my bandgap is 1.2v, but has a upper temp curvature , it seems there is
something wrong with mirror Ma & Mb, but i am not understand why this happens and how to solve it? And is this curvature bandgap works well?

Channel length modulation on Mb? What is the length for Ma and Mb?

Who knows, but this often happens when you run the bipolars at too high a current density. The drop across the base resistance causes the upward curvature. You can fix it by using bipolars with larger area.

Opamp offset temperature dependence could also have similar effects. Since you divided down the "signal" the circuit will be very sensitive to opamp offset. Run a simulation with the difference between the two terminals of the opamp and see the variation. Most of bandgap design is choosing the proper current density and building a VERY good opamp. Most bandgaps fail because people underestimate the importance of these things.

rg

PS: I am assuming you are using the same resistors throughout the design. Different resistors mess everything up.

PS2, I came up with that circuit about 1 month before it was published in the red-rag! Isn't that funny how often that happens?

xwj623 wrote on Jul 27^th, 2009, 6:26pm:

There is nothing wrong with using 4:32 or 1:8, but too high or too low a current density will cause errors like you have shown. A proper test would be to bias both the 1x and the 8x transistor with the same current and check the difference in voltage over temperature. The difference in voltage between the two bipolars should not depend on the current (it should be nkT/q*ln(8)). So start at a low current (like 0.5uA) and keep increasing it until the difference in voltage starts to change from nkT/q*ln(8). Bias the transistors somewhere close to where the voltage starts to change.

[/quote]
1. i use the same res type whihc has a negative temp coefficient.
[/quote]
a very large negative temp-co resistor will turn the curvature upward. To test this, try setting the temp-co to zero or use a different resistor type

Quote:


Plot the difference in voltage between the input terminals of the opamp over temperature and post it here. Also tell me what the ratio of R2a to R2b is. If you do that I might be able to explain it to you.

rg

RobG wrote on Jul 28^th, 2009, 7:30am:

1. i use the same res type whihc has a negative temp coefficient.
[/quote]
a very large negative temp-co resistor will turn the curvature upward. To test this, try setting the temp-co to zero or use a different resistor type

Quote:


Plot the difference in voltage between the input terminals of the opamp over temperature and post it here. Also tell me what the ratio of R2a to R2b is. If you do that I might be able to explain it to you.

rg [/quote]

Power supply is 1.2v ,vref is 0.6v.
R1 is about 28k,
R2a=R2b is about 140k.
Ma:Mb=1:1.

i want to know the upward curvature whether due to the high current density of PNP or due to negetave  resistor?
the negetave resistor  coefficient is about  -145 ppm/C.

thank you.