The simulation I’m talking about is the use of computers to numerically model the behaviour of a system, particularly an electronic or electromechanical one. Typically this is done using a tool called Spice, although often MATLAB, FreeMat or even a spreadsheet can be used effectively. This method allows detailed exploration of system performance over a wide range of conditions prior to building a prototype. It is especially useful to test conditions that occur infrequently or ones you can’t easily replicate with the available test equipment.
The problem comes when you forget about the real world and assume the simulation is absolutely correct. Simulation models are frequently wrong, incomplete or inappropriate for your particular purpose, whether you made them yourself or got them from your part library or even a manufacturer’s website or datasheet. For instance, the models rarely take into account actual physical limitations: you can get valid looking simulation results from a situation which would vaporize parts in a microsecond. Similarly, the system you modeled might ignore some part’s non-ideality or a parasitic signal path that occurs in reality that critically affects performance. I simulated a circuit recently and toyed with what I thought could be a tremendous improvement, only to realize that it required completely impractical matching of two components.
The trick is to always think like you’re breadboarding. Imagine what you would do in the lab and what you might see. Pay special attention to the operating parameters that you simulate and how the parts would behave in those conditions, especially with respect to their absolute maximums. Simulation complements real lab experimentation: model it, make it and repeat as necessary.
A situation I always simulate is to validate the choice of switchmode power supply inductors. If the inductor is poorly chosen the circuit will smoke. I find that the typical equation-based methods for determining the peak current requirements often are optimistic. A quick simulation regularly shows that under certain conditions the current will go just a bit higher. Simulation is also very useful in developing feedback systems. Even in medical devices it’s not unusual for a control system to have hundreds of Watts at its disposal. To rush to experimentation in such a system is likely to result in damaged parts or even injuries.
The key though is to always go to the lab and build stuff. Even if you have simulated a system and think you’ve cracked a particularly tricky nut, you will still learn more when you really do power it up. I find frequently that I missed some key part or characteristic of a system once I actually solder it together and start scoping. In that case I go back to the simulation, make the change and then return to the lab. Simulation allows me to quickly test out a ton of circumstances to ensure the system will behave and then I double check these in the lab with real testing.
With a careful balance of simulation and hands-on lab work, I save a bunch of time and prototyping cost and end up with much more confidence in my designs.