Temperature coefficient of the resistor in ic design

[wing0]

Can someone tell me how do i know which resistor has low TC and which has large positive TC?

From the resistor model document file there are 2 types of TC (TC1 and TC2).
So how do i know the resistor TC?

And how do i know which types of resistor i should choose?
the resistor i will use it to design high precision bandgap.

Thank You.

 

[karesz]

Hi,
Check pls for pp9---11!:
>"Temperature accuracy (base on W/L=2/100)"

You will find such tolerances between i.e.
+3%...and 0.3% or -2.3% and -0.8%...
I think these is clear what kind of a value is smaller as TempCo?
What your circuit need, must know the designer of circuit, also you!_
sorry, & only one way to know what is correct tempco= lot of calculations, or use simulator...
K.

 

[wing0]

what i mean is that based on the resistor model document, how do i know which resistor have negative TC or low positive TC or large positive TC?

i found the formula in bsim file:
tfac = 1 + TC01 x (CT - 25) + TCO2 x (CT - 25) x (CT - 25)
CT = temp
izzit by using this formula i able to know either the resistor have negative TC or low positive TC or large positive TC?

Thank You.

 

[erikl]

how do i know which resistor have negative TC or low positive TC or large positive TC?

The Model Parameter table (p. 4) tells you the TC values.

In your resistor models description, there is only one model with a negative first order TC1, the Ppoly Unsilicide, and it is that one with the lowest TC1 of all. So if you want a low TC resistor, I'd recommend to use this one.

If you want to do even better, then I'd suggest to use a resistor combination which cancels this TC1 as good as possible, i.e. use a serial combination of Ppoly Unsilicide (sbpi) and Npoly Unsilicide (sbni) with the inverse relation of their (absolute) TC1s (about 56 in this case), i.e. R(sbpi) = 56*R(sbni). By this you can achieve a relatively good first order TC cancellation.

Forget about the TC2's ; the models never are accurate enough to render a second order cancellation possible - at least not in the first shot. A lot of other local and asymmetry effects and process tolerance would prevent that anyway.

Nor will you have to care about the VC dependencies, because in a bandGap circuit you will have a rather stable voltage across your resistors.