Question for BANDGAP experts:

[jcpu]

It's about die to die varaiation of VBG:

We know major VBG variation come from
(1) MOSFET mismatch from current mirror & error amp input stage
(2) Vbe mismatch
(3) resistor mismatch

Question is: some trying to reduce the effect of Vbe mismatch
by increasing the BJT ratio up to 12:1 or more.
Does that idea really useful?
Anybody has experience about this?

Thanks in advance.

 

[chungming]

bjt:
8:1 for layout matching

 

[eng_Semi]

to decrease the effect of the vbe mismatch, use 2Vbe as ur reference. It is written in Razavi book

 

[sissi]

also to decrease the effect of the Vos of amp ,It is written in Razavi book.12:1 is bad,

Added after 5 minutes:

we usually use 8:1,24:1,64:1. BJTs is 3*3, 5*5 and 13*5 array

 

[jet_xiang]

Hi,jcpu
which structure of your BG? I have designed many bandgaps for different process and difference structures. if you use error amplifier, the mismatch of AMP is most influence. do mente analysis you can get the result that the error amplifier input devices mismatch is the bigest. so you coult increase the L and W.
the current mirror, BJT and layout mismatch should be included.
BTW you can use BJT 1:8 or 1:15 not 1:12 and the different process will bring out different results.

 

[lunren]

One guy told me that, bigger bjt ratio really improves precision of the bandgap output. His experience is: with 8:1, 1 sigma voltage is about 3mV, with 16:1, 1 sigma voltage is about 1mV.

 

[smartdream]

using 8:1 or 16:1 are better for layout.

 

[szekit]

the larger vbe area ratio together with bigger resistor ratio will give you less systematic mismatch since it can minimize the effect of the input offset voltage of the error amplifier to the bandgap voltage.

put an offset voltage across the two input of the opamp and write up some equations to analylze the bandgap voltage as a function of area ratio of the bjt and resistance ratio of the resistors.

meow