This depends on what you are actually trying to study through the DOE framework , if you are interested in only screening for variables significantly affecting your response from those not, 2 level full factorial/fractional factorial experiments are adequate, however if you are looking forward to probe the response surface deeply you need to go for 2nd order or higher designs where each factor is studied at more than two levels, if you are trying to search for optimal performance settings of your process you need to couple such surface modeling techniques with suitable search direction choice and updating algorithms , these techniques come under RSM(Response Surface Methodology),however if you are trying to find set of process parameters that will make your process robust with respect to.noise variable levels fluctations, you can go with Taguchi's robust parameter design methodology.
This depends on what you are actually trying to study through the DOE framework , if you are interested in only screening for variables significantly affecting your response from those not, 2 level full factorial/fractional factorial experiments are adequate, however if you are looking forward to probe the response surface deeply you need to go for 2nd order or higher designs where each factor is studied at more than two levels, if you are trying to search for optimal performance settings of your process you need to couple such surface modeling techniques with suitable search direction choice and updating algorithms , these techniques come under RSM(Response Surface Methodology),however if you are trying to find set of process parameters that will make your process robust with respect to.noise variable levels fluctations, you can go with Taguchi's robust parameter design methodology.