Now that Analog Devices models are included in the standard LTspice libraries I was curious to check out a few opamp models. The AD8031 model is interesting in that it can output a negative voltage even when having only a positive supply . Have you come across any other oddities?

 

The AD8031 model is interesting in that it can output a negative voltage even when having only a positive supply .

That doesn't surprise me-- not even a little bit.

Have you come across any other oddities?

Far too many to count, especially with op amps as these are usually represented in Spice with behavioral models (macromodels) that may do a fair job of acting more or less like the real thing, but only over a limited range of "typical" application circuits: try to do anything "clever" or "odd" with them, and the model behavior can depart markedly from that of the physical device or even do bizarre things such as what you cited. (And some op amp models are so crude or bug-ridden they do bizarre things no matter what.)

I've been using Spice for 25+ years now, and anymore I view simulation results with a healthy dose of skepticism: although they often mimic real circuit behavior fairly closely, the incidence of "weirdness" is high enough that I usually cross-check them against conventional circuit analysis, measurements taken from actual circuits, long-time experience, and especially common sense.

People who expect Spice to replicate actual circuit performance in minute detail are setting themselves up for a lot of wasted time and head-scratching.

</rant>

 

Agreed. Spice is a useful tool but one needs to be aware of its limitations.

 

Hmmm...

You just emphasized the importance of testing simulation models before using them..

eT

 

I wonder if there would be a market for an op amp that could do that. I recently saw a headphone amp IC that had a negative dc-dc converter built in so that it could output negative voltage without needing a negative supply. Assuming there were no design tradeoffs or compromises () it would be pretty handy to have that in an op amp.

I know when I was even more of a noob than I am now, I was thwarted several times when I realized that my brilliantly simple, clever solution to a problem would require a negative supply that would've been enough extra work/cost that it made the whole idea worthless. An op amp that took care of its own negative voltage needs would've made my day!

 

Now that Analog Devices models are included in the standard LTspice libraries I was curious to check out a few opamp models. The AD8031 model is interesting in that it can output a negative voltage even when having only a positive supply . Have you come across any other oddities?

Hi,

Yeah as others have said, the models are not exact and sometimes the differences will show very unreal behavior.

One of the most interesting differences is their lack of stochastic behavior. Every model of the same part number we use in a simulation will result in an exactly similar response, right down to the nanosecond, nanovolt, etc. This results in circuits that wont be perfectly symmetrical in real life will always be perfect in simulation.
What this tells us is that if we expect to be able to determine how a circuit will behave in real life vs simulation, is that we have to know what might change and make those changes ourselves in the model(s) being used in order to get a better idea what might happen in real life.
This of course means for a transistor for example we might vary the Beta a little in order to see if the circuit still works. If we vary it in an amount that mimics real life then we get more useful results.
Another example is to change the input offset of an op amp so we can see how the real life op amp might cause a problem if we get a package that has a worse than typical input offset.

So that's the bottom line then. We have to make some variations of parameter manually in order to be more sure a circuit will work. Also, most circuits must be tried in real life but there are some that absolutely must be tried such as power circuits and of course RF circuits.

Interestingly, there are secondary issues that come up too such as EMI which i dont think any spice versions support yet.