The Designer's Guide Community
Forum
Welcome, Guest. Please Login or Register. Please follow the Forum guidelines.
Mar 28th, 2024, 5:24pm
Pages: 1 2 
Send Topic Print
The History of SPICE (Read 213 times)
analogspiceman
Junior Member
**
Offline



Posts: 23

The History of SPICE
Aug 05th, 2013, 4:02am
 
I have put together a bullet point history of SPICE leading up to LTspice over on the LTwiki.  There are and have been many varieties of SPICE over the years, but this history is mainly about the road to the popularization of SPICE, so not all flavors of SPICE are covered.

In addition to basic research, I have contacted some of the primary sources directly in order to get their first hand recollection of the historical facts (Larry Nagel, the original father of SPICE, Paul Tuinenga of PSpice fame and Mike Engelhardt, the author of LTspice).  The history is now quite accurate (no doubt mistakes still exist), but it is lacking cites (which I plan to add).   I also plan to preserve the technical history portions of the emails the aforementioned sources were so gracious to send me.  See:

http://ltwiki.org/index.php5?title=LTspice_Genealogy_-_The_Heritage_of_Simulatio...
Back to top
 
 
View Profile   IP Logged
analogspiceman
Junior Member
**
Offline



Posts: 23

Re: The History of SPICE
Reply #1 - Aug 7th, 2013, 9:23am
 
To all,

For the purposes of gathering accurate historical data for the LTwiki's SPICE history page I have been in contact with Larry Nagel (the father of SPICE), Paul Tuinenga (a PSpice founder) and Mike Engelhardt (LTspice).  For the sake of posterity, I will be posting copies of the pertinent portion of those exchanges here (as well as on the LTwiki) -- a.s.



To Paul Tuinenga, on Fri, Aug 2, 2013 at 10:33 AM, <analogspiceman> wrote:

Hello Paul,

I am currently writing a bullet point history of the popularization of SPICE in the engineering community.  The emphasis is on the path SPICE has taken to arrive on the most engineering desktops.  Because of this emphasis, my history begins with the original Berkeley SPICE variants, continues onto PSpice (its limited, but free student version made SPICE ubiquitous) and culminates with LTspice (because, at over three million downloads, it has reached many more users than all other SPICE variants combined).

I am lacking solid information about the beginnings of PSpice, but I understand that you may have been one of the founders of MicroSim and was hoping you might be able and willing to help me to complete the section about PSpice.  Please see:

http://ltwiki.org/index.php5?title=LTspice_Genealogy_-_The_Heritage_of_Simulatio...

Specifically, I really have no idea who were the people initially responsible for creating PSpice.  Also, I could not find any information as to when and at what revision Probe became a part of PSpice.  (Perhaps at the initial release?)

Then there is the meaning of name itself.  I vaguely recall that PSpice was at some point called uPspice (the 'u' being a micro symbol), thus the acronym may have stood for micro-Processor SPICE (others suggest it meant "Personal SPICE" or "Personal-computer SPICE").

Last of all, I would like to list the timing of the introduction of the most important and innovative features of PSpice (a very weak start at this is up on the LTwiki).

I hope I have the facts straight and would be happy to take an suggestions for corrections that you might have (if you are so inclined).  The wiki is a work in progress and any useful feedback and helpful information provided regarding PSpice would be greatly appreciated.

Best Regards



On Fri, 2 Aug 2013 at 7:40 PM, Paul Tuinenga wrote:

I will be delighted to help [... jump to historical recounting ...]

PSpice was launched in Jan 1984, with a short paragraph in Electronics (McGraw Hill) magazine.  I have a photocopy of that announcement, but it's 700 miles away from me just now.

P stands for Personal, as in Personal Computer (PC).  MicroSim (an invented name... having first checked the advertising list in Byte magazine) stands for simulation on microprocessors.  As it turns out, the main product idea of the company was a mistake.  Between the summer of 1984 and the spring of 1985, we built an accelerator based on multiple Intel 8086s (with 8087 floating-point coprocessors), two complete computers per Multibus board, and systems with up to 12 boards.  The external box attached to an IBM-PC/AT and provided over 2x the speed of a DEC VAX-11/780 minicomputer (which many companies were using 24x7 to run UCB SPICE).  It was called PSpice Turbine.  We sold one a week after debuting it at an EDN industrial conference.  Then the industry went into a capital-spending slump and we didn't sell another for the rest of that year.  In the meantime, PSpice the "shrink wrap" software, the sales of which were to "keep us in baloney sandwiches" until the real product, the Turbine hardware accelerator, took off.  It never did, but every month the sales of PSpice grew.  So we shelved the hardware development to re-focus on the software e.g. re-writing it from Fortran into C (then C++ much later).

