To simulate or not to simulate, that is the question

Sep 30, 2014

To simulate, or not to simulate, that is the question-
Whether it is wiser in the Engineer’s mind to suffer
The ridicules of colleagues and wrath of management,
Or to hasten accuracy and time to market,
And by using SPICE, end all doubt?

The use, or overuse, of SPICE simulation is an age old controversy. Many engineers rely heavily on simulation and do relatively little hand calculation. Other engineers have distain for the simulator and only solve problems by hand calculations and measurements. Personally, I think the balanced approach is the best approach. I like to try to confirm circuits with calculation, simulation, and measurement. If you can get agreement of all three methods you can have more confidence in your solution.

It is common for new engineers to wonder why they shouldn’t just rely on simulation. In fact, several common pitfalls should be considered when simulating. The first pitfall is that SPICE will often give an answer whether the circuit is connected correctly or not. For example, some simulators may not notice if you reverse the amplifier’s power supply connections. Running a noise analysis with the supplies reversed will give incorrect results because the amplifier is not biased. However, the result may seem believable, leading the engineer to the wrong conclusion.

In general, it is easy to avoid this pitfall if you run several different simulations to confirm the circuit operation. For example, do an ac sweep and calculate a dc operating point before you run the noise analysis to confirm that the circuit is operating correctly. More importantly, perhaps, you should do a hand calculation to confirm that the simulation results are “in the ball park” or close to expectations.

Another common simulation pitfall is that your key parameter for analysis isn’t necessarily covered by the simulation model. For example, if you are doing a noise analysis, the model needs to include noise behavior. There are two approaches to overcoming this pitfall. First, open the netlist of the model. Many model netlists show a list of all the parameters covered (see Figure 1 ). Another approach is to develop a very simple test circuit to see if the model covers the parameter in question. For example, configure the amplifier in a buffer configuration to test the spectral density.

Figure 1: Check to see what is modeled

Another common issue with simulation is numerical convergence. Sometimes the simulator refuses to give you an answer because the internal matrix mathematics exceeds the software’s ability. This is a pretty complex subject as there are many tricks that you can do which will help the software converge. Note that you can find many web resources covering SPICE convergence problems. Furthermore, the solutions may be dependent on which SPICE simulator you are using.

Considering all the various pitfalls of simulation, one might wonder why we should simulate at all. In fact, SPICE simulation can enable you to solve many problems that would be very difficult to solve with hand analysis. For example op amp stability analysis depends heavily on the open loop output impedance. Real devices have complex open loop output impedance (see Figure 2) that makes hand calculations extremely complex. In practice most engineers will do hand analysis by approximating the output impedance as a constant then use SPICE to get a more accurate result. Also, SPICE models cover a wide range of parameters which you may not have considered in your hand analysis. SPICE also provides many useful tools like Monte Carlo analysis and parameter stepping that help you ensure that your design is robust. Finally, it is quick and easy to do experiments in SPICE. You can use SPICE to quickly run a wide range of tests then use hand calculation to confirm some key tests.