regarding the impact test, spend some time bumping the machine in different locations while moving the accel. Watch the response in all three axis and eventually you will be able to determine a good driving point for the entire test. Once you find the driving point that excites all your measurement points in each planer you can use that for the fixed location.
I usually use 400 lines for a 200hz fmax. that gives you frequency resolution of 30 lines which is more than adequate resolution and reduces collection time considerably. use more resolution if your impact response data suggest closely spaced natural frequencies.