Is this the level of detail you are looking for?

Best regards,

- paul


Back to top
 
« Last Edit: Aug 7th, 2013, 10:30am by analogspiceman »  
View Profile   IP Logged
analogspiceman
Junior Member
**
Offline



Posts: 23

Re: The History of SPICE
Reply #2 - Aug 7th, 2013, 10:01am
 
[ ...continuation of email interview with PSpice founder, Paul Tuinenga... ]

On Sat, Aug 3, 2013 at 12:31 PM, <analogspiceman> wrote:

Hi Paul,

Thank you for your response.  [... cut to further historical query ...]

Some additional questions:  When did you first begin working on what was to become MicroSim/PSpice (and when was MicroSim first officially launched - 1983)?  Also, were you the sole (major) founder or did you have other equal partners?

Here is what I have about you and MicroSim/PSpice (btw, what is your preferred capitalization format for PSPICE, PSpice, Pspice)?

First the wiki page introduction, then the section just about MicroSim PSpice: [cut most of quote of LTwiki page]

[...] If as its primary original author, Larry Nagel is deservedly considered "The Father of SPICE" then as its essential enabler, advocate and general all-round visionary for open source electronic simulation software, Don Pederson most certainly is SPICE’s godfather. [...]



On Sat, Aug 3, 2013 at 3:30 PM, Paul Tuinenga wrote:

Yes, I scanned the LT page but purposely ignored it for my response to let you glean what you find notable.

What's just caught my eye is the verbiage around Dr. Don Pederson.  If you don't have access, I'll find a copy of the IEEE Spectrum cover article on him.  He is both the reason for never creating a circuit simulator as well as the reason SPICE got created.

PSpice is double caps.

The creator of PSpice is Wolfram Blume (BS75 Caltech).  He sat in the office next to mine at Silicon Systems, Inc. (SSi),  In 1983, SSi bought several IBM-PC/XTs.  Blume was the in-house maintainer of UCB SPICE on the Prime minicomputers used for IC design and layout at that time.  SSi developed most of the tools it needed (and we all take for granted today) because none existed at the time.  For example, SSi was the first to make color plots of IC layouts, and using stipple patterns for the layers (these overlap in a visually pleasing way).  We also created a layout-versus-schematic (LVS) that predates by several years anything from the ECAD vendors.

Knowing the innards of SPICE, and where the time was spent for transient simulations, Blume got the idea to see how fast the IBM-PC was.  He tested the speed of MOS level -2 code using the Microsoft Fortran compiler.  As is turns out, that test gave an optimistic result for Intel processors to accelerate simulations and thus propelled the idea of MicroSim.

The problem with SPICE being shoehorned into the PC (640KB max memory) is SPICE2 was five(?) overlays and no IBM-PC Fortran compiler had overlay capability.  Blume figured out how to replace those as two programs run in sequence, with the first doing read-in and checking, then leaving a data structure in memory for the analyses to operate on in the second program.  This is why MicroSim PSpice had no competition for about two years, until later IBM-PC Fortran compilers handled overlaying.

SSi was officially upset with Wolfram and "Blume Engineering" selling PSpice for $495, and demanded royalties.  Blume quit SSi.  Shortly after, he and I met to discuss his situation.  He described this idea for building a hardware accelerator.  He's not a hardware guy, but I am and I knew exactly how to build it.  So I quit SSi, too, and we formed MicroSim in June 1984.  Blume was the major partner.

Later, when the Turbine fizzled, I turned to software and rewrote all the device equations into C.  Then I created the Parts option and built up the several thousand standard components library.  Also, I wrote "the book" which eventually went to three editions (additional material in each), was translated to Japanese and French, and sold well over 100,000 copies.  That last sounds small, but for technical books, which normally sell in the mid 4-figure volume range, I am Steven King.  OK, I wrote a pretty good book, but Prentice Hall gets all the credit for pushing it far and wide.  Dr. Richard Newton told me he saw a copy in a Moscow university office (long before the Soviet empire fell).

What else?

- paul



On Sat, Aug 3, 2013 at 3:33 PM, Paul Tuinenga  wrote:

BTW, initially I was "Vice President" then later "Executive Vice President" and CFO.  Blume was always President (and CEO). - paul



On Sat, Aug 3, 2013 at 1:11 PM, Paul Tuinenga wrote: [Don't know why the time of this email is out of sequence. -- a.s.]

