I've just started circuit analysis. All the examples, I've seen so far involve using using KCL/KVL or nodal analysis. They all consist of setting up a system of equations and solving them which gives you all the voltages and currents.

But systems of equations can have no solutions, infinite solutions or one solution. It's impossible for the system of equations to have no solutions in this case, because current must be flowing through the circuit.

But how do we know that the system of equations (assuming you have n equations for n variables) won't have infinite solutions?


But how do we know that the system of equations (assuming you have n equations for n variables) won't have infinite solutions?

You try to solve it and see. It's not hard to construct circuits that aren't solvable, in some ways or all ways. Perhaps they never reach equilibrium, so can't have a DC solution. Perhaps they contain contradictory constraints. Here are just a few examples:


simulate this circuit – Schematic created using CircuitLab

Of course, when you try to actually build these circuits, something happens, so there is some solution. The issue is that you build the circuit with real components, while the model assumes ideal components. The model doesn't include, for example, the resistance of the wires in the circuit, or the internal resistance of the current and voltage sources, or the leakage current or capacitance to Earth. For a model to have a solution that reflects reality, the model must include all significant parameters of the real circuit.

For very many circuits, the circuit can be solved with the usual techniques of nodal analysis and such if we are careful to include everything in our model that's significant. For example, if we model the series resistance of L1, that circuit becomes solvable.

There are a smaller number of circuits that remain unsolvable. These circuits depend on random processes, or processes so complex that we can only feasibly model them as probabilities. For example, consider the electronic detector that kills (or doesn't kill) Schrödinger's cat. Or for a less exciting example:


simulate this circuit

This is a resistor, on your desk. There is some voltage, at each end of the resistor, relative to Earth. There is some current flowing in it, too, and these things can be measured. Unfortunately, they are mostly due to the Brownian motion of the charges inside the resistor. That is, mostly random noise. We have models to characterize that noise, like Johnson–Nyquist noise. Once you model the noise, you can model (with nodal analysis) its effects on the rest of the circuit. However, the fact will remain that you still have a solution that is some set of probabilities.

Fortunately for us engineers, circuits that do random things or are strongly influenced by unintentional effects (capacitive coupling to surroundings, for example) are generally not useful. We also try to make our circuits relatively immune to parameters subject to wide variation in temperature or manufacturing. When these normal practices are taken into account, the number of circuits you encounter that can't be reasonably solved with nodal analysis are quite few. Nodal analysis is, after all, a simplification more complex, more fundamental physics, and it's our job as engineers to assure these simplifications are valid at each step.

  • \$\begingroup\$ The last scenario is one that many circuit solvers won't accept, though I'm not quite sure why they shouldn't be able to adapt to solvable form (use a common 2uF cap, and divide the current in that cap between the two connecting wires). \$\endgroup\$ – supercat Mar 18 '14 at 15:24
  • \$\begingroup\$ @supercat: The reason you can't solve the final circuit is there is no DC path to the top of the capacitors so it could be literally any DC value. Real capacitors have some leakage so we know the voltage at that node will eventually be zero but if you don't model the leakage there are infinite solutions. \$\endgroup\$ – Warren Hill Mar 18 '14 at 15:52
  • \$\begingroup\$ @WarrenHill: Gotcha. A single cap with nothing connected or an LC with no resistance would have the same issue (any initial conditions could work, but none of them would converge a unique state). I though the problem you were trying to illustrate with two directly-connected caps was that if the initial conditions aren't specified as both caps being equal, the current between them could be infinite (your other diagrams would cause either unbounded current to flow, or unbounded voltage to accumulate). \$\endgroup\$ – supercat Mar 18 '14 at 15:56
  • \$\begingroup\$ @supercat well, that is a problem also. I was only trying to demonstrate that there are problems, and the capacitors is a good one because the problems are more subtle than the other examples. \$\endgroup\$ – Phil Frost Mar 18 '14 at 19:08
  • \$\begingroup\$ Regarding the specific case of parallel capacitors, see Can a charge-pump be 100% efficient, given ideal components? and connecting two capacitors in parallel \$\endgroup\$ – Phil Frost Mar 18 '14 at 19:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.