Again, it sounds like the fault of the operator, using a tool incorrectly.

You've never used software that had bugs? You've never used models that weren't correct?

 

I have never seen a model that is absolutely correct. As per wikipedia a model is a simplified and idealized representation of physical reality. If someone assumes that a simulated circuit using very low complexity models will behave the same as a physical circuit would then he is severly misguided.

Like someone in the forums (no idea who) often says about simulation, garbage in, garbage out. A person using a simulator needs to realize that the simulator only knows what is in the circuit and nothing about what is not in there but in reality would.

Back to the original question of the topic, simulator can be very much relied on, but only to the point of how much we can rely on the accuracy of the circuit that is being simulated. So I think the only people who rely excessively on simulators are those who have no clue what they´re doing.

 

"So I think the only people who rely excessively on simulators are those who have no clue what they´re doing." You must at some point have had garbage in because that is certainly garbage out. Simulators allow you to do calculations in an organised way and cover possibilities that analysis and excel would never cover. Back in the old days I might write a Fortran programme; today I use a simulator.

 

Probably a bad choice of words, by "excessively" I meant people who rely on simulators so much that it is detrimental to them, in the spirit of the thread title. What did you think I meant by that sentence?

 

Even with simulators, which can't be 100% comprehensive, mistakes will still escape detection.

1 (Pentium Floating Point Division Bug) x (Keep It From The Public For Six Months) = $475,000,000.

ak

 

In general, all ICs are simulated, but not with something as crude as LTSpice. The IC simulators include layout, connection resistance, feedthrough capacitance, etc.

Yes and no. There are certainly high end simulation packages that do all that -- and that cost serious money per seat per year. We designed our ICs, which were generally pushing the limits (not necessarily the same limits as "cutting edge" designs) of the processes and were mixed-signal full-custom designs, using PC-based tools in which all of the tools -- schematic capture, simulation, layout, and verification -- totaled about $7000 per seat (to own, not to lease). The tools were pretty basic (PSpice and HSpice, for instance) and they were not integrated so there was a lot of grunt work and a few in-house written utilities to get them to play nice with each other. We had to decide when and where parasitics were important and then we had to model them and we had to add them to the schematics manually. But as a result we knew the tools and what they could do and what they couldn't do and, more importantly, we had a much better understanding of the designs and the second and third order issues because WE had to deal with them. We often did designs that the people using $250,000/seat systems were sure were impossible -- but that was because there was no obvious way for them to do it with their tools which were optimized for a certain type of product and not the "lunatic fringe" type of designs we were noted for (which is why some of our customers were, in fact, those same companies). For instance, despite being "the company that invented the integrated circuit," a certain very large company came to us to design a product that they had dumped a few millions of dollars into over a few years and, for about fifty grand, we delivered a functional chip within a few months. They simply could not compete with us in that realm -- just as there would have been no way that we could have even thought about competing with them in theirs. So a very productive relationship began that, despite divisions getting split off and sold off, continues today.

 

I'm guessing the cost of the die justifies all that prior work... and also, that afterwards thorough testing is done in the real world and compared to the original simulation

Yes, the cost of the masks and prototype fabrication just keeps going up. As a result, the cost of spinning a design to get it right is so high that you rely heavily on simulation and you demand that the simulation results match the real world performance. So you demand that the models be damn good. For the IBM 130 nm process I did a design on about a decade ago each transistor model was actually a subcircuit containing about 300 devices. But it pays off when you find that the performance of the chip once it gets back is almost identical to what the simulations predicted. Our first "is it alive" test was always to see if the bias voltages where close to what the sims indicated and it was seldom that they were more than about 10 mV off and, if they were much more than that, it often indicated a problem.

 

1 (Pentium Floating Point Division Bug) x (Keep It From The Public For Six Months) = $475,000,000.

After Grove's initial response, Intel changed it's perspective and we came to consider that an advertising campaign.

In reality, the bug was really obscure. I kept using my system with a flawed processor and was never impacted. You were more likely to be affected by high energy particles causing a bit flip in memory which, at the time, was rarely error corrected.

The problem was caused by an engineer who changed one of the scripts used to generate the look up tables and hadn't verified it properly. After that, the engineers working on that portion of the chip became extremely reluctant to change anything.

 

You've never used software that had bugs?

Has your brain never made a mistake? Have your hands never missed nailing a plank? Have your feet never tripped you? ...