Correction: the minicomputers at SSi were Perkin-Elmer not Prime. - paul


Back to top
 
 
View Profile   IP Logged
analogspiceman
Junior Member
**
Offline



Posts: 23

Re: The History of SPICE
Reply #3 - Aug 7th, 2013, 10:20am
 
[ ...continuation of email interview with PSpice founder, Paul Tuinenga... ]

On Sat, Aug 3, 2013 at 6:51 PM, <analogspiceman> wrote:

Hi Paul,

Perhaps a couple more questions if you are willing to indulge my persistence.

Was Probe available from the very beginning along with PSpice or was there a period where the only output was ASCII graphics (which went to the printer or monitor or either)?  When was a pixel graphics Probe first released?

How did the student version of PSpice come about?  It seems to me that getting universities to switch to PSpice versus running mainframe jobs was a stroke of marketing genius.  All those newly minted engineers would come off the academic assembly line with a preexisting predilection for PSpice.  Was this the plan?

[ ...regarding Paul Tuinenga's remark about Don Pederson... ]:

I assume that you referring to his influence to go "open source" as having enabled a flood of SPICE vendors to just take the Berkeley code and wrap it engineering eye candy, thereby more-or-less killing the already marginal economic incentive for commercializing the very expensive ongoing support and development of SPICE in what is a rather meager specialty niche market.  Mike Engelhardt of LTC has commented that there is no way he could have spent the last two decades refining LTspice if it had had to have supported itself independently through sales profits.  He already was making the basic tool for internal consumption at LTC and just tweaked it in order to give it a parallel life as a marketing lever for their rather large number of rather functionally complicated power control ICs.  LTspice allows the engineering zombies [ugh, u-hhg, brains... I need brains...] to access built-in design smarts for every part in LTC's product line.

[...]

I think [...] your historical recounting should be preserved and would be fascinating, at least to a certain significant portion of the engineering community.  So many engineers cut their simulation teeth on PSpice, although it seems that role has now pass over to LTspice.  If you don't mind I will collect together the historical sections from you emails and create a page for them on our LTwiki.



[Note, Paul's response is too long to fit in this message, so... ]

Back to top
 
 
View Profile   IP Logged
analogspiceman
Junior Member
**
Offline



Posts: 23

Re: The History of SPICE
Reply #4 - Aug 7th, 2013, 10:28am
 
[ ...continuation of email interview with PSpice founder, Paul Tuinenga... ]

On Mon, Aug 5, 2013 at 11:13 AM, Paul Tuinenga wrote:

I don't mind these questions, just surprised anyone is interested in ancient, narrow history.

Blume developed Probe mostly during the time I was designing and building the engineering version of the Turbine accelerator, which was convenient since he had nothing to develop code on until I got that working.  Probe was available later in 1984 or the beginning of 1985, I forget but can look that up when I get back [...]

In my opinion, if PSpice was "good" then Probe was "great" in that it made the purchasing decision a "no brainer" for most.  So no only was the simulation under the control of the user (e.g. no waiting for a batch queue; sure, maybe the mainframe sim took only 30 seconds however you had to wait in line for 2 hours to have it run), but viewing the results was interactive and high quality. I have always thought Probe was, for most customers, the real product.  Certainly, PSpice was leading edge and convenient but it has a boring status screen, and just sits there working.  Probe was an oscilloscope -- the visible face of PSpice and MicroSim.  The early version supported CGA, VGA, and SVGA graphics, as I recall.  A few times a week, we would field a call for a customer ordering PSpice/Probe and asking for opinions of computer and display to purchase -- they did not have a PC, yet, and were already buying PSpice as the app they were getting the PC to run!

Also, being first, Blume had complete freedom to make a clean user-interface.  Everybody after had to purposely not-copy Probe.

The demo version of PSpice was available nearly from the start, and its limited circuit size was a delicate balance: showing it "had the goods" but not enough to get real work done.  About once a month someone would call to see if we would increase it by "one more transistor".  As if...

Recasting the demo software as a "student version" came about in late '85 or early '86 (as I recall, but I can check that later) when we found there was a mailing list of college "electrical engineering" professors available for purchase.  We thought it would be perhaps a few thousand names/addresses but it turned out to be over 20,000 (as I remember and that was just domestic). To make the disc, label it, and mail it with a cover letter, for that volume, was a "break the bank" expense for MicroSim at the time.  But we forged ahead realizing that if it was a good idea for say 2,000 names it was an even better idea for 20,000.

