Hi All,

I'm trying to design a driver with 3 states (on, off and Hi-Z). I have a PMOS on the high side and an NMOS on the low side. If I have my PMOS threshold voltage at a lower point than my NMOS threshold voltage then I have a high-Z point in the middle of the voltage range. I simulated this in LTSpice and it works correctly. For example with a supply of 5V and PMOS threshold of -3V and NMOS threshold of +3V then I have 3 states based on a drive signal of: 0-2V (on), 2-3V (high Z), 3-5V(off).

The problem is when searching for a PMOS and NMOS candidate the threshold voltages specified fall into a very wide range (1-4V or 2-3V or 3-5V etc). Is this range representative of manufacturing tolerances or temperature effects or something else or all of the above? Should I assume that the min and max values are +/-3 standard deviations of manufacturing tolerances (and follow a normal distribution) and will be relatively fixed for the life of the mosfet or is it going to drift around all over the place?

Are there more precise mosfets threshold voltages out there (I haven't found any yet, perhaps they have a different name?) or is it an intrinsic limitation when making/using mosfets?

Thanks in advance for any info you might have.

 

It's a normal manufacturing tolerance and you should not try to design to that.

A better way to make a Hi-Z output is to have a separate logic circuit to put the PMOS gate at the high voltage rail and the NMOS gate at ground, insuring that both transistors are fully off.

 

Tri state gates require complicated circuits compared to bi state gates, so tri state gates aren't popular. They can be made, but they will be complicated because the threshold problems. Why do you want to make a tri state gate?

 

Tri state gates require complicated circuits compared to bi state gates, so tri state gates aren't popular. They can be made, but they will be complicated because the threshold problems.
........................

Tri state gates may be more complicated, but the standard designs do not suffer from threshold problems, whatever that is.