Where is the right place to get in touch with a KiCAD Dev?

1 Like

So i take your link as ‘yes’ to

In case of ngspice the right contact person is @holger.

Thanks, so in that case now that you mentioned him his awareness most likely has been raised :sweat_smile:

Launchpad has been retired. If you would have clicked on my link then you would have discovered that it now points you to gitlab.

That was the idea :wink: however, be aware that it could take a while (you know, personal life, work out of love, volunteering and all that)

I did check your link and it takes me to their report issue thing which points to this:


for me that’s the same thing…but you’re still right…

On the KiCAD website you will have to follow this link:

It took me a bit also to see it.

Yes, thats what i did.
I pulled the above link out of my history so it points to the helpdesk section :sweat_smile: but if you select issue on the left sidebar you’ll end up in the exact spot you point to.

Hi.

Not a bug. And almost certainly not in KiCad.

The primary issue is probably that your opamp models are not equivalent. Picking the generic items for both does not ensure that they are the same.

1 Like

Thank you very much for your reply. In the NGspice forum was already confirmed that it has to be the opAmp’s Model.

Just to make it absolutely clear, to prevent misunderstandings: The symbol and spice model that was used is the default spice model that does come with the KiCAD libraries,Its reference is: Simulation_SPICE:OPAMP . To be traceable it did setup a clean Fedora 31 VM with freshly installed KiCAD 5.1.5-1.fc31 , did no changes to it and did the setup with only default supplied components.

Over here @marcel_hendrix pointed out that:

If there is no opamp or B-source, the simulation will produce the characteristic of a R-R-C phaseshift network.

It would have been nice that NG or KiCAD raised a warning telling me that it found no OpAmp Model for simulation before replacing it with something.
So i would say it is arguable if its a bug or not but i am also not yet sure if in KiCAD or in NGspice.

Correct. This is not the same as the default SPICE model that comes with LTSPICE. Nor should you ever expect it to be.

It didn’t. You requested the generic OPAMP model and it gave you the generic OPAMP model. That is not the same as the LTSPICE generic model.

1 Like

To make matters worse: the KiCad setups in the simulation lib are not meant to represent any particular part. They are just there as an example how to fill out the parameters without getting a critical error. (we kind of argued if we should fill it with something illegal to ensure users know that the need to fill the info with their parts specifications but opted against it because we did not want to deal with the bug reports we would get that way)


@Evan_Shultz can probably give more insight into this as he was the one who made most of the simulation symbols.

2 Likes

I did not expected it so i tested it. But i do not see how i could have tested better to prevent me from assuming it is fine. If i put the amp in the schematic and look at it’s values there is no reference to any simulation model whatsoever neither in the ‘values’ nor in the ‘edit spice model’ is listed a single parameter nor a pointer to where i can find them. Even now i yet dont know that.

If that is the intended behaviour thats fine with me. You guys have to decide how your software interacts with the users. I just gave my opinion.

That is also totally fine with me if you guys decide that’s the way it should be. But again the amp has no value in its mask whatsoever from which i can assume a phaseshift network.
And it can’t be that easy to find either since it took me three topics to get to know what it does.

Ok sorry i think i confused the symbols for ngspice sources with the opamp. Yes the opamp is not assigned a model. I am not sure why there is a special opamp symbol in the simulation lib then.

Typically there are no simulation models available in KiCad symbol libraries. So the user has to take care to find and attach a suitable model.

The symbol Simulation_SPICE:OPAMP however has a bug. The entry for Spice_Primitive is V in the current version. This is wrong because an OpAmp never will be just a voltage source (aka V).

If the entry was X, that is an OpAmp subcircuit model is required (as is usually the case with OpAmps), and the user did not attach such a ngspice model, the simulator would have answered:

Circuit: KiCad schematic
Error: unknown subckt: xu1 nc_01 nc_02 nc_03 nc_04 nc_05 opamp
Error: there aren't any circuits loaded.

or simular, just bailing out and telling what is missing.

With Spice_Primitive V, ngspice is told to add a voltage source, which it does, and it moves on (with difficult to understand error messages because of the strange voltage source parameters, so setting them to 0) and thus yielding strange simulation results.

2 Likes

So it is a bug after all.

Sounds good to me and would have been all i needed to notify that there is something wrong.

Is this something i should raise an issue for or is it already covered?

I created https://github.com/KiCad/kicad-symbols/pull/2459 to fix this issue.

The OPAMP symbol was submitted by me, and we were in a rush to submit the symbol because the existing opamps symbols typically have their power pins in a separate unit and don’t simulate. I had always picked a SPICE model and didn’t come across this before. Thanks for catching this issue.

The DIODE symbol was a copy and paste of the existing diode symbol and didn’t have a proper SPICE primitive assigned so I fixed that too.

4 Likes

Thanks for your effort and letting me know.

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.