HELP! Bandgap reference 0.18um

[yitch]

I am having trouble with this circuit.

From the paper 'A 2V 23 uA 5.3ppm/C Curvature compensated CMOS bandgap reference' Leung Mok Leung 2003

I cannot seem to get it to work in 0.18um and I have no idea what to change... nothing I do seems to bring the curve to the correct inverted U shape..



Please help

 

[erikl]

I do seems to bring the curve to the correct inverted U shape..

The temperature curve looks like a first order temperature shift of a resistor. You use resistors of different types, which all possess different TCs, so they cannot match.

 

[erikl]

But from the paper, my understanding is that you have to use the two resistors on the right hand side to cancel out the effects of higher order variation

(the Vbe equation where it is the Taylor series VBE = a0 +a1T + a2T^2 + .... anT^n)

is my understanding of it wrong?

No, it is correct. Just the schematic picture isn't very well legible ...

... so I thought R4 is an rpoly.

Still, I think the resistors don't match well. Did you double check your TCs and resistor ratios?

 

[yitch]

Hi Erik,

That is kind of the issue, I do not know where to look for the TC. And I have no idea how to calculate the resistor ratios. I know that the resistor next to the two BJTs are supposed to affect delta VBE which affects the PTAT and that the left hand resistor and the two right hand resistors are supposed to equal, apart from I am just lost.

Thank you so much for your patience.

 

[erikl]

Hi Erik,

That is kind of the issue, I do not know where to look for the TC. And I have no idea how to calculate the resistor ratios. I know that the resistor next to the two BJTs are supposed to affect delta VBE which affects the PTAT and that the left hand resistor and the two right hand resistors are supposed to equal, apart from I am just lost.

Hi Yitch,
as you know, you can optimize the bandGap's temperature behaviour only for a special process; for a different process - even a similar one - the BG voltage output vs. temperature curve would be different. So you'd need the TCs for all the used resistors. 2 possibilities: Either you get them from the foundry whose process you use (normally these data can be found in the process PDK), or you simulate the temperature dependencies of these resistors and extract the associated TCs. I think a first order evaluation will do for the beginning. Then calculate the series resistors to obtain a zero TC as good as possible.

I could give you the TCs for these resistors for a similar process, but it wouldn't work well for your process!

For the purpose of trying, you could also use resistors from the analogLib, which have a default TC=0, but you can assign to them any TC you like, even 2nd, 3rd order (which I wouldn't recommend to use).

BTW: for our bandgap reference (different structure), even for one process, we use trimming, not only for the absolute output voltage value, but also in order to shift the "inverse U" shaped curve to its maximum at 25°C. So we get 1.250V -6mV between -40°C and +125°C. But this trimming has to be done for every single chip! Actually, the trimming values are rather the same for all chips from the same wafer.

Now have fun with the BGA. You can put it to work!
Regards, erikl