Characterizing short-term variability of generated solar power is important for the integration of photovoltaic (PV) systems into the electrical grid. Using different kinds of high frequency, in-situ observations of both irradiance and generated PV power, we quantify insights on temporal averaging effects on the highest observed peaks and ramp rates, which closely relate to grid stability. We use measurements obtained at three specific spatial scales; a single point pyranometer, two household PV systems and a PV system typical for small medium businesses. We show that the 15-minute time resolution typically used for grid calculations significantly underestimates key dynamics at high temporal resolutions, such as ramp rates and maximum power output, at the local grid level. We find that absolute power peaks in the order of seconds are up to 18% higher compared to a 15-minute resolution for irradiance and up to 22% higher for a household PV system. For the largest PV system, the increase is limited to 11%. Furthermore, we find that the highest peaks solely occur under mixed-cloud conditions. Additionally, we show that the time interval-dependency of the largest power ramps is similar for all systems under research, ranging from similar to 20% at a 5-second interval to stabilizing at 70-80% between 5 and 10 min, which we can explain based on meteorological arguments.