Author Topic: Best SPICE software?  (Read 47067 times)

0 Members and 1 Guest are viewing this topic.

Offline TimNJTopic starter

  • Super Contributor
  • ***
  • Posts: 1654
  • Country: us
Best SPICE software?
« on: May 26, 2014, 10:05:55 pm »
I try to do most of my simulation with real hardwired circuits...but sometimes if I need to do it really quickly, I'll use LTSPICE.

However, maybe I'm just not very well versed in LTSPICE, but sometimes I get lost and frustrated with it. Maybe this is true of all circuit simulation programs.

Do I just need to learn LTSPICE better or is there something more intuitive/better out there, that also won't cost me one of my limbs or vital organs?
 

Online T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 21638
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Best SPICE software?
« Reply #1 on: May 26, 2014, 11:16:18 pm »
All the middle to low end simulators I have ever seen are based on either Berkeley 3f5 or XSPICE.  This includes LTSpice (with proprietary mods like multi-core support and apparently some convergence stuff), Multisim, PSpice (with added functions and syntax), Altium (some PSpice support) and many others.

Supposedly there are super mega freaky simulators out there which can timestep subcircuits independently, which I should think would be excellent for speed and convergence, but I have seen very little discussion of them; probably the ultra expensive ASIC simulators have it... so of course they're keeping it on the DL. ::)

For the average design engineer, it seems one is stuck with the basics.  Suck it up, I guess. :(

For usability, Multisim isn't bad for just poking around with components; the schematic capture is reasonably easy to use.  Of course, it's kind of expensive to just get on a whim.  The lower end dedicated simulators (where the only purpose of the schematic is to generate a netlist, nothing leading to PCB design) like Beige Bag, or NGSpice, and LTSpice of course, are cheaper but have varying usability in the various aspects (schematic capture, component properties and models, simulator settings, convergence issues...).

Tim
« Last Edit: May 26, 2014, 11:21:19 pm by T3sl4co1l »
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline Marco

  • Super Contributor
  • ***
  • Posts: 6704
  • Country: nl
Re: Best SPICE software?
« Reply #2 on: May 26, 2014, 11:54:47 pm »
IME it's just way too easy to get into trouble with convergence with Spice (QUCS too). It seems you need an indepth knowledge of the underlying math to even get it to simulate in the first place.

Simetrix is much better in that regard (they have a demo for circuits with limited number of nodes).
 

Offline TimNJTopic starter

  • Super Contributor
  • ***
  • Posts: 1654
  • Country: us
Re: Best SPICE software?
« Reply #3 on: May 26, 2014, 11:57:52 pm »
Hmmph well alright. I guess I should either A.) Learn LTSPICE better and/or B.) Use a breadboard
 

Offline IanB

  • Super Contributor
  • ***
  • Posts: 11831
  • Country: us
Re: Best SPICE software?
« Reply #4 on: May 27, 2014, 12:42:05 am »
When you go to engineering school the first thing you learn is how to predict what your design will do by calculation before you build it. For professional engineers everywhere the best way to predict what a design will do is to run a computer simulation. So in theory, you should always put anything non-trivial through LTspice or some other simulator before you breadboard it.

That said, I have spent my whole career working with and designing computer simulation programs. One thing I have learned is that simulation is hard. Every engineer out there wants a simulation tool that "just works". And every engineer out there is doomed to be disappointed. Simulation tools cannot read your mind, and so they need some degree of precise instruction about what to calculate and how.

I am afraid, therefore, that the best advice is to learn more about how to use LTspice to get the best results from it. There is no magic to be obtained by jumping to a different program. You would still need to learn how to provide precise input instructions to the program to get the output you are looking for.
 

Offline TerminalJack505

  • Super Contributor
  • ***
  • Posts: 1310
  • Country: 00
Re: Best SPICE software?
« Reply #5 on: May 27, 2014, 01:00:35 am »
You might try Tina-TI.  I like its user interface better than LTspice's.

