about design of precision current reference

[wonbef]

hi,all,

now i want to design a precision all-cmos current reference, appicated in a larger temperature range (0~200°C). a low temperature drift required. pls recomment some paper for me! need your helps. thanks in advance.

regards,

 

[steer]

now i want to design a precision all-cmos current reference, appicated in a larger temperature range (0~200°C). a low temperature drift required. pls recomment some paper for me! need your helps. thanks in advance.

I doubt CMOS made of Si works at temperature 200C. You might try silicon-carbide (SiC) to get it work, but this is a non-standard process.

 

[OpAmp]

Hi,

Design of a precision voltage reference is a trivial work by using a bandgap voltage reference. But, there hasn’t been a good solution for designing a very precise current reference. Usually designers use a off-chip resistor to adjust the current of the on-chip current reference. I previously tried to find a solution for designing a very precise on-chip current reference, but I wasn’t successful.
Recently, a solution has been proposed in the following IEEE paper. I haven’t tried it, but it seems to be a very nice solution.

OpAmp

 

[wonbef]

hi, OpAmp,

thank for your reply! i had read the paper you attached before, i think it is complecated. in fact, in my design, the low temperature drift is the most important, current can range from 35uA to 45uA. can you recomment a simple all-cmos design for me! thanks.

kind regards,

 

[wonbef]

can the micropower all-cmos design structure (current < 1uA) be used here? but i failed.

 

[OpAmp]

Hi,

If you use an all-CMOS current reference, like gm-constant biasing, depending on the process variations of resistors in your circuit, the reference current will change considerably. For example if the current setting resistor reduces -20%, the reference current will increase about 56%, And also I think the current is dependent on temp.
A better approach is using the bandgap circuit described in the following paper. In the circuit, current is independent of temperature and just dependent on resistor value. But if current setting resistor reduces -20%, the current will increase just 25%. I have used that circuit in a CMOS technology and it works very well. The problem is that the circuit is patented and if you work for industry, you might have problem for using the circuit.

 

[okguy]

Hi,

Design of a precision voltage reference is a trivial work by using a bandgap voltage reference. But, there hasn’t been a good solution for designing a very precise current reference. Usually designers use a off-chip resistor to adjust the current of the on-chip current reference. I previously tried to find a solution for designing a very precise on-chip current reference, but I wasn’t successful.
Recently, a solution has been proposed in the following IEEE paper. I haven’t tried it, but it seems to be a very nice solution.

OpAmp

This paper is nice, but to avoid opamp offset to disturbe the accuracy of your source accuracy, you need several mA ... not uA :!:

 

[michaelhan]

i agree with the off chip resistor solution for process error;
many OLED chip take the off chip resistor as a solution for precsion current

 

[avt]

hey guys - I tried to elaborate a little bit the approach suggested in the paper mentioned in posting 3 (of Opamp).

Anyway - when I assume a current mirror ratio of 1:1 and just use formulas (1) and (2) in order to eliminate (Vgx - Vt) - I get :

Iout^2 + (1-2*K)*beta1*Vr^2*Iout + beta1^2*Vr^4/4 = 0

I mean there are only a slight differences :

- (1-2*K) instead of 2*(1-K)
- beta1^2*Vr^4/4 instead of beta1^2*Vr^4

but htis makes a huge difference - because now there is no way to get easily rid of the sqrt-term when solving the equation - and there will be an ongoing dependence on K. Furtheron there will be too valid non-zero operating points.

OK - could someone please tell me under which assumptions one/the author was able to calculate (3) and afterwards (4) (which would be no problem then ...)

thank you (otherwise this would be one "funny" IEEE-paper and the remark at the bottom of the first page would be quite interesting )

 

[edwintsu]

If you assume the current mirror ratio (Id1/Id2) of 2:1, when K=2

u get the differe formulas (4) Iout=¼•β1Vr²

But the ongoing derivation is right, and the same.

B/R
Edwin.