Mike Engelhardt's new SPICE simulator QSPICE is out

[ledtester]

If you like to hear Kiss Analog talk, here's his take on Qspice:

LTspice is dead but QSPICE is born - A Great New FREE Circuit Simulation Software
https://youtu.be/u3xkLTgoQec

Note: it is 43 minutes long (only 22 minutes if you watch it at 2x speed).

I learned a few things about Mike Englehardt and the history of LTspice. I also learned that if you've only designed DC/DC converters don't call yourself a power supply designer (according to Kiss Analog).

[Veteran68]

If you like to hear Kiss Analog talk

Ha, that's funny. Poor Eddie, he's a knowledgeable guy no doubt, but but he speaks like the world's worst introvert had a gun put to his head and told he has one chance to deliver a speech to 1M people to save his life. As long as he's been on YouTube you'd think he be a little smoother by now. He's so awkward, it's easy to think he's clueless. But obviously given his work experience and accomplishments he's a smart guy, especially about power supplies.

[iMo]

"This program is specifically not licensed for use by semiconductor manufacturers in the design, promotion, demonstration or sale of their products." 

Yep, banned at my job too. It is very annoying, but for my limited needs random online simulators do the job.

The license does not prohibit to use it by an employee of a semiconductor manufacturer at his home

[Warhawk]

If you like to hear Kiss Analog talk

Ha, that's funny. Poor Eddie, he's a knowledgeable guy no doubt, but but he speaks like the world's worst introvert had a gun put to his head and told he has one chance to deliver a speech to 1M people to save his life. As long as he's been on YouTube you'd think he be a little smoother by now. He's so awkward, it's easy to think he's clueless. But obviously given his work experience and accomplishments he's a smart guy, especially about power supplies.

I was thinking about how to describe my feelings after watching the linked video. You described it perfectly.

[nctnico]

If you like to hear Kiss Analog talk

Ha, that's funny. Poor Eddie, he's a knowledgeable guy no doubt, but but he speaks like the world's worst introvert had a gun put to his head and told he has one chance to deliver a speech to 1M people to save his life. As long as he's been on YouTube you'd think he be a little smoother by now. He's so awkward, it's easy to think he's clueless.

He is doing videos a whole lot better than I would, so I'm not complaining!

[Veteran68]

If you like to hear Kiss Analog talk

Ha, that's funny. Poor Eddie, he's a knowledgeable guy no doubt, but but he speaks like the world's worst introvert had a gun put to his head and told he has one chance to deliver a speech to 1M people to save his life. As long as he's been on YouTube you'd think he be a little smoother by now. He's so awkward, it's easy to think he's clueless.

He is doing videos a whole lot better than I would, so I'm not complaining!

His production quality has certainly increased and he does cover some good content. I've been a subscriber for quite awhile and appreciate his work. But he can still be exhausting to listen to. He just comes across as so uncomfortable on camera.

[mawyatt]

Had the pleasure of meeting Dr Larry Nagle (his PhD dissertation at Berkley was SPICE) at the ISSCC awhile back, a colleague worked with Larry at Bell Labs and introduced us. Larry was honored on the cover of IEEE Circuits and Systems magazine that year, and he is the foundation behind all SPICE based simulators.

Recall the Berkley SPICE as it was called back then, was licensed for $25 to anyone, so many licenses were issued. With all these SPICE licenses (was in FORTRAN) around folks began to try and port to the early PC recently introduced by IBM. Vaguely remember Microsim (PSpice) was founded to develop hardware with multiple processors (multiple intel 8088) to speed up SPICE, so a specific Spice hardware engine. This failed as did many other folks attempts at utilizing multiple processors and they then focused on their version of SPICE which became the widely known PSpice. The issue with multiple processors or selective simulations of various circuit aspects was how to intelligently divide up the computational workload on various simulation circuits and how to link the results, remember a company that showed a significant improvement in simulation speed by simulation of DRAM, but memory was relatively easy to divide as it's just same circuit (memory cell) replication over and over.