It doesn't seem to have as many components as LTspice but you can import standard spice models into it so you can usually find what you need out on the net.

I've attached an example of what it can do.  It's from a circuit that I'm playing around right now.  The circuit is playing around with me as well since I'm struggling to get it to be stable under all conditions.   :D

Edit: Had a duplicate image posted.
« Last Edit: May 27, 2014, 02:05:08 am by TerminalJack505 »
 

Online T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 21638
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Best SPICE software?
« Reply #6 on: May 27, 2014, 03:48:15 am »
That said, I have spent my whole career working with and designing computer simulation programs. One thing I have learned is that simulation is hard. Every engineer out there wants a simulation tool that "just works". And every engineer out there is doomed to be disappointed. Simulation tools cannot read your mind, and so they need some degree of precise instruction about what to calculate and how.

Expanding on that:

- Having only done numerical solutions a few times (simple stuff like polynomials, or guessing the constant for a transcendental (of the form x = e^x) equation), I can safely agree, simulation is hard.  That SPICE is able to achieve a fraction of what it does is impressive.

Yes, it's still a PITA.  But that's what you have to deal with.

- There is no such thing as a perfect representation of reality.  DO NOT allow yourself to think of the SPICE circuit as fact (and whether it works to your expectations or not, or converges at all!).  When you are working with a computational model, first and foremost you are working with a computational model, period.  It is your responsibility, as the thinking entity driving the simulator, to come up with a model that's reasonably descriptive of the particular aspect of reality you're trying to figure out.  It is very easy to make an "impossible" model.  It is also very easy to "burn up" components willy nilly, without a clue what's going on.  You must watch circuit parameters vigilantly, install current and power meter objects (or watch the power dissipation variable, if the simulator produces it -- often, .MODELs have power, but .SUBCKTs do not, and a lot of transistors come as .SUBCKTs so you have to add the power meter yourself).

The breadboard is always truth, because it exists in reality.  But it takes a long time to change things around, and you certainly can't sweep a parameter (component value, supply, temperature, etc.), let alone find a hundred ways to blow a transistor and one way not to.  Let alone if the transistor is $200 each, or something.  You'll be at it forever.  These are good reasons why we use SPICE.  Other good reasons are design verification: assuming you can implement the model in a representative fashion without tearing up the schematic too much, you can have a valid simulation, final schematics, and PCB-ready netlist all in sync.  (And for packages that integrate PCB more tightly, like Altium, you have even better than just schematic-to-PCB-via-netlist linkage.)

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Tac Eht Xilef

  • Guest
Re: Best SPICE software?
« Reply #7 on: May 27, 2014, 05:30:10 am »
There is no such thing as a perfect representation of reality.  DO NOT allow yourself to think of the SPICE circuit as fact (and whether it works to your expectations or not, or converges at all!).  When you are working with a computational model, first and foremost you are working with a computational model, period.

This is a very good point. As something of a statistical modeller IRL I keep having to remind students that (a) "modelling" is a continous process of refinement, with (b) several broad types/steps (conceptual, logical, simulation, etc) that are (c) part of an ever-ongoing iterative refinement process before they can even hope to be called 'representative'

And no representative model is ever accurate, or right, or even complete. A "good" model can only ever be "good enough" - it's up to the user to know enough about the model and the domain it's modelling to know whether it truly is "good enough". And even if you decide it is, you've still got to test your modelled results against reality.

As a really simple example: SPICE will happily let you put 100,000v across a spiral carbon resistor, or switch a 100W light bulb with a BC549, and will even simulate a perfectly logical result. Most implementations, however, won't tell you what'll actually happen when you do...
« Last Edit: May 27, 2014, 05:31:46 am by Tac Eht Xilef »
 

Offline Bored@Work

  • Super Contributor
  • ***
  • Posts: 3932
  • Country: 00
Re: Best SPICE software?
« Reply #8 on: May 27, 2014, 06:10:11 am »
The circuit is playing around with me as well since I'm struggling to get it to be stable under all conditions.

