So I setup a very quick example in the lab last night,
* CLASS D Automotive amplifier.
* Regulated DC power supply. ( Almost completely ripple free )
* Non regulated Automotive power simulator ( Rectified 3Phase AC to DC )
* 1/3F Non pulse ratted Electrolytic capacitor
* Signal generator
The capacitor was diode isolated so the current going to charge the cap only could be measured.
NOTE: the cap will only discharge when the supply voltage is pulled below what is stored in the cap, and the cap will only charge when the supply voltage is greater than what is on the cap.
This means you will only see a current draw ( green) when the cap is 're charging'
The red voltage scale is the the fluctuation from the 14.4v baseline.
NOTE: the red Voltage line has been inverted to allow better measurement between spikes. This means as the voltage is pulled down the red line will go UP.
So first up here is the best case scenario, this is a 14.4VDC Regulated power supply. Almost perfect ripple free with very close to zero noise on the line. ( nothing like you would find in an automotive environment.)
See how the cap will only charge when the amp is not pulling power, this means a nice smooth load on the alternator. and if you add the both together, little to no ripple will be generated.
This is the 'ideal' lab world where almost everyone does there formulas and 'theory' So in this environment yes the cap would work great!
The real world is full of noise.
Here is my un regulated 14.4VDC automotive power simulator.
Look how the cap is following along with the amp draw.
Also look how much more current is being pulled by the cap. the reason for this? non filtered rectified 3Phase is not pure DC it is pulsed DC, and the cap is now trying to smooth out the power supply as well as power the amp.
In this setup the cap is now acting as an additional load in the system.
for the above to work you would have to know on a micro second to microsecond basis what the load on the system is as well as the current voltage and resistance, and they dynamically vary the capacitance, to keep things on track like the above Regulated system.
I hope this helps illustrate why this is such a heated debate.