Also remember another SPICE flavor where they advertised a circuit that compared simulation times with other popular SPICE based simulators, we caught them an alerted others as they had changed the convergence criteria RELTOL, ABSTOL and some other parameters to allow quicker less precise convergence.

Dr Ken Kundert at Cadence created Spectra which was an extension of SPICE for RF, ironically Ken's PhD is also from Berkley!! Ken's creation Spectra and prior work really extended SPICE not only for RF use but in general. Remember the "look ahead algorithm" in time domain simulations was significantly improved and "found" details that typical SPICE based simulators would step over and miss. Also included some of the things we had prior developed before Spectra for simulation of noise effects in the time domain, the speed up of simulations of high "Q" oscillators and so on.

Anyway, looking forward to having some time to play around with QSpice, altho wish it was available for Macs!! Our only "serious" computer is a loaded up Mac Pro (Can type), and not a fan of Windows.

BTW don't think the video is all that bad, yes a little hard to follow, but Eddie has a brew alongside. If the viewer follows Eddie and has a couple brews, then it's much easier to follow  

Best,

[Smokey]

How did I not hear about this?  Exciting.  Installing now.... 

My big question is.... how is the model library support doing?  My biggest problem with LTSpice was 3rd party component models, or lack thereof.  Yes, I had the mega-library of parts someone made, but it was still missing common parts and the mechanism to get it installed where it gets overwritten when you upgrade was a huge pain.  I get it.  It's LT's program, so why make it easy to simulate other companies parts.. But that problem should be gone now!

A spice sim with a huge library, or at least an easy library sharing mechanism will be so much more useful. 

[Warhawk]

How did I not hear about this?  Exciting.  Installing now.... 

My big question is.... how is the model library support doing?  My biggest problem with LTSpice was 3rd party component models, or lack thereof.  Yes, I had the mega-library of parts someone made, but it was still missing common parts and the mechanism to get it installed where it gets overwritten when you upgrade was a huge pain.  I get it.  It's LT's program, so why make it easy to simulate other companies parts.. But that problem should be gone now!

A spice sim with a huge library, or at least an easy library sharing mechanism will be so much more useful.

Mike E. mentioned that they put a lot of effort to have QSPICE compatible with standard SPICE libraries. I am going to make the ultimate test. Downloading TL431 from TI and trying it 

[Smokey]

How did I not hear about this?  Exciting.  Installing now.... 

My big question is.... how is the model library support doing?  My biggest problem with LTSpice was 3rd party component models, or lack thereof.  Yes, I had the mega-library of parts someone made, but it was still missing common parts and the mechanism to get it installed where it gets overwritten when you upgrade was a huge pain.  I get it.  It's LT's program, so why make it easy to simulate other companies parts.. But that problem should be gone now!

A spice sim with a huge library, or at least an easy library sharing mechanism will be so much more useful.

Mike E. mentioned that they put a lot of effort to have QSPICE compatible with standard SPICE libraries. I am going to make the ultimate test. Downloading TL431 from TI and trying it 

Got everything up and running (Build Aug 13).  And it looks like out-of-the-box the included libraries has support for..... only Qorvo JFETs, SiC Fets, and SiC Schottkys....   That's it.  It has generic behavioral stuff, but no other specific parts.

Yes, I understand that I can go find the spice model text definitions and paste them in.  They have a video demoing that here:

Code: [Select]

https://www.youtube.com/watch?v=WGJkZSDY9iA
But I really don't want to make symbols for everything with a sub-circuit... I wonder what the roadmap is for the default library...
I would think this would be something that could be hosted on github. Contribute with pull requests.  Merge libraries.  One community library.

[Warhawk]

Well, I do not expect libraries for everything. That would be barely maintainable. For basic components such as diodes, mosfets, BJTs you typically just take the generic breakout model, change the value (e.g. 1N4148W). Then, you place a directive somewhere in your diagram

Code: [Select]

.MODEL 1N4148W D  ( IS=10.4n RS=51.5m BV=75.0 IBV=1.00u
+ CJO=2.00p  M=0.333 N=2.07 TT=5.76n )
This way it works in pSpice (@PSpice: .model. .....) and LTspice. I expect same from QSPICE.

