Would you like to react to this message? Create an account in a few clicks or log in to continue.

You are not connected. Please login or register

Non Linear Components.

2 posters

Go down  Message [Page 1 of 1]

1Non Linear Components. Empty Non Linear Components. Sat Dec 19, 2020 4:21 pm



This simulator aims to be  as "real time" and interactive as possible.
For that, simulation speed is essential, and non linear components, specially transistors are a big chalenge.

Making fast models means that they will fail to converge in some cases.
Making models that converge in most cases means they must be slow.

I'm working right now to solve this issues. or at least to make it work in as many cases as possible.

I find that in some cases accuracy is essential but not speed, for example in this case (Constant Current Circuit):
In this case, the pair of BJTs can converge setting the accuracy a bit higher that default and letting them iterate as much as they want (>50K iterations).
If you don't let them iterate enought it just wont work at all (the "solution space" is very narrow).

In other cases accuracy is not so important and speed is very important, for example in this case (MCU controlling LPO):
In this case user interaction with the circuit and keep the MCU running is esential.
And after a few hundred iterations the LPO has aproximated a solution that can be good enought for the purposes of the circuit.

There are other cases, like transistor as switches that should be ultra-fast.
This would be possible if we know that there are not feedbacks or emitter-follower (or similar) configurations.

The current model can detect emitter not tied to fixed voltages and do smoother steps.
There is also the "non-linear accuracy" property to set accuracy for the circuit.

But to solve as many cases as possible the only option i can think is that each transistor model is  configured for the specific task.
My idea now is to add some properties to non linear components, so the user can configure it for each case, including accuracy, maximum number of iterations, perhaps a "fast mode" for transistor as switches, and may be some other.

I would like to know what do you think about this and what other ideas come to your mind.

2Non Linear Components. Empty Re: Non Linear Components. Sun Dec 20, 2020 5:02 pm



I think this would be a good idea. On the one hand, it gets more complicated to tickle the simulation in the correct way. On the other hand, this enables one at least to change something to find a solution.

Aside from BJT and FETs, the op amp might also get some parameters in order to manipulate the convergence.

arcachofo likes this post


Back to top  Message [Page 1 of 1]

Permissions in this forum:
You cannot reply to topics in this forum