Physics of Semiconductor Electronics

No device has an infinitely fast response to an input, so the rise and fall times are specified for each one. As the variable frequency drive signals increase in frequency and shorten in duration, more time proportionately is spent "in transition" rather than at one or the other power rails. While the device is in transition, it will need to dissipate additional energy as heat, and raise the operating temperature locally.

Pretty much all semiconductor devices are manufactured through masking a dopant of opposite polarity to the semiconductor layer or substrate, and then fired at relatively high temperature (around 700C) to allow the dopant atoms to diffuse into the semiconductor. After cooling the dopant material is removed, and the diffused atoms remain "frozen" in place in the crystal structure because diffusion has essentially stopped at room temperature.

This sets the stage for the effect frequency and pulse duration have on the lifetime of a power semiconductor. Higher frequency leads to higher temperature leads to greater diffusion of dopant because there is no longer a source for dopant atoms, and the original dopant never got past the thickness for the semiconductor characteristics desired. There remains additional semiconductor layer depth that has not been doped, so a natural force will cause the diffusion to extend the doped depth deeper and to a lower concentration.

Both of these will change the device characteristics, usually for the worse when it comes to speed and power dissipation, and are irreversible if and when the temperature is lowered. This thermal degradation along with other diffusion processes like oxidation eventually leads to device failure, the MTBF discussed above being an average.

There are two major effects that follow: one is that in continuous operation the device can reach a point of "thermal run-away" where decreasing performance leads to increased power dissipation that feeds upon itself until the semiconductor becomes a conductor and results in catastrophic system failure. The other is a predictable decrease in overall system lifespan, which may be lower than the manufacturer's specified MTBF.

For designers, the bottom line is two-fold: insure that the maximum frequency of the system remains well below the device manufacturer's switching specifications, and double-check the thermal power dissipation path to ensure that it is sufficient to cool the power devices under worst case conditions that get specified for the system as a whole.