[iMo]

Until Qspice can read the existing .asc files and libraries I doubt it comes to a significant move from LTspice to Qspice..

Frankly, I can hardly imagine someone who has been working with - for example LTSpice - will switch to the Qspice when all the existing LTSpice libs and .asc files are not readable and compatible (not talking here ADI and LT chip models, but general stuff). The Qspice has been created exclusively for Qorvo (as he did for LT in past), and until ADI acquires Qorvo  the reading in the .asc files will not happen, imho.

[Warhawk]

Until Qspice can read the existing .asc files and libraries I doubt it comes to a significant move from LTspice to Qspice..

Frankly, I can hardly imagine someone who has been working with - for example LTSpice - will switch to the Qspice when all the existing LTSpice libs and .asc files are not readable and compatible (not talking here ADI and LT chip models, but general stuff). The Qspice has been created exclusively for Qorvo (as he did for LT in past), and until ADI acquires Qorvo  the reading in the .asc files will not happen, imho.

The question is if there's somebody at ADI willing to update and understand LTspice for future. There's a good reason why TI has now pSpice for TI as the main simulation platform instead of Tina TI. If LTspice has not been updated for a while and has some limitations that QSPICE can address. Time will tell us.

[iMo]

ADI has released a new version recently (17.x.y) and the older XVII is no more supported (last update in Jan 2023).
Afaik MikeE is not working with ADi for 3 years already, thus it seems ADI does it alone.
And as wikipedia says he has written simulators since 1975, thus most probably ADI and Qorvo would insist on having sources and some know-how transfer to younger staff at their hand..
And worst case ADi will aquire the Qorvo, imho (and perhaps that is the Mike's business model behind)..

[SiliconWizard]

Until Qspice can read the existing .asc files and libraries I doubt it comes to a significant move from LTspice to Qspice..

Frankly, I can hardly imagine someone who has been working with - for example LTSpice - will switch to the Qspice when all the existing LTSpice libs and .asc files are not readable and compatible (not talking here ADI and LT chip models, but general stuff). The Qspice has been created exclusively for Qorvo (as he did for LT in past), and until ADI acquires Qorvo  the reading in the .asc files will not happen, imho.

Yeah, probably not. I'm suspecting that he did reuse a significant chunk of LTSpice's code (I can't imagine he would have rewritten a Spice simulator entirely from scratch), I already don't know for sure how he managed the IP matters with AD, or if there's going to be any down the line. Just wondering.

[gnuarm]

Until Qspice can read the existing .asc files and libraries I doubt it comes to a significant move from LTspice to Qspice..

Frankly, I can hardly imagine someone who has been working with - for example LTSpice - will switch to the Qspice when all the existing LTSpice libs and .asc files are not readable and compatible (not talking here ADI and LT chip models, but general stuff). The Qspice has been created exclusively for Qorvo (as he did for LT in past), and until ADI acquires Qorvo  the reading in the .asc files will not happen, imho.

Yeah, probably not. I'm suspecting that he did reuse a significant chunk of LTSpice's code (I can't imagine he would have rewritten a Spice simulator entirely from scratch), I already don't know for sure how he managed the IP matters with AD, or if there's going to be any down the line. Just wondering.

I doubt the option existed to use the LTspice code.  I seem to remember Mike saying, some time back, that it would not be much harder to code from scratch, than it would be to update the code to the way he wants it to be. 

[iMo]

..
Yeah, probably not. I'm suspecting that he did reuse a significant chunk of LTSpice's code (I can't imagine he would have rewritten a Spice simulator entirely from scratch), I already don't know for sure how he managed the IP matters with AD, or if there's going to be any down the line. Just wondering.

I would bet the IP stuff with ADI has been managed by Mike properly (he is an entrepreneur as well) and his Qspice is from 98% the LTSpice source.
He just threw out all ADI models (therefore it is not compatible with ADI's libs and models and it should not be till ADI agrees with Qorvo somehow) and traded it with Qorvo as well, imho.