Genius?  In retrospect, but that's always the way.  We were confident it would be tried by a large percentage of professors or they would give to to someone to try (i.e. not throw it away).  We knew it would give students with access to PCs something to play with and enhance their lab-bench efforts.  And there was not much else useful to run on PCs back then ("Lotus 123", etc.), but engineering departments had invested in PCs as "the future" even without a plan for software that supported their course offerings.  We were in the right place at the right time.

If anything, this knitted PSpice into the fabric of EE education.  A few years later, we read in an industry rag (ED, or EDN, etc.) an article about circuit simulation that alluded to "...PSpice and all the other PSpices..."  Then we knew we had made it -- PSpice had either replaced Berkeley SPICE, in mind share, or perhaps many people had never heard of Berkeley SPICE and assumed PSpice was the original thing.

We often could track the career of many customers.  They would take a new job, or new consulting client, and order PSpice again.  "Oh, it seems Joe is working at Floobydust Systems now."  Some did it a half-dozen times.

Another competitive barrier-to-entry we erected was having a growing library of standard parts, e.g. 1N4001, 2N2222, uA709,...  Immediately after Microsoft's Fortran compiler started supporting overlays there was a flood of alphabet-soup-SPICE-on-the-PC products.  Not only did we make a library, but pushed it to over 1,000 components.  A nice round number, and enough that even if a particular part was missing a suitable substitute could likely be in that library.  And we released the Parts option, the software we used to make the library, so customers could make anything missing from the library.  That broaden the market to board-level engineers without a clue as to Gummel-Poon, Shichman-Hodges, etc. models.  It took the "IC" out of SPICE.

In the process of that, we upended the market for SPICE models for hire and fostered the expectation that models be unencrypted ("open").  I recall a printed interview with one of the Hailey brothers (the "H" in H-SPICE) bemoaning how Parts had ruined their $6,000/piece modelling business.  Boo hoo.

Anything else?

- paul


Back to top
 
 
View Profile   IP Logged
analogspiceman
Junior Member
**
Offline



Posts: 23

Re: The History of SPICE
Reply #5 - Aug 10th, 2013, 9:08pm
 
The Designer's Guide has a relatively recently updated copy of The Life of SPICE by Laurence W. Nagel.

http://www.designers-guide.org/perspective/life-of-spice.pdf

This was one of my primary sources for the Berkeley SPICE portion of the LTwiki bullet point history culminating in LTspice.  It's a great read, but some details were still unclear to me, so I wrote Larry Nagel with a few questions, which he most graciously answered.  The email exchange follows below (also now on the LTwiki).


On Fri, 19 Jul 2013 at 10:47 AM, <analogspiceman> wrote:

Hello Dr. Nagel,

I am a user and fan of SPICE (LTspice, to be precise) and a major contributor to the LTwiki and the LTspice users' group.

I have compiled a bulleted history of SPICE and the road to its popularization (culminating in LTspice).  Please see:

http://ltwiki.org/index.php5?title=LTspice_Genealogy_-_The_Heritage_of_Simulatio...

I hope I have the facts straight and would be happy to take any suggestions for corrections that you might have (if you are so inclined).  The wiki is a work in progress and I intend to fill it out with references to your website and pertinent publications.  Also, with your permission, the owner and host of the LTwiki would likely be willing to directly archive copies of some of your publications.

Best Regards


On Tue, 30 Jul 2013 at 12:22 PM, Larry Nagel wrote:

I apologize for taking so long in responding to your inquiry.  I've had a lot of consulting work lately, and then I just finished with a visit from grandchildren from Boston.  The dust hasn't yet settled!

First of all, let me congratulate you on doing an excellent job on the history of SPICE.  I don't know whether you were ever connected with the SPICE effort at Berkeley, but you certainly seemed to sort out most of the facts in a very clear and concise manner.  Job well done!

