcome.
I suggest you to worry more about system performance but no doubt simulation is also important because it could save a lot of time if the simulation was fast. All you can do is just compromise.

Regards

[alistu]

Hi all,

I agree if some measures are taken, the system performance is somehow compromised for shorter simulation time to be achieved, but we cannot say this for using shorter sequence length. In this case, if only the sequence length becomes too short, the trade-off is between “accuracy of results” and simulation time.

Regards

[Aabid Baba]

Hi Alistu,
I presume that shortening the sequence length would only reduce the simulation time but in doing that we may have to compromise the accuracy of the system design. As i mentioned above that what matters most at the end is the system performance. Rarely you would be asked about the simulation time though its important too. But yeah i agree with you there is a compromise between accuracy of results and simulation time.

regards

[alistu]

I also agree that accuracy is important unless the simulations are possible to carry out. Mathematical approximations used to model many component verify that sometimes for calculations to be shorter, it is OK to get less accurate results. But generally, I would also rather get accurate results as much as possible.

[Aabid Baba]

Yes Alistu , you made a very good point here. I second you on that. It must also be the mathematical modelling of the components that is also responsible for their simulation times. Indeed very very important point you mentioned.
Ans as far as simulation time is concerned , i would again suggest to bother more about accurate results. In fact, It is common for all of us , as long as we are getting more accurate results we should be least bothered about the simulation time.

Thanks anyways
regards

[Author]

Posts