resistor for current reference

[jutek]

hello

The picure presents references currents depending on temperature.

Typically it sould be close to 10uA.

One uses a pdiff resistor 2um wide and 150u long, the other uses rpolyh resistor 2um wide and 15u long.

As we can see the second dependence is stronger. Isn't that a problem to use pdiff resistor? I'm using AMS 0.35um pdk.

I've peeked at model and rpdiff has only one temperature coefficient for all corners (TC1=1.500e-03). It's seems to be strange for me

 

[pkalavakuru]

Hi
use both type of resistors and run ur sim u'll see better results

 

[jutek]

Hi
use both type of resistors and run ur sim u'll see better results

as you can exactly see i did so, but my question was different. Won't that be any problem if i use this pdiff resistor and why in the original model TC1 is the same for all corners

 

[pfd001]

it si a simple resistor model

 

[elantra]

check your model.
normally fabs provide very simple model for res and cap, so if possible you should add some other parameters for more accurate results such as related diodes and caps.

by the way you result seems quite good for me, you know, sometimes an accurate bias current is very differcult to get inside the chip.

 

[DenisMark]

It's normally describe pdiff resistor like jfet (many fabs do it), because resistance of pdiff depend on n-well potential where pdiff is located. Verify model pdiff. May it be jfet.

 

[johnsmith101]

It seems to me from your plots that the pdiff resistor varies much less over temperature, but it is much more nonlinear. If you need a constant current over temp, I would use the polyh resistor current and sum it with a current that varies inversely with temperature. The inverse with temperature current can be generated using a basic bipolar circuit. If you're using a pure CMOS process, you can use parasitic pnps.