Here are a few comments/corrections:
  • CANCER wasn't a derivative of the class project program, it WAS the class project program.  After the class project was over, CANCER became my Master's Project, and Ron Rohrer was my research adviser for my Master's program.  You are correct that there was controversy between Don Pederson and Ron Rohrer regarding whether CANCER code was proprietary.
  • After I finished up my Master's degree with Ron Rohrer, and after Ron decided to leave Berkeley, and after a summer job working with Ron Rohrer at SofTech in Boston, I started on my PhD program with Don Pederson in the fall of 1971.  It was then that CANCER was renamed SPICE and version 1 of SPICE was released to friendly users shortly thereafter.
  • You are correct that CANCER and SPICE 1 did not use Modified Nodal Analysis, but they did implement both ideal voltage sources and inductors using a row-swap algorithm in addition to Nodal Analysis.
  • SPICE was officially released to the world in May 1972 and was officially announced at the 16th Midwest Symposium on Circuit Theory in April, 1973.  By this time SPICE contained the both Ebers-Moll and Gummel-Poon bipolar models, Shichman-Hodges JFET and MOSFET models, flicker noise models, sensitivity analysis using the adjoint network, distortion analysis of weakly nonlinear bipolar circuits using Volterra series models and the adjoint network, and numerous improvements intended to make the program more user-friendly and bulletproof.
  • There is this persistent rumor that Ellis Cohen added dynamic memory allocation to SPICE 2.  Both Ellis and I will tell you that it was I who added dynamic memory allocation to SPICE 2.  It's in my thesis.  Ellis was my roommate at the time and we certainly had many discussions about how to best implement dynamic memory allocation, but it was I who punched the cards and debugged the monster.  I will never again try to implement dynamic memory allocation in a FORTRAN program!
  • SPICE 2B had subcircuits (it's in my thesis) but Ellis Cohen reworked the implementation in his version of SPICE.  Ellis was solely responsible for the addition of poly sources and transmission lines.
  • I think that Ellis Cohen would agree with me that SPICE 2G6 was mainly the work of Andrei Vladimirescu.  Much of the MOSFET modeling work in SPICE 2G6 was due to Sally Liu.  The statement that "Ellis Cohen becomes primary contributor with later help from Andrei Vladimirescu" may be a little unfair to Andrei.
  • The P in PSPICE stands for Personal Computer, not micro Processor.  PSPICE was the first commercial simulator that ran on a personal computer.  PSPICE may have originally evolved from SPICE2, but quickly migrated to the SPICE3 code base.  I don't know if there ever was a FORTRAN version of PSPICE.

I hope this helps you in your history exercise.  It's always fun delving back forty years ago in my lifetime!

Thanks!!!

Larry

Laurence Nagel
Omega Enterprises Consulting
251 Stanford Avenue
Kensington, CA 94708
(510) 558-0842
lwn@omega-enterprises.net
http://www.omega-enterprises.net


On Fri, 2 Aug 2013 at 10:01 AM, <analogspiceman> wrote:

Thanks ever so much for your comments!  It really is best to get the facts directly from "the horse's mouth" so to speak. Smiley  I have now incorporated your comments and corrections into the wiki's history page.  In case something got garbled in the process, please don't hesitate to provide further corrections.

Best Regards
Back to top
 
 
View Profile   IP Logged
Ken Kundert
Global Moderator
*****
Offline



Posts: 2384
Silicon Valley
Re: The History of SPICE
Reply #6 - Aug 24th, 2013, 10:56am
 
You can find more on the history of SPICE in the Spring 2011 edition of the IEEE Solid-State Circuits Magazine. Be careful when you search for this. There were actually two versions released. In the first version, several of the articles were accidentally left out. This was corrected in the second version. One way you can tell which version you have is to look for my article, which was missing from the first version.

I have also added my article to this website. It can be found at Life After SPICE.

-Ken
Back to top
 
 
View Profile WWW   IP Logged
analogspiceman
Junior Member
**
Offline



Posts: 23

Re: The History of SPICE
Reply #7 - Aug 25th, 2013, 2:42pm
 
Hello Ken,

I read both the interesting paper you linked and the 2007 paper from the Proceedings of the IEEE: http://www.kenkundert.com/docs/procIeee2007.pdf.

A theme interwoven throughout your historical recounting is your surprise at the inertia of the engineering community at embracing various improvements in simulation software technology.  I wonder if the traditional expensive simulators are today's simulation dinosaurs - big, awesomely powerful, but slow to adapt to their changing environment (the wants and purchasing abilities of the majority of the engineering community).

Over my career, I have used many flavors of SPICE, some by desire, some by dictate, but I have never used any of the really expensive behemoths (I am not an IC designer - my needs have always been analog and switched mode power).  My original favorite was PSpice (which Cadence effectively neutered by raising the price and lowering the support), so the mantle was passed on to LTspice.  It currently is by far the best general purpose simulator out there that is easily available to the masses, both now and for the foreseeable future.