And after you get it stable in spice and after you really build it, reality will teach you a lesson. Just because you have something stable in spice doesn't mean it is stable in reality. E.g. directly driving a capacitive load is a no-no for most op-amps (I didn't check the OPA140).
I delete PMs unread. If you have something to say, say it in public.
For all else: Profile->[Modify Profile]Buddies/Ignore List->Edit Ignore List
 

Offline ivaylo

  • Frequent Contributor
  • **
  • Posts: 661
  • Country: us
Re: Best SPICE software?
« Reply #9 on: May 27, 2014, 06:52:30 am »
Get the book - http://www.we-online.com/web/en/electronic_components/produkte_pb/fachbuecher/ltspice_iv_simulator.php (careful to select English if this is what you want). If you really want to learn it, worth every penny.
 

Offline John Coloccia

  • Super Contributor
  • ***
  • Posts: 1208
  • Country: us
Re: Best SPICE software?
« Reply #10 on: May 27, 2014, 07:02:33 am »
Another vote for Tina.  I've found it to be a lot more useable than LTSpice.
 

Offline free_electron

  • Super Contributor
  • ***
  • Posts: 8515
  • Country: us
    • SiliconValleyGarage
Re: Best SPICE software?
« Reply #11 on: May 27, 2014, 07:07:23 am »
By far , the gold standard is ELDO. But unreachable unless you are a gigantic semiconductor maker.

Simulatos stand or fall with the wuality of the models. Most of them abaolutely suck in that respect.
Professional Electron Wrangler.
Any comments, or points of view expressed, are my own and not endorsed , induced or compensated by my employer(s).
 

Offline LvW

  • Frequent Contributor
  • **
  • Posts: 282
  • Country: de
Re: Best SPICE software?
« Reply #12 on: May 27, 2014, 07:32:24 am »
A lot of good information and recommendations .....my add another one?

Don`t blindly trust simulation results. It happens rather often that in your simulation arrangement
* you forget to ground a particular node, or
* you forgot to allocate a suitable value to one of the power supplies, or
* that you - erroneously - have more than one single ac source switched on, or
* that the two opamp inputs are interchanged, or............

Hence, my advice: You always should have a rough knowledge how the "exact" simulation result should look like.   
 

Online T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 21638
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Best SPICE software?
« Reply #13 on: May 27, 2014, 08:57:54 am »
* you forget to ground a particular node, or

This is one that can sometimes help: it's rarely a good idea in actual practice to leave nodes floating.  SPICE specifically forbids it -- it will generate a "singular matrix" error, which is to say, it can't find a solution for some set of nodes (e.g., series capacitors) or loops (parallel inductors) because there is an excess degree of freedom in the equations (the voltage on the floating node(s), or the current in the loop, could be anything, so when it tries to find *one* solution, it fails).

But this, too, can be gummed up.  Typically one sets RSHUNT = 1e9 or so, which is usually enough to "bleed" floating nodes towards ground.  But the simulation crawls slowly (the matrix is near-singular, forcing it to do extra work) and you don't know why, until you spot the mistake.  As usual, there is no substitute for a vigilant eye.

In general, a more physically realistic circuit is also better to simulate (besides producing more realistic results).  Wires have inductance.  Nodes have capacitance.  Inductors have resistance (series and parallel -- both dependent on frequency of course..).  Capacitors have ESR and ESL.  It can be helpful, for instance, to take a diode model (typically, manufacturers provide just the .MODEL for diodes, which only represents a certain ideal approximation of the die alone) and pack it in a SUBCKT with realistic package parasitics (I usually go for series lead inductance, in parallel with a damping resistance, and a small capacitor across the diode to aid convergence -- wide swings in circuit parameters, like CJO, make for slower simulations too).

Your overall model is only as good as the sub-models in it; scrutinize them well.  One of the most notorious types are the three-terminal op-amps -- without supply rails to work from, they simply produce an output corresponding to the input stimulus, no clipping whatsoever.  40kV from a "LM741" seems strange, but it's quite logical for a three-terminal model.  Such models may not include slew rate limiting either, so beware, not just of nonlinearities, but also their potential absence -- which, of course, unless you're familiar with the device, you won't know offhand whether it's right or wrong.  Confusing business for a beginner.

