Closed tomhajjar closed 1 week ago
First issue was the 4-pin Blue BJTs were removed from Qucs-S.
It is not removed. It is hidded for Ngspice by mistake. I will revert this device in the next patch.
Can you spot why the seemingly identical files have different Rout? I didn't see anything in a text editor
It is not a bug. The result form Qucsator and SPICE may be different. You have ideal BJT model. It acts as the ideal current source and you have dI=0
in the R=dV/dI
equation. The dI=0
may give an unstable result and random noise when approaches to machine epsilon. The SPICE model may act differently and give different result. Set the VAF
parameter of the BJT or use the real SPICE model instead of ideal and get the different result. The same is valid for the initial part of the Rout
curve. The BJT again acts as the ideal source.
I have to redo the Nutmeg equations to see what Rout will be under the latest Qucs-S/ngspice.
It is again not a bug. You have to specify the simulation name instead of dc
in the Nutmeg equation. The DC
is a valid simulation name for DC OP analysis. See #463, #513, and #124. You have to specify the SW1
to attache the equation to DC OP. After making these corrections everything works as expected.
The PR #983 makes BJTsub device available.
Updated ngspice and qucs versions of the project. Might be useful for someone.
The SPICE behavior depends on GMIN setting. The default GMIN=1e-12
works fine, but random noise starts from GMIN=1e-14
. The GMIN=1e-16
fails to converge. The Qucsator have another behavior. I didn't find GMIN setting for Qucsator DC.
I ran into an issue testing an old Qucs-S project that was previously converted from Qucs. First issue was the 4-pin Blue BJTs were removed from Qucs-S. I then converted to 3-pin Blue but then the "old" Nutmeg equations didn't work because of other changes to Qucs-S. It wasn't obvious what changes were needed to the Nutmeg equations so I went back to Qucs and the original project.
This is when I found I couldn't get "Rout" for "mirrorwilson.sch" in different project folders to be the same. I renamed each file and placed them in the same folder "Mirrors_qucs_prj". I confirmed the Rout data for each version of "mirrorwilson.sch" was different when tested in either Qucs or Qucs-S so qucsator/qucsator-rf wasn't the issue.
1) Can you spot why the seemingly identical files have different Rout? I didn't see anything in a text editor. 2) My "names" for the various files "correct", "Wrong1/2" may be wrong. ;-) 3) I have to redo the Nutmeg equations to see what Rout will be under the latest Qucs-S/ngspice. I found a PC running Qucs-S 0.0.24 and ngspice 39+ and the Rout data is different but maybe correct?
TODO list