[JohnG]

Don't forget that a significant chunk (most) of LTspice's code is highly likely to be Spice3/Xspice, which I believe is open source. So, Mike Engelhart would not necessarily need to have any agreement with ADI to start from that codebase.

John

[Smokey]

...
Got everything up and running (Build Aug 13).  And it looks like out-of-the-box the included libraries has support for..... only Qorvo JFETs, SiC Fets, and SiC Schottkys....   That's it.  It has generic behavioral stuff, but no other specific parts.

Yes, I understand that I can go find the spice model text definitions and paste them in.  They have a video demoing that here:

Code: [Select]

https://www.youtube.com/watch?v=WGJkZSDY9iA
But I really don't want to make symbols for everything with a sub-circuit... I wonder what the roadmap is for the default library...
I would think this would be something that could be hosted on github. Contribute with pull requests.  Merge libraries.  One community library.

I found where some of the built in models live (build Aug 25).  When you drop in a diode or a MOSFET or something like that, right click it and go to "selection guide".  That's where the built in models are.
So there is some stuff.  But it doesn't look like there are any opamps. 

[SiliconWizard]

Don't forget that a significant chunk (most) of LTspice's code is highly likely to be Spice3/Xspice, which I believe is open source. So, Mike Engelhart would not necessarily need to have any agreement with ADI to start from that codebase.

Almost all Spice-based simulators use more or less of the Spice3x source code, but that's far from making a complete, modern, marketable simulator.

Besides, in his interviews, Mike explicitely talks about how the original source code is a bit of a mess and relatively inefficient, so many have re-implemented a significant part of it.
And, there's also a lot of features implemented on top of pure Spice (like all the mixed-signal stuff).

But, anyone's free to try. You can start with the original Spice source code and see how it goes as far as making a simulator that can "compete" with LTSpice or the like.
These days, it would probably make more sense to start with ngspice, which is already a gigantic work past the original Spice. But even so, there would be a lot of work needed to make something that would remotely look like commercial software.

[iMo]

I got my first Spice in around 1988/9, on a single diskette (afaik 3G5, it was around 350kB large binary), for my Atari 520STM (8MHz 68k, 1MB ram). We tested it against the "mainframe" versions and we get identical results, to our surprise..
I ran a lot of sims at that time (a postgrad) and also I wrote a GFBasic (or what was the name of that Basic compiler) graphing app which read the long columns of numbers (the Spice generated) parsed them and draw nice auto scaling pictures (similar to today's graphics) I could printout on my matrix printer. I can remember a simulation of a CMOS logic element consisting of say 6 transistors with a transient analysis like 1us long (looking at single input and output voltages) took me perhaps 30minutes at that time.. 
I cross my fingers the Qspice and LTSpice will find a common path finally..

[JohnG]

But even so, there would be a lot of work needed to make something that would remotely look like commercial software.

For sure.

It is also a lot of work to recreate libraries.

I have played with QSPICE a bit. The ability to handle C code for a controller looks really powerful for some future applications of mine. But, at the present, I don't need it, so I find myself going back to LTspice because I have a substantial collection of libraries and circuits that I don't feel like recreating. It will be years before QSPICE has a decent set of libraries, I fear. I hope Qorvo and others have the patience to stick it out.

I'm a little leery of spending a lot of time on libraries and such just yet. I've been burned more than once having tools go away on me. Anyone remember the MathCAD debacle?

John

[b_force]

Just a pure practical question, but what are the benefits with yet another SPICE program?
Besides the many free alternatives that are already available?

[gnuarm]

Just a pure practical question, but what are the benefits with yet another SPICE program?
Besides the many free alternatives that are already available?

Mike used to be the "Spice guy" at LT.  LT was bought by ADI.  Don't know if there is a connection, but sometime later, Mike leaves ADI.  Now Mike is the "Spice guy" at... whatever his company is called.  So, yet another Spice.  At least, it shouldn't suck.

[SiliconWizard]

Another question is, what's going to happen to LTSpice? Sure they probably have other people to maintain it, but now that Mike is doing the same thing somewhere else, AD may not even have the right to get any help from Mike, just in case.