Simulation (grumble grumble)

[throbscottle]

Having been slowly getting back into electronics over the last few months, one of the big changes I've discovered is the possibility of simulating circuits before building them.

Well, I've been trying to use some free simulators for Linux for a while now, and it seems such a lot of effort to get any kind of sensible result at all. Most luck I had was with Qucs. I like the Geda suite for drawing, but completely lost with simulation on that. I was hoping to be able to just cobble something together to see if it works, only virtually, and see what's going on with virtual instruments vastly better than my very limited real ones. Well, I'm about ready to give up on simulations now and just use the old fashioned way of creating a scary ball of components with wires sticking out of it - it's quicker and easier! Time to invest in a breadboard...

throbscottle

 

[Sceadwian]

What about LTSpice? It's interface has been kept deliberately simple to allow support for an Linux machine that supports WINE.

 

[throbscottle]

Oooohhh! Okay, that's more like it Thanks!

 

[Sceadwian]

You'll find a lot of help on the Yahoo LTSpice group, and it's the free simulator of choice outside of students.
It's components library is constantly updated to support linears product line, yet it's general usability makes it bar none the choice for most people.

 

[Mikebits]

I second that!

 

[strantor]

I use LTspice when I simulate. I don't simulate much, only when I want to see how a part works that I don't have. Most of the stuff I do involves things that aren't in the program, like relays for example. It's a royal pain to add new components, I tried once. Either that, or when I simulate something and I ask about it on the forum, someone says "it doesn't work like that in real life." So I don't get much added value out of simulators. Others do though.

 

[Sceadwian]

The learning curve of using Spice in general is quiet high strantor. It's very easy to do some simple things with programs like LTSpice, which try to put some sane default's to help mimic the real world a but if you want to simulate some more complicated (and I mean physics complicated not component complicated) it get's hard fast.

Keep in mind, all the hardware you're using right now was designed on a simulator front to back before the first silicon was ever etched. Chip makers models actually simulate the cmos structures that are made and then they sick the computers on it to make it all work.

 

[kubeek]

The learning curve of using Spice in general is quiet high strantor.

That is why I like Microcap, it has a spice simulator under the hood but it has much nicer UI so you don´t need to do stuff like defining a simulation by writing a chunk of text into your schematic.

 

[Winterstone]

That is why I like Microcap, it has a spice simulator under the hood but it has much nicer UI so you don´t need to do stuff like defining a simulation by writing a chunk of text into your schematic.

Yes, I agree - however, I like PSpice as well. It`s very handy and there is no unnecessary stuff (e.g. virtual instruments in Multisim).

 

[duffy]

Even the best simulators start giving you nonsense after more than a handful of analog circuitry.

 

[crutschow]

Even the best simulators start giving you nonsense after more than a handful of analog circuitry.

Only if the analog circuitry is nonsense.

 

[Sceadwian]

There is only one limitation to a simulator, the amount of effort spent inputting the boundaries of the simulation as they occur in the real world. It's all about the parasitic components that the user does not input.

Real world CMOS level simulations are quiet accurate, they have to be to test out new designs, and any other parasitic issues previously unknown are worked out by empirical measurement from first silicon.

 

[Winterstone]

There is only one limitation to a simulator, the amount of effort spent inputting the boundaries of the simulation as they occur in the real world. It's all about the parasitic components that the user does not input.

Yes, I agree - and I have a good example:
Take the model for an opamp (ideal VCVS or real model) with positive resistive feedback.
The operatiing point as well as an ac magnitude analysis will indicate stability with gain as expected.
Only the phase response vs. frequency exhibits an unexpected behaviour. A TRAN analysis will reveal the error (dc instability).

Question: Is this a simulation error? Who is guilty? I think, the simulation engine was correct. Without any environmental disturbing influence, without noise and without any power switch-on effect there would be a stable operating point with gain.
This can even be verified by hand calculation.
Fazit: In most (if not in all) cases the user has made an error if some unexpexted (or even false) results appear.

 

[duffy]

Only if the analog circuitry is nonsense.

No, very straightforward designs start showing significant differences with empirical results after you add more than a couple of dozen analog components. Changing the update speed and adding digits to the precision settings doesn't improve it beyond a certain point - and that point is in the dozens, not hundreds, of analog components. If you had any experience with it, you would be painfully familiar with the problem.

 

[Sceadwian]

Again duffy, this isn't a fault of the simulator but of the person that creates the simulation, in order to get true to life simulation you'd have to add every parasitic that exists in a real world circuit, and there are a massive number of them and possible sources, everything from atomic events to a failure to simulate shotnoise/johnson noise etc.. etc..

 

[duffy]

There is only one limitation to a simulator, the amount of effort spent inputting the boundaries of the simulation as they occur in the real world. It's all about the parasitic components that the user does not input.

Again duffy, this isn't a fault of the simulator but of the person that creates the simulation, in order to get true to life simulation you'd have to add every parasitic that exists in a real world circuit

No, there's also "simulation speed" and "precision settings" (among others) that have a greater effect. Anyone who actually works with them on a regular basis, using more than a handful of components, becomes very familiar with the inherent weaknesses of simulators.

 

[Roff]

No, there's also "simulation speed" and "precision settings" (among others) that have a greater effect. Anyone who actually works with them on a regular basis, using more than a handful of components, becomes very familiar with the inherent weaknesses of simulators.

Can you post a .ASC file that illustrates your point?

 

[crutschow]

No, very straightforward designs start showing significant differences with empirical results after you add more than a couple of dozen analog components. Changing the update speed and adding digits to the precision settings doesn't improve it beyond a certain point - and that point is in the dozens, not hundreds, of analog components. If you had any experience with it, you would be painfully familiar with the problem.

I've had considerable experience with Spice simulators and have used one to simulate a full buck switching regulator, including the loop response, with good success, for example. What are these "significant differences" and "inherent weaknesses" you are referring to that you are so painfully familiar with?

 

[ericgibbs]

hi duffy,

If you are aware of any short comings in LTSpice, it would be helpful if you could post examples.

E.

 

[duffy]

I can dig something up, but you would have to actually build a circuit with - like I said - more than a couple dozen analog components and compare signals between the two. After a certain level of complexity the simulator waveforms just don't match what's on the scope.

crutschow - The analog part of your "buck regulator" reduces to what - a coil, a diode, a drive transistor, couple of caps and a voltage comparator? I've had no problem simulating coil drivers very much like that, and from an analog standpoint, also very simple.