Bandgap Current Reference

[wjaachen]

bandgap current reference

Hi,

I have question about bandgap.

How to design a bandgap current reference with resistors with negative temperature coefficient?


Thank you!

 

[mpig09]

bandgap current reference

dear wjaachen:

do you mean you want a negative temperature coefficient of current?

mpig

 

[wjaachen]

bandgap current

no, I need a current independent of temperature.

It is easy to have a current relatively constant with temperature because resistors typically have a positive TC. (in the book of Thomas H.Lee)

But I have only resistors with negative TC available.

The current of Bandgap is heavily dependent of temperature.

What can I do?

 

[PaloAlto]

bandgap current reference

The current of Bandgap is heavily dependent of temperature.

That is not true. It is the case in the typical bandgap structure, but if you understand how a bandgap works, you can make anything you want. Remember: A bandgap is a linear combination of vbe voltages You can add these voltages any way you want it such that you get the temperature dependency that you want either for V and I.

OK. The most typical configuration is to subtract two vbe voltages with a certain ratio. That gives you a voltage that is PTAT, Proportional To Absolute Temperature. Then you add to that another vbe voltage (negative temp coefficient) and you end up with a voltage that is independent of T. You can see that in your circuit there are voltages dependent on T, on 1/T and independent. Of course, if you put the flat voltage over negative T resistors, the current that you get acts the same way vs T.

Now, what do you want?

* A current that is independent of T and you don't care about the voltage
* A current that is independent of T and a voltage that is independent too

If it is the first case, what you have to do is to make the linear combination such that the current is flat with T and not the voltage. You make that changing the ratios of the resistors (Rptat, Rvbe, Rref...)

If you want both to be independent of T, you first make usual bandgap. Then you have a voltage that is flat with T and a current that depends on temp. Now look for a voltage in your circuit that has opposite t coeff than your current. Use a small amplifier to force that voltage over a resistor. Add the two currents. What you have to do is to change the resistor value until the current is also flat

 

[rfsystem]

band gap current

I would like a little different approach but i depend how big is the negative "tempco"

If the instance the negative temperature coefficient of your high ohmic poly resistor is -800ppm/k you need to trim a specific bandgap for this purpose to -800ppm/K too. That is simple by adjusting the kT-multiplier which add the positive temperature voltage Vt to the negative temperature coefficient voltage Vbe.

Then use an opamp which regulate the output voltage through the -800ppm/k resistor to this -800ppm/K bandgap voltage. Use a mirror current source to get the "zero tempco" output current.

 

[wjaachen]

bandgap vs current

[quote="PaloAlto"
If you want both to be independent of T, you first make usual bandgap. Then you have a voltage that is flat with T and a current that depends on temp. Now look for a voltage in your circuit that has opposite t coeff than your current. Use a small amplifier to force that voltage over a resistor. Add the two currents. What you have to do is to change the resistor value until the current is also flat [/quote]

Thank you very much!

I have tryed it, but the current is now heavily dependent of corners (e.g. slow or fast).
It would be helpful if there are some papers about bandgap current reference.

Thanks again!

 

[PaloAlto]

poly resistor tc negative

No matter how many papers you look for, that'll always be the case. The current is V/R and R is internal with at least 10% variation over corners, therefore, your current will always have 10% variation.

The only way to solve that is to use one extra pin and use an external resistor to generate the current. That's common practice too.

 

[ytliang]

poly resistor negative temperature coefficient

You would need a voltage that is CTAT. If your voltage and resistance both decrease with temperature at the same rate, you would get a temperature independent current.
To generate a CTAT voltage, VBE of a BJT would be a good start.