Temperature response of BGR

[diemilio]

I'm designing BGR using a very well known topology (Fig 1). When I look at the current vs. temp behavior through the branches connected to the positive and negative inputs of the correction amp, I get the expected response (red trace in Fig 2), but when I check the mirrored current to the output the temperature performance is severely degraded (blue trace in Fig 2)!!

I'm using ideal resistors with no tempco and the size ratio of the top PMOS is 1:1. The design is being implemented in 0.18 um and I've implemented a very similar topology in 0.5 um and didn't see this happening.

Can someone give me an idea of where this is coming from??

Thanks,

diemilio

 

[keith1200rs]

I would guess it is due to the poor performance of a simple current mirror, assuming I understand your question correctly. A simple current mirror will only give exactly 1:1 when the source-drain voltages are identical. You can probe the drain currents of the two transistors of the current mirror to check this. You could try a cascode mirror.

Keith.

 

[diemilio]

Yeah, thanks... I just noticed that... the problem is the high impact the temperature dependence of vds has on my currents... In the two branches tied to the amp I don't see the effect because their vds match, but at the ouput is a different story...

I didn't noticed this was the problem cause I was matching the output's vds by adjusting the output resistor, but obviously as temperature drifts the condition won't be valid anymore.

 

[Yanhui]

R load too large?
Maybe too much voltage consumed at resistor.