Automated Tool for Modeling Impedance for Spice (ATMIS)

Description

The ATMIS is a tool designed to automatedly generate an LTSpice model, given the frequency dependent impedance data of a measurement.

Data needs to be provided to the software in Touchstone (.sNp) file format. The tool will then calculate the impedance data of the device under test (DUT) and proceed with a series of least squares optimization routines to build a suitable LTSpice model.

The models obtained consist of serially chained parallel resonance circuits for inductors or parallel series resonance circuits for capacitors.

The tool is specialized in fitting passive components with DC bias and generating LTSpice models that are current or voltage dependent.

System Requirements

• Python version 3.10 or newer.
• Around 2 Gb of available Memory
• A good CPU (any CPU will work fine, but the calculation time can become quite high)

If you find yourself running into memory issues, you can adjust the MULTIPROCESSING_COUNT variable in the config.py file. This determines how many processes will be started for fitting; less processes will go easier on RAM usage at the cost of time.

Usage

Start-up

Run ATMIS via the main.py file.

When the tool is run, the user will see a GUI, consisting of the Specification area, a file list that will be filled and is empty at first execution and a log window which is also empty upon first execution.

To fit a component, one must first select the type of component in the dropdown menu in the left upper corner (NOTE: currently only capacitors and inductors are supported). Furthermore, the calculation mode for impedance has to be selected in the mid-right area spelling "Z Calculation Method". Select "Series Through" or "Shunt Through" depending on how the DUT was attached to the measurement setup.

Loading Files

After that click on the "Select s2p File(s)" button to load the files for the DUT. The files will be loaded and displayed in the file list pane on the right, together with a radiobutton and a small entry box.

Getting Things Ready

To tell the tool the DC Bias for each file, write the DC Bias values into the entry boxes right of the filename in the file list pane. The units for the DC Bias are Amperes or Volts for inductors and capacitors respectively (NOTE: engineering or scientific notation is not yet supported).

Also, a reference file needs to be selected, this can be done by clicking the radiobutton left to the file which is the reference file. The reference file is the one, which does not have any DC Bias.

The nominal value of the component can be provided in the specification area on the top left of the GUI in Henry or Farads for inductors or capacitors respectively. However, the tool has a built-in calculation for the nominal inductance/capacitance which works quite well.

The series resistance of the component can also be input at the specification area. The tool has a calculation method for that as well though. NOTE: It is highly recommended to provide the series resistance for any inductor you might want to fit, since the series resistance calculation does not work well for inductors. It does, however, work fairly well for capacitors.

The prominence for the fit can also be adjusted. This parameter needs to be input in dB and determines the prominence a peak has to have in order to be fit as a resonance circuit.

Further notes

Config File

Certain parameters can be adjusted in the config.py file: - FREQUPPERLIMIT: The highest frequency to which the fitter will handle the data. - CAPTYPE: if you want to fit the acoustic resonance of an MLCC, set this constant to constants.MLCC, otherwise set to constants.GENERIC. - FULLFIT: if you want the model to be fully parametric, i.e. all resonance circuits to be DC bias dependent (as opposed to only the main resonance parameters), set to 1. - MULTIPROCESSINGCOUNT: determines how many processes will be started for the fit.

Known issues

• The GUI does not have responsive design, so depending on your screen resolution, things might look a bit weird.
• Some coils/capacitors might not be detected by the tool, that is because the inductive/capacitive range is determined by the phase of the dataset. If you desperately need to fit a part that runs into this issue, you can go to the constants.py file and set the PERMITTEDMINPHASE to a lower value. However, do not expect perfect results from that workaround.