LTSpice ......the only fast simulator for PWM controller IC's ?

[eem2am]

Regarding circuit simulators capable of simulating SMPS circuits...........

Is LTSpice the only circuit simulator which can simulate circuits with PWM Controller IC's?
(-albeit the range of PWM Controller ICs available in LTSpice is limited to the linear.com range of controllers)

For example.......
I am under the impression that supposing i designed a SMPS circuit with say an Infineon ICE2QSO3G Qusai-resonant flyback controller, ......
-then i would not be able to simulate this circuit in, say, Simetrix, for example.

...since it would not be possible to get a model for the ICE2QSO3G into Simetrix?

.....i appreciate that simetrix could be used to synthesize a model for ICE2QSO3G, -but the simulation that included this model would run too slowly?

ICE2QSO3G DATASHEET:
https://www.infineon.com/dgdl/Datas...d24f8&fileId=db3a304324fc7f9a01250689233a7046


So basically, its my understanding that they guy who invented LTSpice, had a special software algorithm which allows LTSpice to simulate
linear.com PWM controller IC's *quickly*

Its also my understanding that no other simulator in the world can simulate circuits containing PWM Controller ICs with anywhere near the simulation speed of LTSpice?

..is this true?

 

[keith1200rs]

There is nothing special about LTspice. It is not fast. Linear Technology don't provide their models in a form that can be used with anything else and most other manufacturers don't provide models of power supply chips at all.

Keith

 

[FvM]

The trick in achieving acceptable simulation speed with complex systems, e.g. PWM switchers, is to replace transistor level by simplified behavioral models. Unfortunately, some circuit properties also get lost when changing to this kind of functional simulation. As far as I have seen, the Linear models offer a reasonable compromise. But a designer should be aware of the model's limitations.

I'm under the impression, that other semiconductor companies mostly focus on convenient design tools, that promise to design a switcher circuit when you enter a set of parameters.

 

[eem2am]

Keith1200rs:

You say there is nothing special about LTSpice.....but your subsequent words make me think that there *is* something special about LTSpice....because you say that most manufacturers dont provide PWM chip models for simulation at all.

..at least linear.com does provide these models for its chips

It is this very thing, the availability of a variety of PWM chips in LTSpice, which makes LTSpice so wonderful in simulation.......because a PWM chip is valuable to have in a simulation.....even if its not the control chip one will eventually use...........one can set it up to work in an equivalent topology and check out capacitor ripple currents etc.

Your posts above seem to confirm to me that the ICE2QSO3, for example, is not available anywhere in the world as a time-domain, transient simulation .model.

....i sense that one would have to do a behavioural simulation of it....and this is not as good as the linear.com PWM chip models that are available.

It is earthmovingly significant, that in these days of high technology, and wonderful computing power , that we are not able to do transient type simulations of most PWM Controller IC's that are in the market...........unless its a linear.com IC

 

[keith1200rs]

FvM has already given a good explanation of why LTspice PWM simulations are quick. It is something which can be done with any Spice simulator. Again, FvM has explained that other manufacturers prefer to provide other solutions to the design problem. If you to prefer to think that LTspice is something special or wonderful then that is up to you. I know it isn't!

Keith

 

[eem2am]

I think LTSpice is wonderful.....its just unfortunate that it only contains linear.com PWM Controller IC's.

It really would be brilliant if ti.com, onsemi.com and national.com etc etc would follow suite with a similar simulator for their PWM Controller IC's.

I have worked in a number of SMPS companies, and with a lot of SMPS engineers far senior to myself.

Each and every one of them would always accompany any design with a transient simulation that was as near to their design as possible.
-unfortunatley they couldn't be exact with eg Infineons ICE2QSO3, since there is no model for it....and an SMPS engineer cannot be expected to have the software and math skills to be able to synthesize a model that would run sufficiently quickly.

I am definetly not alone in my belief that having a transient simulation , as near as possible to what you are designing, is about the best way to design any SMPS circuit....-if nothing else , it acts as a check on your calculated values.

It is a great shame that the guy who sorted out and designed the LTSpice program can't sell his idea to ti.com, onsemi.com national.com etc etc, so that they can do their own simulator.

 

[andre_luis]

The main problem at SMPS simulations is the magnitude of values involved.
It have infinitesimal increments added to larger values.
So, it happens loss of the precision.

I think that an trustable simulator to SMPS should work with float point numbers.
In fact, time processing would increase too much, but output result would become reliable.

+++

 

[keith1200rs]

Spice simulators all use floating point anyway. All Linear Technology have done is modify a version of Spice so they can use behavioural models which are not in a simple text format to prevent other people from examining their models. Another way of doing the same thing is to use encrypted Hspice models like Fairchild and a few other manufacturers are now doing. The difference there is that no-one is giving away Hspice (although I assume encrypted Hspice models can be used on Cadence, but I am not sure). I have some models of SMPS chips which are unencrypted and run fine on SIMetrix and I have behavioural models I have created myself which run very quickly. The problem with creating you own models is the datasheet doesn't give sufficient information about the whole of the IC internals. That is the whole point of encrypting the models - to hide things about the IC workings. It is the IC design that may have clever ideas LT want to hide. The simulator is just designed to allow that to stay hidden to protect their IP.

Keith

 

[eem2am]

Thankyou Keith1200rs,

From your last post i gather that SIMetrix has the ability to take in PWM Controller IC .models and use them in transient analyses.

...of course, as you seem to imply, the problem is that other than linear.com, virtually no semiconductor companies actually produce .models of their PWM Controller IC's for transient analyses....

...i have no problem with linear.com encrypting their .models, since after all, they provide a free simulator.

-After many enquiries, SIMetrix is the only other realistic simulator for SMPS..............but costs £3000.
(the free version of SIMetrix is too limited to simulate SMPS circuits containing PWM Controller IC's)

-Even SIMetrix does not contain the LTSpice "superb" integration method which is called the "modified trap"..........
-the "modified trap" being , apparently the best simulator algorithm for SMPS on the earth today.

I am still amazed that most semiconductor companies (eg ti.com, national, onsemi etc etc) do not provide .models of their PWM Controller IC's for use with a version of Spice which they could synthesize and offer-up themselves, (so that prying eyes could not gather the secrets of their control chips)

I was based at a large Electronics Company in northern UK.....and their were many superb electronic engineers there.......but none of them were able to synthesize their own .models of PWM Controller ICs (eg for offline flyback).

-Their most brilliant engineer (and this guy was seriously clever) managed to come up with something, but it ran ridiculously slowly, and was not worthwhile.

 

[andre_luis]

Hi,

I found a paper, covering this subject.
**broken link removed**

They present an overview to most critical issues in simulation :
> Loss of convergence due to truncation error.
> Device modeling