Why RDSon has postive temp coefficent

[Aranion]

Hi. I am quite confused why Rdson has positive temp coefficient when semiconductors has negative temp coefficient. What's happening to the channel when temperature increases?

 

[cabraham]

With semiconductors, as temp increases, more electron-hole pairs, ehp, are generated due to increased thermal energy. Hence more electrons/holes are available for conduction. Hence for a given E field, more charges are out into motion at higher temp due to greater abundance of carriers. So semiconductor resistivity decreases as temp increases.

A FET, however does not rely on thermally generated ehp for available charge carriers. A FET relies on inversion. An n-channel FET is actually a p-channel substrate w/ n-channel drain & source. When the gate-source terminals are biased w/ an external source, charge polarization occurs. The p-channel substrate is literally flooded w/ n-type carriers, namely electrons. Hence electrons are the majority carriers & have high mobility, & move easily w/ an E field.

When temp increases, more collisions take place between electrons & lattice structure, inelastic in nature. This is resistance which increases w/ temp. This occurs in other semiconductors, but this property is small compared w/ the large non-linear increases in carriers due to ehp thermal generation.

Also, as temp increases, the substrate, naturally being p-type, generates more ehp, resulting in additional hole mobility which recombine w/ electrons. But electrons are the majority, so extra holes neutralize the effective number of charge carriers & resistance goes up due to net conduction electrons decreasing.

Normally, semiconductor conductivity is due to thermally generated ehp. But FETs use inversion by transporting electrons into the p substrate, overnumbering the holes, making electrons the new majority charge carrier. Hence the temp characteristic is markedly different.

Claude

 

[skeptic2]

Here is a reference that has additional material. Note in the third (indented) paragraph that FETs may have a nearly zero temperature coefficient. http://coe.uncc.edu/~dlsharer/ECGR3132/ReviewStuff/F8.pdf

Anecdote:
The company I was working for hired an electronic engineer from Russia. Shortly afterward we needed a circuit to indicate when a 5.3V mercury battery was nearly discharged. Mercury batteries have a characteristic of providing a very constant voltage over about 98% of their life. In the last 2% the voltage dies very quickly. The circuit we needed had to be able to detect and react to a voltage drop of about 0.1V.

All the engineers had the same idea, an opamp with a resistor and zener as a reference on one input and a resistance voltage divider on the other. It didn't work. The circuit had to work over the full industrial temperature range of -30C to 85C and the zener voltage varied too much over that range to work.

After all the engineers had given up, the Russian redesigned the circuit using an FET as a current source for a resistor. Since the FET had a nearly zero temperature coefficient, it succeeded where the zener had failed. Later he explained to me that in Russia at that time, zeners were expensive and the engineers commonly used FETs as constant current sources for voltage references.

 

[Aranion]

thanks guys. now it's making more sense..