Questions about ring modulator model

[mteng]

Dear Lumerical staff,

I have a number of questions about ring resonator/modulator model used in interconnect.

You may refer to the following example:
https://kb.lumerical.com/en/pic_circuits_ring_modulators_dc_response.html

In this example the waveguide loss is setup as 7dB/m, which is 4e-4 dB/round trip intrinsic loss for a 60um long ring.
However I do suspect the unit should be 7dB/cm while 7dB/m means the ring is almost lossless.
Then the power coupling ratio of 1% is used, which after interconnect simulation shows almost maximized extinction ratio. However this does not make sense since only at critical coupling condition could max extinction be obtained and 1% coupling loss is over coupling this lossless ring.

I do feel this is problematic and please confirm if the waveguide loss unit is correct. Also I feel more comfortable with round trip loss like dB/round trip rather than dB/cm. Is there any way to define like that without converting roundtirp loss to dB/cm or dB/m?

2. For a ring modulator, most people choose to use single-bus ring rather than double-bus ring. However I saw most interconnect example used double-bus ring for modulation with only one bus waveguide connected to ONA. In that case does the frequency response give the spectrum same as single-bus ring (one coupling loss) or double-bus ring (two coupling loss)?

3. I see the coupling coefficient and waveguide loss as constant in the ring model. In reality they can be strongly wavelength dependent. Is there any way to specify wavelength dependent parameters (decomposing ring model with Mode waveguide can help but not sure how to define coupler properly)?

4. I saw many interconnect example specify non-equal neff and group index. It’s well known that this happens when where is dispersion but those examples all set dispersion to 0 as default. Do you think it’s necessary to find D and convert to proper unit to fill it or leave it as zero is also fine?

Best
Arthur Teng

[gwang]

Hi @mteng,

Regarding your questions:

In this example the waveguide loss is setup as 7dB/m, which is 4e-4 dB/round trip intrinsic loss for a 60um long ring.
However I do suspect the unit should be 7dB/cm while 7dB/m means the ring is almost lossless.
Then the power coupling ratio of 1% is used, which after interconnect simulation shows almost maximized extinction ratio. However this does not make sense since only at critical coupling condition could max extinction be obtained and 1% coupling loss is over coupling this lossless ring.

I do feel this is problematic and please confirm if the waveguide loss unit is correct. Also I feel more comfortable with round trip loss like dB/round trip rather than dB/cm. Is there any way to define like that without converting roundtirp loss to dB/cm or dB/m?

There are three options for the loss unit, namely 1/m, dB/km and dB/m, you can choose among them. This example file is just a toy model and you can key in the property values that fit your own design. I agree that the loss value and coupling ratio are not typically practical, please do use your own design values to test the model.

For a ring modulator, most people choose to use single-bus ring rather than double-bus ring. However I saw most interconnect example used double-bus ring for modulation with only one bus waveguide connected to ONA. In that case does the frequency response give the spectrum same as single-bus ring (one coupling loss) or double-bus ring (two coupling loss)?

The inner structure of the ring modulator is based on a double bus ring and the frequency response is for double bus ring.

I see the coupling coefficient and waveguide loss as constant in the ring model. In reality they can be strongly wavelength dependent. Is there any way to specify wavelength dependent parameters (decomposing ring model with Mode waveguide can help but not sure how to define coupler properly)?

This primitive model doesn’t consider the frequency dependency on the loss and coupling ratio, but it does take into consideration of the effective index and group index frequency dependency. To take this into consideration, you can build your own compact model of the ring modulator. The loss and dispersion dependency on frequency can be simulated in MODE Solutions and imported to the MODE Waveguide element. The coupling coefficient frequency dependency can be simulated in FDTD Solutions and its s-parameter can be imported to INTERCONNECT into the Optical N Port S Parameter element.

This is a ring resonator compact model in the Lumerical Compact Model Library that uses the similar building approach, please refer to the following post for the implementation details:

LCML - Ring resonator with thermal tuning | lcml_ring_tunable_1550

Hi @aya_zaki, Sorry that I overlooked your first question… The Fork and the Mix are just to convert the voltage to power and then the 0.02 gain is the resistance. We do this to match up the table we have in the modulator, which is neff changing versus power changing. I hope this could help. Let me know if you have further questions

I saw many interconnect example specify non-equal neff and group index. It’s well known that this happens when where is dispersion but those examples all set dispersion to 0 as default. Do you think it’s necessary to find D and convert to proper unit to fill it or leave it as zero is also fine?

It is not necessary to key in the dispersion, but it is better to include this information.

[mteng]

Dear Lumerical staff,

Thanks for your answer and I do have additional questions on ring modulators.
I have viewed a number of pages about ring modulators and Lumerical KB and I also build my onw ring modulator models with couplers and waveguide.

freq_domain.icp (480.4 KB)

neff_vs_voltage_negative.txt (75 Bytes)

1. With optical modulator component I put in, I need to specify its length and lookup table for neff changes vs. voltage. However I wish to know if the length of optical modulator will contribute to total optical pathlength of the ring resonator. My gut feeling is no since no optical property (neff,ng,disperion…) is specified and the length here is only used for calculating the accumulated index changes. Please correct me if my undertanding is wrong.

2. With ring modulator component in the model library, modulator is defined with double bus waveguide . but I notice we can specify coupling 1 and coupling 2 for that model. Then does it mean if I choose coupling 2 =0 then I can effectively simulate a single-bus ring resonator or modulator?

