Agree that theory and hands on are the best approach, maybe not in that order tho
Recall an incredibly complex system we were developing back in 80s, so advanced that we had to develop out own test equipment! We had created a new type clocking driver for some CCD chips we had developed, these chips were not imagers but real time mathematical convolvers that formed part of a Fourier Transform called "Chirped Z Transform". Our goal was to do spectral analysis, equivalent to 20,000 point 16 bit FFTs, plus a bunch of proprietary stuff, in real time. The equivalent computing power was 1/3 of Cray 1 Supercomputer but our system was the size of a small textbook and consumed 3 watts total!! The clock circuit was based upon a reactive clock scheme that utilized the CCD capacitance in resonance with an external inductance for energy exchange. This was all worked out theoretically and worked beautifully (much later (~10 years) the concept was patented by NASA, but we never brought prior art up since our sponsor was the USG, not a good idea to tick off the Wookey
) While doing some hands on measurements in the lab I knocked off one of the power supply wires, the system should have stopped but it didn't, and the total system power consumption dropped
After checking everything out I went back to my desk and developed the theory behind what had just transpired. The reactive clock driver circuit was a unique circuit that forced itself into the lowest energy consumption, compensating for power supply and component variations, including temperature and even producing one of the supply voltages it required from the higher supply voltage. We kept this circuit as a Trade Secret until the NASA patent.Anyway, that was a case where hands on preceeded some theory.
Todays chip design is almost totally based upon theory, since measurements with such small devices and features, as well as the amount of devices involved, prevents traditional hands on type experimentation. Even if you could make hands on measurements of a test board the results would be questionable since the parasitic elements in the chip and test board are so much different. Add to this the cost of a chip design and you can see why this is a highly specialized field where a mistake or inferior design can cost many Millions $. Very stressful indeed!!
Think about the work of James Maxwell, or Albert Einstein, or Issac Newton, they had no means to do hands on. Maxwell's 4 fundamental equations describing all EM phenomena, Einstein's Relativity & Gravitional Waves, Newtons laws, and so many more concrete contributions to human understanding, all done without any hands on....amazing indeed!!!
Best,