I am not sure about the exact answer, but I can tell you this: ConductionFiniteDifference march through time, solving one timestep after another, relying on the solution found for the previous and/or following instant. This 'and/or' will depend on the solution scheme used. (I guess this is controlled with *difference scheme* in the **HeatBalanceSettings:ConductionFiniteDifference object in EPlus**)

For example, you start at t=0, solve an equation and find the values for t=1. Then, use t=1 to find values of t=2.

These methods usually involve a critical timestep that assures convergence (if timestep is bigger than that, your solution will explode in plusHUGE or minusHUGE numbers). Fulfilling this criteria, however, does not assure accuracy, so it is a common practice to divide the critical timestep by some constant, and use that as the timestep in your solution. This constant, I think, has something to do with the **space discretization constant** in the **HeatBalanceSettings:ConductionFiniteDifference object in EPlus**.

**CRITICAL TIMESTEP IS SMALLER IF YOUR MATERIALS ARE LIGHT** (it is actually a relationship between thermal mass and conductivity --> thermal diffusivity)

Finally, (and here I am guessing) if you have some "light" materials (i.e. plywood, gypsum boards) the critical timestep might be smaller than what you asked for EVEN IF THE MATERIAL IS NOT SUBDIVIDED, thus forcing EnergyPlus to ignore the input option and making it 3 minutes or so.

I would love to know if I am correct.