# Subthreshold Slope

#### goodbyegti

Joined Apr 28, 2004
59
Hi,

I'm having a hard time getting a good definition of the subthreshold slope, in terms of a MOSFET.

So far the best explanation is:

"measure of the voltage required to switch the device on and off"

This is a bit vague, anyone know better?

Also, I have seen units in mV/dec

What the unit "dec"?

and..

Am i right in thinking a steep subthreshold slope is a good thing, i.e the FET switches on/off in a more discreet way?

Any help much appreciated!

#### n9352527

Joined Oct 14, 2005
1,198
If you plot log(Ids) against Vgs at a constant Vds voltage, for Vgs < Vth region the line would be fairly linear. The slope of this line is the subthreshold slope. It represents how fast the Ids ramps up for an increase of Vgs. Steep slope means the device gets out of the subthreshold region faster or in other words has lower Vth and therefore conduct better.

It is usually given as kT/q*log(1+Cd/Ci). Where Cd is the depletion layer capacitance and Ci is the gate oxide capacitance. Typical value is 70mV/decade. Different processes have different slopes.

The decade refers to a 10 times increase of the Ids.

• InterRupt

#### goodbyegti

Joined Apr 28, 2004
59
wow! thanks for that explanation - it makes perfect sense to me now!

• InterRupt

#### vishal sharma

Joined Oct 14, 2010
1
If you plot log(Ids) against Vgs at a constant Vds voltage, for Vgs < Vth region the line would be fairly linear. The slope of this line is the subthreshold slope. It represents how fast the Ids ramps up for an increase of Vgs. Steep slope means the device gets out of the subthreshold region faster or in other words has lower Vth and therefore conduct better.

It is usually given as kT/q*log(1+Cd/Ci). Where Cd is the depletion layer capacitance and Ci is the gate oxide capacitance. Typical value is 70mV/decade. Different processes have different slopes.

The decade refers to a 10 times increase of the Ids.
vishal sharma: i think that subthreshold slope is not given as kT/q*log(1+Cd/Ci), but it is given as n*kT/q*ln(10)=(1+Cd/Ci)*kT/q*ln(10) and the approximate value is given as 60mV/decade for n(subthreshold slope factor)=1.