what's the prolem of my bandgap?

[winsonpku]

I have taped out a bandgap circuit and the test results shows that the bandgap is failed.
Test reports show that the bandgap drifts about 200mV on one wafer
i mean test all dies on one wafer, the voltage drift about 200mV.
The bandgap is amp based.
the amp is a two stage amp, the first stage is folded-cascode and the second stage is a simple drive stage, current source load stage. Pmos is the drive stage.
The current flows the two bipolar is about 2.5uA seperately
and the amp consumes about 6uA current

so what is the problem about my bandgap?
The simulation results of my bandgap shows it could work properly
please help me and any advice is welcom!

 

[MikeR]

Are you can show some scheme and test point diagrams ?

 

[surianova]

Are you can show some scheme and test point diagrams ?

tis problem mostly due to the input offset of the opamp. The length of the input transistor need a few times bigger than minumim size to reduce the input offset.

other /might be layout.. do your layout have matching for the transistor suppose to match ?

 

[hung_wai_ming@hotmail.com]

Some of the possible failure points
(1) input mismatch from error amp
(2) resistance ratio variation, due to process and layout, most of the time, from layout
(3) BJT corner difference
(4) Too small size in length from error amp

Did u simulate all corners from different RES and different BJT models?

 

[winsonpku]

I have simulated all the corners for different RES and BJT models.
The channel length of the input transistors is 3u with 0.18u process.
Resistance ratio is 28:2. The 2 resistors are located in the center and other 28 resitors are matched also.

Some of the possible failure points
(1) input mismatch from error amp
(2) resistance ratio variation, due to process and layout, most of the time, from layout
(3) BJT corner difference
(4) Too small size in length from error amp

Did u simulate all corners from different RES and different BJT models?

 

[niezimei]

1. mismatch
2.input offset

 

[cliffzheng]

You can refer to the attached paper.
Device mismatch and opamp offset are usually some major issues for too large output voltag drift.

 

[dwayne22]

Bandgap circuits usually magnify the amplifier offset by a factor, usually 8 to 12. That means if your resistors are well matched (by using unit resistors), the amp offset is around 25mV.

Perhaps the layout is not truly symmetrical. Pay attention to the metal 'environment' over your amplifier input mosfets and also the resistors. The metal can have a big effect on the matching. If possible, avoid metal2 and above in the area of the amp and resistors.

 

[hung_wai_ming@hotmail.com]

winsonpku

I guess from your description, it comes from the mismatch of the two ratioed resistors, where the centre part does not match with the rest of the resistors. Most of the time, we may increase the small resistor by paralleling them. For example, you may have a resistor of 10k matching with 200k, a ratio of 20, in order to match them well, we will increase the finger number of 10k resistor. Say, each finger has 10k, and originally, you will have 1 finger in the center and 20 fingers surrounds it, but we will pair up the 10k by splitting them into two parallel 20k (parallel sum=10k) so that you have 4 fingers of 10k interdigitated in those 200k resistors to minimize the process profile effect.

I have faced similar problem before as well

 

[winsonpku]

I still don't think the misamtch will generate so much influnce on the circuit
The bandgap drift about 300mV on one wafer!!!!!!
Maybe so many small problmes produce the bad test results

winsonpku

I guess from your description, it comes from the mismatch of the two ratioed resistors, where the centre part does not match with the rest of the resistors. Most of the time, we may increase the small resistor by paralleling them. For example, you may have a resistor of 10k matching with 200k, a ratio of 20, in order to match them well, we will increase the finger number of 10k resistor. Say, each finger has 10k, and originally, you will have 1 finger in the center and 20 fingers surrounds it, but we will pair up the 10k by splitting them into two parallel 20k (parallel sum=10k) so that you have 4 fingers of 10k interdigitated in those 200k resistors to minimize the process profile effect.

I have faced similar problem before as well

Added after 2 minutes:

I will appreciate your help if you could explain the metal effect in details
How the metal could have a effect on the matching?

Bandgap circuits usually magnify the amplifier offset by a factor, usually 8 to 12. That means if your resistors are well matched (by using unit resistors), the amp offset is around 25mV.

Perhaps the layout is not truly symmetrical. Pay attention to the metal 'environment' over your amplifier input mosfets and also the resistors. The metal can have a big effect on the matching. If possible, avoid metal2 and above in the area of the amp and resistors.

 

[WindRay]

Hi

First of all, ensure that the probe result is not effect by any light (photo) emision because bandgap circuit is sensitive to light. If the result is correct, then I guess the problem is your resistor ratio.

With a high resistor ratio of 28:2, that means you are amplifying your opamp input offset voltage by (1+28/2).That means if you have 10mV offset at the input of the opamp, you will see a variations of 150mV at the bandgap voltage. That make your measurement very close.

 

[lxcpku]

Hi

First of all, ensure that the probe result is not effect by any light (photo) emision because bandgap circuit is sensitive to light. If the result is correct, then I guess the problem is your resistor ratio.

With a high resistor ratio of 28:2, that means you are amplifying your opamp input offset voltage by (1+28/2).That means if you have 10mV offset at the input of the opamp, you will see a variations of 150mV at the bandgap voltage. That make your measurement very close.

i did not understand why bandgap circuit is sensitive to light ? could give me more details ? thx a lot

 

[WindRay]

I may not have a good explaination, I guess light energy will absorb by electrons in silicon and they may change the bandgap voltage. I observed this issue when working on wafer probe test results and also in some cases when we do debugging on IC (that have problems... when we need to probe some reference voltages generated by bandgap).

Maybe someone can have a better explaination.

 

[babbage]

some guys' words are maybe right. But maybe u should give out the more detailed information about ur circuit and layout. Based on these, more better suggestions can be gave out.

 

[protose]

The OPAMP input offset influences the reference output greatly, if your BJT pair ratio is 8:1, then your Vptat=56mV, if the input offset is 6mV, then the ptat current may drift by 10%, this will induce ur bandgap output drift a lot. If your BJT pair ratio is smaller, the BGR will drift larger. As hung said, the unwell matched resistor also makes the BGR worse.
For ur reference.