- This topic has 10 replies, 4 voices, and was last updated 1 year ago by
.
-
Posts
-
For your information, QucsStudio is not only a very good free simulator, it is also a very good tool for modelling devices (extraction of model parameters, model analysis and validation, …). For more information, see the documents presented at the last 2023 HICUM workshop.
Apparently, named node variables can not be used in equations. E.g. U=1-VBE.V-VBE.V and similar throws error in the new revisions therefore we are restricted to use v3.3.2. Originally, the same issue existed in v3.3.2 as well until we were sent an updated EXE file (to D. Celi) by Michael. Hopefully, the issue would be fixed from the next release on.
This is a v4.3.1 simulation trial with the QucsStudio implemented hicumL2V2p33 (same result with v4.4.2). There are 3 filter boxes on the schematic but neither works. Eqn. filter3 runs fine with the updated v3.3.2 EXE file of Michael.
Attachments:
Tomhajjar
It is not so easy for several reasons:
1- I’m using QucsStudio v3.3.2 with a patch sent by Michael M. to fix a bug (see ZHU’s comments). Unfortunately, this fix is not available in version 4.3.1. I hope it will be available in the next version of QucsStudio.
2- To better automatize the extraction steps, I use additional OCTAVE routines developed by ZHU. I am not owner of these routines.
3- The data used in the papers are company confidential.
My proposal is to post, as soon as I have time (over the Christmas period), a simple extraction example, based on the SPICE Gummel-Poon model, using non-confidential data.
In the meantime, for more information, please consult
This is the linear regression package for using in QucsStudio
https://www.iee.et.tu-dresden.de/iee/eb/forsch/AK-Bipo/2019/1-AMS_zhu1_BipAK19.pdf
– Zoltan
Thank you so much Zoltan for sharing your linear regression routine. This is a key point for extracting model parameters directly (when it is possible) without using an optimizer.
Didier
The QucsStudio Owner’s optimization tutorial extended by the alternative linear regression option.
– solution can be obtained in one-shot w/o iterations
– pointwise error reduction by one order of magnitude
– easy configuration
Full or partial application of linear regression i/o optimization is strongly advised whenever it is feasible.
Zoltan
Attachments:
- You must be logged in to reply to this topic.