LTSpice model of a pure sine SPWM inverter

Thread Starter


Joined May 23, 2011
I am currently working on a LTSpice fully parametric model of an inverter, using SPWM bipolar modulation.
It works well and i am testing it on a variety of load levels and types.

The goal is not to make a cicuit out of it because ASIC boards such as EGS002 or EGS005 already do the work well.
However it seems a good way to teach myself and others all the intricacies of inverter design.

I used the EGS002 application schematics as a starting point to create this model.

For now it features also dead time insertion and an output LC filter.
and the first stages of voltage feedback for a precise voltage control, the loop is not yet complete.
It seems that most designs use peak ac value detection instead of using an averaged value coming from a rectifier stage.

I've gone for an averaged value (rectifier+voltage divider+filter) method with a differential output to a Howland current pump to transfer the signal to an optocoupler. Now I think that I should have put the voltage divider before the rectifier to use lower voltage diodes, but I digress...

I don't know if that will work well enough for a control loop response that is not too slugish or should i implement peak detection instead ?

I also had to reference the Opamp grounds in the howland current pump to the 'neutral' of the output stage instead of DC GND. It seems that bipolar SPWM has a varying offset voltage with reference to GND, and voltage feedback needs to use a differential method. That is why i also used an optocoupler, since the logic part is referenced to DC GND.

Anyway, if you see any possible improvement, please comment.
I am sorry for the horrible routing in the schematic, I am currently working on fixing things.

The design does not have convergence problem with "uic startup" options and is faster with the ALTERNATE solver for MOSFETs and IR2110 compatibility.

The design and more information is also published on my website, and will be updated.