Performance, availability (it's free) and the perception of an ongoing persistence in those two features are the keys to LTspice's pervasive success, in my opinion.  But this formula runs counter to traditional business sense (who's gonna pay all the ongoing software development bills?).  I think the real brilliance of the folks at Linear Technology was in inventing a new funding paradigm - once you get your simulator on nearly every engineer's desktop, you don't have to hit them over the head with your IC product sales pitch (which they would abhor anyway), you just make designs with your ICs all unobtrusively available within your simulator so that the lazy and the incapable can't resist going there.  And if your IC products are high quality and work well in their products, it's a win-win and they will barely notice that your parts cost more than your competitors'.  Equally important is that this funding is almost all gravy, because your free marketing tool is really just a for-public-consumption version of your proprietary in-house simulator, the development costs for which you had to carry anyway.

Some of the expensive simulators do a lot more hand-holding than LTspice.  A lot of this falls into the category of trying to do your thinking for you, which I find to be an annoying layer that just gets in my way and obscures what is really going on under the "eye candy."  A small portion is probably useful, an example being automatic setup of Monte Carlo or worst case analyses (LTspice has the hooks for these, but the user has to figure out how to set them up - or look to the LTspice Yahoo group).

A lot of simulators have added POP/PSS (Periodic Operating Point/Periodic Steady-State) solvers to their repertoire.  Micro-Cap recently added POP capability.  They discuss it here: http://www.spectrum-soft.com/news/winter2013/pss.shtm.  Notice the beneficial effect it has on the resolution of the FFT example.

Many other simulators take this further with the addition of a time domain based fast running FRA (Frequency Response Analyzer) capability (SIMPLIS, SpecteRF, PLECS, PSIM and NL5 come to mind, but there may be others as well).  This capability allows the simulator to directly produce Bode plots and loop gain analyses for switched circuits such as switched-mode power supplies.  SpectreRF obviously is tailored for RF type circuitry and as such it provides six types of periodic analyses: PSS (Periodic Steady State), PAC (Periodic AC, aka FRA), PXF (Periodic Transfer Function), Pnoise (Periodic Noise), Pdisto (Periodic Distortion) and an Envelope Analysis.

I've read that these can be easily applied to switched mode power supply type circuits.  Is this actually true?  Does this work as well as with some of the piecewise linear sequential topology simulators mentioned above?  How does the SpectreRF method compare to that used by SIMPLIS?  The simulators mentioned above carry a per seat cost of $500 to $10k+.  For reference, what does a full single seat of SpectreRF cost today?
Back to top
 
 
View Profile   IP Logged
analogspiceman
Junior Member
**
Offline



Posts: 23

Re: The History of SPICE
Reply #8 - Aug 25th, 2013, 3:32pm
 
The application of small-signal frequency-domain analysis to periodically switched piecewise-linear systems presents tremendous challenges when confined to traditional Modified-Nodal-Analysis Newton-Raphson methods.  Some years ago I made a FRA completely within LTspice using its special a-devices and behavioral functions.  About that time, Mike Engelhardt also added some FRA examples to LTspice that use post processing (.meas statements) to complete the analysis and plot the results.  Mike writes about this at length in the FAQ section of LTspice's Help file topic ("How to get a Bode Plot from a SMPS").  He argues that it is not worth the trouble because it is generally not needed in order to be able to compensate a design using an LTC IC because most of them use current mode control (this rationalization is a bit of a "cop out" in my opinion).  Both of the above LTspice FRA approaches (mine and Mike's) are painfully slow and suffer numerical dynamic range noise problems.

What most interests me about SIMetrix is that its SIMPLIS module seems to have a well refined built-in time domain FRA (frequency response analyzer).  LTspice has some of the hooks required to build its own FRA, but it looks like SIMPLIS has done it from the ground up, adding any and all the taps into the basic simulation software needed to make the process seamless and the graphical output very well behaved and presented.

The results look very believable - especially at and beyond the switching frequency, very much like actual measured output from an HP4194A Gain-Phase Analyzer (the SIMPLIS DFT windowing must be well optimized and perhaps some sort of graphics smart filtering has been employed to eliminate the alignment noise between the switching frequency and the FRA DFT sample frequency points).

