How to design a current reference with zero TC?

[staric]

where can find a current reference circuit with zero TC? I read some papers. The method is to find the sum of PTAT current and invert PTAT current. And Is there any other method?

 

[nxing]

Most of the circuit use that way, otherwise, you need an off - chip resistor, which is very precise.

 

[Humungus]

Current references with zero TC is nearly imposible. Why? Because you always generate a current based on a resistor which has a finite non-zero TC. You can try making the resistor with two materials, one with positive TC and the other with negative one. Then, preview a triming mechanism and trim your circuit at the test stage.

 

[xiongshoufen]

i agree with Humungus very much!!

 

[bastos4321]

Current references with zero TC is nearly imposible. Why? Because you always generate a current based on a resistor which has a finite non-zero TC. You can try making the resistor with two materials, one with positive TC and the other with negative one. Then, preview a triming mechanism and trim your circuit at the test stage.

I would say it is realy impossible. You always have some temperature dependency in the circuit using any kind of components even if very precise.

Bastos

 

[staric]

Current references with zero TC is nearly imposible. Why? Because you always generate a current based on a resistor which has a finite non-zero TC. You can try making the resistor with two materials, one with positive TC and the other with negative one. Then, preview a triming mechanism and trim your circuit at the test stage.

But when the two type resistor are be changed to different corner(fast and slow) in the process, the current ref. value will shift a lot.

 

[qutang]

using a curve TC compensation.

 

[Ruritania]

I think Zero TC is just like a wish, a TC about 10ppm might be more realistic.

Ruri

 

[rfsystem]

Take this route:

1. Build up a bandgap reference. Typical VBDG=1.25V

2. Scale the VBDG with the VT multiplier resistors so that the resulting VBDG>1.25V or VBDG<1.25V. If the voltage is below the temperatur coefficient is negative. If it is above it is positive.

3. Select a resistor from the process with low temp coefficient.

4. Scale the VBDG so that both temp coefficents match

5. Use a mirror configuration to feed the selected resistor and your wanted output current

5. Built a regulator which drive the mirror so that the volatge drop across the selected resistor match the scaled bandgap voltage

Some principles are so easy that you can concept/draw/dimensioning/verify circuits within hours.

 

[Hughes]

Current references with zero TC is nearly imposible. Why? Because you always generate a current based on a resistor which has a finite non-zero TC. You can try making the resistor with two materials, one with positive TC and the other with negative one. Then, preview a triming mechanism and trim your circuit at the test stage.

But when the two type resistor are be changed to different corner(fast and slow) in the process, the current ref. value will shift a lot.

So trimming is needed, which Humungus mentioned above.

 

[JPR]

Generating a current with zero TC (over a given range) is not too bad. As mentioned by other posters, you can use a bandgap type circuit that is intentionally off from the sweet spot in order to get a voltage with a tempco that matches your resistor tempco. (Or match resistor tempco to a voltage tempco. I have seen references where resitstors of differing tempcos are stacked to get a PTAT tempco, then a PTAT voltage is applied, and you have zero tempco...)

The difficulty is that the zero tempco sweetspot is not the same from part to part. This means that without VERY careful design (I have not seen one), you must select between either zero tempco OR a specific current at a specified temperature.

Unless you are very careful in your design (and probably need a trim, too), you will not be able to get BOTH a zero tempco and a precise current level. If you always get zero tempco, you will likely have a wide distribution of current levels. If you always get the same current, you will likely have a wide distribution of tempco.

This is why bandgap voltage references are so nice. If you hit the right voltage, you will also be at the tempco sweet-spot.

If you can include either a trim (or two?) or an external resistor (which, in essence, is a resistor that has been trimmed by sorting), then obtaining a zero tempco at a precise current level becomes fairly straightforward. You are simply matching tempcos between a resistor and a voltage (tempco of the voltage and the resistor can be positive, negative, or even zero). When you have the right tempco, then you can have a current trim to get the proper current level out.

 

[layes2]

use bandgap generate zero TC voltage
and you can use a res from pin to get the current