Mike Engelhards next simulator after LTSpice

Anyone understand how Qorvo is involved? I'm not getting it.

That's great. LT Spice digital simulation is awful.

Free is even better. He could have sold it for big bucks, but I suspect he's crazy rich already.

The real breakthrough would be if there will be Spice models for their RF transistors.

Awesome. I'm looking forward to trying it.

Cheers

Phil Hobbs

I'd settle for built-in support for (1) diffusive transport, e.g. forward recovery in diodes (triple credit for modelling SRDs correctly), and (2) plot scaling that sits still when you re-run the sim.

BTW Infineon has Spice models for their SiGe parts.

Cheers

Phil Hobbs

It can't know in advance how high the current is gonna go on this run!

So does EPC for their GaN.

I think that s-params and Smith charts and load pull data are all relics of the graph paper and slide rule days. Spice makes so much more sense, especially as the world gets more wideband and more nonlinear.

Am 12.05.23 um 04:58 schrieb John Larkin:

No . they are not relics. Network analyzers, SmithCharts and s-params look at a circuit from the frequency domain just as spice, scopes and TDRs look from the time domain.

I learned Spice 2G4 on a Telefunken TR-4 computer that still ran mostly on core memory and Ge-Transistors, the first commercial microprogrammed machine. A 48 bit machine and its

96 bit double reals were a joy. VAX and X87 was a major step backwards for spice. There are no longer punched cards and output is no longer on a chain printer.

Other than that, Spice itself has not made a lot of progress, the biggest step was the transistion from Fortran to C but since Berkeley has left the boat - that's it. 30 Years?

A number of companies have sculptured their private user interface on the spice kernel and sold the result as a new product. p-spice, microcap, H-spice for timesharing service, Kevin's Superspice. Where are they now? They all may have made small improvements, that are headed to the gutter. No central authority anymore that keeps things together and forever. NG-spice perhaps, but I cannot remember any news since years.

As nice as LTspice was, it was a disaster for further improvements. Now with Qorvo there is at least some free competition. Free as in free beer.

Still it cannot model carrier lifetime for PIN diodes, it cannot do nonlinear noise analysis. Noise analysis of a chopper amplifier? Don't make me laugh so hard!

Noise and frequency response is computed in Spice by linearizing the circuit around the operating point and then doing small signal analysis. Is that any better than using s-parameters for 5V / 4mA right from the start?

Which is the stable operating point in a chopper amplifier, or in an oscillator with pulse feedback in Lee-Hajimiry style to suppress phase noise? How does Spice compute phase noise? There's no harmonic balance simulation.

Keysight ADS & Genesys and MWO have it and they have the RF transistor design kits in their libs. And they can ask an insane amount of money for it.

Even ham simulators like QUCs-Studio start to get it. <

>

Cheers, Gerhard

Am 12.05.23 um 03:55 schrieb Phil Hobbs:

and (3) running freq response, timing and noise in one job without editing the .tran / .noise / .ac line, where the old parameters are used as defaults for completely unrelated things.

That one is a wart, for sure. It’s reasonable to work around, though—once you get used to the syntax, you can just put .tran, .ac, and .noise in one block, and comment out the ones you’re not using. That way you can have all the defaults you like.

Having the simulator ignore prominent effects like carrier transport is more of a problem.

For time-saving, it would also be nice to have better treatment of board strays. A full method-of-moments EM sim would be unwieldy, but something as simple as trace capacitance wouldn’t be that hard.

Cheers

Phil Hobbs

But the RF tools do one frequency at a time. The data sheets provide s-params for a few selected frequencies where I guess the money is.

I wrote my own sims that ran on a PDP-8 and graphed on a teletype. And that sold a lot of stuff. My first attempt was 32,000 horsepower with

10-minute time constants.

I understand that LT Spice is not based on the UCB code. It's actually an x86 compiler.

LT Spice killed them. It's great and it's hard to compete with free. It's like LabView, give it to the kids in college and they'll use it when they grow up.

Qspice will be free too. Smart.

It did get continuous improvement as long as Mike ran it. I don't know what ADI will do with it. They seem to be struggling to build part models just now.

Noise is always small-signal. Frequency response can involve serious nonlinearities that Smith Charts can't do.

I work time domain, DC to GHz all at once, nonlinear as hell, so I use Spice and have to derive my own models for RF parts. RF data sheets are ghastly.

I use it to estimate picosecond jitter in LC and crystal oscillators. How do you use frequency domain tools to quantify the Johnson and device and power supply and temperature contributions to phase noise?

Yes.

What do you call insane? Microwave office (pre ANSYS) and Silvaco SmartSpice/SmartSpice RF were under $10K not too long ago.

That's not insane?

I'll settle for a very powerful simulator that is free even with some shortcomings. There is a reason LT SPICE is a de-facto standard.

BTW, should be *Englehardt's* in the subject.

No, the correct name is: Mike Engelhardt

Bernd Mayer

*In the subject is Englehards*. That is wrong! His name is mike Englehardt. The possive is Englehardt's!

Show me wrong, *Berndt*.

Am 14.05.23 um 01:30 schrieb John S:

Subject may be wrong, but his name is Engelhardt and not what you say.

Hanno

hello

Justice Mike Englehard knows almost nothing about physics, electronics and programming of a simulator for electronic circuits.

Mike Engelhardt does.

Open your eyes wide and read all those infos available on the internet.

Bernd Mayer

Oh! I understand. Your first language is not English. Sorry.