One thing I've had trouble with is getting good switching behavior out of a MOSFET.  Some models consider Cdss, some don't.  The problem is, when the MOSFET turns off, that capacitance has to be charged, and how fast it charges strongly controls how much power is dissipated in the transistor.  So, you can set up a toy simulation where the MOSFET switches a current source; when it turns off, drain voltage rises proportionally to the capacitance.  Often, they'll be lazy and put in a constant capacitance.  This is pretty awful when you consider the datasheet says the capacitance varies from 10nF from under 10V, to less than 500pF at 100V -- a huge difference!  Or they'll miss breakdown voltage, so over time, the voltage just keeps on rising happily -- 10kV on a 500V device, anyone?  Some of these are easy to fix by hand (add BV and IBV parameters to the body diode model), others not so much (good luck fitting some of the more extreme capacitance curves).  Sometimes the datasheets even lie -- I have one example of an ST transistor which I measured the Cdss of; the datasheet graph is significantly steeper than the real article.

On the other hand, some manufacturers go to great lengths to provide accuracy; I believe it was an Infineon model I once tried, which was, sadly, completely useless to me -- it ran too slow to get numbers out of!  I don't doubt the accuracy of their model, but there is ultimately a speed-accuracy tradeoff which one must consider in the process.  And that includes how broad a range a model covers.  This particular example, I went with a similar Vishay/IR device (which uses a Level 3 MOSFET model; average quality) and did the boring usual engineering process: overestimate losses as a worst-case, and optimize to that, so that the real thing probably works much better than, but hopefully no worse than, the model.

Things that -- alas -- you can do on paper anyway, with the understanding that, hey, you're doing it on paper, so of course the numbers are going to suck -- so you overestimate by a moderate factor, and it probably works first time.  One would hope simulators can avail us of such process, but that is not the case.

Tim
« Last Edit: May 27, 2014, 09:09:26 am by T3sl4co1l »
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline GK

  • Super Contributor
  • ***
  • Posts: 2607
  • Country: au
Re: Best SPICE software?
« Reply #14 on: May 27, 2014, 10:38:42 am »
I try to do most of my simulation with real hardwired circuits...but sometimes if I need to do it really quickly, I'll use LTSPICE.

However, maybe I'm just not very well versed in LTSPICE, but sometimes I get lost and frustrated with it. Maybe this is true of all circuit simulation programs.

Do I just need to learn LTSPICE better or is there something more intuitive/better out there, that also won't cost me one of my limbs or vital organs?


The "help" files in LTspice are comprehensive, though not quite tutorial material. However there is plenty of excellent third party material available if you search for it and even a dedicated LTspice users forum. You'd be using your time much more efficiently pursuing these avenues than starting a "SPICE" thread here, which is unlikely to evolve into much more than the typical anti/pro philosophical crap, that one can read on these forums ad nauseam already; replete with dummies thinking that they utter words profound by stating in 50,000 words or more that SPICE is only simulation. Like, duh. Just IMO, of course.



     
« Last Edit: May 27, 2014, 10:43:31 am by GK »
Bzzzzt. No longer care, over this forum shit.........ZZzzzzzzzzzzzzzzz
 

Offline Marco

  • Super Contributor
  • ***
  • Posts: 6704
  • Country: nl
Re: Best SPICE software?
« Reply #15 on: May 27, 2014, 02:09:27 pm »
By far , the gold standard is ELDO.

For ASIC level design I guess you mean? (I'm doubtful though, NXP for instance seems to use other simulators and they were the major industry partner in developing the current gold standard MOSFET model.) For board level design most models are released as HSPICE, so isn't that the gold standard as a direct result?
« Last Edit: May 27, 2014, 02:11:40 pm by Marco »
 

Offline eecook

  • Regular Contributor
  • *
  • Posts: 118
  • Country: ar
Re: Best SPICE software?
« Reply #16 on: May 27, 2014, 03:45:04 pm »
I personally recommend NL5. It is not SPICE (I like it better), but it works like a charm, is really intuitive and fast and the manual is very clear. You can get a student's license for 1 year for free.

