LT Spice

[kc3ase]

Does anyone know if LT Spice's models are assuming the worst case specifications for their products, the typical case or the best case?

[tggzzz]

Does anyone know if LT Spice's models are assuming the worst case specifications for their products, the typical case or the best case?

That entirely depends on who wrote the model and what their objectives were. All models are simplifications; many spice models (LT Spice included), are too simple - they just indicate what the device manufacturer wanted you to simulate.

[SaabFAN]

+1 to what 'tggzzz' said.

However, I did notice that LT does a pretty good job on it's own parts [with a few exceptions], but I still think that they put in the data-sheet 'typical' values.

The very best computer and program than you can find to simulate your analog designs is in between your ears.  Bob Pease was once asked "What is your favorite analog simulation program?", and he replied: "Solder!"...

Nothing can be simulated 100%.  If you did this, you would end up spending more time on building and testing your simulation models than it would take to physically build and test multiple versions of your design.

I use LTspice to do some quick tests on my ideas, and to play around with different circuit configurations to see how they behave.  With a simulator, if it tells you that something is *not* going to work, then 99.99% of the time, your design [in the real world] will not work.  BUT-- if it tells you that your design *will* work, well maybe it will, and maybe it won't-- [i.e. you can get a false positive].  At that point you still have to build and test the real thing to be certain that it will work in real life.

In analog electronics, the simulators are not the "be all, end all" tools that many people have been led to believe.  Sure, some of the simple components [some passives, some simple semiconductor devices] are modeled fairly well for most use-cases, but when you get into more complex devices the models are usually not modeled very well.  This is in complete contrast to digital simulation tools, where the device physics are usually well known and are modeled very well.  You would not be wise to ship out an FPGA design unless you have simulated it, and tested the design in the simulator.  In this use-case, simulators are mandatory and work very well.

True dat!
Designed a circuit that worked beautifully in LTSpice, but oscillated like crazy when built on the actual boards. Blew 30€ on that.
So don't believe everything the computer tells you.

[tree]

I spoke with some guys from LT how they design their models for their switching converters and the answer: it varies. They try to capture the behavior as closely as possible. If I recall correctly they use the Ridley model for their converters which captures most of the behavior pretty accurately.

[Bassman59]

Designed a circuit that worked beautifully in LTSpice, but oscillated like crazy when built on the actual boards. Blew 30€ on that.
So don't believe everything the computer tells you.

In fairness to LT, if we are talking about using LTSpice to simulate their DC-DC converters, I've used it to simulate several power-supply designs based on their parts, and the designs worked correctly on the bench. Of course, you really have to pay attention to the dire warnings about layout on the data sheet. You can't just plop parts down on the board in any old fashion and expect it to work. One can extend this caveat to all circuits. The board must be considered a component.

And as we well know, a simulation is only as good as the models used, and finding accurate models of inductors and transformers is not easy.