3. I also try to do time domain simulation and set CW laser wavelength located at 193Thz (resonance). However I see the modulated optical signal with strong overshoot and distortion and electrical signal out of LP filter cannot truly recover the electrical signal. I am not sure if any setup is wrong.

time_domain.icp (1.0 MB)

4. I would like to know the sampling setup for this model. If this modulator is operating at 5Gb/s, then should sampling rate should be orders of magnitude higher? In addition, for higher modulation speed (~ 50Gb/s), would even higher sampling rate required? Also for waveguide components, would like turn digital filter on with delay compensation?

5. Does the overshoot in time domain transient limit the modulation speed? I believe this dynamic effect is real in physics and what can I possibly do to reduce this overshoot? I also wish to know if Lumerical interconnect can generate a modulation depth vs. modulation speed plot.

Best
Arthur Teng

[mteng]

Dear Guanhui,

I am currently evaluating Interconnect and our group is going to
purchase the license in two weeks.
However I would like to know how can I request the PDK from a certain
foundry?

I think our group may choose to work with HHI, IMEC or smart photonics
and I would like to see some PDK model can be imported into interconnect
from those foundries. Are those PDKs free to request or we need to purchase?

Best

Arthur Teng

[gwang]

Hi @mteng,

The CML reader license need to be purchased and the PDK/CML need to be requested from the foundry. We recently worked with IMEC and developed a CML with them and you can download the library from IMEC’s website. You can also email sales@lumerical.com for licensing inquiries.

I am happy to discuss more with you if you have further questions on this.

[gwang]

Hi @mteng,

Regarding your questions:

With optical modulator component I put in, I need to specify its length and lookup table for neff changes vs. voltage. However I wish to know if the length of optical modulator will contribute to total optical pathlength of the ring resonator. My gut feeling is no since no optical property (neff,ng,disperion…) is specified and the length here is only used for calculating the accumulated index changes. Please correct me if my undertanding is wrong.

The length of the Optical Modulator is not calculated to be the optical path length since it doesn’t contribute to optical delay or so (the reason is like you said, it has no optical physical properties) however it represents the optical path length (or part of it if the active region is not the whole ring). The length of the Optical Modulator should be set to the same length as the ring’s active region so that it can correctly represent the effective index changing in the active region. You can think of that the Optical Modulator and the (part of the) Straight Waveguide are overlayed (is one piece of component) and represent the active region.

With ring modulator component in the model library, modulator is defined with double bus waveguide . but I notice we can specify coupling 1 and coupling 2 for that model. Then does it mean if I choose coupling 2 =0 then I can effectively simulate a single-bus ring resonator or modulator?

Yes, if one of the coupling coefficient is set to 0, then this ring will be single bus.

I would like to know the sampling setup for this model. If this modulator is operating at 5Gb/s, then should sampling rate should be orders of magnitude higher? In addition, for higher modulation speed (~ 50Gb/s), would even higher sampling rate required? Also for waveguide components, would like turn digital filter on with delay compensation?

The sampling rate should definitely be higher than the operating frequency. It is better to use higher sampling rate if you don’t enable the delay compensation in the waveguides (so that the delay in the waveguide will be integer numbers of samples). Please refer to this page for more information on delay compensation.

Does the overshoot in time domain transient limit the modulation speed? I believe this dynamic effect is real in physics and what can I possibly do to reduce this overshoot? I also wish to know if Lumerical interconnect can generate a modulation depth vs. modulation speed plot.

I modified your file a bit and now the received signal and the transmitted signal are match (opposite). There still is some overshoot effect in the received signals, and it represents the process to reach the steady stage in time domain. To reduce this effect, I think we can use some powerful low pass filters?

Here are the modified file:
time_domain_modified.icp (841.5 KB)
freq_domain_modified.icp (494.9 KB)

Could you please provide some more information about the modulation depth vs. modulation speed plot? We can discuss this in a separate post.

[mteng]

Dear Guanhui,

Your modified file seems to be complied by a newer version of
interconnect and I cannot open it.I am not sure what is happening since
I am also using the latest trail version.However, whenever I open your
file, interconnect get crashed. Can you briefly explain what changes you
made to the old file and where mistake is?

Yesterday your college Xu Wang arrived in MERL and we had a very
discussion on interconnect product. He actually mentioned that the
overshoot effect can be reduced by using ring with lower Q. My
understanding is that overshoot is like a oscillation with damping and
damping term may be round trip loss. However I am not sure what
determine the oscillation frequency and if there is any time constant
that dictates it.

Currently I am interested to learn how to design a ring modulator with
hypothetically high modulation speed (~50GhZ) with just optical
modulator element and imported lookup table. Then I would like study
more practical limitation (RC constant, carrier drift time…) and
non-ideality at such high modulation speed.

I also attached a figure of optical response vs. modulation response. I
suspect it means insertion loss will go up as modulation speed
increases but I may be wrong. I am interested in how to generate such
plot.Eventually I would like to know if remote meeting is offered by
Lumerical support team upon request where customers can directly talk to
one of Lumerical support staffs?

Best

Arthur Teng

[gwang]

Hi @mteng,

Great that you had met Xu and I believe you guys had a good conversation! I am using the Beta Release that published a few days ago, you can get it here (Select “Beta Release” from the “Select a version” drop down menu).

The figure plots the optical small signal response, which also highly depends on the modulation bandwidth (electrical and optical), you can also understand it as the insertion loss goes up along the increasing of the signal frequency. This plot cannot be directly generated by the analyzers in INTERCONNECT but we can use a script to get it.

I’d like to discuss more with you if you have further questions.