Is there something wrong with PMOS model in the LTSpice XVII?

You didn't observe both forward overshoot and reverse recovery?

You're so set in your ways you go about proclaiming, "Can't be done", thus I tease >:-} ...Jim Thompson

I saw an edge with a peak.

If you thought I was claiming that there was no way to make a waveform with overshoot in SPICE, you weren't reading very closely. That's all you've got so far.

I'm interested if you have something more substantive, but I'm not holding my breath waiting.

Cheers

Phil Hobbs

I have a Metcal, and it's worth it. Temperature controlled, gobs of heat when you need it, heats up in seconds, and you can change the tip in about 15 seconds.

Of course I use 63/37 leaded, rosin-core solder.

Didn't someone make a monolithic generic linear IC? The connections were top layer metalization, or maybe even wire bonds.

The analog equivalent of the FPGA has been attempted but not very successfully.

Poor baby! It's a diode. ...Jim Thompson

All my G-jobs are thru-hole, no surface-mount stuff, so I don't need anything fancy. (The 60/40 IS rosin-core.) ...Jim Thompson

Yes. They had different size chips with different arrays of devices. I did a few designs for an audio-phoolery company in San Jose _many_ years ago. It was really tough to get a high-percentage of utilization with only one metal layer... you could use N+ pass-under's, but that wasted an NPN in the process.

It just came to me, the array-chip provider was Interdesign, Hans Camenzind's company.

Yep. Too many variables. ...Jim Thompson

g my breath waiting.

And that in a nutshell is why I don't bother following your links. A pity-- you'd have a lot to contribute if you could just get over yourself.

However, life goes on either way.

Cheers

Phil Hobbs

I was just demonstrating that it _is_ possible to Spice model forward over-shoot and reverse-recovery... which modeling you have claimed is impossible without "hacking" the simulator code.

You are wrong, but go ahead and stick your head in the sand if you want.

I will release the model in encrypted LTspice format. ...Jim Thompson

You didn't demonstrate squat. You posted a screen shot and followed it up with a lot of bluster.

Where I come from we know a trick worth two of that one.

Cheers

Phil Hobbs

Bwahahahahahaha! ...Jim Thompson

I gave the Metcal away when I got a Weller WX2. Metcal is stone age stuff.

No socialist solder for me, either.

Well, there is the PSoC and IIRC Maxim has some sort of programmable mixed signal thing.

You can't solve diffusion*, but you can make a damn good approximation of it (say with an R+L or R+C ladder), and if it's close enough for gov't work, who cares?

As you're rather more in-depth with physics than I am -- could you illustrate with some examples of systems that are impractical to approximate in this way?

Why not a diode? Under what V, I, t conditions would the approximation fail?

Or approximated, such as LC ladder sections. Shitty, but practical for some cases -- as is the condition for all approximation methods, including anything else in SPICE.

Tim

te

Systems with memory, diffusion, or delay elements such as transmission line s, for a few. You can use a bunch of elements to model the internal state o f the process, as you say, but it only works over a restricted parameter ra nge and is not usually very satisfactory.

It takes a metric buttload of LCs to model coax at all accurately over more than a fraction of a wavelength.

on fail?

Basically any except the ones you used to fit the model. ;)

That's a bit unfair of course, but I invite you to consider how you'd model an SRD accurately in SPICE. It sure wouldn't be a lot of little RCs.

Some people run numerical models in Excel, because that's the only tool the y know. Putting bandaids on SPICE may sometimes be better than writing your own simulator, but it's not a universal tool even for circuit design.

Cheers

Phil Hobbs (simulator writer and very occasional SPICE bandager)

You don't even need that... think charge accumulation as a trigger mechanism.

...Jim Thompson

Well you could simulate a computer that could run a better model... :)

Been there. The number of LC sections goes as the square of Td/Tr. That gets painful fast, especially running ECA on an 8086 DOS machine.

LC delay line models ring terribly, too.

One wouldn't likely need a full model, just something that behaves about right under the expected bias and such. A full model would be nasty to write.

It would in Spice, for sure, but it ought to be possible to do it with an auxiliary 1-D transport model bodged into the main SPICE code.

In my EM simulator I have something like that to handle materials with Lorentz and Debye poles (e.g. real metals). I didn't invent the idea, I cribbed it from Taflove & Hagness, "Computational Electromagnetics". (Great book btw.)

Cheers

Phil Hobbs