Does that mean you shouldn't use your brain / hands / feet?

Just common sense, isn't it?

You've never used models that weren't correct?

I would take it one step further:

1) every model I have ever used is wrong;
2) no rational person will ever use a model that's 100% correct.

Think about that and you may learn a thing a two about "models".

 

Has your brain never made a mistake? Have your hands never missed nailing a plank? Have your feet never tripped you? ...

Does that mean you shouldn't use your brain / hands / feet?

Just common sense, isn't it?

Getting a little personal now, are we?? I fail to see how your comment is relevant to this thread.

 

Those are just examples that help you think more vividly.

 

Those are just examples that help you think more vividly.

My cognitive skills are performing nominally. I design for intellectual stimulation and my primary use for simulators is to better understand things that I don't understand well enough to do on my own.

 

My cognitive skills are performing nominally.

I don't know what you meant by "normally".

If you meant that you have "humanly" cognitive skills, that means you make plenty of mistakes plenty of times. If making a mistake were a valid reason not to use a tool, or a body part, how would you have lived your life?

If your body has never failed you, well, you certainly live in a different world from the rest of us and whatever we have discussed doesn't apply to you.

 

my primary use for simulators is to better understand things that I don't understand well enough to do on my own.

That does not make anybody else's use of simulators, however different, less valid.

 

There's little point taking the discussion to ridiculous extremes. The TS pretty clearly was not asking whether every use of any simulator only makes people dumber. Similarly, people responding that they are causing people to become dumber are not claiming simulators should be banned or that they aren't a useful tool. Similarly, the "pro-simulator" crowd isn't claiming that it is impossible to misuse simulation software or the results they produce. As many of pointed out, simulators are a tool and any tool can be used for either good or bad, including by the same person. The question is really about whether simulators are being too misused by too many people.

 

That sounds like the fault of the operator, not the tool the operator is unable to use correctly.

So you agree then.
As you see, the simulator can not run itself, it is always a combination of user and simulator. Knowing how and when to use the simulator what the user has to know.

Any approach is going to be limited in one way or another, because no tools can ever be perfect.

No one or any tool can be perfect...stating the obvious...does it help?
On the other hand, theory is exact; perfect. That's how we can get exact answers to problems to prove or disprove ideas where it is hard to do with only numerical results. A formula from theory tells us a whole story that would take countless numbers to show.

And many times, you don't need to be "exact".

And many times, you do

You cannot be 100% comprehensive any any approach, simulator or not. The fact that a tool (simulator or a human brain, or actual circuit) isn't 100% accurate or comprehensive should be a reason not to use it.

I dont think anyone ever said not to use it.
Theory, can be 100 percent comprehensive, but simulators dont deal directly with pure theory it is part theory and part numerical. The human can however deal with theory as well as numbers and thus has an advantage.
We could also talk about circuits that are very difficult for simulators to run.

 

Probably a bad choice of words, by "excessively" I meant people who rely on simulators so much that it is detrimental to them, in the spirit of the thread title. What did you think I meant by that sentence?

..."those who have no clue what they´re doing" will be incapable of using a simulator, they are difficult to use and you do need more than a clue to get them to work. However for people learning electronics, simulators can be a useful "playground" in which to learn in parallel with making and testing circuits. In the professional world simulation is essential, I've worked on projects that simulate the electronics, the electro-magnetics, the system mechanics and the digital control.

 

Personally I have seen a lot of cases where someone just cobbles up a circuit in a simulator, puts in some random values and says great it works, but he doesn´t have enough experience to see that he is using an ideal opamp that exceeds its power supplies and a whole lot of ratings of components like mosfets etc. is way exceeded, and the circuit will blow up when first turned on.

 

Personally I have seen a lot of cases where someone just cobbles up a circuit in a simulator,

As long as we have humans, we will always have people making mistakes and not knowing what they are doing.

That's no reason to not use a particular tool.

 

As long as we have humans, we will always have people making mistakes and not knowing what they are doing.

That's no reason to not use a particular tool.

@dannyf
You are right. Everything you say is right. It appears there is no need for any of us to say anything else on this subject because you are so right. I agree that anyone who wants to add a comment is an idiot and there was no chance for them to view the question differently than your correct way. So, now that you know you are right, maybe you can sit back for a bit and watch how ridiculous the rest of the members are as they continue discussing this topic.