SIMPLIS simplifies diodes, transistors and other nonlinear elements into piecewise linear approximations (for maximum simulation speed, as few as two pieces may be specified, just on and off).  It seems SIMPLIS may further speed things up by a posteriori predicting the endpoints in time between which it "connects the dots" with straight lines.  If so, this is something LTspice cannot do.  LTspice must back up after it notices a state change at a switching edge in order to find the time point of the edge.  Although LTspice does this very efficiently, it will necessarily calculate both a number of rejected data points and excessively closely spaced points before it gets back up to speed to zoom to the next switching edge.

SIMPLIS claims a speed up of 10x to 50x over standard SPICE.  From personal experience, I would estimate that LTspice can run SMPS simulations 2x to 5x times faster than standard SPICE when using LTspice optimized models.  Based on these rather loose estimations or run times, this gives SIMPLIS definite speed advantage, but this may be at the cost of far less realistic waveforms at the switching edges (which would be just fine for certain applications, such as frequency response analysis).

When further adding in the speed up from the way SIMPLIS runs its FRA (minimizing unnecessary FRA overhead during the time domain simulation portion of the analysis - the final FRA calculations are a post process), I wouldn't be surprised if the SIMPLIS FRA runs >1000x faster than LTspice's add-on time domain FRA methods that don't have all the optimum software "hooks" available.
Back to top
 
 
View Profile   IP Logged
Ken Kundert
Global Moderator
*****
Offline



Posts: 2384
Silicon Valley
Re: The History of SPICE
Reply #9 - Aug 25th, 2013, 10:32pm
 
I quite surprised that you characterize the IC simulators as dinosaurs. I look at them and think they can be so much better, and am completely frustrated at how slowly they are improving, but I always felt most of the big changes in simulation came out of the IC world.

I don't actually know that much about Simplis, but from what I understand it uses highly idealized models. Spectre, of course, is a full transistor level simulator. Thus, SpectreRF performs a detailed transistor level RF analysis, whereas Simplis is faster but much more approximate.  SpectreRF is certainly capable of simulating switching power regulators, and it has the PSTB analysis which should provide the best accuracy available for determining the stability for switching regulators. However, it can be challenging to apply SpectreRF on an abstract representation of switching regulators because it can be tough to build RF-compatible behavioral models.

-Ken
Back to top
 
 
View Profile WWW   IP Logged
Frank Wiedmann
Community Fellow
*****
Offline



Posts: 677
Munich, Germany
Re: The History of SPICE
Reply #10 - Aug 26th, 2013, 1:12am
 
analogspiceman wrote on Aug 25th, 2013, 2:42pm:
SpectreRF obviously is tailored for RF type circuitry and as such it provides six types of periodic analyses: PSS (Periodic Steady State), PAC (Periodic AC, aka FRA), PXF (Periodic Transfer Function), Pnoise (Periodic Noise), Pdisto (Periodic Distortion) and an Envelope Analysis.

I've read that these can be easily applied to switched mode power supply type circuits.  Is this actually true?

I would say yes (see http://www.cadence.com/community/blogs/rf/archive/2009/04/22/spectre-rf-by-any-o...). You also forgot PSTB (Periodic Stability Analysis) in your list (as already mentioned by Ken). You were probably looking at a pretty old document that was written when PSTB did not exist yet (Pdisto was renamed QPSS a long time ago).
Back to top
 
« Last Edit: Aug 26th, 2013, 4:54am by Frank Wiedmann »  
View Profile WWW   IP Logged
analogspiceman
Junior Member
**
Offline



Posts: 23

Re: The History of SPICE
Reply #11 - Aug 26th, 2013, 10:09am
 
The behemoth simulators are not dinosaurs because of performance.  It is their old school revenue paradigm that is relegating them to a relatively tiny niche market of users.  LTspice has over 3 million copies distributed.  LTspice has a web based bug report system and web based updates.  Incremental revisions are released dozens of time a year or more.  When a true bug is reported, it is often fixed immediately with a revised update available within hours.  

Most new EE students will probably be cutting their simulation teeth on LTspice (this is a role that PSpice used to fill).  Start ups and smaller companies generally are restricted by budgets such that the behemoth simulators are not an option.  And the behemoths generally don't like to fill this niche because they must charge enough per seat sold in order to support their business (this creates a small niche positive feedback loop).

