Why 3dB bandwidth keep the same as the detector length （Ge length） increase？

Member Posts: 20
edited May 28

Dear lumerical support,

When I increase photodetector length(Ge length) from 5um to 50 um, I found performance behave normally except for 3dB bandwidth, which remain the same, which isconflict with what it should be.

The performance calculated is shown below.

I really want to know what happened.

The simulation profile of 3dB bandwidth for length 5um and 30um is in attchment.

I really wish to get help from you!

• Posts: 148Ansys Employee

Hi @zhenjiehan,

Thank you for your post. I'm not allowed to check your files, but I was wondering, did you change the normalization length in the CHARGE simulation?

This parameter is what will define the length of the device since we only run a 2D simulation.

• Posts: 40Member

Hi.@greg_baethge ,

Thank you for your respond. Acturally, I have checked the normalization length in device simulation. They are displayed correctly.

What else could cause this situation?

Can I send the file to you via Baidu Cloud Disk? Can you receive files from Baidu Cloud Disk?

Best regards!

• Posts: 148Ansys Employee

Thanks for the additional information, @zhenjiehan. As I mentioned, Ansys employees are not allowed to download any file posted here (see this post), so I can't check your files. Could you give some further information on the flow you used to get the response of the PD for different lengths? Regarding the bandwidth, did you use the ssac and/or the transient simulation? If both, do they give the same results?

• Posts: 40Member

Hi,

Thank you very much for your respond and sorry for not replying for a long time.

The simulation file is got from online course.

The photon absorption and average generation rate are simulated by 3D FDTD, but the cw generation exported are averaged in x direction (the direction of detector length),and then exported to charge simulation.

# Add CW Generation Monitor

set("use relative coordinates",0);

set("x",0);

set("x span",l);

set("y",0);

set("y span",w_rib);

set("z",(th_si+th_ge)/2);

set("z span",th_si + th_ge);

set("source intensity",1/yspan_mode/zspan_mode);

set("average dimension","x");

The charge simulation for darkcurrent and responsivity is 2D simulation.

This simulation uses  transient simulation to get frequency response.

I tried another struture, the 3dB bandwidth for 5um and 10um photodetector is 139.16 and 140.381 GHz.

For 5um length detector,

For 10um length detector,

I am more unfamiliar with this method（ssac）you provide，how to operate it?

I am new to photodetector simulation. Could you help see what is wrong?

If you need additional parameter information, please let me know

Thank you in advance and looking forward to hearing from you!

• Posts: 148Ansys Employee

Hi @zhenjiehan,

If you change the length of the device, you will have to rerun the FDTD simulation as the absorption will depend of the length as well. This is mentioned in the Vertical photodetector example (although we use the EME solver in MODE for this). In this example, you will also find some comparison between using the small signal AC analysis and transient analysis to get the bandwidth. Typically, the ssac analysis is easier to setup and analyze. It is valid for a small signal. I would suggest to check this example.

• Posts: 40Member

Thank you very much for your assitance.

I have changed the norm length in FDTD as I increase detector length.

Folloing the vertical photodetector example (https://support.lumerical.com/hc/en-us/articles/360042957893), the f3db bandwidth I get is still about the same, which doesn't drop as photodetector increase.

Here is the performance for length of 20um and 50um in this vertical photodetector example.

The f3dB bandwidth from ssac and transient  are about the same.

What could be this caused by? I was confused by this for a long time.

Looking forward to hearing from you!

• Posts: 148Ansys Employee

Apologies for the long wait, I've been out of the office and, I guess, my colleagues missed your post. I'm not too sure what is the issue here. I guess it all depends on what is the limiting factor in the bandwidth. If the carrier transport is the limiting factor, I guess it makes sense the length doesn't matter much since we run a 2D simulation so the carrier distribution is uniform along the detector length. In this case, a 3D simulation could be more adapted (but more costly in resource). I'll check with my colleagues, in case I missed anything. We'll keep you posted.

• Posts: 148Ansys Employee

Thanks for your patience. I discussed this with my colleagues, it happens that if there is no external series resistance, the effect of changing the norm length is not seen in the simulation. The reason is that device R reduces with increasing norm length and device C increases, so they cancel each other and the bandwidth does not change much. But if external series resistance is included it will effectively be added to device R and then the bandwidth will be affected by changing the norm length. This external R can be set in the electrical boundary condition:

By default, no resitance is defined.