Blind simulation without understanding

Yes :-( ...Jim Thompson

That sounds more like laziness to me. Maybe they do not want to put in the time to learn and truly understand a circuit.

But there were some disturbing trends even all the way back at my university in the early 80's. A professor said, literally "Well you have to learn all this analog circuit level stuff for the exam but in a few years you'll never need it again". What a fool that was. But he was professor so most students believed that nonsense. I didn't :-)

Spice has that effect on some people. It happens here, too.

John

Back in the late 60's, I had ONE, count 'em, ONE prof who I thought knew what he was doing. He was an engineer for TI. Teaching was some kind of outreach program to teach/acquire new engineers for TI. He taught a practical approach to engineering with the math reserved for fine tuning what you had designed based on your UNDERSTANDING of the problem and the devices used in the solution.

I had a thermodynamics prof who's approach to teaching was to discuss a concept at the first part of the class then spend the bulk of the time painstakingly working thru solving the differential equations with us. That was the ONLY class where I thought I understood how to do the math involved in a practical context.

All the rest of the profs had a teaching style that could be described as RTFM.

A couple of years later when I was hiring new grads, I experienced the same results. High GPA students who could solve equations, but had no idea how to decide which equations to write.

I was lucky enough to land a job with a mentor. I learned more from him in a few months than I did in 5 years of school.

Mentors are rare. If you don't get one, you're not likely to be able to mentor your charges. Don't get me started on management and the Peter Principle...

So, the more computing power became available to engineering students, the stupider they got. I remember standing in line to access one of the few calculators the university had for our use. Good times.....

Simulation is a great thing for prototyping your design. It's a lousy substitute for a synthesis process.

Simulation is a great deal cheaper than building bread-boards. Lab space is expensive, and has to be supervised. Students learning to solder burn themselves from time to time, which leads to acrimonious complaints from parents and guardians.

Virtual breadboards are much less trouble.

Beginners are always going to play with circuits that they don't fully understand - that's one of the ways that one develops understanding. Part of my job as a senior electronic engineer was to field questions about circuits that weren't doing what the more junior engineers expected them to do - some of them were remarkably easy to answer, which did arouse some resentment.

I once got got a question about a circuit that was going wrong, with the error going one way - and I came up with a plausible explanation - only to be confronted with the same junior engineer telling me that he had got the sign wrong and it was going the other way, for which I came up with an equally plausible (different) explanation, after which the junior engineer involved went around claiming that I didn't know what was actually going wrong because I could explain anything (which wasn't entirely fair).

I did point out that I was just giving him hypotheses to test, rather than laying out how the universe worked, but he still thought that he had scored a point.

-- Bill Sloman, Nijmegen

And you learn a great deal less. Building stuff and doing your own math is the way to wizardhood. Poking simulations will _not_ get anyone there.

I use Spice a fair amount these days--a lot more than I used to--but that's primarily to give clients a warm fuzzy feeling that the final circuit will do what I say it will. It's also very useful as a sanity check and for optimization, once a circuit topology and most of the component choices have been made on other grounds.

Real circuit design is a mental exercise, and mostly doesn't require computers.

Cheers

Phil Hobbs

-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics

hobbs at electrooptical dot net

Absolutamente.

I (and a client) had our SPICE comeuppance almost right after I started self-employment for the first time. Ultrasound front end boards were deteriorating. Some channels would become "weaker" and it kind of depended on how many hours each machine had been running and at which pulse power level. Since this was, of course, not a tube design but all semiconductors they could not explain it. They were at wits end.

Probing around the unit I couldn't see anything suspicious. Then I put part for the front end on PSPICE. That was always an adventure back in the days of DOS. Lo and behold this showed a spike of such short duration that even really expensive scopes from that era could not capture it. From there it was only minutes until I had figured out that there was some race conditon in their logic circuitry. The guys were highly skeptical about my findings. All this voodoo and witchcraft and whatnot, and then blaming it on the digital part. The chief digital dude was a guy similar to Hulk Hogan so I had to be careful :-) ... But, they decided to fix this race condition. The number of front end boards that degraded in the reworked machines dropped to ... zero.

But it does sometimes require good beer :-)

I find it's often easier to brainstorm a circuit, hash it out on paper to get the implementation correct, then do the layout, and finally, tweak the simulation until it produces realistic results.

Simulation is only a design tool in as much as verifying circuit performance, based on educated guesses of real world parameters; it will never assist you with circuit synthesis. Especially to the untrained student, it can be downright dangerous because real transistors explode, most often from circuit elements you didn't realize are there.

Tim

Hogwash. When we went to engineering school, we learned the behavior of parts at the basic physical and mathematical level. Then we combined Rs and Ls and Cs into basic circuits and understood the math of that, still in the classroom. Then we had labs where we *verified* the math with real parts.

Resentment? About Sloman's kindly guidance? Who woulda thunk it?

John

It's the new trend of Social Engineering. Take someone else's work and expect it to work for you.... Then take credit.

