Some discussions about PyTFT

[xzou0718]

Hi Biswas,
The project I am working on is similar to PyTFT, and I have some questions that I would like to discuss with you after reading your project.
#1 According to my understanding, a good spice model should be designed for your manufacture process, which should be simulating different W/L devices well with same extracted parameters. However, your PyTFT seems only designed for single TFT ?
#2 During Usat calculation, saturation region of Ug~Id was used. Which is reasonable using average of Usat as U0? I think array of Usat should be fluctuates violently.
Thanks a lot
xzou

[ArpanBiswas99]

1: You are correct that PyTFT is currently designed for a single TFT device. The focus of this project is to create a basic model for one IGZO TFT using SPICE Model 3 to simulate its behaviour and estimate its parameters. This allows for straightforward validation and matching with real device data. Extending this model to handle arrays with varying W/L ratios or process variations is a logical next step.

2: The calculation of u0 in PyTFT uses the principle of field-effect mobility in the saturation region. This approach provides a good approximation of u0 , which is then used as an initial estimate in the SPICE Model 3. The process involves identifying the saturation region for each set of Id measurements, calculating the mobility based on channel width, length, and gate oxide capacitance, and then averaging the results across different constant drain voltages. This method provides a solid starting point for the SPICE model, and further refinement of u0 is done by fitting the simulated data to the measured characteristics. While this approach may not fully address fluctuations in the raw mobility data, it works well for building a basic and functional model.

[xzou0718]

Thanks for you reply. After some research and attempts , I found your method maybe more reasonable for exploring TFT's performance initially. It require great effort to make a spice model with the ability to cover all process or physical information. I am really looking forward to your next work. And I have another question about PyTFT. How Id model in above_threshold region connected with Id in sub-threshold region? Generally, Id_above and Id_sub are separately modelled, and the merged Id model has large differences in transition zone around Vth with Id_Vg curve measured from experiment. Looking forward to hearing from you. Best regards xzou Arpan Biswas ***@***.***> 于2025年1月6日周一 21:05写道：

…

1: You are correct that PyTFT is currently designed for a single TFT device. The focus of this project is to create a basic model for one IGZO TFT using SPICE Model 3 to simulate its behaviour and estimate its parameters. This allows for straightforward validation and matching with real device data. Extending this model to handle arrays with varying W/L ratios or process variations is a logical next step. 2: The calculation of u0 in PyTFT uses the principle of field-effect mobility in the saturation region. This approach provides a good approximation of u0 , which is then used as an initial estimate in the SPICE Model 3. The process involves identifying the saturation region for each set of Id measurements, calculating the mobility based on channel width, length, and gate oxide capacitance, and then averaging the results across different constant drain voltages. This method provides a solid starting point for the SPICE model, and further refinement of u0 is done by fitting the simulated data to the measured characteristics. While this approach may not fully address fluctuations in the raw mobility data, it works well for building a basic and functional model. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[ArpanBiswas99]

Hi @xzou0718,

Thank you for your insights and for exploring PyTFT! The sub-threshold region wasn’t modeled due to measurement limitations and because it was out of scope for the SPICE-based parameter extraction.

If needed in the future, a sub-threshold model could be added to improve accuracy near Vth.