NE-2 I-V Curve ??

Yes ;-)

See...

for possibilities. ...Jim Thompson

It usually _is_ the parts, or rather, the model/subcircuit of the part... discontinuities with no derivatives... the very topic I've been harping about... making better models.

LTspice runs faster because, by default, the .TRAN command leaves off filling in the "Maximum Timestep" amount. Thus, on an "active" circuit with a parasitic oscillation, it skips over that "grass" and _you_ think your circuit is gorgeous and you chide me when I point out that it's oscillating, claiming I'm wrong... your standard retort... when you're wrong, claim the truth is wrong :-(

What you seek is right under your nose...

_Simulate | Edit Simulation Cmd | Transient | [Maximum Timestep] ...Jim Thompson

ISTR that just makes the plot window catch up with the transient.

There's a list of convergence suggestions at

You can set a minimum capacitance from each node to ground

.options cshunt=10f

Cheers

Phil Hobbs

I don't think the library parts have problems. Sometimes Spice converges slowly. I don't use it often enough that hitting esc now and then is a big deal. When I esc out of the initial convergence stall, the rest of the sims seem fine.

Sometimes that helps, often it doesn't. What I really want is a *minimum* time step. The convergence hangup is almost always at the beginning of a transient run; once that's over, after a while or after a couple of esc hits, the rest runs fine.

If the RC neon thing is oscillating at the beginning of the run, it's not real.

It's no big deal; I don't design many neon oscillators.

Ah, I see--it depends on what the solver is doing at the moment. Usually when mine get hung up it's early in the damped pseudo-transient stage, where that doesn't help.

A lot of TI's op amp models don't play at all well in LTspice--try the OPA140, for instance. None of the tricks on the webpage below help that one at all, whereas it converges instantly with an LM358 macromodel (fwiw).

Cheers

Phil Hobbs

Easy to see why...

.MODEL S_VSWITCH_1 VSWITCH (RON=1 ROFF=100MEG VON=100M VOFF=-100M) .MODEL S_VSWITCH_2 VSWITCH (RON=1 ROFF=100MEG VON=100M VOFF=-100M) .MODEL S_VSWITCH_3 VSWITCH (RON=1 ROFF=10MEG VON=100M VOFF=-100M) .MODEL S_VSWITCH_4 VSWITCH (RON=1 ROFF=10MEG VON=100M VOFF=-100M) .MODEL S_VSWITCH_5 VSWITCH (RON=10M ROFF=100MEG VON=150 VOFF=0) .MODEL S_VSWITCH_6 VSWITCH (RON=10M ROFF=100MEG VON=150 VOFF=0) .MODEL S_VSWITCH_7 VSWITCH (RON=1M ROFF=10MEG VON=-10M VOFF=0) .MODEL S_VSWITCH_8 VSWITCH (RON=1M ROFF=10MEG VON=10M VOFF=0) .MODEL S_VSWITCH_9 VSWITCH (RON=1 ROFF=10MEG VON=1 VOFF=-1) .MODEL S_VSWITCH_10 VSWITCH (RON=1 ROFF=10MEG VON=1 VOFF=-1) .MODEL S_VSWITCH_11 VSWITCH (RON=1 ROFF=1G VON=10 VOFF=-10) .MODEL S_VSWITCH_12 VSWITCH (RON=1 ROFF=1G VON=10 VOFF=-10) .MODEL S_VSWITCH_13 VSWITCH (RON=1 ROFF=1G VON=10 VOFF=-10) .MODEL S_VSWITCH_14 VSWITCH (RON=1 ROFF=1G VON=10 VOFF=-10)

Whoever writes TI's models doesn't know how to apply bounding conditions except with a switch.... all those switches cause static convergence problems.

You could substitute Mike's TANH-based switch and see if that helps.

Of course the model works quite well in PSpice in spite of those switches >:-}

Often what works around all these obstacles is to have your power supply be a pulse- or PWL-based source... start up from zero (in a .TRAN situation) seems to work every time. ...Jim Thompson

Not with that mess, it doesn't. It toddles along till the power supply starts to ramp, and then pukes.

The VSWITCH device has a gradual turn-on of some sort as well.

Cheers

Phil Hobbs

Have you tried forcing a maximum timestep?

Also, I think LTspice supports gmin stepping... which is well worth trying. (I tend to make gmin stepping my default method.) ...Jim Thompson

...snip... Often what works around all these obstacles is to have your power supply be a pulse- or PWL-based source... start up from zero (in a .TRAN situation) seems to work every time. ...Jim Thompson

yes, it seems TI purposely causes trouble for LTspice, to encourage the use of TINA.

On most of those, I simply use a PWL as the power supply, usually works.

As you know I'm building a business making models that work properly, match the real device (or at least the data sheet), and a number of TI parts are on the wish list... not only because the models are obtuse, but many new ones are encrypted. ...Jim Thompson

Well, good luck then...

I've just sold my first major manufacturer modeling job. So your good wishes paid off ;-) ...Jim Thompson

I've never had to do that, but I'm running it under wine which may make a difference.

In other words, it works better if _you_ turn it on :)