Object of this game is to keep just enough pressure on the valve seat with the valve face to keep a reliable seal without bouncing the valve against the valve seat above the absolute max engine RPM. Any more results in power loss in the cam drive system and numerous other areas where friction increased with loading (bearings and such).
Concept behind double springs, conical shaped springs, Beehive shaped springs, friction spring dampers and such are to reduce or control the self resonant frequency of the springs-valve mass, opening-closing forces.
Keep in mind when the bomb goes off inside the combustion chamber-cylinder that is going to be a LOT of force driving that valve hard against the valve seat to aid in sealing.
Each time the cam lobe begins contact with the valve follower is about the same as a hammer beginning to tap on the valve follower. This pounding action is some what similar to hitting a bell with numerous self-resonate frequencies and wave nodes.
Imagine trying to drive a vehicle with springs in the suspension but no dampers? Tires, driver and .... are NOT going to be happy at all.
The idea of increasing spring rate is to increase self resonate frequencies to prevent the valves and all related from bouncing. This sort of works, but there are other ways that help reduce this.
Increasing the spring rate of he suspension to where the whole chassis-suspension-driver and all chatters with very little suspension movement. The vehicle essentially hops and jumps over bumps rather than following the road contours and bumps in a controlled manner (tires have a spring rate that is significant once the chassis spring rates get obscene, Tires also have hysteresis and damping issues added to what is happening).
Lowering the overall mass of the moving valve train parts will shift this self resonant frequency up. This is why so many current race motors use titanium valves-valve locks-spring retainers, small diameter valve stems (helps to reduce intake port obstructions too) and such.
Using two different spring rates results in two different self resonant frequencies. As these two different self resonant frequencies work against each other, they loose energy fighting each other. This effectively helps to damp or control the system and goes a way to prevent the valve from bouncing against the valve seat.
The ramp rates on the cam lobes also makes a BIG difference in valve behavior. This was the very problem Keith Duckworth had to deal with when the first Cosworth V8 race motor was designed. Lots of messy math later, they had to go by trial and error along with all that messy math to get the cam ramp rates correct for reliable valve train operation at high RPM.
Cams also have a quiet side or as the valve is closing against the spring, this rate of change can also be controlled by the lobe of the cam. Good performance cams have non-symetrical lobe shapes to account for rate of change for the valve opening and rate of change for the valve as it is closing.
IMO, use the lowest possible valve spring rates that causes no problems at some margin beyond the max engine RPM. Double valve springs are proven to do fine in the Lampredi SOHC engine. Even stock as delivered, the valve train has little problems above 7,500 RPM constant if all parts involved are in good condition.
I'm not sure how effective conical valve springs might be, There is hard science behind the idea, it is the real world results that matter. This needs to be balanced against real world experience of many who have built lots of performanized motors that live fine above 8,000 PRM in endurance race duty.
If it works, why fix it?
BTW, when Renault introduced the pneumatic driven valve trains, more than a few believe the French were off their croissants and butter. Except, that idea worked and worked really well. This was part of what ushered in the 20,000 RPM Formula One engine era due to the ability to control valve train self resonant frequencies, valve operating time and all that related stuff.
Bernice