Hope it helps,

Simon.
Nullius in verba
 

Offline Marco

  • Super Contributor
  • ***
  • Posts: 6704
  • Country: nl
Re: Best SPICE software?
« Reply #17 on: May 27, 2014, 06:27:07 pm »
I personally recommend NL5. It is not SPICE (I like it better), but it works like a charm, is really intuitive and fast and the manual is very clear. You can get a student's license for 1 year for free.
I've always been intrigued by this one, mainly because of the promise of making convergence problems go away.
 

Offline dentaku

  • Frequent Contributor
  • **
  • Posts: 881
  • Country: ca
Re: Best SPICE software?
« Reply #18 on: May 27, 2014, 07:28:00 pm »
I really like Circuit Wizard. It's lacking quick keyboard shortcuts like LTSpice but otherwise I really love the interface and the different views that show you current and voltages across our entire circuit in an intuitive way.
Like everything else it's missing some components but for basic stuff I prefer it over anything else I've used.
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 3014
  • Country: gb
Re: Best SPICE software?
« Reply #19 on: May 27, 2014, 08:23:33 pm »
I try to do most of my simulation with real hardwired circuits...but sometimes if I need to do it really quickly, I'll use LTSPICE.

However, maybe I'm just not very well versed in LTSPICE, but sometimes I get lost and frustrated with it. Maybe this is true of all circuit simulation programs.

Do I just need to learn LTSPICE better or is there something more intuitive/better out there, that also won't cost me one of my limbs or vital organs?
I'm another happy user of Simetrix and I'd certainly recommend it as 'worth a try' to see if you like it. The free (unlimited time) demo version is very generous on circuit size.

I've been using Simetrix at my place of work since the 1990s. However, I'm not an advanced Simetrix/SPICE user by any means. I do RF design so I usually use other simulation tools at work for RF stuff. But I find Simetrix to be extremely useful for any general tinkering about with analogue circuits. It has some impressive analysis tools built in and I don't know if LT SPICE can compete in this respect. (but maybe it can?)

Note: The GUI and circuit entry method in Simetrix hasn't really changed much in over 15 years (despite going from V2 to V7) so it will look a bit dated and uninspiring at first if you do try it. 

If you find that Simetrix is too dated or clunky then maybe you should go back to LT SPICE and try out a few tutorials because an awful lot of people use LT SPICE and are very happy with it :)
« Last Edit: May 27, 2014, 08:27:35 pm by G0HZU »
 

Offline rbola35618

  • Frequent Contributor
  • **
  • Posts: 298
  • Country: us
Re: Best SPICE software?
« Reply #20 on: May 28, 2014, 12:25:20 am »
I am a Topspice user.  It has a very large library close to 30k of components in its database and it convergences very well. I cost about 800 dollars and their technical support is top notch. I have been a costumer of Topspice for a very long time and I am very happy with it.

Robert
 

Offline TimNJTopic starter

  • Super Contributor
  • ***
  • Posts: 1654
  • Country: us
Re: Best SPICE software?
« Reply #21 on: May 28, 2014, 12:27:45 am »
Thanks guys. I'm quite aware of the limitations of SPICE software. Simulation is just what the name implies. It tries to simulate the real world, but of course that doesn't completeley happen. I guess I'll just go sit down and try to figure out LTSPICE a little better. Maybe I'll give Tina a try.
 

Offline IanB

  • Super Contributor
  • ***
  • Posts: 11831
  • Country: us
Re: Best SPICE software?
« Reply #22 on: May 28, 2014, 02:59:34 am »
With transient simulations in LTspice I often get the best results by starting external DC supply voltages at 0 V and skipping the initial operating point solution. This brings the simulation closer to a real life power on situation.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf