Very strange problems about bandgap

The first tapeout ICs hrpoly resistor's unit value is 1.7k/um, the bandgap value is about 1.20V to 1.23V; and the second tapeout I only changes the metal 2 and metal1 mask, the bandgap block doesn't make any changes, but unluckily the process changes mainly that hrpoly resistor's unit value changes to 2k/um, so all the resistors' value used in bandgap block increase. When the second tapeout ICs comes back, the bandgap's value changes to about 1.23V to 1.26V.
I really can't understand why the result is so because the R1/R2 ratio can cancel the resistor's unit variation.(The simulation also shows that). Of course, the bandgap PTAT changes smaller, but does it cause the problems? Or any other reason? Thank you for your instruction.

For untrimmed bandgap 5% vref variation due process variation is normal. R1/R2 is not a reason for variation of "bandgap's value about 1.23V to 1.26V" for ur case. U must keep in mind that parameters of diodes changes too, that is real reason.

I think also that contact resistance value and tempco will have its effect on the bandgap voltage. Mismatches can also be change the bg value

It should be the diode parameters variations

to sum up . bandgap use parastic Pnp device
and consider Resistor variation , no trim bandgap will drfit some voltage ..

another problem is "density or concentration" I ever met , from WAT restsor data is Ok , but really bandgap design use 2 long channel resistor . and cause resistor value mismatch . WAT only measure little square resistor
but in general , we use "many square" for bandgap resistor circuit .

I think there are two reasons:
1. Parasitic PNP characteristc change.
2. PTAT current change due to resistor value change. This will lead to little change in Vbe.

laglead said:

I think there are two reasons:
1. Parasitic PNP characteristc change.
2. PTAT current change due to resistor value change. This will lead to little change in Vbe.

Click to expand...

This brought me another question about BGR. It's said that the size of diode to form Vbe in Vbgr can be arbitary. But with same current, different size of diode will generate different Vbe, and this adds up with PTAT current across resistor, the final Vbgr will be generated slightly different. Is this correct? Or I missed some points.

Thanks,

bageduke said:

laglead said:

I think there are two reasons:
1. Parasitic PNP characteristc change.
2. PTAT current change due to resistor value change. This will lead to little change in Vbe.

Click to expand...

This brought me another question about BGR. It's said that the size of diode to form Vbe in Vbgr can be arbitary. But with same current, different size of diode will generate different Vbe, and this adds up with PTAT current across resistor, the final Vbgr will be generated slightly different. Is this correct? Or I missed some points.

Thanks,

Click to expand...

Hi, Bageduke,
When diode area is different, their current density will also change (same current). This may change the Vbe(T) curve a little and will lead to different R1/R2 ratio. This effect togather with Vbe wmay change the final Vbg.

Thanks, laglead

So final Vbgr does relate to the area of diode? Then why doesn't this relationship reflect in the equations we derived for bandgap reference? I mean dVbe/dT+d(nVt)/dT=0.

And as I remember, dVbe/dT is independent of diode area, true or not?

Thanks,

hi, bageduke,
The final d(Vbe)/d(T)@T0=(Vbe-Vgo)/T0+(α-γ)(k/q).
Current density Jc≈T*exp(α). This α will affect Vbe value and then final Vbg value.
Regards