Difference results from optical power monitor and optical network analyzer in Interconnect?

needattention

#1

Hi,
I meet a weird problem when using Interconnect. The .icp file is attached.
test.icp (373.3 KB)
Smatrix_data.dat (3.6 KB)

There is a S-matrix element. Its data is calculated by FDTD. I use two configurations to get the transmission of this element. One is optical power meter, another one is by optical network analyzer.

The problem is the results for two configurations are different. For power meter, the output power is -3.10dB, while for ONA, its transmission is about 0.497524 (-3.03dB). I have no idea where the difference comes. Could anyone help me find it ?


#2

Hi @lishifeng,

Sorry for the delayed reply! I believe the difference you noted is due to the fact these 2 configurations model the element in completely different ways.

When using the ONA, you are running a S-parameter simulation, while in the other case, we’re running a transient simulation. This might not be obvious, as we have a CW laser as a source, and no time varying signal in the circuit. The consequence is, SPAR_1 element is described by a FIR filter, and SPAR_2 is described by its S matrix.

Although the FIR filter is calculated from the S-matrix, there can be some discrepancies. In this specific case, modifying the “Digital filter” settings can improve the results. For example, if I set “initialize filter taps” to true, the output power gets to a much closer value.

I hope this will help!


#3

Thank you very much for your reply. I have some further questions.

What is purpose of setting “initialize filter taps” as true ? Let the S-parameter elements do not work like a filter, or something else?

And in what situation do we need to change this setting?


#4

You’re most welcome! :slight_smile:

When setting this parameter to true, the initial input signal will be used to initialize filter state values. The main difficulty is, the S parameter is defined in the frequency domain. In the time domain, INTERCONNECT will use it to determine the impulse response of the element, and a digital filter will be used to fit this impulse response.

There are various settings that will affect the fit used. Typically, the optimum settings depend on the application and what is most critical.
For example, in your case, you can reduce the error by increasing the sample rate, but it increases the number of samples and, therefore, the simulation time.
You can turn on diagnostic for the element (set “run diagnostic” to true). This will give the ideal transmission/gain and the error due to the filter.


#5

Ok, I will try. Thanks a lot