How to make it temperature-independent??

[Question]

I want to design a bandgap circuit what can generate 2.1V voltage. But the core of bandgap can only generate 1.27V voltage, so I use a output buffer as the schematic below.

The output of bandgap core is temperature-independent. It vary only 0.002V (from 0 to 70 centigrade). But the output of buffer will vary 0.04V (from 0 to 70 centigrade).

How can I design a buffer with better temperature-independent performance?

Thanks a lot.

 

[leebluer]

i am not so sure. but i think the problem maybe come from the temperatrure coefficent of the threshod voltage of M1 & M2, assuming that the buffer is ideal. perhaps the TCs are not uniform, which arouse a big temperature-excursion. you can simulate it, then find whether it is the reason. i also want to get the answer.

 

[kyrandia]

I don't think it comes from M1 & M2. I think it must comes from the opamp. As all know, the opamp have systemic offset, this will vary with the temperature. Also the gain vary with temperature too. The above two aspects can make your output temperature dependent.

 

[nkchenliy]

I think it comes from M1 & M2.If you replace M1 & M2 for resistor, temperatrure coefficent will improve.

 

[sunking]

The offset and M1 M2 may all the reason.

 

[Humungus]

I think that both temperature coeeficients in M1 and M2 and offset voltage of OpAmp are responsible for the variation of the voltage reference with temperature.

As somebody else said, use resistors instead of transistors in your amplifier. Minimize "systematic" offset. Random offset can also be reduced by some chopping techniques.

Another approach would be to design a current bandgap, so that your output voltage may be whatever you want.

 

[kyrandia]

I made a mistake by thinking M1 & M2 are uniform. Taking place of them with resistor is a good idea. Systematic offset can be reduce by carefully design. The random offset can be reduce by design and layout. Also chopper can be adopted(but is it need for just a bg?)

 

[Question]

I think the problem come from the M1 and M2. Becaues if I ues the Op Amp as a unity-gain buffer, the output of buffer is temperature-independent. So I will replace M1 & M2 for resistor.

Who can give me some advice of which kind of resistor is better? I use TSMC 0.18um process. I think here need big resistors, but they also need more area.

Thanks!!

 

[wylee]

If you are using PMOS transistor as voltage divider (which it is for your case), make sure their sizes are identical.

Example: W1=W2, L1=L2 hence M1=M2 to minimize the effect of mismatch

 

[nuiscet]

I think the problem is just the temperatrure coefficent M1 & M2 . To acquire the output of buffer temperature-independence, the best way is to get another contrary temperature-dependence signal (For example, replace M1 & M2 for resistor in another similar circuit), Then the sum of two signal can have good temperature-independence.

Added after 1 minutes:

For the resistor has contrary temperature-dependence contrasting the MOSFETs.

 

[suria3]

Hi,

If you use resistors to replece the PMOS, make sure the value of each of them is the same. This is important in order to keep the same ratio as the resistor value may vary to process and so on.

So, as we know the Vout = ( 1 + R1/R2) * Vin, when we keep the same ratio, we will get almost the same voltage at the output even for the process variation as long the opamp gain not vary much which result in offset error.

Thanks,
Suria3

 

[bageduke]

I think the problem come from the M1 and M2. Becaues if I ues the Op Amp as a unity-gain buffer, the output of buffer is temperature-independent.

Even if the opamp configed as unit gain buffer, the offset voltage can still reflect at output side. I think you'd better simulate your buffer at unit gain configuration and to see how big the offset is and how offset changes with temperature sweeping.