I read some where it is common for those labeled 'Generation X'

Cheers

Wanna hear something worse? Ok. At the Vaughan College of Aeronautics near LaGuardia airport, they used to have labs with engines and stuff. Now they have computers in every room to practice aircraft maintenance in simulation.

Which falls down because no mathematical model is either perfect or perfectly comprehensive. This approach is exactly what Spice embodies, and you have to do enough work with real components to know when Spice is going to mislead you, and why.

Sometimes the errors can be rather subtle. If you simulate the Baxandall class-D oscillator with bipolar transistors and use an extra- large inductor to feed the centre-tap of the tank circuit inductor/ transformer to minimise the harmonic content of the output, the circuit eventually settles down to produce a constant output amplitude sine-wave signal.

In real life, such a circuit "squegs" and never settles down into a stable state. Apparently this doesn't happen if you build the circuit with MOSFET switches.

The problem with the simulation seems to be in the Gummel-Poon model of the transistor which doesn't model the behaviour of the inverted bipolar transistor entirely correctly.

LTSpice also offers the VBIC transistor model with "Weak avalanche and Base-emitter breakdown model:Improved Early Effect modelling; Physical separation of Ic and Ib" which might work better, but no transistor manufacturer will release VBIC model data - I tried to get example data from a guy who works for NXP and is a member of my trim-hockey team, but apparently NXP regards this kind of information as a serious industrial secret.

It wasn't all that kindly. If you don't mildly discourage that kind of question you can spend all day answering them. And you can't use that technique to discourage management, who can be very thick-skinned. In consequence I tended to stay a work for an hour or two after most people had gone home to dinner to get time when I could work uninterrupted. Since my wife had adopted the work-late strategy long before we got around to getting married it didn't disrupt our domestic life.

-- Bill Sloman, Nijmegen

There's nothing new about it. A great deal of research has always been "me too" studies that follow up original breakthroughs. Nobody in their right mind re-invents the wheel - except people who don't read the literature.

-- Bill Sloman, Nijmegen

Obviously. Mathematical models are always over-simplifications of reality, and you have to test them against reality from time to time.

re.

They certainly won't get you all the way there, but the wizard is just somebody who has decent mathematical models inside their own head, and simulations can help to build these models.

But computers can do the arithmetic a lot faster than our mental machinery. We are much better at pattern recognition than computers, and consequently much faster at coming up with configurations of parts that may do what we want, but computers are a handy tool for exploring what a particular configuration of parts might do in reality - if mathematical models of the parts do a good job in the region we are trying to explore.

-- Bill Sloman, Nijmegen

If you're merely claiming that computers can be useful, I agree. If you're claiming that spending more time on computers and less time doing algebra and bench work produces wizardly engineers faster, better, cheaper, or whatever, I disagree 100%.

I especially disagree that a wizard is merely somebody with better mathematical models in his head. Where on earth did you get that idea?

Cheers

Phil Hobbs

-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics

hobbs at electrooptical dot net

If you're merely claiming that computers can be useful, I agree. If you're claiming that spending more time on computers and less time doing algebra and bench work produces wizardly engineers faster, better, cheaper, or whatever, I disagree 100%.

I especially disagree that a wizard is merely somebody with better mathematical models in his head. Where on earth did you get that idea?

Cheers

Phil Hobbs

-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics

hobbs at electrooptical dot net

If you're merely claiming that computers can be useful, I agree. If you're claiming that spending more time on computers and less time doing algebra and bench work produces wizardly engineers faster, better, cheaper, or whatever, I disagree 100%.

I especially disagree that a wizard is merely somebody with better mathematical models in his head. Where on earth did you get that idea?

Cheers

Phil Hobbs

-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics

hobbs at electrooptical dot net

there.

ity

u're

and

er, I

More time spent on computers certainly translates into a cheaper education, and in so far as you can test an idea faster on a computer than you can by soldering stuff together, it can speed up the process.I did go to the trouble of pointing out that " Mathematical models are always over-simplifications of reality, and you have to test them against reality from time to time" and any adequate education has to include at least some of that testing - ideally enough to allow the students to enjoy the experience of seeing the mathematical model fall over.

You have just set up a straw man which everybody with any sense can see is 100% wrong, but it doesn't have all that much to do with what I was saying.

atical

It is certainly one aspect of wizardry. Knowing about lots of different ways of doing lots of different jobs is actually more important, but that's actually maintaining a lot of not-particularly- detailed mathematical models too, at the level of "a transistor base has zero input impedance and always sits about 0.7V away from the emitter and the collector current is a fixed multiple of the current going into the base".

-- Bill Sloman, Nijmegen

I see more questions that involve a misunderstanding of the simulator's function, the graphical user interface, and the models 'provided'.

Lack of understanding of circuit function is universal to SW and hardware implimentations - that's why the 'student' going through either exercise.

A dumb question is usually just a dumb question, but not always. That's why its important to ask them.

RL