Slow LTSpice model using LTC3115-1

[Rusttree]

I have a simple LTSpice model that is mostly based on the circuit in the datasheet of the LTC3115-1. It's running painfully slow (about 50uS/S). If any LTSpice gurus can show me how to improve the simulation speed, I would be very appreciative. The .asc file is attached.

Thanks!

 

[WTP Pepper]

When you use a SPICE model from the likes of LT based on their design, it models absolutely everything. It has been derived by the engineers who designed the part and used in their extensive testing before being released to the markets. Before the design goes to the chip foundry it has to be spot on design wise. Otherwise it would not be a SPICE model and could cost them a fortune if they got it wrong.
This is a relatively complex part so the simulation can take time. Not like a simple transistor with a transfer function depending on input and output parameters. It has feedback and transient properties to be taken into account.

I recently did an LT SPICE simulation using several of their TIMERBLOX ICs. What appeared to be a simple 'logical' design took over three hours to run the simulation. However, when I finally built the electronics I had designed, it worked first time to as far as I could tell with absolute precision. Just like the simulation had shown. Voltage levels, pulse widths and frequencies to better resolutions what I could measure.

That's the price you pay and it would have been a damn more expensive design had I had to redo the PCB with different build standards regarding components and different topologies.

LT SPICE is probably one of the best simulation packages I have come across (without layout type of restrictions) and don't forget...it's free.

 

[Rusttree]

Understood. I've used Linear's models for a few previous designs and never saw it go this slow, so I was hoping it was something in my external parts selection. But I guess the LTC3115-1 just has more complex guts than previous regulators I've used. No matter, I'll just be more patient. Thanks!

 

[ronsimpson]

Your run time is 16mS.
The IC is switching at 300nS or about 3mhz.
The part turns on/off 50 000 times.
I can't find the real number but I think the software 'loops' about 100 times/cycle.
So the software must do its study of the circuit 5,000,000 times.
Huge files are created of all voltages and currents every xnS.
Your computer is working very hard.

 

[simonbramble]

Please reload the file. I don't seem to be able to save the file above (maybe a result of the server crash). My jig file in LTSpice works OK