I am not dissing the performance of the behemoths (in fact, I lust after some of the features they offer).  I just wonder if the old school marketing strategy will eventually relegate them to the dustbin.  If Mike Engelhardt decides to add in some of the features now missing from LTspice that could be very bad for the behemoths.  Personally, I would love to see a sophisticated simulation house team up with one the other major semiconductor vendors and provide LTC and LTspice with some healthy competition.
Back to top
 
« Last Edit: Aug 26th, 2013, 1:11pm by analogspiceman »  
View Profile   IP Logged
analogspiceman
Junior Member
**
Offline



Posts: 23

Re: The History of SPICE
Reply #12 - Aug 26th, 2013, 1:34pm
 
Hi Frank,

Yes, I know next to nothing about SpecteRF as I have never used it and only know the very little I have picked up from various IEEE papers (mostly on power electronics) that I have read over the years.  Lately I have specifically gone looking for information about SpecteRF.  By the way, do you know what a single user, full seat costs?

It seems that Ken has done a lot of groundbreaking work advancing the state of the art in electronics simulation.  Other simulators offer POP/PSS and transient domain FRA, but they do so by simplifying the circuit into a succession of endpoint continuous linear topologies.  Apparently SpectreRF employs other techniques to achieve its speed (however relatively fast that is, I really have no idea).

It would be very interesting to run a loop gain analysis of a benchmark switched mode power supply in all the various interesting simulation platforms - the free and low cost ones versus the behemoths.  Of course, to be fair, each simulation should be optimized by an expert user for each software platform.  They all should be run on the same PC, probably I7 based would be typical.  Output quality and run times would be of prime interest, with ease of set up being a distant second (even cumbersome user interfaces become transparent after enough hours "in the cockpit").

I would very much like to convince Mike E to add POP/PSS and FRA to LTspice.  Perhaps the best ploy would be to appeal to his pride in maintaining LTspice as the very best power supply simulator.  It seems that SIMPLIS et al provide some very useful features that LTspice is lacking in that regard.  If within the LTspice community a buzz to this effect started to grow, I don't think Mike could resist not responding. Smiley
Back to top
 
 
View Profile   IP Logged
Frank Wiedmann
Community Fellow
*****
Offline



Posts: 677
Munich, Germany
Re: The History of SPICE
Reply #13 - Aug 26th, 2013, 1:58pm
 
analogspiceman wrote on Aug 26th, 2013, 1:34pm:
By the way, do you know what a single user, full seat costs?

I could find out but I don't think that I am allowed to pass on this information. It's probably best if you ask Cadence for a quote. I would guess, however, that SIMPLIS is much better suited for your needs.
Back to top
 
 
View Profile WWW   IP Logged
analogspiceman
Junior Member
**
Offline



Posts: 23

Re: The History of SPICE
Reply #14 - Aug 26th, 2013, 3:37pm
 
I am not really looking to buy anything as LTspice works more than well enough for almost all I ask of it.  The lack of POP/PSS and FRA/PAC is not a problem for me personally because I am very adept at creating and comfortable with using averaged models on the spot when needed, but this remains a serious deficiency in LTspice for the typical user, in my opinion.  For the very few occasions when I need an exact simulation, it seems that the free, crippled version of SIMPLIS likely would handle most situations - or else I might just buy a personal copy of NL5 if it works well enough (haven't tried it yet).  One of my coworkers has PSIM and I probably could get my own copy of that (or of SIMPLIS) if I requested it, but I would much rather have something I could take home for personal use and that produced files that were useably sharable with others.

But my interest here on this forum is more in learning about the latest and greatest in simulation technology.  Perhaps the behemoths (anything costing from $5k/per seat up) really are able to produce and offer performance worth the price.  (Personally affordable software would be anything from free to costing up to about $500, in my opinion).  In order to evaluate value (performance versus price), it would be very interesting to run the simulation benchmark contest I mentioned in a prior post.  

Attention Benchmark Volunteers!

I wonder if anyone would be willing to help with this?  If not, I might download several trial version from the various simulator vendors in order to run such a benchmark - but the problem is I might not be expert enough in all the packages to run a fair comparison.  Maybe to keep things simple, it would be best to run the benchmark on a basic buck converter with an input filter.  A power factor correcting boost design also might be interesting because it involves both line frequencies and switching frequencies.  I also have a simple, but very unique power supply topology of my own in mind that might make it interesting for those running the tests.
Back to top
 
 
View Profile   IP Logged
Pages: 1 2 
Send Topic Print
Copyright 2002-2024 Designer’s Guide Consulting, Inc. Designer’s Guide® is a registered trademark of Designer’s Guide Consulting, Inc. All rights reserved. Send comments or questions to editor@designers-guide.org. Consider submitting a paper or model.