OptiBPM simulation results seems to be very pixelated

[Narendra Putra Dipta]

Hi sir/madam. My name is Narendra Putra Dipta. I am an undergraduate student from Universitas Indonesia
I have an issue regarding OptiBPM 13 simulation. Recently, OptiBPM 13 in my PC behave really weird, the simulation results seems to be very pixelated. This issue just come 2 days ago, before this program worked just fine without any problems.

I have attached the screenshot along with this post. Please help me to resolver this problem.

Best Regards
Narendra Putra Dipta

[Scott Newman]

Hello Narendra,

I would like to ask a few clarifying questions/requests in order to look into this for you.

1. To be clear you ran this same design a few days ago and got acceptable results and now you run the exact same design with no changes and now are getting this behaviour?
2. Can you please provide me with the version number of your OptiBPM found through the help/about menu item. I would need the full version number such as 13.1.1.1492.
3. Have you performed a windows update between now and when the results were acceptable?
4. Are you able to provide the bpd file for this? If not publicly on the forum then through support@optiwave.com.

Scott

[Narendra Putra Dipta]

Hi sir, thank you for your reply

1. Yes. Even more, i can never make my own design however simple it is. The only model having acceptable simulation result is the ‘Getting Started’ demo model, if it’s designed in my PC it’s definitely going to be pixelated. I even tried to build exactly the same model as the ‘getting started’ demo and the result for my copy is pixelated while the demo model is just fine
2. 13.1.1.1467
3. I checked my windows update history and all i can see is the windows defender definition update, no significant update between the period
4. Sure.

I am looking forward for your reply.

Best regards
Narendra Putra Dipta

[Narendra Putra Dipta]

Here is my .bpd file

[Scott Newman]

Hello Narendra,

I found your problem in examining your bpd file. In your design you have the background of your simulation (Set in Simulation Domain / 3D Wafer Properties) to air while your waveguides are 1.5. If you will note in the example design it is in fact set to a cladding value of 1.44 with a core of 1.46. Did you mean to setup a simulation with such a high dielectric contrast? Note that with new designs these settings are defaulted to the previous design.

OptiBPM is a numerical algorithm that is designed for low-index contrast applications that do not have reflections. Modelling a high-index contrast y-branch would prove to be very troublesome using this numerical algorithm. This is something more in line with the OptiFDTD application.

Scott

[Narendra Putra Dipta]

Hi sir, I really appreciate your reply.

Thank you for your advice and solution. It is becoming crystal clear now.
Yes, I wanted to design/model power beam splitter on Silicon wafer with SiO2 buffer layer. For such design, is it better for me to use OptiFDTD instead of OptiBPM?

Best Regards
Narendra Putra Dipta

[Scott Newman]

Hi Narendra,

It is not necessarily that simple a question as a lot would depend on your preliminary design which you need to have an idea of before getting into the products. Let me touch on a few items for both products to see if I cannot clarify a few things for you because OptiBPM may be able to support your structure but you cannot simply make the sample a high contrast as it complicates things.

OptiBPM

• By simply changing the materials so you have 1.5 to 1 those 4 um waveguides which were single mode in the sample (1.44 to 1.46 contrast) are now most definitely multimode so when you hit the split your fundamental mode and couple to other modes as well which leads to some of the issues as you are now going to see the beating between a number of modes.
• The full vector OptiBPM (Simulator option in 3D Isotropic simulation parameters) is more tolerant of high refractive index contrast.
• The resolution in the sample is not adequate for a high contrast simulation.
• If you drop the waveguide widths to 1um (single mode), increase your mesh to 10 points/um, propagation step to 1, and run a Vector simulation you will see that OptiBPM is able to obtain what appear to be reasonable results.

OptiFDTD

• Provided you use adequate resolution (minimum of lambda/(10n) OptiFDTD can readily model high index structures.
• One of the problems with OptiFDTD is that due to the resolution requirements simulations can be resource intensive, especially in 3D. You need to keep an eye on the size of the domain as the resources required for a 3D simulation that is orders of magnitude larger than the source wavelength may be more than you have.
• One thing to look at is that the sample you are working with is 800 um long because it is low index contrast and therefore low confinement and requires that distance to do the split. Moving to high index results in higher confinement which means your split can be accomplished over a shorter distance. Typically SOI photonics based structures are not on the scale of millimeters because they do not need to be.

Regardless of what application you ultimately end up needing I strongly encourage you to have a very good understanding and expectation of the parameters of your structure. This expectation will guide you in the decision to the nature of the numerical algorithm you need for your research.

Scott

[Narendra Putra Dipta]

Hi Mr. Scott, thank you for your help and explanation. I really appreciate it and I definitely will learn more about this matter.

Anyway, I have one more question regarding your other product, OptiMode.
I tried to create a model that consists of 4 layers: Silicon, SiO2, and GaN with air as the cladding (see ‘Structure_error.png). But The OptiMode can never run the simulation. I believe the error is caused by the refractive index of silicon which is higher than all the layers in the model and so the OptiMode expects the silicon layer to be the desired waveguide which is not the case. I want the light to propagate in the GaN layer.

My question is: Is it possible to set the GaN layer to be the waveguide layer? I tried to look into the doc but couldn’t find any information

Best Regards
Narendra Putra Dipta

[Scott Newman]

Narendra,

This can be a difficult structure to model as the mode solver works on the whole structure and you are correct most of your dominant modes are going to be found in the Si layer. A way of working around this would be to use the FEM mode solver as it allows you to set a “Mode Index Estimate”. As your layer of interest is 2.25, you would expect the fundamental mode to be slightly below this. You can set that to 2.25 and request a large number of modes (30ish) and sort through them to find the one you are interested in.

Based on a quick test it looks like your fundamental mode is at 2.2405825. Please note that if you are injecting this you must have you number of modes and settings selected such that the mode you want injected is the last one in the list.

Scott

[Narendra Putra Dipta]

Hi Scott

Thank you for your response and your patient.
It’s all resolved now. I learned a lot from you. Thanks

Best Regards
Narendra Putra Dipta

[Author]

Posts