What Analog Circuit Simulation Software is Best?

[SebG]

Hi.
I'm interested to know what Analog or Mixed Signal Circuit Simulation software is most popular to simulate general analog circuits such as those with transistors, op-amps and other active and passive elements.  What types of analysis does it support? Is it free or low cost? Does it have a intuitive GUI and is it easy to use?

[T3sl4co1l]

Anything SPICE that's between $0-1000 is pretty much the same.  And components in much more expensive packages (Altium uses the same damn XSPICE that everyone else does).

SwitcherCAD (?) or SuperSpice may be some of the more interesting ones, but I've never used them.  Moderately priced I think (~$500?).

Free, LTSpice.  Intuitive, easy, no.

Moderately priced, plus SCH/PCB?  Multisim.  Intuitive, moderately priced (as these things go... which is still $1-2k).

Tim

[brabus]

One of the best quotes from my Mentor: "Remember, YOU have to master the simulator, don't let the simulator master you!"

I ended up using MATLAB (or Scilab!) for most of the things; in the end it's just a way to speed up calculations that one can do by hand, especially loop compensation and stuff like that.

I believe that simulators begin to be very useful when it comes to RF stuff: finite element modeling, field distribution, parasitics estimation from the 3D design, etc. Awesome stuff, but we're talking about something very different from SPICE.

For any other use, I quote Teslacoil: pretty any simulator does the job. My favourite? LTspice of course!

[nfmax]

Another MATLAB user here. Recently I had to modify an existing L-C AAF design to work with a different cutoff frequency, and re-optimize for zero-overshoot transient response. I used a basic design from Williams' book and modeled it using the nodal admittance matrix technique, written from scratch in MATLAB over the course of the morning. I even remembered the technique from those university lectures 30+ years ago! (I was really interested in the technique at the time & wrote my own simulator in BASIC - which helped it to stick). Easy to do things like Monte Carlo exploration on the effect of tolerances on the transient response, and quicker than learning the UI on yet another package. Why are they ALWAYS so awful?

[katzohki]

I like the free version of Orcad Capture CIS Lite.

[liquibyte]

I've been having really good success lately with LTSpice IV thanks to the help of members here.  I'll have to agree that it's not that intuitive but there's a ton of information out there (and here) for it.  Runs under WINE nicely too.

[free_electron]

the most commonly used simulator for 'real' work is probably ELDO. That thing is maintained , gets updates, and has a massive backing. Thousands of integrated circuits have been designed on this simulator and they work right.

ELDO combined with the technology libs gives an almost 100% exacting result as the real deal.

drawback : price. you can't buy it. It is leased per user, per year and needs a farm to run properly.

[ovnr]

I prefer LTspice - I don't even find it difficult to deal with. There's some learning curve, but it's also faster to get going than, say, Multisim.

Also, when I do end up using Multisim, I get super annoyed that you can't even simulate an inductor with real-world characteristics without adding them as discrete components. And a lot of things do not work as expected if you simulate them with ideal components vs real-world ones!


So at the end of the day I sim things in LTspice, do schematic layouts in Multisim, and boards in Ultiboard. It works well enough for me.

[janengelbrecht]

I Prefer Labcenter Proteus/ISIS and NI Multisim.
Sometimes I use LTSpice also.

Very expensive stuff though:
http://www.labcenter.com/products/platinum.cfm
http://sine.ni.com/nips/cds/view/p/lang/da/nid/203438

Well company pays...and in the end the State pick up half the price (Tax reductions )

[T3sl4co1l]

Also, when I do end up using Multisim, I get super annoyed that you can't even simulate an inductor with real-world characteristics without adding them as discrete components. And a lot of things do not work as expected if you simulate them with ideal components vs real-world ones!

Well, there are some ferrite bead components in there, which still are only modeled as R + RLC circuits (roughly correct, but not very accurate at any particular frequency).

FYI, note that models returned from e.g. KEMET Spice are the exact same thing: a lumped equivalent model valid only at one frequency, and only grossly similar at others.

