Simulating simple electric circuits

[Bjoern Schliessmann]

Dave Baum wrote:
> Are you trying to do logic simulation (digital) or analog circuit
> simulation?

Mh, a mix of both :) I want to simulate (in principle simple)
magnetic relay circuits. The only "evil tricks" that are used are 

- shortcuts (e. g. a relay coil is bypassed and thus the relay gets
  no voltage and it relaxes)
- low currents (so a relay has current but doesn't switch)

> I don't want to go into too much more detail about this one
> because I have a hunch you are really looking at digital, but if
> you're interested in the analog approach let me know and I'll fill
> in more of the details.

Good hunch :) But thanks for the explanations. The fully analog
network approach is definitely overkill here.
 
> For digital:
> 
> Event based simulation is typical here.  These simulations
> generally are concerned with voltage, not current.  A circuit
> consists of signals and devices.  At any given time, each signal
> has a certain state (high/low, on/off, 9 level logic, whatever). 
> Devices are connected to one another by signals.  You'll also need
> events (a signal, a time, and a new state), and an event queue
> (events sorted by time).  This is easier if each signal is driven
> by at most one device: 
>  
> 1) pop the next event off the queue
> 2) if the event's signal's state is the same as the new state, go
> to 1 
> 3) set the event's signal's state to the new state 
> 4) for each device that is attached to the signal, run the
> device's code, which should look at all of its inputs, and post
> new events to the queue for any outputs (do this even if the
> computed output is the same as the current output).  These events
> are usually posted for some time in the future (1
> simulation 'tick' is fine).  
> 5) go to 1
> 
> This approach is pretty simple to do in Python.  I wrote a sample
> digital simulator a while back and the core of the simulator was
> around 50 lines of code.  Rounded out with some basic logic gates
> and helper functions to probe the simulation, it was around 150
> lines.  It was only 2 level logic and signals could only be driven
> by a single device.
> 
> The devices that you want to model (switches, loads, etc) don't
> have explicit inputs and outputs, and you'll need to deal with a
> signal being driven from multiple sources, so it will get a bit
> more complicated. You will probably also need 9 level logic (or
> something equivalent) to deal with the fact that ground beats a
> source through a load when determining a node's state.
> 
> The basic idea is that each signal has multiple drivers, each of
> which has an internal state.  When a device wants to set an
> output, it only changes its driver's state.  The signal then has
> code that looks at the state of all drivers and combines them in
> some way (this is called a resolution function).  That combined
> state is what devices see when they read the signal.
> 
> It isn't *that* complicated to implement, but if you can turn your
> problem into one with 2 level logic and no multiple drivers, then
> it will be easier to write and debug.

Sounds more familiar than the analog approach. Maybe I misunderstood
something ... but I can't transfer my problem to this way of
thinking yet. My biggest problem is the fact that relays aren't
really interested in voltage, but current. 

Also, I find it difficult to transfer this circuit logic to boolean
logic I can contruct logic gates from. Sometimes, electric circuits
are used in different directions.

This is how far I've come with my first approach yet:

I set up the mentioned "controller" which, at the beginning, tries
out all possible ways through the network and saves them. So, for
every possible circuit it knows which switches must be closed and
which relays will work if it's "on". In theory, it should now be
possible to try out every path, tell the relays if they have
voltage/current, and let the relays report back in to the
controller if their status changes so it can again test the
circuits that may have changed. I haven't tried out the last step,
but I will in the next days. Is there any logic error in my
strategy?

Regards,


Björn

-- 
BOFH excuse #122:

because Bill Gates is a Jehovah's witness and so nothing can work on
St. Swithin's day.