Do the "G"s connected between Vdd and ground simulate its current consumption?

Yes.
'Typical' values from the datasheet are used in the model. Any variation of these with supply voltage are not modelled.

 

Many thanks, all of you.
@eetech00 Could you enlighten me on UIC (Use Initial Condition) and why I should use Gear for integration? (New to me - I looked them up on the wiki)

Different Integration methods handle convergence issues and circuit malfunctions differently. Many research papers have been written on this subject and can explain them better than I can. Please continue your google search.

The .UIC tell the simulator to skip finding the operating point. There are probably other circuit conditions that are not accounted for in the model design. I didn't analyze what is causing or why its happening. Like I stated earlier, Microchip models are known to cause problems with LTspice and i don't wish to spend alot of time on this, since I don't work for Microchip. However, none of the mods I made will affect the behavior of the original model.

When deciding on a model, one thing you should ask yourself is "what do I want from the model?". For example, do I need it to simulate rise/fall times accurately? Do I need to simulate power supply effects? This will help decide how accurate a model you need. Sometimes, a real part is more favorable than attempting to model the required behavior.

 

To that point, I would not have believed it, until I saw it, but you can model an entire buck converter with dependent voltage sources, dependent current sources, transfer functions, and no actual components. It is true that the values for the R's and the L's and the C's are embedded in the coefficients of the transfer functions., but gone are the diode and the MOSFET.

Simulations run blisteringly fast to boot with not even a whiff of a convergence problem.

 

What I usually want from SPICE is an answer to the question "Would it be a complete waste of time to build this to see if it works?" If there's an inkling that it might work, then I'd rather be on the bench with the soldering iron and the scope, than behind the computer waiting for SPICE to converge.

By the way, what are "fill-in"s, and why can I stop them by pressing [ESC]?

 

This was the response from Helmut on the LTspice users group from 2012. If he didn't know precisely, then the only other person who might is M.E. and he hasn't said much recently.

Helmut Sennewald 05/26/12 #54779 --- In LTspice@yahoogroups.com, Jim Thompson <yahoo@...> wrote: I keep getting messages at the bottom of the pane saying "fill-ins=<number> What does that mean? Is it interpolating? Hello Jim, I don't really know what it does, but I think LTspice has trouble with convergence of this circuit and thus it fill-ins estimated values in the matrix. You could try with the Alternate solver or other settings to help LTspice. Best regards, Helmut

It does square with what I intuituvely think it means. There are many additional hits on "fill-ins" but I found this on pretty quickly.

 

What I usually want from SPICE is an answer to the question "Would it be a complete waste of time to build this to see if it works?" If there's an inkling that it might work, then I'd rather be on the bench with the soldering iron and the scope, than behind the computer waiting for SPICE to converge.

Yes....we all want models that work and converge quickly.

 

To that point, I would not have believed it, until I saw it, but you can model an entire buck converter with dependent voltage sources, dependent current sources, transfer functions, and no actual components. It is true that the values for the R's and the L's and the C's are embedded in the coefficients of the transfer functions., but gone are the diode and the MOSFET.

Simulations run blisteringly fast to boot with not even a whiff of a convergence problem.

That is true. You can even model an opamp with a single primitive device. But I think that good convergence is more of an exception than the rule. It is highly dependent on the periphreal circuit complexity. Most model designers will stay away from that type of design because it will eventually cause convergence problems when used with other circuits.