Background
When measuring s-parameters of a passive element for statistical modeling, it is often difficult to tie the variations in the s-parameters to some physical parameter, e.g., waveguide width or thickness. With the new feature added to the sparsweep_pcell photonic model in the 2021 R1.4 release, users can now use their raw measured s-parameter data directly to create statistically enabled compact models for their passive elements. When running a statistical simulation, the model will randomly select a different set of measured s-parameter data in every trial. For an example model check the Grating Coupler (Statistical Using Measured S-Parameter Data) element in lumfoundry_template.
Workflow
- Collect as many s-parameter measurements as possible for your device
- Label each measured element (sample) using two numbers. One number labelling the GLOBAL sample, i.e., the wafer number or lot number or die number. A second number labelling the element itself within the GLOBAL samples. The numbering for both labels will be integers and will start from 1. If all measurements are from the same GLOBAL sample, then the labelling for the GLOBAL samples will be a single number, 1.
For example, if the measurement was done over 10 different wafers, each containing 100 elements, then the GLOBAL samples will vary from 1 to 10 and the LOCAL samples will vary from 1 to 100.
[[Note:]] The model only supports the same number of LOCAL samples for each GLOBAL sample.
- Using the GLOBAL and LOCAL sample numbers as parameters, combine the measured s-parameters into a single text file format of which you can find in the S-Parameter file formats page under the optical N-port s-parameter sweep option.
[[Note:]] When creating the parameterized s-parameter file, the GLOBAL sample/parameter should always be the first parameter.
- The source .json file is identical to any other statistical passive model that uses the sparsweep_pcell photonic model with one key difference. Instead of defining the ‘stat_parameters’ variable, define the newly added ‘stat_measurement_data’ variable (see following section).
measurement_data
The ‘measurment_data’ variable is used when creating a statistically enabled s-parameter element using raw measurement data with the sparsweep_pcell photonic model. It contains information about the parameterization of the raw data in terms of the GLOBAL and LOCAL sample numbers. The variable is defined as a struct with the following fields in the “statistical” section:
GLOBAL: Defined as a struct with fields
- N_sample: Total number of GLOBAL samples.
- nominal_sample: The GLOBAL sample number for the nominal s-parameter data.
- param_name: The name of this GLOBAL statistical parameter that will get added to the compact model. The name must match with the first parameter name in the s-parameter data text file.
LOCAL: Defined as a struct with fields
- N_sample: Total number of LOCAL samples per GLOBAL sample.
- nominal_sample: The LOCAL sample number for the nominal s-parameter data.
- param_name: The name of this LOCAL statistical parameter that will get added to the compact model. The name must match with the second parameter name in the s-parameter data text file.