The howstuff works ain't gonna get you there. There are two turbines, one compresses, one expands. Each one is depended on the other. One makes work (the expander hot side) and the other consumes the work (the cool side air compressor). To perform the calculations is an iterative process to match work generated to work consumed. The work consume by the compressor is just the polytropic compression where you have pounds of air per minute at some pressure ratio or head at an effiency (the air temperature is another input and you can also solve for the temperature of the air coming out). Use the equations for polytropic compression and assume 2 to 1 ratio (14.65 psi boost) and a 68% effiency and how many pounds of air you need at that pressure to fill the displacement of the engine at some RPM. You solve for wor or HP required. For a 1 liter engine this might be 10HP.
Next, you know you put in some number of pounds of air plus about 14 to air to fuel ratio, so you know how many pounds of exhaust you have. So at this point you know the puonds of exhaust and the pressure the expander is going to let the exhaust down to, so you will have to iterate what was the pressure of the exhaust entering the expander to gererate the same amount of work the air compressure used. You'll need to assume some properties of the exhaust, like its temperature and composition. You know its going to be about 65% N2, 25 % CO2 and 10% water. The expander will run about 65% efficient. (close enough eh?). So now you've matched the expander with the compressor. If you make lots of calculations you can draw a map of each side and match everything up. You can also calculate how cool the exhaust is coming out of the expander.
If you have access to process similators that handle all the ethalphies, entropies, heat and material balnces it will go much quicker.