I don't know of any other packages or libraries with accurate inductor data (if such things exist, I would be interested!).  If you mean in comparison to LTSpice allowing the arbitrary specification of parasitics (which I don't like*), that can be abused in all sorts of ways that aren't at all apparent from the schematic.  And anyway, with a little SPICE code, you can build the same thing in Multisim -- a .SUBCKT with PARAMs is one thing, but they can be one step even more convenient with active / dialog controls (take the non-ideal transformer component, for instance).

*They're hidden away in a dialog box, so unless you take the time to download the .ASC and inspect every line, you don't have a clue if any of the R, L, C or V/I sources are anywhere near what they claim to be.  For simulation purposes, I would much rather see the explicit components, or at least an "arbitrary impedance box" with the parameters labeled on the schematic.

As for actually really accurate inductor models, they're not easy to make, and that's given actual data and specifications from the manufacturer.  Take this example,
http://seventransistorlabs.com/Images/CurveFit1.png
http://seventransistorlabs.com/Images/CurveFit2.png
The impedance isn't a constant inductance, it kinks to about f^0.5, probably a "skin effect" effect.  The rolloff corresponds to winding capacitance, but the resonances could be better modeled, if it were necessary that is.  And this is just common mode; who knows what differential or normal modes look like.

Easily 95% of all inductors have no published data whatsoever (beyond DCR / temp rise, and Isat if applicable), making the whole procedure orders of magnitude more expensive (test equipment required, etc.).

Tim

[ampdoctor]

Not sure what exactly you're looking to do with the simulator, but for ease of use I'd google circuitmaker 2000 and dig up a copy somebody has archived somewhere. It's a dinosaur but it's also easier than falling off a log and you can run ac dc, transient, monte carlo analysis, parameter and temperature sweeps and a ton of others. It's got a huge though somewhat dated parts library and includes many of the "jellybean" components you're likely to run into even by modern standards.

If you're looking to investigate and understand analog basics then this is it because you wont spend all your time getting frustrated fighting the simulator.

[free_electron]

Also, when I do end up using Multisim, I get super annoyed that you can't even simulate an inductor with real-world characteristics without adding them as discrete components. And a lot of things do not work as expected if you simulate them with ideal components vs real-world ones!

try doing mutual inductance, or a loaded transformer for some real fun... or a motor with permanent magnet rotor ...

there are tools that can do that. the name escapes me right now...

[G0HZU]

For use in a linear simulator you can obtain small signal inductor or transformer models in the frequency domain quickly and easily using a full 2 port (or n port) VNA. Often this type of inductor model is included in the relevant manufacturer's library that gets bundled with the high end RF simulation tools anyway. eg there will be models like this for every inductor in their range.

[VEGETA]

I've used Proteus ISIS for quite some time, well, it is good in general. However, it does bad stuff sometimes especially for specific chips (like LT1968 I wanted to buy and couldn't), but general stuff and even more, it is good.

I tried LTSpice but due to 0 experience in it, not much time to learn it properly... I don't think I can or should give an opinion.

Regarding MATLAB, this is the software that I really had 0 intent or purpose to learn! I may be wrong but this is how I feel. I am a fresh graduate of Mechatronics engineering, I like PIC MCU and want to dig deep in the field... maybe product design in the future If I was fortunate enough.

Now, why do I need MATLAB? what to do with it? or better yet, someone like Dave who is a pro engineer... We used it in college to simulate Electrical drive circuits and it was good, but what about the over 9000 other fields within one electrical/electronics/mechatronics engineering major?

We spent a lot of time in college using it and learning it, but if it is up to me, I would have removed it and put some PCB or product development courses/labs instead. It is just my opinion which can be wrong but I am open.

I see people even buy that PIC MCU simulation blocks! there is Arduino ones too! SERIOUSLY, who uses MATLAB to develop MCU projects?! And NO, "develop projects with no programming" or "bring your project to life faster and better" will not be a convincing answer to me.

How about we suggest this topic in "Suggestions" forum? << xD

[janengelbrecht]

MATLAB is nice too... Im into MathCAD instead though
But essentially to model things with your own equations seems to be the right way if you are a student.....when youre a pro....time is of essense...you just want quick results...SPICE it is then