-Discovery is essentially what you suggested: observing a phenomenon to get research.
The key is examination tools or blocks to measure and record the discovery, from something as simple as a book and quill to cameras to advance stationary spectrometer. Using these give knowledge about items or blocks which decreases the score needed for relevant research, gives a bonus inspiration check, or just give some nice trivia.
This way, the game only needs to check your discovery when you use the tool on something, the player has a pretty good control on what he discovers, and the tools can be used to tier the progression with research and material cost.
The thaumometer and the deconstruction table are a perfect example- scan things to get research points, specific things unlock certain researches, and put items in the machine to get research points.
-An experiment is a structure, process or crafting that have an assigned research, and gives you more advance knowledge for performing, in the same way as discoveries. It is an expansion on discovery: rather than examining things in the world, the player has to create the circumstances.
Simple experiments can be using a book and quill on water while having a bucket, while more advance experiments might require making a multiblock structure, such as a redstone circuit.
As a continuation of discovery, experiments must be activated using a specific tool or machine.
Even if you make a particle accelerator, most of the requirements are complete dummies- using the tool/activating the machines is what really checks for the structure, needed research, other tools in your inventory, consumes materials, and make the thing run, by either a light show or forming a real multiblock machine. Again, this is so the checks for the experiments only happen when needed, and so any structure, phenomenon or crafting could be used as an experiment, without the need to change the blocks themselves.
Experiments give a cost of time and effort to advance in research, engage the player and give him control over his progress (since he can do experiments that lower the score requirements for research he wants).
You could make it so players who already know what they're doing can make the experiments and get it the moment it's research score is below 1 (which decreases it's score to 0).
As the player advance, he"ll have less and less things to easily discover in the world and more and more experiments to do: Think of thaumcraft- in the beginnings, you scan everything, but later you really have to scrounge for things to scans. This is, I personally think, one of the very few things thaumcraft doesn't do well.
-Prototyping is a sub-type of experiment. While experiments mainly meant to advance the player's research, prototyping is meant to give a very specific technological advancement, though they are not exclusive. Prototype require having their specific research, unlike experiments which only need to have their research score below 1.
Rather than making a unique phenomenon and examining it, the player has to craft, build, or use a prototype- a basic, crude, and inefficient version of the technology using tools such as sticks, hammers, or welders to be able to realize his research into actual technologies- everything from basic gears, which are easily crafted, to complex machines.
Besides giving another way to make research cost and engaging the player, prototyping creates more methods to exclude and share knowledge. Unlike regular experiments, which require more basic knowledge to do and only have a specific research to help, players can't make a prototype without it's research- the finalized tech is a black box for anyone else. On the other hand, a working prototype can be examined, giving it's specific research (if it's possible for you to unlock), whether made by another player or naturally generated in the world. This ties to the earlier idea of a museum- players on a server can make a museum with discoveries, experiments and prototypes to allow other players to quickly gain that research.